DJing Discussion

This area is for discussion about DJing in general. Please remember the community rules when posting and try to be polite and inclusive.

MP3Gain and SSL Auto Gain comparison

msoultan 10:42 PM - 27 May, 2013
Hi,
I have some technical questions about the MP3Gain and SSL's auto gain feature that I was hoping the Serato guru's could help out on.

When using MP3Gain, it analyzes the song and based on the target volume, it will tell you if the track is going to clip. What I wanted to know is if SSL's auto gain feature is using the same (or a similar) algorithm to end up at a particular target volume.

The reason I ask is because I currently have my target volume in both SSL and MP3Gain set to 92. Let's say I have a particular track that I've analyzed with MP3Gain and it's telling me that a track will be clipping at 92, does it mean that same track processed by SSL with the auto gain set to 92 will also be clipping?

What I'd like to know is if MP3Gain is going to yield the same target volume as SSL's auto gain feature. What I don't want is that MP3Gain's target volume of 92 is pushing tracks towards clipping and SSL's auto gain is not, or vice versa.

Thoughts?
msoultan 1:32 AM - 28 May, 2013
So I did some research on a test track. I had a track that before applying 1.5db of gain, it was at 90.8 (not clipping), and after applying the track gain, it was now at 92.3 (clipping). Before applying the 1.5db of gain, I did some tests in sound forge using the "find clipping tool" using the following presets starting at 90.8 db:

detect all 0 db clipping: 0
detect only audibly noticeable clips: 0
detect only extreme clips: 0
detect small clips above -3 db: 0

after applying the 1.5 db of gain to get 92.3 db with mp3gain reporting clipping:

detect all 0 db clipping: 0
detect only audibly noticeable clips: 0
detect only extreme clips: 0
detect small clips above -3 db: 0

Well that's interesting - sound forge reports no clips but mp3gain does - let's bump it up another 1.5 db to 93.8 db:

detect all 0 db clipping: 100+
detect only audibly noticeable clips: 33
detect only extreme clips: 2
detect small clips above -3 db: 86

I listened to the "audibly noticeable" clips and I couldn't hear anything odd, so I bumped it up another 1.5 db to 95.3 db:

detect all 0 db clipping: 1000?+ (sf stops counting markers after 99)
detect only audibly noticeable clips: 600?+
detect only extreme clips: 100
detect small clips above -3 db: 1000?+

Still couldn't hear the clips, but they were all kicks and snares, so it was mostly transients that you wouldn't really notice anyways.


Now, the question is how SSL scans the tracks and if it ends up with the same clipping results. Unfortunately I don't really have way of getting the raw output from SSL so there's really no way of knowing unless we get some input from the Serato folks. I hope they're willing to chime in on this otherwise we're really just flying blind on whether we're clipping or not.

Next I'm gonna test this on a track with less transients (kicks and snares) and see what happens.
msoultan 2:05 AM - 28 May, 2013
Ok, testing with a classical track (Pachelbel's Canon In D Major (piano)) with longer sustained notes and less transients so hopefully I should be able to hear the clips:

93.0 db (no clipping in mp3gain):

detect all 0 db clipping: 0
detect only audibly noticeable clips: 0
detect only extreme clips: 0
detect small clips above -3 db: 0


94.5 db (clipping in mp3gain):

detect all 0 db clipping: 43
detect only audibly noticeable clips: 30 (noticeable!)
detect only extreme clips: 0
detect small clips above -3 db: 41 (noticeable!)

The clips above were definitely noticeable - not horrible, but there was a definite click and it didn't sound natural.


96.0 db (clipping in mp3gain):

detect all 0 db clipping: 200+?
detect only audibly noticeable clips: 300+?
detect only extreme clips: 10
detect small clips above -3 db: 300?+

clipping much more pronounced and not desirable at all!


Well, there you have it - very noticeable results with very little change in volume (only 1.5 db!)... but as the example goes to show, the results depend on the program material. If you're dealing with dance music, it's most likely not going to be noticeable over the noise of the instrument that's causing the transient peak, but if you have a classical song, be careful!


So, that being said, I'm actually thinking of turning off the SSL auto gain because you have no idea if a track is clipping (after SSL auto gain has been applied) unless you hear it, and by then it's probably too late and you heard the clips in your recording. Unfortunately there's no indicator to tell you if there are any clips like MP3Gain has.

At least with MP3Gain you can bump the target value up or down and see when tracks start to clip and depending on the program material, decide how much further you want to go. Plus, you can apply a value and see how it affects your entire library, something you can't do with SSL's auto gain. Another perk to MP3Gain, and a huge one at that, is that your waveforms will all be a consistent size, which is not compensated for using SSL's auto gain.

Lastly, it seems like MP3Gain errs on the conservative side because it was reporting clipping and sound forge wasn't, but sound forge was requiring at least 3 samples held at 0 db to report a clip, and MP3Gain might be using less - which is fine - the less clips the better and erring on the conservative side is going to result in better quality output!


As for what target value, I'm gonna start with 92 and see how it affects my library - I'm going to do a track analysis on 1000+ songs of varying styles (house, trance, d&b, trip hop, chill, etc) and see what I get. If I see dance tracks clipping, I'm not going to worry, but if I see ballads and other songs with long-sustained sounds clipping, I will lower the overall value.

Hope some of you found this useful because it definitely helped me clear up some of the mysteries of using MP3Gain and SSL's auto gain feature. I'll report back shortly when MP3Gain finishes analyzing all 1031 tracks.

Oh, and if I remember correctly, if you do use MP3Gain after you've analyzed your tracks, you'll have to reanalyze them.

Mike
Laz219 2:15 AM - 28 May, 2013
I use auto-gain as a rough start, then manually tweak my tracks from there as I play them.
It's a good base-point, but I wouldn't consider it a perfect solution.
msoultan 2:25 AM - 28 May, 2013
Honestly, I'd consider auto-gain to be far from a perfect solution for the reasons listed above. And if you're adjusting it by hand, then you really have no reference at all other than your ear, and that's only for the one particular portion of the song that you're listening to. For all you know you're bumping other portions of the track into the clipping zone and/or creating an overall imbalance between that track and others that have been adjusted with auto gain. But yeah, sometimes auto gain doesn't work quite right, and I've adjusted it, but now I'm going to be *very* cautious about adjusting up, especially if 92 is going to yield tracks that are close to clipping.

Btw, it looks like I've just found out that SSL's detection is in-fact different than MP3Gain and is definitely less conservative. I have both SSL and MP3Gain set to 92 and "Pachelbel's Canon In D Major (piano)" noticeably clips in SSL using only SSL's auto gain. This means that the detection algorithms between the two programs are different as clipping occurs with SSL's auto gain and not with MP3Gain's gain adjustment. This has pretty much sealed the deal for me to use MP3Gain instead of SSL's auto gain.

... still waiting on the MP3Gain track analysis to finish - about 1/4 of the way done...
msoultan 2:35 AM - 28 May, 2013
As another interesting sidenote, so far probably 98% of my tracks are show as clipping in MP3Gain and about 7% or so are shown to be clipping at the 92.0 db target value. That should be a good indicator that if you don't hear it clipping at a higher value (as it came from the artist), then you won't hear it at the lower level. The ones to watch out for are where it pushes the value up - most of my tracks are getting pulled down. That's another nice feature of MP3Gain is that you can sort by the clipping or Track Gain fields and instantly see what's going on. I can't believe I didn't play with this a loooooong time ago.
msoultan 8:23 PM - 28 May, 2013
So, after scanning all of my tracks (2156), I noticed something interesting that will affect what I set my target volume levels to. I sorted the list of tracks from loudest to softest and I'd say 5/6 of my entire libary is already clipping, with "No Doubt - Just a Girl" peaking out at a volume of 103 db. I wasn't the one that encoded that file so someone really cranked it up because it's got easily over 1000 audible clips - oddly, I didn't really notice the clips, but that's because the track is highly compressed with a lot of transients and like I mentioned before, whether a clip is audible or not depends on the program material.

Also, about 5/6 of the way is where songs with an original volume of 92 db ends and gets softer from there. I mention that because a lot of people use 92 for their SSL auto gain value (myself included), so if I were using 92 db for auto gain, I'd be pulling down 5/6 of my tracks and adding gain to 1/6 of them. That being said, songs that MP3Gain reports as originally clipping (before any adjustments are made) start to thin out around the 90 db mark and below that are less and less tracks that are clipping at their original volume. That said, those clips might not be audible - MP3Gain is just reporting that a number of samples hit 0 db and it thinks its a clip.

So the ultimate goal is to have all of the tracks end up at the same loudness but try and keep the volume as loud as possible without introducing NEW clipped tracks. Remember, if a track is already clipped, it doesn't matter because once a file is clipped, turning it down just lowers the volume, but if the clip was audible at 92, there will still be an audible sound at whatever lower volume you pick. MP3Gain is only looking to see if your waveform is hitting 0 db, not what it sounds like. In other words, if the track sounds like crap at 92 (clipped, distorted, whatever), it will sound like crap at any volume lower than that - the damage is already done. Fortunately, if you're messing around with MP3Gain and you cause your tracks to noticeably clip, it's not permanent. MP3Gain is only just turning the volume knob up and down and not editing the waveform. As long as you have recorded your perfect mix (with the included clips), you can just readjust your MP3Gain target value to something lower and be on your way.


Now it's time for some testing - we need to figure out what to set the target value to so that we avoid clipping tracks that weren't originally clipped while trying to keep the volume of all tracks as loud and consistent as possible. After a track analysis (I did NOT apply gain), I did some creative sorting at various target volumes and here are the results:

@ 86.0 db - 1 total clipping - 0 already clipping - 1 new clip
@ 87.0 db - 4 total clipping - 2 already clipping - 2 new clips
@ 88.0 db - 11 total clipping - 5 already clipping - 6 new clips
@ 89.0 db - 24 total clipping - 13 already clipping - 11 new clips
@ 90.0 db - 60 total clipping - 35 already clipping - 25 new clips
@ 91.0 db - 152 total clipping - 99 already clipping - 55 new clips
@ 92.0 db - 320 total clipping - 204 already clipping - 116 new clips

The first number is the target volume and the next numbers are as follows - the first number shows how many will be clipping after that target gain is applied, the second is how many of those tracks were already clipping, and the last number shows how many previously non-clipping tracks will now be clipping. So, for example, if I pick a target of 92.0 db, that means I will introduce clipping into 116 tracks that weren't previously clipping! I'd say that's a pretty significant amount, but again, it depends on the program material.

What's scary is the amount of tracks that are already clipping without any adjustments - in my case that was over 3/4 of my entire library. I didn't know this before, but after some research I found that when you encode an MP3, it will tend to increase the volume and could cause clipping due to the algorithms used to encode the MP3. It's not necessarily audible clipping, but there are parts of the track's waveform that are railing at 0 db for a few samples.


For kicks I decided to do another test on 850 different tracks - these were all ripped from CDs and are a very wide variety of tracks (Radiohead, Basement Jaxx, Orbital, Orb, William Orbit, FSOL, Jimmy Page, Michael Jackson, classical music, etc.). After analyzing them in MP3Gain, 2/3 of the tracks are already clipping. I then did the same tests as above and here are the results:

@ 86.0 db - 7 total clipping - 1 already clipping - 6 new clips
@ 87.0 db - 18 total clipping - 3 already clipping - 15 new clips
@ 88.0 db - 29 total clipping - 4 already clipping - 25 new clips
@ 89.0 db - 57 total clipping - 14 already clipping - 43 new clips
@ 90.0 db - 101 total clipping - 30 already clipping - 71 new clips
@ 91.0 db - 167 total clipping - 60 already clipping - 107 new clips
@ 92.0 db - 302 total clipping - 115 already clipping - 187 new clips

The results aren't horribly different from the other set of tracks, but it is scary to see how many new clipping tracks are introduced at the 92 db setting! That said, those clipped tracks might not have audible clips, but it is important to note is that the 1 track that was already clipped at 86 db is also part of the group at 92 db and has 6 db of gain added to it. I took that track (Mouse on Mars - Hallo), set it to 92 db and used Sound Forge to see if those clips were audible - MOST DEFINITELY!!

So right now I don't really mind new clips so much as the clips most likely won't be audible. What worries me is adding gain to tracks that are already clipping - as in the example above with the Mouse on Mars track, you have potential to really screw things up. For that reason, it looks like I'm gonna go for a target volume of 89.0 db. For the tracks that are clipping, it's not by a huge amount and the waveform in ScratchLive still looks nice. I will also be turning off SSL's auto gain feature as it's redundant and potentially harmful. I'm already doing the processing with MP3Gain and you really don't know the results that you're getting with SSL's auto gain feature until it's too late.

Hopefully all of the above helped - I know it helped me out immensely and I'm looking forward to some much better sounding mixes!!

Thanks,
Mike
Jeff Scott 8:42 PM - 28 May, 2013
I'm sure Seratos auto gain only applies to the setting in the program and the levels by each deck will show you whether or not clipping is happening, obviously you should be to hear it aswell.

You'v done a lot of testing and seem to be interested in getting the best sound possible, my advice is if you still have a SL1 as is showing in your hardware info is to get any of the newer interfaces as the sound quality and dynamic range are greatly improved!
msoultan 9:11 PM - 28 May, 2013
Quote:
I'm sure Seratos auto gain only applies to the setting in the program and the levels by each deck will show you whether or not clipping is happening, obviously you should be to hear it aswell.


Serato has an algorithm to detect the perceived loudness of a track and MP3Gain has an algorithm to do the same thing - so yes, SSL's auto gain only applies to the program, but it still is *in addition* to any pre-processing done in your DAW or via MP3Gain. If I remember correctly they vary very slightly in their detection methods, but ultimately they're doing the same thing. That said, I'd recommend using one or the other, but not both as you're just adding one more layer to your gain structure and causing yourself more confusion - in other words, if you really understood what was going on, you would know to only use one tool ;)

I'd also go as far to recommend that you use MP3Gain primarily because you get consistently sized waveforms (which SSL's auto gain won't do), you have notification of clipping at processing time whereas you only hear or happen to see the clip in SSL, you can see how the change affects your entire library in one glance with MP3Gain, and lastly it seems like MP3 gain errs on the conservative side which is definitely a good thing.

Quote:
You've done a lot of testing and seem to be interested in getting the best sound possible, my advice is if you still have a SL1 as is showing in your hardware info is to get any of the newer interfaces as the sound quality and dynamic range are greatly improved!


Changing the hardware has nothing to do with clipping and I'm pretty happy with the sound quality of my hardware. What I don't want is the distorted sound of clipping, and that's something that can be easily avoided without any added expense (other than time). I could have the best hardware in the world and if a track clips, it clips - the MP3 is still going to clip at 0db...
msoultan 5:48 AM - 29 May, 2013
So I'm not really feeling it at 89 as I'm raising my noise floor. I'm debating going up a db or two to 90 or 91 and see how that feels. It seems like a lot of the songs that are clipping tend to be ones with weird dynamics or very quiet with rare loud sections. As for all of my house music, it can handle a few clips here and there without consequence, and most of them aren't clipping anyways. I was looking at Platinum Notes as it actively adjusts the dynamics in the file, but then I'll have to re-add all of my cues and that's not something I'm willing to do right now.

I will post more when I have a better idea of what I"m going to do. For now I need to make a mix so auto gain it will be for the near future.
Serato, Support
Jamie W 1:48 AM - 11 June, 2013
Hey Msoultan,

Thanks for getting in touch,

Firstly, nice work with all your tests and research!

I personally would advise you use the SSL built in Auto Gain.
-- For any file that it does not calculate correctly, manually adjust the gain.

If you would like I can move this thread into the "General Discussion" area so other users out there can give you there feedback / recommendations?

Let me know if that sounds good to you.

Thanks :)
msoultan 1:57 AM - 11 June, 2013
Yes, please move it there as it would probably be more appropriate.

And besides the fact that you are a Serato employee, why would you recommend one over the other? The SSL autogain feature is essentially a black box as you really don't get a great overview of clipping tracks until you actually play the song back and *happen* to see it peaking - I think this is a huge disadvantage. Also, you don't get consistent waveform sizing. It would be nice if SSL took care of that.

Mike
Serato, Support
Jamie W 2:08 AM - 11 June, 2013
Hey Msoultan,

I have now moved this thread into the general discussion area.

I recommend the SSL Auto Gain because its built into the application so there is less chance of files becoming corrupt / unsupported.
-- This can happen when you put music files through other applications to re-encode / re tag the file etc.

Quote:
The SSL autogain feature is essentially a black box as you really don't get a great overview of clipping tracks until you actually play the song back and *happen* to see it peaking - I think this is a huge disadvantage. Also, you don't get consistent waveform sizing. It would be nice if SSL took care of that.


We are always working hard on improving our software.
In regards to those two feature ideas, they may benefit other users out there.
I recommend you jump into our forums "Feature Request" area. Start a thread about these features, if enough people out there support your idea, it may be something we look at implementing further down the track.

Thanks :)
2:09 AM, 11 Jun 2013
Discussion moved to Rane Mixers General Discussion
2:11 AM, 11 Jun 2013
Discussion moved to DJing Discussion
msoultan 8:41 AM - 13 June, 2013
Quote:
I recommend the SSL Auto Gain because its built into the application so there is less chance of files becoming corrupt / unsupported.
-- This can happen when you put music files through other applications to re-encode / re tag the file etc.


First, MP3Gain does not re-encode the MP3 and it's changes are non-destructive and completely reversible. Is it Serato's stance to not use applications such as Mixed in Key, Platinum Notes, MixMeister BPM counter, or any other pitch detection or tagging program because they edit the MP3? That just doesn't seem like a realistic argument.

I would venture to guess that MP3Gain has been out longer than than the above applications and has a huge installation base, especially because it's free. It is compatible with all MP3 players as it adjusts the volume of each frame in the MP3, so it doesn't rely on custom tags requiring the player to make the volume adjustments.

I still think MP3Gain provides more value to the DJ than Serato's autogain feature.
phatbob 8:55 AM - 13 June, 2013
All that faffing about, all that effort.

Then uses an SL1.

Oh dear.
msoultan 3:05 PM - 13 June, 2013
You'll get the same exact results with any audio interface - clipping is clipping, and it will sound bad even if you have the best interface ever.
s3kn0tr0n1c 3:57 PM - 13 June, 2013
Quote:
I use auto-gain as a rough start, then manually tweak my tracks from there as I play them.
It's a good base-point, but I wouldn't consider it a perfect solution.

+1
DJ DisGrace 4:02 PM - 13 June, 2013
Quote:
Quote:
I use auto-gain as a rough start, then manually tweak my tracks from there as I play them.
It's a good base-point, but I wouldn't consider it a perfect solution.

+1

+1
Unless you are using another DVS, I don't see the value in changing the volume of your tracks. The SSL auto-gain gets me close enough.
msoultan 4:11 PM - 13 June, 2013
The funny part about the above post is that it completely defeats the purpose of auto-gain. Auto-gain/MP3Gain change the volume based on apparent loudness in an attempt to get all tracks sounding the same volume and do a pretty good job of it. One of the problems with SSL's auto-gain is that it doesn't tell you anything about clipping caused by its process so the only way you find out about that clipping is during playback - which is a horrible time to find this out. At least MP3Gain will tell you immediately whether you're clipping the track.

If you really wanted to have consistent volume levels that don't clip, the above quoted post should really read:

"I use SSL auto-gain as a rough start, then if a track is clipping, I turn down the SSL auto-gain master adjustment to keep this track from clipping while keeping the track's volume consistent with all the other tracks in my library."

Were you to use SSL's auto-gain or MP3Gain in this manner, you wouldn't be tweaking individual tracks unless there was a problem with the SSL auto-gain or MP3Gain algorithm. The problem is not everyone really understands what's they're doing, hence the +1s above.
DJ DisGrace 4:15 PM - 13 June, 2013
My understanding is that SSL auto-gain doesn't actually change the track, it only adjusts the "input gain" of the track. If you see red on the track's output meter, just turn it down....

On the other hand, with a third party software that changes the actual file, you can turn it down all you like in SSL and it's still churning out flattened waves.
DJ DisGrace 4:17 PM - 13 June, 2013
Quote:
Were you to use SSL's auto-gain or MP3Gain in this manner, you wouldn't be tweaking individual tracks unless there was a problem with the SSL auto-gain or MP3Gain algorithm. The problem is not everyone really understands what's they're doing, hence the +1s above.

No. The algorithm isn't perfect, obviously, so some tracks still need to be turned up or down slightly in SSL after running auto-gain on them. In this case I only need to change the gain on a single track, not the whole library.
msoultan 4:20 PM - 13 June, 2013
Quote:
My understanding is that SSL auto-gain doesn't actually change the track, it only adjusts the "input gain" of the track. If you see red on the track's output meter, just turn it down....


Correct, but again, you've now negated the whole purpose of SSL's auto-gain feature. You should really be changing the master level so that the track's volume level stay's consistent with all the other tracks. The problem is that you're arbitrarily picking a volume level to which you're adjusting all of your tracks to. Unless you listen to every single track in your library and watch the meters, you have no idea which tracks are clipping.

Quote:
On the other hand, with a third party software that changes the actual file, you can turn it down all you like in SSL and it's still churning out flattened waves.


Turning up SSL's auto-gain too high will have the exact same effect. Whether you turn up the volume in SSL or MP3Gain, you're still running the same risk of clipping tracks. The nice part about MP3Gain is you can see the changes right away and choose whether or not you want those changes. In regards to MP3Gain, it adjusts the volume of each MP3 frame and is non-destructive, so if you went too hot, just go lower. But again, you won't be spitting out clipped tracks all the time as you wouldn't have made the mistake of picking the wrong master level from the get-go.
DJ DisGrace 4:25 PM - 13 June, 2013
Quote:
You should really be changing the master level so that the track's volume level stay's consistent with all the other tracks

No the master level changes everything, not the individual file that is still too low/hot after running through auto-gain.

Quote:
The problem is that you're arbitrarily picking a volume level to which you're adjusting all of your tracks to. Unless you listen to every single track in your library and watch the meters, you have no idea which tracks are clipping.

No. I do these adjustments at the club, using headphones and PFL meters on the mixer, then listening in the monitors. It's not rocket science to tell if a track is too quiet or loud. I managed to mix vinyl for years and keep consistent volume from record to record, it's a lot easier now that I only have to make that adjustment once, if ever.

Quote:
In regards to MP3Gain, it adjusts the volume of each MP3 frame and is non-destructive, so if you went too hot, just go lower.

And how would I do this in the middle of my set? At least with SSL I can adjust the level on the fly and be done with it forever.
msoultan 4:27 PM - 13 June, 2013
Quote:
No. The algorithm isn't perfect, obviously, so some tracks still need to be turned up or down slightly in SSL after running auto-gain on them. In this case I only need to change the gain on a single track, not the whole library.


Assuming that SSL didn't make a mistake, you're now making the apparent volume level of that track different from all of the other tracks in your library, which is defeating the whole purpose of auto-gain. If it did make a mistake, then you need to now listen to the current track and other tracks to make the volume levels consistent. It would be really nice if SSL's analyze feature spit out a spreadsheet with the volume levels so you could see how many tracks you're clipping by using a particular master auto-gain level.

That being said, I did realize another thing - MP3Gain only works on MP3s where I'm *guessing* that SSL auto-gain works on other formats - is that correct? So if you are mixed-format, then using MP3Gain would probably be more problematic than helpful.
DJ DisGrace 4:27 PM - 13 June, 2013
The Master level is totally separate from the individual track gain (which is adjusted by auto-gain). You seem to be confusing the two and talking about them interchangeably...
serato.com
DJ DisGrace 4:30 PM - 13 June, 2013
Quote:
Quote:
No. The algorithm isn't perfect, obviously, so some tracks still need to be turned up or down slightly in SSL after running auto-gain on them. In this case I only need to change the gain on a single track, not the whole library.


Assuming that SSL didn't make a mistake, you're now making the apparent volume level of that track different from all of the other tracks in your library, which is defeating the whole purpose of auto-gain. If it did make a mistake, then you need to now listen to the current track and other tracks to make the volume levels consistent. It would be really nice if SSL's analyze feature spit out a spreadsheet with the volume levels so you could see how many tracks you're clipping by using a particular master auto-gain level.

That being said, I did realize another thing - MP3Gain only works on MP3s where I'm *guessing* that SSL auto-gain works on other formats - is that correct? So if you are mixed-format, then using MP3Gain would probably be more problematic than helpful.

But some tracks do "sound quieter", regardless of what a computer algorithm or spreadsheet tells me. I adjust these tracks to 'sound' the same volume as the track that is playing, which sounded the same as the last track and the one before that. Now the track's volume in question is more in line with the rest of my library.

Changing the master will change all of them up or down, so the individual difference between tracks will still exist.

And yes, I mainly use mp4 at this stage....
msoultan 4:33 PM - 13 June, 2013
Quote:
Quote:
You should really be changing the master level so that the track's volume level stay's consistent with all the other tracks

No the master level changes everything, not the individual file that is still too low/hot after running through auto-gain.


Sorry- I should have said "you should really be changing the SSL auto-gain master level".

Quote:
No. I do these adjustments at the club, using headphones and PFL meters on the mixer, then listening in the monitors. It's not rocket science to tell if a track is too quiet or loud. I managed to mix vinyl for years and keep consistent volume from record to record, it's a lot easier now that I only have to make that adjustment once, if ever.


Yes, and like I just said in my previous post, if you adjust the level of that one track, you're now changing the apparent volume level of that track compared to the apparent volume level of the rest of your library. You can't use a peak meter on a mixer as it's showing peak level, not apparent loudness - these are two different things. You can have transient peaks all over the place and still have a relatively soft apparent loudness.

Quote:
And how would I do this in the middle of my set? At least with SSL I can adjust the level on the fly and be done with it forever.


That's the beauty, you don't need to adjust it in the middle of your set as you did the work ahead of time. You can still adjust the levels in real-time, but you're just going to potentially clip a track that MP3Gain told you was going to clip, but you can do it anyways.
msoultan 4:33 PM - 13 June, 2013
Quote:
The Master level is totally separate from the individual track gain (which is adjusted by auto-gain). You seem to be confusing the two and talking about them interchangeably...
serato.com


Yeah, I meant to say the auto-gain master - sorry!
DJ DisGrace 4:37 PM - 13 June, 2013
Quote:
Yes, and like I just said in my previous post, if you adjust the level of that one track, you're now changing the apparent volume level of that track compared to the apparent volume level of the rest of your library. You can't use a peak meter on a mixer as it's showing peak level, not apparent loudness - these are two different things. You can have transient peaks all over the place and still have a relatively soft apparent loudness.

OK, I kind of see what you are saying. But the thing is once I make this adjustment once, live, in the actual setting of where I use it, I don't ever need to make it again. I only need to worry about this with new tracks that I've just added to the library. Even then, SSL auto-gain has already done half the work for me.

Quote:
That's the beauty, you don't need to adjust it in the middle of your set as you did the work ahead of time.

And everyone has time for that? Headphones at home and nightclub systems don't really compare. I would bet you still need to make adjustments on the fly at the club.
msoultan 4:43 PM - 13 June, 2013
Quote:
But some tracks do "sound quieter", regardless of what a computer algorithm or spreadsheet tells me. I adjust these tracks to 'sound' the same volume as the track that is playing, which sounded the same as the last track and the one before that. Now the track's volume in question is more in line with the rest of my library.


Yes, I agree that adjusting to the previous track does work, but you're still potentially causing clipping on the current track (if turning it up) and you won't know unless you watch the SSL meter the entire time. It would be nice if they at least had a resettable clip indicator that reset every time you loaded a new track.

I'm just saying that we should tend to be making our changes to the auto-gain master instead of tweaking individual tracks. The problem is that everyone wants their music to be LOUD so they pick a high value. If you notice, the default value is 92db and based on my research in MP3Gain, that causes a LOT of tracks to clip. If it's dance music, you probably won't hear the clips, but other types of music can suffer greatly from this.

The only real way to fix this problem is to do a lot of mastering work on the tracks (difficult and time-consuming) or use a program like Platinum Notes to do it for you. I don't use PN because it would be hugely time-consuming to redo my entire library, but I've considered it. SSL auto-gain and MP3Gain are not perfect solutions by any means, but MP3Gain has some great advantages over SSL's auto-gain as stated towards the beginning of this thread.
DJ DisGrace 4:45 PM - 13 June, 2013
Quote:
Quote:
The Master level is totally separate from the individual track gain (which is adjusted by auto-gain). You seem to be confusing the two and talking about them interchangeably...
serato.com


Yeah, I meant to say the auto-gain master - sorry!

I still think that's the same thing. The Auto-Gain master in setup adjusts every song up or down a little based on the overview analysis data. Tweaking each individual track is the only solution.

I do see your point about clipping. If the Auto-Gain was set to 98dB, all my tracks would be clipping. Better to keep that low and then turn up the SSL Master output or the channel inputs on the mixer. Some tracks will still need to be turned up/down to 'sound' the same level, and if that causes it to clip, then you need to adjust the auto-gain level or the master level accordingly.
phatbob 4:45 PM - 13 June, 2013
Quote:
All that faffing about, all that effort.

Then uses an SL1.



And only plays mp3s, not lossless.

Quote:
Oh dear.
DJ DisGrace 4:47 PM - 13 June, 2013
Quote:
Quote:
All that faffing about, all that effort.

Then uses an SL1.
And only plays mp3s, not lossless.

Quote:
Oh dear.

haha +1
msoultan 4:51 PM - 13 June, 2013
Quote:
OK, I kind of see what you are saying. But the thing is once I make this adjustment once, live, in the actual setting of where I use it, I don't ever need to make it again. I only need to worry about this with new tracks that I've just added to the library. Even then, SSL auto-gain has already done half the work for me.


Yeah, but you could have processed your tracks with MP3Gain and then make your tweaks in SSL in real-time as well - there's no reason you can't still turn it up in SSL - you'll potentially still cause the exact same damage you did with SSL's auto-gain - gain structure really hasn't changed. The nice part about pre-processing with MP3Gain is that you saw what was going on clipping-wise ahead of time and all of your waveforms are a consistent size - not a bad thing.

Quote:

And everyone has time for that? Headphones at home and nightclub systems don't really compare. I would bet you still need to make adjustments on the fly at the club.


Heh, it's all time-consuming one way or another - it really sucks when you hear the clipping in your recording!! ;)

The nice part about MP3Gain is you're getting a great visual representation of what's going on, so it's actually saving you time and headache down the road. Believe me, I understand how time-consuming all of this is!
msoultan 5:01 PM - 13 June, 2013
Quote:
Quote:
Quote:
All that faffing about, all that effort.

Then uses an SL1.
And only plays mp3s, not lossless.

Quote:
Oh dear.

haha +1


Like I said above, clipping is clipping, so if you're gonna laugh at me just because I have an SL1 box, then you're not paying attention to the discussion and really just being a distraction to the actual issues... that, or you don't understand the concepts being discussed. Clipping will affect the best equipment to the worst equipment. It would be nice if you appreciated that I'm taking the time to help everyone here, and all of my research is being done before it ever hits the SL1, as you would have noticed if you read through my initial posts. Who's faffing about now ;)

Oh, and if you're going to give me a hard time for using MP3s, feel free to back up your chides and see if you can *audibly* tell the difference between MP3s and lossless files:

serato.com

Good luck!
msoultan 5:17 PM - 13 June, 2013
Quote:
I still think that's the same thing. The Auto-Gain master in setup adjusts every song up or down a little based on the overview analysis data. Tweaking each individual track is the only solution.


Tweaking is the only solution for tracks that SSL's auto-gain (or MP3Gain) gets wrong, adjusting the SSL master auto-gain is a great solution to avoid a LOT of tracks from clipping, something you'd only know if you played each and every track. MP3Gain will tell you immediately whether a track is clipping. Download MP3Gain and run an analysis on your library - you'll be pretty amazed to see what is already clipping. Then put the volume level to your current SSL auto-gain level and you'll probably be pretty shocked to see the resultant output - it's kinda scary.

Quote:
I do see your point about clipping. If the Auto-Gain was set to 98dB, all my tracks would be clipping. Better to keep that low and then turn up the SSL Master output or the channel inputs on the mixer.


Unfortunately that won't work either - the gain structure in SSL is kinda weird so if you turn down the track and then turn up the master, you'll still clip the output (if I remember correctly). We had this discussion a few years back and my recommendation was to get rid of the master volume knob entirely because it really is confusing and pointless. I think the master volume value is just added or subtracted to the track's volume and then applied to the audio, so there isn't any headroom there. If it clips it clips - I'll have to double check this or have someone correct me on this one... it was a long time ago.

Quote:
Some tracks will still need to be turned up/down to 'sound' the same level, and if that causes it to clip, then you need to adjust the auto-gain level or the master level accordingly.


Yes, some tracks will need to be adjusted if SSL's auto-gain (or even MP3Gain) screwed up, but if you turn it up, you run the risk of clipping portions of the track. Adjusting the SSL auto-gain or master volume after the fact just complicates the problem. Leave the master volume be and adjust the auto-gain to a conservative value and then fix any problem tracks. But don't go changing both all the time - you'll drive yourself nuts!
Laz219 10:40 PM - 13 June, 2013
Quote:
Quote:
Quote:
I use auto-gain as a rough start, then manually tweak my tracks from there as I play them.
It's a good base-point, but I wouldn't consider it a perfect solution.

+1

+1
Unless you are using another DVS, I don't see the value in changing the volume of your tracks. The SSL auto-gain gets me close enough.



Because even after auto-gain, I still find some excessively loud/quiet tracks when I'm cueing that need to be matched up more accurately. Just refining really.
Serato, Support
Jamie W 10:43 PM - 13 June, 2013
Hey Msoultan,

Quote:
First, MP3Gain does not re-encode the MP3 and it's changes are non-destructive and completely reversible. Is it Serato's stance to not use applications such as Mixed in Key, Platinum Notes, MixMeister BPM counter, or any other pitch detection or tagging program because they edit the MP3? That just doesn't seem like a realistic argument.


Its just my recommendation. Serato has a built in Auto Gain feature that works really well, so I recommend you simplify things and use whats built into the software already.

I am not familiar with the program MP3Gain.
I was referring to third party applications in general.
-- They CAN cause issues for files, making them become corrupt.

Mixed in Key is a great program and its key detection works really well.
-- Metabliss is also a great ID3tag editing program.

Thanks.
msoultan 10:45 PM - 13 June, 2013
Quote:
Because even after auto-gain, I still find some excessively loud/quiet tracks when I'm cueing that need to be matched up more accurately. Just refining really.


Which is really what most people are doing because the algorithms aren't perfect. This is especially true with tracks that have a very sustained loudness and then spike or dip - the algorithms don't seem to handle that very well.

If you glance at the results at the beginning of this thread, you'll notice that very minute increments in gain can cause you to clip. That said, the clipping is probably inaudible, but you're still causing clipping, and potentially causing more clipping than you might have already had. This is why I recommend that people at least try an analysis using MP3Gain to see what tracks are already clipping as SSL's auto gain feature doesn't give you a "big picture" view of what's going on with your tracks.
DJMark 10:50 PM - 13 June, 2013
Quote:
First, MP3Gain does not re-encode the MP3


No, but it does re-tag them.

I'm not personally aware of any Scratch Live issues from files with MP3Gain tags, but the general advice given was solid.

No gain-finding algorithm (Serato's, MP3Gain or anything else) is going to 100 percent "normalize" perceived volume levels. There's a lot of factors that play into "perceived volume level"... and the comparative differences will vary depending on the playback system.
msoultan 10:55 PM - 13 June, 2013
Quote:
I am not to familiar with the program MP3Gain.


I would *highly* recommend that you try it. It's a free download and it's been around for ages and has a huge install base. While you enjoy the use of SSL's auto gain feature, you too might be surprised at how many of your tracks are already clipping, and how many more you're causing to clip (or cause further clipping) if you're pushing a higher value on SSL's auto gain. Do you know how many of your MP3s are currently clipping? Do you know how many previously non-clipping tracks you're now causing to clip by using auto gain? These are really important questions and Serato is making a very broad recommendation to use something that could be quite damaging if used incorrectly, and if people are setting it at 92db as the default, they're probably causing a lot of tracks to clip and degrade audio quality (just look at my results).

Quote:
I was referring to third party applications in general.
-- They CAN cause issues for files, making them become corrupt.

Mixed in Key is a great program and its key detection works really well.
-- Metabliss is also a great ID3tag editing program designed by Mixed In Key.


But those programs can potentially cause corruption as well ;) I'm just bustin' your chops on that one, but in all seriousness I just didn't want you to discount MP3Gain because it's a third party app while endorsing other programs such as Mixed in Key or Metabliss. It is a legitimate program and has been discussed widely on the forums as well.

Thanks!
msoultan 11:00 PM - 13 June, 2013
Quote:
No, but it does re-tag them.


True, and it does edit the volume for each frame - I really was just trying to emphasize that it's a non-destructive program in how it works, but yes, it does edit the file.

Quote:
I'm not personally aware of any Scratch Live issues from files with MP3Gain tags, but the general advice given was solid.

No gain-finding algorithm (Serato's, MP3Gain or anything else) is going to 100 percent "normalize" perceived volume levels. There's a lot of factors that play into "perceived volume level"... and the comparative differences will vary depending on the playback system.


True! But clipping is still clipping, and regardless of the playback system, if the clip is audible, you'll hear it (unless it's a horrible audio system). And unfortunately with SSL's auto-gain feature, you won't know it's clipping until you play the track, and a club is far from the ideal location to be listening for clipping audio. I prefer to be proactive and get a head's up before the track even gets imported into SSL.
DJMark 11:06 PM - 13 June, 2013
Quote:
And unfortunately with SSL's auto-gain feature, you won't know it's clipping until you play the track, and a club is far from the ideal location to be listening for clipping audio. I prefer to be proactive and get a head's up before the track even gets imported into SSL.


If you actually watch the meters ***when DJ-ing*** (not just using your computer in no-interface mode), you'll see on the more modern interfaces and mixers that any "clipping" is really infrequent. Also if you're extra-worried about that you can always turn the master-gain down a bit.

Plus so much aggressively-mastered music in the last 15 years already has a shit-ton of "clipping" regardless of where you set playback levels.

I really would suggest a newer SL interface if you're this concerned with quality.
msoultan 11:13 PM - 13 June, 2013
Quote:
If you actually watch the meters ***when DJ-ing*** (not just using your computer in no-interface mode), you'll see on the more modern interfaces and mixers that any "clipping" is really infrequent. Also if you're extra-worried about that you can always turn the master-gain down a bit.


Or, you can pull it into a program like MP3Gain and see that it's a lot more frequent than you think ;) Orrrrr... you're smart and happen to be using a lower SSL auto-gain value.

Btw, turning down the master gain is the same as turning down the SSL auto-gain, something I recommended a few posts earlier.


Quote:
Plus so much aggressively-mastered music in the last 15 years already has a shit-ton of "clipping" regardless of where you set playback levels.



I don't agree - any legitimate audio engineer will master their music with a limiter to around -0.2db or somewhere in the vicinity. I have ripped a bunch of CDs and haven't seen the WAVs clip. If you're looking at your MP3s and noticing they're clipping, that's due to the encoding process, not the original master.

Quote:
I really would suggest a newer SL interface if you're this concerned with quality.


So I can have higher quality clipping? ;) Again, if it's clipping, it's clipping. I don't understand why people think a higher quality interface will stop the clipping - it won't - if anything it will make it more apparent!
DJMark 11:36 PM - 13 June, 2013
Quote:
I have ripped a bunch of CDs and haven't seen the WAVs clip.


Of course you won't see them clip since the maximum possible level from a CD rip would be 0dbFS, but hard clipping is very often used in the mastering process (and often in the production process) as a means of gaining loudness.
msoultan 11:47 PM - 13 June, 2013
Quote:
Of course you won't see them clip since the maximum possible level from a CD rip would be 0dbFS, but hard clipping is very often used in the mastering process (and often in the production process) as a means of gaining loudness.


Hard *limiting* is a means of gaining loudness. Hard clipping means that your program material's waveform railed hard against the 0db limit and you're now essentially generating a squarewave - those are two very different things, one of which sounds much worse than the other (depending on the program material, of course).

So, just so everyone is clear, do NOT try and induce clipping in your tracks to gain loudness - you will most likely get weird noises and it will not sound good. If you want to increase the apparent loudness of your track while keeping it as descent sounding as possible, use a high quality limiter in your DAW. This is essentially what Platinum Notes does, along with some other neat stuff such as EQ, multiband compression, and they've probably got a few other tricks up their sleeve.

Honestly, I'm really starting to lean towards using Platinum Notes, but you lose all of your cue points in the process... not something I am looking forward to!
DJMark 12:02 AM - 14 June, 2013
Quote:
Honestly, I'm really starting to lean towards using Platinum Notes, but you lose all of your cue points in the process


You also unavoidably transcode any lossy-encoded files going in.

As far as clipping...whether you are aware or not, it's been used as a "mastering" technique for a long time now. The levels on a given CD rip hitting 0dbFS or not isn't an indication. Looking at the waveforms will tell you a lot more. Clipping is also a commonly-used function in broadcast-audio processing to gain loudness, also is incorporated into some DAW limiter plug-ins (the Oxford limiter one such example).
msoultan 12:54 AM - 14 June, 2013
Quote:
As far as clipping...whether you are aware or not, it's been used as a "mastering" technique for a long time now. The levels on a given CD rip hitting 0dbFS or not isn't an indication. Looking at the waveforms will tell you a lot more. Clipping is also a commonly-used function in broadcast-audio processing to gain loudness, also is incorporated into some DAW limiter plug-ins (the Oxford limiter one such example).


Well, upon further research, I stand corrected! I didn't know that they use clippers that way. I just did a bunch of reading through various threads and a lot of people say to use your ears for the sound you want to get, so if you like the sound of your audio clipped (and I don't mean that in a bad way), then it's definitely a tool people here can use. My worry is that it's going to color the track in a way that doesn't make it sound pleasing to my ear - but then again, it might not be that apparent, either. That's a tough one and a really good point and I think I mentioned it towards the beginning of this thread, the clipping may or may not be audible...

That said, if people are using clipping as a tool, they should be listening to all of their tracks ahead of time to see how much clipping they'd really like to apply to their tracks and make sure it's the sound they want. I would still find MP3Gain, along with any other descent DAW, to be of great help in this process. Good post!
DJMark 1:17 AM - 14 June, 2013
I *maybe* should have mentioned....I've done quite a bit of mastering (for record labels, including a few major-label releases) and was formerly a broadcast engineer specializing in audio processing. So it takes relatively little motivation for me to get fairly verbose on this subject if time allows.

I really don't advocate that DJ's use clipping as an effect. First off, because of the already mentioned aspect of a club sound system not always being the best reference point to listen for issues. Secondly, because the resulting high-frequency trash isn't the best thing for people's ears or for speakers. Presumably in production/mastering, the effect is controlled and bandlimited to a reasonable degree (ha!) but much harder to get control of that in a "live" situation.

So the initial "quality" concern is something I fully appreciate and support. I just don't think it usually warrants too much concern if using reasonable gain settings and the more modern SL interface hardware (providing more internal headroom).
nik39 12:30 PM - 14 June, 2013
Quote:
The funny part about the above post is that it completely defeats the purpose of auto-gain. Auto-gain/MP3Gain change the volume based on apparent loudness in an attempt to get all tracks sounding the same volume and do a pretty good job of it. One of the problems with SSL's auto-gain is that it doesn't tell you anything about clipping caused by its process so the only way you find out about that clipping is during playback - which is a horrible time to find this out. At least MP3Gain will tell you immediately whether you're clipping the track.

And then? Would you undo the gain adjustment once you have found out that this one track clips, because it has a small clip?

Quote:
Btw, turning down the master gain is the same as turning down the SSL auto-gain, something I recommended a few posts earlier.

Why would you recommend such a thing and at the same time using an SL1? By turning down the master gain in the software you are losing headroom and you are worsening the s/n ratio. Esp. when using the SL1 which has "only" 16Bit D/A's. If you turn down the volume by 6dB you are reducing the resolution to 14Bit.

Striving for best quality (which is good) and then sacrificing quality by using sub-optimal hardware doesn't make sense to me.
msoultan 4:33 PM - 14 June, 2013
Quote:
I *maybe* should have mentioned....I've done quite a bit of mastering (for record labels, including a few major-label releases) and was formerly a broadcast engineer specializing in audio processing. So it takes relatively little motivation for me to get fairly verbose on this subject if time allows.


I think it's great to get an educated opinion on this stuff. Many people aren't audio engineers, nor do they take the time to really delve into these subjects or understand what's going on. I enjoy these kinds of discussions because it really helps me further understand the technicalities of what's going on.

Quote:
I really don't advocate that DJ's use clipping as an effect. First off, because of the already mentioned aspect of a club sound system not always being the best reference point to listen for issues. Secondly, because the resulting high-frequency trash isn't the best thing for people's ears or for speakers. Presumably in production/mastering, the effect is controlled and bandlimited to a reasonable degree (ha!) but much harder to get control of that in a "live" situation.


Upon reading, it did seem like people were using it in a controlled environment and it seemed like they were using soft clippers and other equipment to ease the harshness of the sound. It was also interesting because the first thing I looked at was the Omnia processors as I do broadcast engineering, but I hadn't really paid much attention to the minute details in the processors. Sure enough, there were multiple clippers in addition to the limiters and levelers - interesting stuff!

Quote:
So the initial "quality" concern is something I fully appreciate and support. I just don't think it usually warrants too much concern if using reasonable gain settings and the more modern SL interface hardware (providing more internal headroom).


I disagree that modern hardware is going to change the amount of playback headroom. A computer, CD player, SL1, protools, digital console, and any other digital audio device has the same DAC limits - 0db (for those who don't know, DAC is digital-to-analog converter). And any headroom for further amplification in the digital file being played was taken away during the recording and mastering process, and relatively ignored with further usage of SSL auto-gain. If the waveform hits 0 (by whatever means - mp3gain, platinum notes, SSL auto-gain, gain knobs, master knob, etc), there's nothing left after that. If the artist (or DJ) chooses to rail harder against that 0db limit and clip their audio, having more headroom after the DACs (or wherever it supposedly exists) isn't going to make an ounce of difference - the DAC can only output the most that the digital file gives it, 0db, and beyond that (clipping), it's still only outputting the audio represented at 0db, but now it's a different sound (whatever that clipped audio happens to sound like).

If you run 24 bit hardware, yes, you can now represent a wider dynamic range but you're still pushing the potential amount of volume bits in the *downward* direction, not the upward direction. You're not getting any more output out of the digital file - 0db is still 0db. I would assume these are things you know, but others might not. I am still surprised that you mentioned the hardware, though. Sure, there are different *quality* of ADCs, but that's an entirely different discussion.

As for those that think 24bit is better than 16bit (this should address Nik39's statement about the SL1 as well), let's continue. I found a great article that talks about 24 vs 16 bit and here's an excerpt from the beginning of the article linked below:

www.head-fi.org

"So, 24bit does add more 'resolution' compared to 16bit but this added resolution doesn't mean higher quality, it just means we can encode a larger dynamic range. This is the misunderstanding made by many. There are no extra magical properties, nothing which the science does not understand or cannot measure. The only difference between 16bit and 24bit is 48dB of dynamic range (8bits x 6dB = 48dB) and nothing else. This is not a question for interpretation or opinion, it is the provable, undisputed logical mathematics which underpins the very existence of digital audio."

And the next paragraph really hits it home, especially due to the fact that most of us are playing dance music that is highly compressed and has a *very* limited dynamic range:

"So, can you actually hear any benefits of the larger (48dB) dynamic range offered by 24bit? Unfortunately, no you can't. The entire dynamic range of some types of music is sometimes less than 12dB. The recordings with the largest dynamic range tend to be symphony orchestra recordings but even these virtually never have a dynamic range greater than about 60dB. All of these are well inside the 96dB range of the humble CD. What is more, modern dithering techniques (see 3 below), perceptually enhance the dynamic range of CD by moving the quantisation noise out of the frequency band where our hearing is most sensitive. This gives a perceivable dynamic range for CD up to 120dB (150dB in certain frequency bands)."

This would accurately describe dance music as we're barely using any of the available dynamic range, both 16bit or 24bit. So the 24bit argument doesn't really apply here.

Furthermore, this discussion is not about dynamic range or DAC quality, it's purely about clipping and the results of that clipped output. I think it's safe to say, due to the SSL auto-gain default of 92db, that most of the DJs using SSL auto-gain are clipping their outputs. Now whether it's audible or not is probably one of the most important items of discussion. Like DJMark said, we can use reasonable gain levels, which is precisely what I recommended, but nobody wants to turn it down - they all want to push it into the red.

Nik39 also mentioned this as well:

Quote:
And then? Would you undo the gain adjustment once you have found out that this one track clips, because it has a small clip?


And this is a great one - some of these clips aren't audible, but some are - one of the issues I'm wrestling with is how hard can I push my program material before those clips become audible and should I even allow it to clip? And that also depends on the program material as well. That said, I still come back to my statement that SSL's auto-gain gives you no prior notice of what tracks are going to potentially clip, something that MP3Gain is great for. And furthermore, I'm moving towards the recommendation of turning it down in SSL (via a lower SSL auto-gain setting or lower MP3Gain setting) to have a more accurate representation of the original program material, because that's probably what we're going for, right?

And I will repeat this one more time because everyone *still* likes to compare hardware. Clipping affects everyone! SL1, SL3, SL4, SL5000, computer, CD player, or any digital device. Don't think that because you have a better interface than the SL1 that you're immune to clipping, or that your clipping will sound any better than mine, because most likely it won't - if anything you have better hardware, you also have better DACs and it will probably sound even more obvious! I'm not doing this to prove that mine is better than yours, I'm doing this to help all of you make your music sound as good as possible.
nik39 5:16 PM - 14 June, 2013
Of course 24Bit won't make a difference about the maximum volume which can come out of the output. That's not what I was aiming at.

Quote:
This would accurately describe dance music as we're barely using any of the available dynamic range, both 16bit or 24bit. So the 24bit argument doesn't really apply here.

This is true. As DJMark has said, Dance music nowadays is a block of form (waveform would not apply here ;)).

However - suddenly you are changing your direction. Now suddenly 16 or 24Bit does not make a difference for you, because the material (Dance music) does not use the whole range.

Suddenly you seem to mention that for normal music 12Bit is enough. How about you try your mp3 vs lossless music comparison on 12Bit vs 16Bit? ;) Not that it would matter - but.. again you are changing direction.

Are you looking for the best quality or not?

Fact is, by reducing the internal volume of a track by half you are throwing away 1Bit. Reduce it to 1/4th, you just lost another Bit. If you don't care cause your initial material is sh**, then why would you worry about (minor) clipping!
msoultan 5:59 PM - 14 June, 2013
Quote:
Of course 24Bit won't make a difference about the maximum volume which can come out of the output. That's not what I was aiming at.


That's how this discussion started and I'd prefer to keep the focus on clipping and not dynamic range or interface quality, although those are very important topics.

Quote:
Quote:
This would accurately describe dance music as we're barely using any of the available dynamic range, both 16bit or 24bit. So the 24bit argument doesn't really apply here.


This is true. As DJMark has said, Dance music nowadays is a block of form (waveform would not apply here ;)).


And that really pisses off the audiophiles that want dynamic range back - but that's a whoooooole other discussion ;)

Quote:
However - suddenly you are changing your direction. Now suddenly 16 or 24Bit does not make a difference for you, because the material (Dance music) does not use the whole range.


Bit depth was never part of my discussion, because I'm not making a hardware comparison - if I or you or whoever else has to turn down their audio to avoid clipping, then we'll all suffer the consequences, the 24bit people to a lesser (and most likely inaudible) amount. The key here is what do we do about clipping, which as I will mention one more time, affects *everyone*, regardless of what interface you have. If you push your audio to 0db and I push mine to 0db, 16, 24, 256bits of dynamic range makes absolutely no difference whatsoever. Plus, we're not even using the full 16bit dynamic range anyways, so your 24bit interface isn't helping you, even if we both turn down the volume a small amount to avoid excessive and audible clipping.

I'm happy to branch out this discussion as to how much clipping we should allow, but that's a tough one as you really need to listen to the program material to hear its affects. I'm trying to come up with a broad recommendation which is what SSL's auto-gain essentially is - a broad application of volume control. Unfortunately it has no reporting feature which is why I recommend MP3Gain.

Quote:
Suddenly you seem to mention that for normal music 12Bit is enough. How about you try your mp3 vs lossless music comparison on 12Bit vs 16Bit? ;) Not that it would matter - but.. again you are changing direction.


If you would like to debate the merits of lossy vs lossless audio, I would love to hear your input here:

serato.com

Quote:
Are you looking for the best quality or not?


I don't know how many times I'm going to have to say this, but I'll say it again - this is not a discussion about quality and it never was - it's a discussion about clipping - that affects everyone of all different interfaces. I don't mind that I don't have as good as equipment as everyone else and I am willing to make that sacrifice, but like I keep saying, clipping affects everyone, so this is not a me/you discussion, it's a "how do we make the best compromise between loudness and clipping". Furthermore, it's a discussion of the merits of MP3Gain vs SSL's auto-gain feature.

Quote:
Fact is, by reducing the internal volume of a track by half you are throwing away 1Bit. Reduce it to 1/4th, you just lost another Bit. If you don't care cause your initial material is sh**, then why would you worry about (minor) clipping!


I think my material sounds good and sounds better than if it's clipping, and chances are that anyone playing lossless audio will also sound better than if it's clipping - I'm not saying mine is better than yours because it's not - but clipping affects all of us.

While I didn't want to get into a bitrate discussion, it went there and it's not a horrible thing to briefly discuss it. But as the article describes, it has been show that if I lower the volume by a small amount I'm not losing much of anything because the dynamic range of dance music isn't utilizing the full dynamic range available anyways. Therefore the discussion of bitrate is pretty pointless for dance music in the scope of this thread. It might make a difference for the classical music DJs, but that's not what we're talking about and can be discussed in a different thread for all of the classical music DJs out there.
nik39 6:14 PM - 14 June, 2013
Quote:
Bit depth was never part of my discussion, because I'm not making a hardware comparison

Bit depth is not a discussion bound to any hardware.


Quote:
Plus, we're not even using the full 16bit dynamic range anyways, so your 24bit interface isn't helping you, even if we both turn down the volume a small amount to avoid excessive and audible clipping.

Sure my 24Bit interface is helping.

Quote:
If you would like to debate the merits of lossy vs lossless audio, I would love to hear your input here:

No, I don't care.

Since you put a lot of energy into this thread, I assumed you were striving for best quality overall.

Quote:
While I didn't want to get into a bitrate discussion, it went there and it's not a horrible thing to briefly discuss it. But as the article describes, it has been show that if I lower the volume by a small amount I'm not losing much of anything because the dynamic range of dance music isn't utilizing the full dynamic range available anyways.

Moving in circles, huh? ;)

Not much of anything because you already had sh*t before? You are degrading quality. That's what it is about. If you're argument is "I am not losing much quality" - this applies to this entire discussion. I wont be losing much quality either when allowing that on vinyl crackle which has amplified to distortion to match the overall track volume. :)
Detroitbootybass 8:31 PM - 14 June, 2013
Good to see some old-timers (Mark/Nik) in this thread!

:)
msoultan 2:47 AM - 15 June, 2013
Quote:
Bit depth is not a discussion bound to any hardware.


Nor was it ever meant to be part of this discussion and it still doesn't apply because we're talking about the loud end of the dynamic range - and as that article linked above states, you won't even notice a difference between 24 and 16 bit when the dynamic range of dance music is something like 12db and the entire dynamic range is at least 96db. So even if I drop my volume 6db, I'm still not losing dynamic range because I'm getting nowhere near the -96db floor. That said, I'll entertain the conversation a little longer if you can justify it - please see the next quote below.

Quote:
Sure my 24Bit interface is helping.


I would like to know in what way a 24bit interface is going to help when you turn down the volume down slightly to avoid clipping? The only benefit that I see to having a 24bit playback device is for recorded music that utilizes a large dynamic range. If you're playing dance music, you're not even using that entire range, nor do I really need to worry about bringing up my noise floor as you'll never hear it. So, how would 24bit playback be any different?

Quote:
Since you put a lot of energy into this thread, I assumed you were striving for best quality overall.


You're making a lot of assumptions. I started this thread with the intention of discussing clipping and what volume settings to use - that's it. If you want to talk about best overall quality, you're more than welcome to start another thread.

Quote:
Quote:
While I didn't want to get into a bitrate discussion, it went there and it's not a horrible thing to briefly discuss it. But as the article describes, it has been show that if I lower the volume by a small amount I'm not losing much of anything because the dynamic range of dance music isn't utilizing the full dynamic range available anyways.

Moving in circles, huh? ;)


Are you just going to tell me I'm going in circles because I'm willing to discuss something you wanted to discuss? I'm entertaining something you brought up - at least be polite enough to reply with some justifications of your thoughts instead of giving me a hard time because I'm entertaining something you wanted to bring up. Due to reasons stated above, bitrate doesn't play a role in this particular discussion, but I'm willing to discuss it because you don't seem to fully understand it - or I'm getting it completely wrong, which I'll accept, but you'll at least have to provide a valid argument other than telling me I'm going in circles. I'm trying to work with you here, but snide remarks towards me is far from constructive.

Quote:
Not much of anything because you already had sh*t before? You are degrading quality. That's what it is about. If you're argument is "I am not losing much quality" - this applies to this entire discussion. I wont be losing much quality either when allowing that on vinyl crackle which has amplified to distortion to match the overall track volume. :)


I'm sorry if the intention of this discussion was not properly stated at the very beginning, but so you now know, the intention of this discussion is to avoid coloration of the audio due to clipping and also to find the ideal volume levels to achieve audio clean of clipping coloration while maintaining loudness.

Everything else you're bringing up is really great stuff and I'd be happy to discuss it in another thread, but it really isn't applicable in this thread for the reasons stated above.
nik39 3:29 AM - 15 June, 2013
Bottom line: If your music is Dance crap (which could also use 12Bit) - you should not care about clipping. If you care about clipping, then your material should be worth more than 12Bit. If it is worth more thant 12Bit we can discuss about Bit resolution. Easy.
nik39 3:30 AM - 15 June, 2013
Quote:
Good to see some old-timers (Mark/Nik) in this thread!

:)

DBB! You're alive :)))
msoultan 5:59 AM - 15 June, 2013
Quote:
Bottom line: If your music is Dance crap (which could also use 12Bit) - you should not care about clipping. If you care about clipping, then your material should be worth more than 12Bit. If it is worth more thant 12Bit we can discuss about Bit resolution. Easy.


What is dance crap? If you're referring to music that has very little dynamic range, then pretty much all dance music is crap. That being said, I'd rather not have my crap clip because that will potentially make it sound ever crappier. However, losing 6db to give me some headroom (to avoid clipping my crap), then another 12db of dynamics range for my crap song.. still leaves me a whopping 78db of dynamic range to go before my audience hears the noise floor. Heck, let's say my crap uses 40db of dynamic range, I still have 50db to go before I hit my noise floor.

Also, did you not read that article? Like it says, dance music (crap) rarely uses more than 12db of dynamics - that means even 12bits is *more* than sufficient to represent the dynamics range of a typical dance crap, give it some headroom to avoid clipping, and still leave plenty of room above the noise floor.
nik39 10:01 AM - 15 June, 2013
Why would you care about clipping if your music uses 12dB of dynamic range? That means 2Bit. TWO Bits!

Caring about clipping at two Bits is ridiculous, IMHO.
DJMark 10:02 AM - 15 June, 2013
Quote:
I disagree that modern hardware is going to change the amount of playback headroom.


Two different things here.

1) I believe the newer SL-series interfaces *do* have higher maximum analog output levels.

2) the newer SL interfaces (and mixers) are internally-processing and D/A converting at 24 bits or higher (believe the 61/62/68 mixers are internally 32-bit floating-point). Therefore if you lower gain on a 16-bit file on the more modern hardware, you are NOT knocking off the bottom-most bits of audio as you would be if doing that with a 16-bit output.
DJMark 10:08 AM - 15 June, 2013
Quote:
Like it says, dance music (crap) rarely uses more than 12db of dynamics


The source you're getting that from is wrong. Even aggressively-mastered modern "EDM" typically has a much wider range than that between the low background stuff and the peaks.

I have to say I really don't understand a discussion about "avoiding clipping" that attempts to divorce that one thing from any other concerns about audio quality.
DJMark 10:09 AM - 15 June, 2013
Quote:
Good to see some old-timers (Mark/Nik) in this thread!

:)


Speaking of "old timers", !!!!!!!!!!!

(nice to see you too :-)
msoultan 3:04 PM - 15 June, 2013
Quote:
1) I believe the newer SL-series interfaces *do* have higher maximum analog output levels.


Where is that extra headroom worked in? We do have a relative max and that's clipping the input of the next device. Even if the output of the box is louder after the DAC, a 16 bit unit still has 96db of dynamic range and 24bit unit has 144db. Digital is digital and the DAC is just representing whatever the digital domain is pushing out and it can get no higher than 0db, regardless of whether it's 16, 24, or 32 bit. Furthermore, the analog side of the chain is, well... analog, so anything after the DAC is not affected by the 16 vs 24 bit dynamic range, unless it's got louder amps, but then you still have to turn it down so as not to clip the next component in the chain.

Quote:
2) the newer SL interfaces (and mixers) are internally-processing and D/A converting at 24 bits or higher (believe the 61/62/68 mixers are internally 32-bit floating-point). Therefore if you lower gain on a 16-bit file on the more modern hardware, you are NOT knocking off the bottom-most bits of audio as you would be if doing that with a 16-bit output.


I'd like to understand this one more - the 16 vs 24 issue here has to do with dynamics, correct? We have 96db of dynamics range and even more with 24. I could understand if we were utilizing the entire range, then yes, by all means every single thing that Nik39 says holds true. The issue here is that we're not using the entire range, and far from it. I threw on some orbital tracks and the most that they used was 20db except for the fade-ins and fade-outs. So even if I lower the volume slightly, I'm not losing any dynamic range that the song was recorded with. Say I lower it 6db, song takes up 20db more, there's still 60db left. Am I screwing that part up? At 24 bits with the same song is still gonna be 6db less (because turning it down 6db whether it's 16 or 24), song uses 20db, and I I'm left with even more potential room to the headfloor... which the song isn't using anyways.

Quote:
The source you're getting that from is wrong. Even aggressively-mastered modern "EDM" typically has a much wider range than that between the low background stuff and the peaks.


I just tried and I didn't see much difference between the peaks and valleys. Pull in some of your modern (and older) tracks and see what you get. Dance music is pretty strongly compressed and I do agree that dance music uses very little dynamic range (relative to classical music, let's say).

Quote:
I have to say I really don't understand a discussion about "avoiding clipping" that attempts to divorce that one thing from any other concerns about audio quality.


Ok, how about this - audio quality is such a vague term and nobody has taken the time to define it. It can refer to many different things - noise, dynamic range, sound quality, etc. In this case (16 vs 24 bit), the only thing we're dealing with is dynamic range, correct? If that's the case, I agree that turning down the music is losing potential dynamic range - for that I completely agree. But if a song is using, let's even say 60db of dynamic range and you turn it down 6db, you still have 30db before noisefloor becomes apparent - is there anything else that matters here? That's what this 16/24bit discussion is all about - avoiding bringing up the noisefloor. It seems obvious to me - am I seeing it incorrectly?
deezlee 4:55 PM - 15 June, 2013
dang, this dude is patient.
nice discussion.
lots of good info/points.
yup.
DJ DisGrace 5:03 PM - 15 June, 2013
Quote:
dang, this dude is patient.
nice discussion.
lots of good info/points.
yup.

+1

Was staring at my waveforms a bit last night and it occurred to me that it would be nice if the waveform shown in SSL corresponded to the end result of the auto-gain feature (as opposed to the original file's waveform). It would be similar to the waveform view that the OP is talkin about in MP3Gain, and would allo you to visually see which tracks were clipping
s3kn0tr0n1c 10:04 PM - 15 June, 2013
Quote:
Quote:
dang, this dude is patient.
nice discussion.
lots of good info/points.
yup.

+1

Was staring at my waveforms a bit last night and it occurred to me that it would be nice if the waveform shown in SSL corresponded to the end result of the auto-gain feature (as opposed to the original file's waveform). It would be similar to the waveform view that the OP is talkin about in MP3Gain, and would allo you to visually see which tracks were clipping

+1
Joshua Carl 12:00 AM - 16 June, 2013
Joshua Carl 12:00 AM - 16 June, 2013
Joshua Carl 12:00 AM - 16 June, 2013
just kidding,.... I just invested 30 minutes to schooling myself some more here.
good looks everyone.
Detroitbootybass 8:25 PM - 16 June, 2013
I have to admit that I don't possess anywhere near the same level of knowledge of this subject as other here in this thread... so I can't really add much to the conversation (sorry).

But what I do have is a question (or questions).

But first, a little back story: I used to use Serato's built-in gain feature when adding new MP3s and analyzing the waveforms. I had it set it to 92 DBs. It seemed to work fairly well, but some tracks still seemed way too hot (and clipped) and other way too quiet. To compensate, I would move the gain 'knob' for each track and the software would save it's position for that song. I did this without really any further thought on the matter.

Then, about a year ago, while encoding some vinyl and making some MP3s from some CDs of mine through Audacity, I began to give more headroom than I was typically seeing on purchased downloads or my CDs (at or near 0 DB). I started doing this while making a simple edit for myself. I added an effect and noticed the waveform on the affected section 'get bigger' after doing so. It led me to think that added effects essentially raised the gain of a given track and, by giving some 'headroom', one could possibly avoid clipping.

Question number one and two: Is the above statement correct? Does using an internal effect in Serato (potentially) raise the gain of the track?

So I began to take all new tracks into Audacity to view the waveforms. All MP3s without sufficient headroom were loaded into MP3 Gain one by one and reduced by 1.5 (not sure if the '1.5' in MP3 Gain refers decibels). Then I could double-check by reloading into Audacity and confirming the new level is to my liking. With this, I was hoping avoid clipping in the instances that I used an effect. To be clear, I don't often use effects while playing music. But I can on occasion and wanted to avoid problems.

Question three and four: Am I wasting time with this? Could I be doing something better?

If my thoughts on this are somewhat correct, I will go back to all the stuff in my library and execute the same process on my older tracks. But I need the help of much wiser people than myself. If they'd be willing to weigh in, I'd by highly appreciative.

And sorry for the minor thread-jack. But it is related the original topic (gain) and also involves MP3 Gain and Serato.
Detroitbootybass 8:31 PM - 16 June, 2013
Please excuse all the spelling errors!
msoultan 2:16 AM - 26 June, 2013
Quote:

Was staring at my waveforms a bit last night and it occurred to me that it would be nice if the waveform shown in SSL corresponded to the end result of the auto-gain feature (as opposed to the original file's waveform). It would be similar to the waveform view that the OP is talkin about in MP3Gain, and would allo you to visually see which tracks were clipping


Well, you won't be able to visually see them clipping because a waveform can be railed against the 0db limit (via compression or limiting) and still be fine. But, MP3Gain will tell you if the tracks are actually clipping, something SSL won't (easily tell you).

Quote:
I have to admit that I don't possess anywhere near the same level of knowledge of this subject as other here in this thread... so I can't really add much to the conversation (sorry).


I don't know everything either, hence the reason for starting this thread. There are a lot of issues which are brought up which I then spend further time researching. DJMark brought up clipping as a means of making a song sound louder which I wasn't familiar with, but it is in-fact a legitimate means of making a track sound louder. That said, it is best to utilize this technique in the control room of your studio while mastering a track, not in the club ;)

Quote:
But first, a little back story: I used to use Serato's built-in gain feature when adding new MP3s and analyzing the waveforms. I had it set it to 92 DBs. It seemed to work fairly well, but some tracks still seemed way too hot (and clipped) and other way too quiet. To compensate, I would move the gain 'knob' for each track and the software would save it's position for that song. I did this without really any further thought on the matter.


This is what a lot of people do. You're either doing one of two things - fixing where SSL's auto-gain went wrong or changing that perceived volume of that track relative to the rest of your library. If you're doing the prior, awesome - if you're doing the latter, you should be adjusting the SSL auto-gain value instead.

Quote:
Then, about a year ago, while encoding some vinyl and making some MP3s from some CDs of mine through Audacity, I began to give more headroom than I was typically seeing on purchased downloads or my CDs (at or near 0 DB). I started doing this while making a simple edit for myself. I added an effect and noticed the waveform on the affected section 'get bigger' after doing so. It led me to think that added effects essentially raised the gain of a given track and, by giving some 'headroom', one could possibly avoid clipping.


Headroom really applies to recording or the input device (i.e. mixer, recording input, etc), not playback, especially in this instance where playback though SSL is ultimately limited by the digital domain (don't send your signal higher than 0db). That being said, it makes sense to give yourself some headroom when recording from an LP, tape, microphone, etc but you also need to understand what that means and the repercussions of doing so. Obviously you don't want to clip the recording device otherwise you're pristine LP now sounds like crap when it's digitized. However, you also don't want to give yourself so much headroom to the point that when you raise the volume of the track, you also end up raising the the noisefloor to an noticeable level.

A little on noisefloor - there's always a noisefloor and it depends on your equipment and how noisy your room is. So, for example, you've got a mic setup in your studio and there's an air-conditioning unit blaring in the background. Your input meters show a steady input level at -30db - this means your noisefloor is -30db. A sound emitted below -30db will be unintelligible (masked by the AC unit) and anything louder will mask the noise of the AC unit (these are FAR from ideal conditions, obviously). If you now record continuous vocals never dipping below -25 or so and peak at -5db, you're fine because the vocals will mask the noisefloor. Normalize your track to 0db and the softest part of your vocals will come up to -20db and your noisefloor *will also come up* with it to -25db. As long as there aren't any pauses in the vocals, you'll never hear all that noise. However, if there are any pauses, you will now hear the noisefloor come through.

In the case above, giving yourself 5db of headroom wasn't a bad thing because the softer parts of your vocals never hit the noisefloor - in other words, they never got softer than -25db. Imagine, though, if you gave yourself 10db of headroom during recording with the exact same vocal dynamic range - the softer segments of your vocals are now hitting the noisefloor and will become part of it and most likely become unintelligible.

So, giving yourself some headroom is good as long as the dynamic range (the difference between the loudest and softest segments) of the music does not bring you close to the noisefloor. You can then peak normalize your music (making the loudest segment come up to 0db) which will maximize the volume of the track without clipping, but it also brings the noisefloor up with it. But again, if the dynamic range of your song never gets soft enough to bring you close to the noisefloor, then you're fine.

Quote:
Question number one and two: Is the above statement correct? Does using an internal effect in Serato (potentially) raise the gain of the track?


I'm not totally familiar of the gain structure of the effects system in SSL, but if the effect is adding volume in one form or another, then yes, it can potentially push a non-clipping track into the clipping region.

Quote:
So I began to take all new tracks into Audacity to view the waveforms. All MP3s without sufficient headroom were loaded into MP3 Gain one by one and reduced by 1.5 (not sure if the '1.5' in MP3 Gain refers decibels). Then I could double-check by reloading into Audacity and confirming the new level is to my liking. With this, I was hoping avoid clipping in the instances that I used an effect. To be clear, I don't often use effects while playing music. But I can on occasion and wanted to avoid problems.


This is fine for the most part but you can be negating your work in MP3Gain by using SSL auto-gain because all it's going to do is pull it back up so the perceived volume is back at 92db. Also, you can't just look at the peak in a track and assume that's the loudest part of the song. The perceived volume is different than peak volume, and that's why you can't just apply an arbitrary gain reduction to all of your tracks - you need to let MP3Gain calculate the perceived volume and then lower the target value until as many tracks are not clipping while not pulling the volume so low to bring the soft parts close to the noise floor.

Quote:
Question three and four: Am I wasting time with this? Could I be doing something better?


I think you should use either SSL auto-gain or MP3Gain, but not both. Like you said, bringing in a track that has it's volume reduced in MP3Gain is a step in the right direction, but that is for that particular track. MP3Gain (and SSL auto-gain) are looking to make the perceived volume the same for all tracks and do their best job to accomplish that goal. What you need to do is find a target volume (using one tool or the other) that will yield the least amount of tracks that are clipping, but also keep the softest segments of your music away from the noisefloor of your playback device.

Btw, if you want to measure the noisefloor of your playback device (i.e. SL1, SL3, etc), generate a sine wave at 0db and play it out of SSL and set the input of your recording device so that it's meters peak at 0db. Now stop playback and see what level your input meters now show - that's your noisefloor. Now make sure the softest segments never get close to that level and you're good to go :)


So now that we've explained the noisefloor, this allows us to turn our attention to something that Nik39 keeps bringing up. He keeps saying that turning down the volume by 6db to avoid the track from clipping means we're losing 1bit of dynamic range, and that statement is entirely correct. People also have kept criticizing me for using an SL1 because it's 16bit vs 24bit. What does this all mean?

The SL-1, CD players, and most computer and even professional audio devices represent volume by a 16bit value. This means that the device is able to represent 65,536 distinct volume levels - or another way to look at it is that 1bit can represent 6db worth of dynamic range - 6db * 16bits = 96db. Wow, that's pretty awesome until you factor in noise in your mixer, unbalanced outputs (something all SL boxes use), induced hum, etc. Furthermore, you're most likely never going to play a track that has volume levels that get that go software than maybe -60db - like I've said repeatedly above, most of our dance music has a dynamic range of about 12db from the loudest portion of the song to the softest - heck, let's say you have a song that has a dynamic range of 40db and you turn the volume down by 12db, will you even come close the noisefloor of the SL1? No, of course not!

So, you're next question is why go 24bit? Well, probably the same reason you won't see the manufacturer chime in.... marketing hype. 24bit dynamic range extends the dynamic range of volume levels down to -144db - this is well beyond the dynamic range of the human ear which is complete silence and the loudest volume possible (120db being near instant hearing damage). Obviously in theory 144db of dynamic range is better than 96db, but in practicality you will never be able to hear the difference, especially in the club or in your bedroom. The only place you *might* hear the difference was if you were in the studio and you were playing music that has a large dynamic range (i.e. classical)... but then you probably wouldn't be using SSL ;)
msoultan 6:18 AM - 26 June, 2013
Oh, and before someone calls me out on it, in my example with the AC noise in the background and the vocals on top, the vocals would only mask the frequencies of the noise in the background - obviously the frequencies have to be similar so it was a somewhat bad example. But that said, if you have similar frequencies that are louder than background frequencies, they'll mask the background noise and make it difficult to notice those background frequencies. That said, if at any point you pull away the foreground noise, the background noisefloor may be apparent, depending on its volume level.

Usually this noise takes the form of a general white/pink noise - try turning your amp up all the way (without anything playing) and then listening to the noise compared to when the volume was all the way down - that's an example of the amp's noisefloor, but you never hear that noise while you're playing music because the music masks that noise, especially because it's probably about 80db louder than the noise.
nik39 9:29 PM - 26 June, 2013
Quote:
But, MP3Gain will tell you if the tracks are actually clipping, something SSL won't (easily tell you).

No software will be able to tell you whether a track is clipping after it has been amplified (post). Mp3gain will tell you *before* it has applied the gain whether it is clipping or not.

Quote:
This is what a lot of people do. You're either doing one of two things - fixing where SSL's auto-gain went wrong or changing that perceived volume of that track relative to the rest of your library. If you're doing the prior, awesome - if you're doing the latter, you should be adjusting the SSL auto-gain value instead.

Nope. That's wrong. Adjusting SSL's auto-gain will obviously change the overall volume of *all* tracks. What the OP is doing here: Fine tuning certain, individual tracks.


Quote:
So, for example, you've got a mic setup in your studio and there's an air-conditioning unit blaring in the background. Your input meters show a steady input level at -30db - this means your noisefloor is -30db. A sound emitted below -30db will be unintelligible (masked by the AC unit) and anything louder will mask the noise of the AC unit (these are FAR from ideal conditions, obviously). If you now record continuous vocals never dipping below -25 or so and peak at -5db, you're fine because the vocals will mask the noisefloor. Normalize your track to 0db and the softest part of your vocals will come up to -20db and your noisefloor *will also come up* with it to -25db. As long as there aren't any pauses in the vocals, you'll never hear all that noise. However, if there are any pauses, you will now hear the noisefloor come through.

In the case above, giving yourself 5db of headroom wasn't a bad thing because the softer parts of your vocals never hit the noisefloor - in other words, they never got softer than -25db. Imagine, though, if you gave yourself 10db of headroom during recording with the exact same vocal dynamic range - the softer segments of your vocals are now hitting the noisefloor and will become part of it and most likely become unintelligible.

Why would you record with 10dB headroom? It makes no sense. You would record your voice as loud as possible, without clipping.

Quote:
So, giving yourself some headroom is good as long as the dynamic range (the difference between the loudest and softest segments) of the music does not bring you close to the noisefloor. You can then peak normalize your music (making the loudest segment come up to 0db) which will maximize the volume of the track without clipping, but it also brings the noisefloor up with it. But again, if the dynamic range of your song never gets soft enough to bring you close to the noisefloor, then you're fine.

Again - this is theory. Vocals without any silence? Vocals using the exact same freq spec as the noise? C'mon, this is not a real life example at all, so no need to discuss this at all.


Quote:
So now that we've explained the noisefloor, this allows us to turn our attention to something that Nik39 keeps bringing up. He keeps saying that turning down the volume by 6db to avoid the track from clipping means we're losing 1bit of dynamic range, and that statement is entirely correct. People also have kept criticizing me for using an SL1 because it's 16bit vs 24bit. What does this all mean?

The SL-1, CD players, and most computer and even professional audio devices represent volume by a 16bit value. This means that the device is able to represent 65,536 distinct volume levels - or another way to look at it is that 1bit can represent 6db worth of dynamic range - 6db * 16bits = 96db. Wow, that's pretty awesome until you factor in noise in your mixer, unbalanced outputs (something all SL boxes use), induced hum, etc. Furthermore, you're most likely never going to play a track that has volume levels that get that go software than maybe -60db - like I've said repeatedly above, most of our dance music has a dynamic range of about 12db from the loudest portion of the song to the softest - heck, let's say you have a song that has a dynamic range of 40db and you turn the volume down by 12db, will you even come close the noisefloor of the SL1? No, of course not!

You're completely missing the point. As said before: Why would you discuss extensively about best quality and then leave out this part which has an impact (probably not a big impact)? BTW, mixers such as the 62,61,68 are all internally digital. No loss.

Quote:
So, you're next question is why go 24bit? Well, probably the same reason you won't see the manufacturer chime in.... marketing hype. 24bit dynamic range extends the dynamic range of volume levels down to -144db - this is well beyond the dynamic range of the human ear which is complete silence and the loudest volume possible (120db being near instant hearing damage). Obviously in theory 144db of dynamic range is better than 96db, but in practicality you will never be able to hear the difference

I agree - as long as you're not manipulating music and reducing bit depths. See.. I won't worry too much about effects, gains etc. when using 24Bit. But with 16Bit there might be negative effects, as explained before.

Anyway, you seem to be happy with accepting negative impacts of 24Bit vs 16Bit and reducing the dynamic range/noisefloor etc. So I don't actually see any reason why you should care about small clippings. But hey, I don't need to understand everything, and I don't think there is a good explanation from your side to shed light on this contradiction. Since I don't seem to be the only one not being able to follow you - I assume it is what it is - a contradiction. No need to try to explain it any further as long as you're repeating yourself. ;)
msoultan 10:11 PM - 26 June, 2013
Quote:
Quote:
So, you're next question is why go 24bit? Well, probably the same reason you won't see the manufacturer chime in.... marketing hype. 24bit dynamic range extends the dynamic range of volume levels down to -144db - this is well beyond the dynamic range of the human ear which is complete silence and the loudest volume possible (120db being near instant hearing damage). Obviously in theory 144db of dynamic range is better than 96db, but in practicality you will never be able to hear the difference


Quote:
I agree - as long as you're not manipulating music and reducing bit depths. See.. I won't worry too much about effects, gains etc. when using 24Bit. But with 16Bit there might be negative effects, as explained before.

Anyway, you seem to be happy with accepting negative impacts of 24Bit vs 16Bit and reducing the dynamic range/noisefloor etc. So I don't actually see any reason why you should care about small clippings. But hey, I don't need to understand everything, and I don't think there is a good explanation from your side to shed light on this contradiction. Since I don't seem to be the only one not being able to follow you - I assume it is what it is - a contradiction. No need to try to explain it any further as long as you're repeating yourself. ;)


I'm not ignoring your previous comments as I think you have valid points, but I think we're talking about so many different things that it's difficult to debate each one individually when they're all really part of one main issue that you've brought up which is the "negative impacts of 24Bit vs 16Bit and reducing the dynamic range/noisefloor etc".

You've mentioned a few times that bringing down the volume, let's say 6db, means you're losing 1bit of dynamic range. I completely agree with that statement, no argument what-so-ever. What I'm not seeing are the the negative impacts on *most of the music that we play* because the dynamic range of dance music doesn't get anywhere near the noisefloor.

So, instead of talking theory, let's use a real-world example. My question to you is how exactly does -6db translate into a real-world problem? More noise, less quality, what is the actual problem? To make it understandable for everyone, pick a song in your collection and explain how that track is taking full advantage of the dynamic range available in 24bit audio - or even further, how is that track taking full advantage of the dynamic range in 16bit audio? How, in real-world terms, are we seeing the benefits or the negative impacts?
nik39 10:52 PM - 26 June, 2013
Quote:
You're completely missing the point. As said before: Why would you discuss extensively about best quality and then leave out this part which has an impact (probably not a big impact)?

I meant which "also" has an impact. But I already said: Probably not a big impact. As much as your clipping discussion.

Quote:
Like it says, dance music (crap) rarely uses more than 12db of dynamics - that means even 12bits is *more* than sufficient to represent the dynamics range of a typical dance crap

Quote:
However - suddenly you are changing your direction. Now suddenly 16 or 24Bit does not make a difference for you, because the material (Dance music) does not use the whole range.

Suddenly you seem to mention that for normal music 12Bit is enough. How about you try your mp3 vs lossless music comparison on 12Bit vs 16Bit?

Want to take the challenge?
msoultan 12:41 AM - 27 June, 2013
I haven't changed my stance on things - I want to avoid clipping (which is great for sound quality), and I don't think that turning down the music hurts overall sound quality, something you claim it does.

Let's get back to the very first thing you brought up:

Quote:
"Why would you recommend such a thing and at the same time using an SL1? By turning down the master gain in the software you are losing headroom and you are worsening the s/n ratio. Esp. when using the SL1 which has "only" 16Bit D/A's. If you turn down the volume by 6dB you are reducing the resolution to 14Bit."


Since it's the first thing you brought up and challenged me on the topic, let's start there and discuss it. I agree that turning down the volume brings your softest segments closer to the noisefloor of the device (SL1, SL3, etc) - not great, but my claim is that it doesn't have an affect on the output, especially for dance music. Also, contrary to what you said above, you are actually increasing headroom between the peak of the track and the output limit of the device - that's a good thing. Furthermore, since we've been talking theoretical the whole time, I'm going to use a real-world example to illustrate how you're not losing any sound quality even though you're bringing the sound level of your softest segments closer to the noisefloor of your playback device. Furthermore I will show that it doesn't make a difference if you have a 16 or 24 bit unit which everyone loves to bring up.


To highlight the first issue you brought up, I'm going to go a step further and bring the volume down a whopping -12db and use a real-world example to show how turning the volume down will not have a negative effect on the music being played.

Let's take a dance classic: Orbital - Impact

The track has a dynamic range of 30db - it peaks at 0db and the softest part of the track is at -30db. Awesome - the track is as loud as it can be (without compression or limiting).

I now take that exact same track and load it up in Sound Forge and reduce the volume by 12db - dynamic range is still 30db but it now peaks at -12db and the softest part of the track is at -42db.


Those are essentially identical tracks, each with 30db of dynamic range but they peak at different values (0db and -12db). We don't know where the recorded noisefloor is because we are never able to hear it because the song never gets softer than -30db - so the noisefloor of the track is essentially the softest part of the track.

Now, I could take the second track, save it, then push the volume back up by 12db and it would be identical to the first track. Yeah, I could be bringing up the noisefloor while editing in the digital domain, which would probably be somewhere around -96db, but we can't hear it anyways so it's moot.


So now let's do a device comparison because real-world numbers are never as good as theoretical numbers - according to the specs on the Rane site, the dynamic range of the SL1 (16bit) is 94db and the dynamic range of the SL3 (24bit) is 104db. That means they're saying the noisefloor of their device is somewhere around -94db or -104db for each device respectively otherwise they couldn't really claim to represent audio signals that soft, so we'll give Rane the benefit of the doubt here.

Coming back to the Orbital track that was reduced by 12db, the softest section of the track is at -42db, which is still 52db away from the SL1's noisefloor. That's HUGE! You not only kept that track from clipping, but you're a whopping 52db away from the playback device's noisefloor... and that's on my SL1!


So you probably say, "Mike, that was an unfair comparison because Orbital has no dynamic range"... fine - I pulled up a professionally recorded classical track "Peter Gynt - Suite Nr. 1". Loudest part of the track is -1db and the softest is -50db (essentially the hiss of the hall it was recorded in). It is still nowhere near the noisefloor of the SL1. Heck, even if I reduced the volume of that track by 12db, it's softest part would now be at -62db - that would put the softest section of the song 32db away from the SL1's noisefloor (94-62=32)... on a classical track! In other words, the SL1's noisefloor is still negligible, even for a measly 16bit device.

So really, where's the problem?
DJMark 1:19 AM - 27 June, 2013
Quote:
The track has a dynamic range of 30db - it peaks at 0db and the softest part of the track is at -30db.


Okay this is where you're going wrong and misinterpreting "dynamic range". I don't know how (or with how much time-granularity) you're making these measurements, but even if the very quietest points in the track are measuring -30dBFS, you would almost certainly be hearing program material that's well below -30dBFS at the same time.

To hugely simplify what I just said: if you play a 200Hz tone at -30dBFS, and have (just for an example) a 2,500Hz tone playing at -60dBFS at the same time, you WILL hear the higher tone as well as the lower one.

Here's the issue with lowering gain on a 16-bit system by any significant amount: not only are you reducing the possible dynamic range by one bit for every 6dB, those bottom-most (least-significant) bits that are being dumped are be being replaced with noise and/or distortion (details depend on exactly how software and/or hardware is handling the gain-reduction/truncation/quantization/re-dithering process). Those artifacts, when amplified on a loud high-quality sound system, may well be audible.

The principle is conceptually much the same (though differing in technical details) as how lower-bitrate lossy-compressed soundfiles can sound "fine" at moderate listening levels, but then a/b one of those with a piece of vinyl or a linear-PCM copy of the same track in a club and the differences are readily heard and felt.
msoultan 1:53 AM - 27 June, 2013
Quote:
Quote:
The track has a dynamic range of 30db - it peaks at 0db and the softest part of the track is at -30db.


Okay this is where you're going wrong and misinterpreting "dynamic range". I don't know how (or with how much time-granularity) you're making these measurements, but even if the very quietest points in the track are measuring -30dBFS, you would almost certainly be hearing program material that's well below -30dBFS at the same time.


I don't mind defining things because people often use the same words but have different meanings associated with those words ;) I just scrolled to the softest portion of the song and found that it was around -30db. But like you alluded to below, there could be a softer instrument. I could do a frequency analysis at that soft section and see what frequency is the softest, but I'm guessing that there wouldn't be much program material softer than -40db... and below that the noisefloor of their equipment. And I say that because 40db is a huge dynamic range, something that dance music doesn't normally take advantage of.

But even then, using the classical track is even easier because the noisefloor (and dynamic range) is essentially defined by the hiss of the room, so there's no way that Orbital's dynamic range is going to be anything more than 40db, which is still well away from the SL1's noisefloor, which is where the bitrate comparison actually comes into play, even if I were to pull the volume down a whopping 12db (which yes, is excessive).

Quote:
To hugely simplify what I just said: if you play a 200Hz tone at -30dBFS, and have (just for an example) a 2,500Hz tone playing at -60dBFS at the same time, you WILL hear the higher tone as well as the lower one.


Completely agree.

Quote:
Here's the issue with lowering gain on a 16-bit system by any significant amount: not only are you reducing the possible dynamic range by one bit for every 6dB, those bottom-most (least-significant) bits that are being dumped are be being replaced with noise and/or distortion (details depend on exactly how software and/or hardware is handling the gain-reduction/truncation/quantization/re-dithering process). Those artifacts, when amplified on a loud high-quality sound system, may well be audible.


Ok, you say you're reducing the possible dynamic range... of what exactly? You're not changing the dynamic range of the song unless your applying dynamics processing, which we're not. I agree that you're lowering the dynamic range of the playback device, with the downside of that being that you're bringing the noisefloor up. Like I said above, the noisefloor of the playback device is still well below the lowest volume levels being played back in the track, so you would never hear the difference. Even in the case of the classical track, you still have tons of db before you hit the SL1's noisefloor, and I reference the classical track because you'd never play a track at a dance club that uses that much dynamic range.

Quote:
The principle is conceptually much the same (though differing in technical details) as how lower-bitrate lossy-compressed soundfiles can sound "fine" at moderate listening levels, but then a/b one of those with a piece of vinyl or a linear-PCM copy of the same track in a club and the differences are readily heard and felt.


I'd rather not discuss the differences between MP3/WAVs in this thread because MP3s still have a particular dynamic ranges of their own. I'd love if you chimed in here:

serato.com
msoultan 6:59 AM - 27 June, 2013
Quote:
Okay this is where you're going wrong and misinterpreting "dynamic range". I don't know how (or with how much time-granularity) you're making these measurements, but even if the very quietest points in the track are measuring -30dBFS, you would almost certainly be hearing program material that's well below -30dBFS at the same time.

To hugely simplify what I just said: if you play a 200Hz tone at -30dBFS, and have (just for an example) a 2,500Hz tone playing at -60dBFS at the same time, you WILL hear the higher tone as well as the lower one.


The first paragraph didn't really click until the second paragraph where you mentioned frequencies. By me just looking at the waveform's volume peaks or troughs, that pretty much was showing me the combined frequencies. So I brought Orbital into the spectrum analyzer for very different results!

During the main portion of the track the 62hz freqs were peaking at 0db and the mids to highs were peaking at around -15dbs. Then, however at the soft portions of the track, the 62hz range was still pushing -10db, but the mids were hovering around -35db and the highs sloping off towards -80db at 16khz freqs, where during most of the song the 16hz freq range was seeing an average of -20hz. That means the highs were in-fact seeing a large dynamic range of 60db and getting close to the noisefloor of the SL1. Thank you DJMark, I stand corrected!

I appreciate that you pointed that out because without looking at the track with the frequency analyzer, I would have never seen how large the dynamic range is for the various frequencies.

So, I'm kinda back to square one which is a bit frustrating. I mean, turning the volume down a few db on all the tracks most likely isn't going to be noticeable, and also having a few clips isn't that noticeable either because they're usually happening on the transients. Ultimately so much depends on the program material, but having MP3Gain to give a bit of insight to avoid some clips is still helpful. Running tracks through Platinum Notes would probably also be helpful, but damn that's a lot of work, especially to try and do it on my existing library...

That being said, 92db as a target volume still doesn't seem like a good number because it's pushing quite a few of my tracks into the clipping zone by a few db. I'm thinking I might settle on 90db or so, but I'll probably have to stare at my numbers towards the beginning of this thread. Lemme know you thoughts on this...
nik39 8:10 AM - 27 June, 2013
I am surprised that you're surprised because you brought up the (frequency) band issue before and brushed it off by yourself:

Quote:
Oh, and before someone calls me out on it, in my example with the AC noise in the background and the vocals on top, the vocals would only mask the frequencies of the noise in the background - obviously the frequencies have to be similar so it was a somewhat bad example. But that said, if you have similar frequencies that are louder than background frequencies, they'll mask the background noise and make it difficult to notice those background frequencies.



BTW, I have tried using Orbital - Impact. If you upload the entire song once in 16Bit, and once in 12Bit - I am able to distinguish between those two. Not so easy, cause the 16Bit Original seems to have a bad noisefloor already. I must also admit that it is very difficult distinguishing both version on the loud parts - at least for my untrained ears. But - we're striving for best quality eh? We're already losing lots of quality from the recording to the computer, from the computer to the soundcard, then to the amp, speakers, alcoholized/drunk listeners ;)
DJMark 8:28 AM - 27 June, 2013
"Do as little harm as possible" is a always good practice...especially if you're in a situation where you can't always carefully monitor/measure what you're doing.

As far as the on-screen gain settings...I don't worry about them much unless there's a *serious* level issue (some older DVD rips with average levels around -20dBFS, for example.

I dealt with varying levels/EQ's for years in real time with vinyl records and CD's...I deal with Serato the same way. Even with almost NO use of the onscreen gain knobs it's still easier than before since the waveform can be seen *and* it gives good clues about frequency balance.
Detroitbootybass 6:51 PM - 29 June, 2013
Quote:

I think you should use either SSL auto-gain or MP3Gain, but not both.


I should have made myself a bit more clear - I've dumped Scratch Live's auto-gain and am now adjusting each track individually beforehand with MP3Gain and Audacity (to view the waveforms). Once in SSL, tracks that are a bit louder (or quieter) than the overall mix will be adjusted by the line faders. I keep the up-faders at 3/4ths of the way up so I have some extra room to raise it a bit for quieter tracks. All loudness-related issues will be dealt with via the faders (and in very rare occasions, the EQ knobs).

I also confirmed that my adjustments are only -1.5 dB (never any lower), which is just enough to give myself a tiny bit of wiggle-room in case of using some effects.
DeezNotes 1:31 PM - 30 June, 2013
It's good to see someone else take this whole thing a bit farther. Makes my auto-gain-OCD look like child's play.
msoultan 4:01 PM - 30 June, 2013
I'm gonna take a quick side-note to reply to this fader-issue and then I'll reply to DJMark and Nik39

Quote:
I should have made myself a bit more clear - I've dumped Scratch Live's auto-gain and am now adjusting each track individually beforehand with MP3Gain and Audacity (to view the waveforms).


I'm not sure if I'm misunderstanding you, but you shouldn't be adjusting *each* track individually, you should be highlighting all of your tracks, doing a track analysis/gain on the whole bunch. What that will do is make the apparent volume of *all* the tracks identical - just turning all of the clipping tracks down some amount might stop the clipping but kinda defeats the purpose of using MP3Gain, or SSL's auto-gain for that matter - what they're trying to do is get the apparent volume to be the same among all the tracks.

Quote:
Once in SSL, tracks that are a bit louder (or quieter) than the overall mix will be adjusted by the line faders. I keep the up-faders at 3/4ths of the way up so I have some extra room to raise it a bit for quieter tracks. All loudness-related issues will be dealt with via the faders (and in very rare occasions, the EQ knobs).


You're not really solving the problem by using the faders to adjust channel volumes, you're just moving the problem somewhere else and really creating more of a headache for yourself than you need to. Keeping the faders at 3/4 is one of the hardest ways to mix - my friend was doing the same thing and he was riding the faders and chasing the meters all day... making for a *very* inconsistent output level. You're just increasing workload by a large amount when you really don't need to.

Here's the deal, if you're in the mix and want to cut the volume on a channel really quickly with the fader, you have to bring the fader back *exactly* where you just had it before. If you happen to knock into the fader by accident, then you have to bring it back up or down to the exact point it was before (really bad if you knock it up!). Don't make your life so difficult! Push the channel faders all the way up, use the channel gains to adjust your volumes, and make sure your master levels are good and you're essentially doing the exact thing but you'll be making your life SOOOOOO much easier. You'll notice that most experienced DJs do this (especially scratch DJs and many other DJs as well) as it makes it *really* easy to slam the fader up and down without losing your channel volume settings - I would venture to guess that most DJ mixers are designed with this type of mixing style in mind and this is especially helpful if you use your channel faders to cross-fade because you can focus more on the mix instead of worrying about what the levels are doing and more about the actual mix. Also, if you use SSL auto-gain or MP3Gain, then you'll barely be touching gain settings at all - the key is trying to find the best setting, hence the topic of this thread ;)
msoultan 4:26 PM - 30 June, 2013
Quote:
I am surprised that you're surprised because you brought up the (frequency) band issue before and brushed it off by yourself:

Quote:
Oh, and before someone calls me out on it, in my example with the AC noise in the background and the vocals on top, the vocals would only mask the frequencies of the noise in the background...


Ha.. I still had my blinders on - I should have actually payed attention to what I wrote! I think because I was just looking at the waveform instead of the spectral analysis I was forgetting something that DJMark clued me in on.. I can have a clipping 20hz frequency and I'll still very clearly hear the -60db sound at 1k ;) I was thinking narrow instead of wide - yeah.. stupid.


Quote:
BTW, I have tried using Orbital - Impact. If you upload the entire song once in 16Bit, and once in 12Bit - I am able to distinguish between those two. Not so easy, cause the 16Bit Original seems to have a bad noisefloor already. I must also admit that it is very difficult distinguishing both version on the loud parts - at least for my untrained ears.


Do you remember what kind of things you noticed? Did it tend to be noisier?

Quote:
"Do as little harm as possible" is a always good practice...especially if you're in a situation where you can't always carefully monitor/measure what you're doing.

As far as the on-screen gain settings...I don't worry about them much unless there's a *serious* level issue (some older DVD rips with average levels around -20dBFS, for example.


Actually, it seems one of the most problematic type of tracks are the ones that are very quiet and then have a loud spike - artists such as Mouse on Mars or other weird and experimental tracks - the auto-gain programs have a hard time dealing with those in that they jack the volume up and it clips horribly.

And I honestly don't change my gains very much, but in bringing things into MP3Gain, I realized that lot of my tracks were clipping pretty strongly at the 92db SSL auto-gain setting which spurred this thread.

Quote:
I dealt with varying levels/EQ's for years in real time with vinyl records and CD's...I deal with Serato the same way. Even with almost NO use of the onscreen gain knobs it's still easier than before since the waveform can be seen *and* it gives good clues about frequency balance.


Actually interesting that you mention that because you can't really use the waveform with SSL's auto-gain because it's just showing you the original wave-form size, not the actual waveform amplitude that's being played. This is another area where MP3Gain comes in really handy in that SSL will now show you the actual waveform amplitude.

Quote:
But - we're striving for best quality eh? We're already losing lots of quality from the recording to the computer, from the computer to the soundcard, then to the amp, speakers, alcoholized/drunk listeners ;)


That's the thing - I'm pushing hard on this because I do ultimately want the best quality and I want to make sure that every choice I make has *very* sound advice behind it because the problem we're running into is that we're making very wide-sweeping changes that will have very long-term effects.

A while back (5 years?) when I started using SSL, I settled on 192kbps Fraunhofer-encoded MP3s for my entire library because my drive wasn't huge and I needed to save space. Then I recently got into a discussion with a sound-engineer friend about whether there's a noticeable difference between 192 and lossless, and now the whole clipping thing, and because I'm so OCD about all this stuff, I'm seriously considering going through my library and redoing a bunch of stuff. But if people really can't hear the difference, then I'll make minor tweaks here and there, but the goal here is to have the best quality and make good decisions now that will be good for the long-term.

So, at this point I've learned that minor clipping is ok - it's too bad that MP3Gain doesn't tell you by how much a track is clipping, but you can figure that out relatively easily by changing the target value up and down to see when tracks fall off of the clipping list. Also, turning the volume down too much isn't going to be advantageous because of noise-floor issues so it's really going to have to be a compromise between the two and depends highly on the program material. I just didn't want to accept the SSL auto-gain default of 92db because it was easy without having some justification as to why I was using that value. This thread has been extremely educational and I thank everyone for their patience!!
Detroitbootybass 6:51 PM - 1 July, 2013
Quote:
I'm not sure if I'm misunderstanding you, but you shouldn't be adjusting *each* track individually, you should be highlighting all of your tracks, doing a track analysis/gain on the whole bunch. What that will do is make the apparent volume of *all* the tracks identical - just turning all of the clipping tracks down some amount might stop the clipping but kinda defeats the purpose of using MP3Gain, or SSL's auto-gain for that matter - what they're trying to do is get the apparent volume to be the same among all the tracks.


I know the idea behind both MP3Gain & Scratch Live's 'auto gain' is to make all tracks have the same relative loudness, but it was causing problems on certain songs. The theory is solid on a macro-level, but fails on a micro one. No matter what level I set either program at, some tracks will be too hot and others too quiet. One size does not fit all. Plus, my ears are better than the programs at measuring perceived 'loudness'.



Quote:
You're not really solving the problem by using the faders to adjust channel volumes, you're just moving the problem somewhere else and really creating more of a headache for yourself than you need to. Keeping the faders at 3/4 is one of the hardest ways to mix - my friend was doing the same thing and he was riding the faders and chasing the meters all day... making for a *very* inconsistent output level. You're just increasing workload by a large amount when you really don't need to.


It's not that hard for me - I've been doing it since 1992. It's second-nature for me to mix like that. It's how I learned.

Also, I still mix vinyl records with my digital tracks (somewhere around 40% on vinyl), so the technique would still be used regardless.



Quote:
Here's the deal, if you're in the mix and want to cut the volume on a channel really quickly with the fader, you have to bring the fader back *exactly* where you just had it before. If you happen to knock into the fader by accident, then you have to bring it back up or down to the exact point it was before (really bad if you knock it up!). Don't make your life so difficult! Push the channel faders all the way up, use the channel gains to adjust your volumes, and make sure your master levels are good and you're essentially doing the exact thing but you'll be making your life SOOOOOO much easier. You'll notice that most experienced DJs do this (especially scratch DJs and many other DJs as well) as it makes it *really* easy to slam the fader up and down without losing your channel volume settings - I would venture to guess that most DJ mixers are designed with this type of mixing style in mind and this is especially helpful if you use your channel faders to cross-fade because you can focus more on the mix instead of worrying about what the levels are doing and more about the actual mix. Also, if you use SSL auto-gain or MP3Gain, then you'll barely be touching gain settings at all - the key is trying to find the best setting, hence the topic of this thread ;)



I sometimes enjoy doing micro-cut/pauses of the audio, but I don't use the line faders for that. On my Hak 360, I use the dedicated transformer switches. I wanted to buy, but never never did, a Rane Rotary Empath. On that mixer, there aren't any transformer switches - only a phono/line switch. That won't work with SSL as the audio of the control vinyl would be audible during cuts. In that case, I'd just use the crossfader to do the same thing. For me, the line faders are just for bringing in to, and taking out, the songs with the overall mix.
msoultan 9:11 PM - 1 July, 2013
Quote:
I know the idea behind both MP3Gain & Scratch Live's 'auto gain' is to make all tracks have the same relative loudness, but it was causing problems on certain songs. The theory is solid on a macro-level, but fails on a micro one. No matter what level I set either program at, some tracks will be too hot and others too quiet. One size does not fit all. Plus, my ears are better than the programs at measuring perceived 'loudness'.


I guess I was just saying that I would make those micro adjustments to the gain in SSL - it sounded like you were making those changes in MP3Gain and that's not the most efficient place to make those changes. Plus, if you ever decide to make global changes in MP3Gain and you've made micro changes in SSL, those changes in MP3Gain won't have an effect on your micro adjustments, which is ideal.

Quote:
It's not that hard for me - I've been doing it since 1992. It's second-nature for me to mix like that. It's how I learned.

Also, I still mix vinyl records with my digital tracks (somewhere around 40% on vinyl), so the technique would still be used regardless.


Just because you learned that way doesn't mean it's the most efficient ;) The beauty of using the faders at full-up is that you'll find yourself making very few adjustments to the channel gains on mixer. Yeah, you can do it the way you learned, but you might find that doing it the other way will make you a lot faster on the mixer with less time adjusting volume levels. If you've got the SSL gains dialed in, you'll very rarely be making any gain adjustments with your faders full up. However, if you use 3/4 fader, you'll always be making adjustments.

The other reason the faders are not a good place to make minute adjustments is because lots of times the fader curves aren't linear as many times they mimic the curves available on the cross-fader. Not always, just just something to remember.

As for vinyl, CD, SSL... doesn't make much of a difference - you still have to adjust gain levels regardless. However, if you always have your faders up top, you'll find that you'll be making much fewer adjustments as the method you're using just isn't nearly as precise.
BBN 8:23 PM - 23 October, 2022
I used macMP3gain more than 10 years ago. Because it wasn't updated at some point and stopped working because of an Apple update, I switched to a programm called Smart Gain.

Now Smart Gain isn't supported anymore and can't be used on current Mac OS, so I have to go back to macMP3gain which is called MP3Gain Express 2 these days.
The problem ist I don't remember which settings I used 10 years ago.

Why am I using a tool like this anyway, when there is Autogain in Serato DJ and LED Meters on every mixer? Because only normalized tracks have a big waveform in Serato DJ and low level tracks allways have a tiny waveform where it's hard to see any needed infos like breaks, kicks, snares, whatever.

So if someone in here is using MP3Gain let me know which dB number between 89 and 100 I have to set for a big waveform in Serato DJ and if you allow clipping or not, just let me know.
nik39 10:14 PM - 23 October, 2022
IIRC I've used 92 dB with clipping.
DeezNotes 12:51 AM - 7 November, 2022
I use the aacgain binary on a mac. Tired of using Doug's scripts, I wrote a small python app to do my most common ID3 tasks. I use 95db for main tracks and 93db for instrumentals and acapellas - this gives them an "even" balance to me, rather than the inst/acap being too loud when going between that and the main. Using the binary, the db levels are specified differently (95db = 6, 93db = 4). Clipping enabled. I'll also use lower values (93 or 91 db) on main tracks that are bass heavy.

Example:

1. aacgain -r -f -d 6 -c -s i file.mp3


Prior to using python, I wrote my own Applescript that integrated with iTunes/Music with a menu, but python is faster.