DAWs have an intrinsic sound, why?

[taherbert]

So my studio partner and I were doing some last minute premastering mix tweaks the other night on some stereo mixes prior to sending to mastering. We had recorded them originally to 1-bit DSD files and converted to 96K 24-bit files with Korg AudioGate. It seemed like an ideal time to settle this long running debate we had about whether our two main DAW programs sounded different (Logic Studio 9 vs. Digital Performer 6). So we imported our mixes into both programs and played them both side by side. They sounded different. Logic had this warmer low mid thing going on, DP had this wider upper mid clarity thing going on. But they were different, no doubt about it.

Keep in mind these were the exact same audio files, no difference in sample rate, no DA/AD conversion step, no plug-ins, no change in volume, just one track and master faders at unity. No Rewire application funny business, either.

This obviously has implications about choosing a DAW that have nothing to do with workflow or feature considerations. Now, given that these programs are just playing back identical audio files, why would they color the sound? I would think the programmers would all be using the same algorithms to decode digital audio to make it transparent as possible.

[voodoo4u]

Same converters?

[joeq]

what was your success rate in the blindfold a/b/x?

[taherbert]

Same everything. We took the same sample rate converted files from Audiogate from the same computer folder, imported them into two different DAWs at the rate sample rate, and played them back through the same channels of the same Aurora Lynx converters into the same monitor controller through the same monitors. The only variable was the DAW itself.

[Ethan Winer]

If the wave files really were identical, it could be due differences in each DAW's Pan Law implementation. But before you bother checking that, export both mixes and see if they null against each other.

Also, did you really have each DAW do a SRC on the files?

--Ethan

[taherbert]

Originally Posted by joeq

what was your success rate in the blindfold a/b/x?

Didn't do a blindfold test because there were only two of us, but we went back and forth a couple of times on the same sections of the same songs. I know that would have been more scientific.

FWIW, I was highly skeptical there would be a difference (I wanted to "settle this debate once and for all"), and my partner was totally convinced there was a difference ("I told you so, I've heard this all along."). We both used the same words to describe the differences we were hearing, so I think if our biases were affecting the results, thay'd be offsetting and we'd have more disagreement about the two sounds.

All this being said, I couldn't listen to a radio mix and say "this was definitely mixed in Logic, this one was in Performer, etc." They're very subtle differences we're talking about, but it would have been the case that I would have probably EQ'd the mixes slightly differently if I'd been mastering in one vs. the other.

[DrFrankencopter]

With no plug-ins, and no fader moves I'd say its down to a few possibilities in decreasing order of likelyhood:
1) Pan Law
2) Expectation bias
3) Dithered mixer

This sort of test has been conducted many times, and results in solid nulls once the variables are appropriately accounted for.

Cheers

Kris

[taherbert]

Originally Posted by Ethan Winer

If the wave files really were identical, it could be due differences in each DAW's Pan Law implementation. But before you bother checking that, export both mixes and see if they null against each other.

Also, did you really have each DAW do a SRC on the files?

--Ethan

No, all SRC was done in Audiogate from the 1 bit DSD files. All we did in the DAWs was drag the files into the Audio Bin/Sound Bytes folder and playback. We didn't bounce the mixes or export them or anything like that.

It could be a difference in pan law, DP sounded a little wider than Logic, but we attributed that to more emphasis on the upper mids (which were panned a bit more in the mix) vs. the lower mids (which were a little more centered in the mix).

[joeq]

Originally Posted by taherbert

Didn't do a blindfold test because there were only two of us

oh.

[taherbert]

Originally Posted by DrFrankencopter

This sort of test has been conducted many times, and results in solid nulls once the variables are appropriately accounted for.

Cheers

Kris

I've never done those tests myself, but I wouldn't be surprised that if the audio processing algorithms were different between the different programs, they'd both be written in such a way that when you bounced the audio ITB they would be identical to the original audio. That would be a quantifiable goal you could tell a programmer to accomplish with an algorithm.

Differences in the sound of playback would be harder quantify because you'd have to probably record the audio files from the two DAWs into a third program, probably going through an A-D conversion, then line them up in a DAW again to see if they null. Doing that, you'd probably not get a perfect null even going through the same DAW twice.

[theblue1]

Originally Posted by taherbert

Didn't do a blindfold test because there were only two of us, but we went back and forth a couple of times on the same sections of the same songs. I know that would have been more scientific.

FWIW, I was highly skeptical there would be a difference (I wanted to "settle this debate once and for all"), and my partner was totally convinced there was a difference ("I told you so, I've heard this all along."). We both used the same words to describe the differences we were hearing, so I think if our biases were affecting the results, thay'd be offsetting and we'd have more disagreement about the two sounds.

All this being said, I couldn't listen to a radio mix and say "this was definitely mixed in Logic, this one was in Performer, etc." They're very subtle differences we're talking about, but it would have been the case that I would have probably EQ'd the mixes slightly differently if I'd been mastering in one vs. the other.

If you've got a PC sitting around, you can use the ABX comparator plugin for (the Win-only media player) Foobar to automate the ABX process, including keeping track of your responses and doing the statistics calculatons.

I know people get tired of hearing about ABX testing -- and it is not a be-all/end-all/tell-all solution for all issues -- but once you have a chance to test your assumptions, it can be a real eye-opener. Cognitive distortion in the form of confirmation bias is a very real, pervasive aspect of human perception. It is extremely easy to listen to two very similar sounds and think the difference is subtle but unmistakable -- and then turn around and test those assumptions with true blind testing and find that you really can't tell the difference just by listening.

BTW, the cognitive distortion we call 'confirmation bias' has many aspects... some people tend to hear what they want to hear -- but some people tend to hear what they don't want to hear. (I'm thinking there's an avenue for testing other aspects of the human condition, there, but let's not stray too far afield. )

[Bassmankr]

Pan Law defaults are not the same DAW to DAW and that trips many up so rule that variable out first.

[kstackm]

Originally Posted by Bassmankr

Pan Law defaults are not the same DAW to DAW and that trips many up so rule that variable out first.

Yes Sir !

[Jolly Jimmy]

No reliable testing... no reliable result.

[theblue1]

Originally Posted by taherbert

I've never done those tests myself, but I wouldn't be surprised that if the audio processing algorithms were different between the different programs, they'd both be written in such a way that when you bounced the audio ITB they would be identical to the original audio. That would be a quantifiable goal you could tell a programmer to accomplish with an algorithm.

Differences in the sound of playback would be harder quantify because you'd have to probably record the audio files from the two DAWs into a third program, probably going through an A-D conversion, then line them up in a DAW again to see if they null. Doing that, you'd probably not get a perfect null even going through the same DAW twice.

If you're testing for differences in DAWs, you want to isolate to just the DAWs, to the extent possible.

With regard to null testing -- if you have one file and pass it through two different DACs and track them both in through the same ADC, compare the files, it would be very unlikely for them to null significantly -- unless the two test DACs were running from the same clock. 'Infinitesimally' small differences in timing, when extended across several minutes would cause the two signals to 'drift' and so, in all likelihood, produce a 'difference' file with what would appear to be a large amount of content.

That's just one of the reasons why such 'null testing' is of value only in limited circumstances. But, if two files do null completely, it is reasonable 'proof' that they are virtually identical. Even if the difference file derived from inverting one and summing both has some content, if it's below the nominal threshold of audibility, the two files can be thought of as being audibly equivalent.

[taherbert]

Well, I know all about confirmation bias and the merits of blind testing (I'm a scientist by day). I will try to do a blind test in the next couple of days if I have time and I'll post the results.

As far as trying to null the stereo files: Does anyone know how to record the output of a DAW into another program ITB without doing an SRC or D/A conversion on a MAC?

Update: I found this http://emptymusic.com/software/ABXer.html

[taherbert]

Originally Posted by theblue1

If you're testing for differences in DAWs, you want to isolate to just the DAWs, to the extent possible.

With regard to null testing -- if you have one file and pass it through two different DACs and track them both in through the same ADC, compare the files, it would be very unlikely for them to null significantly -- unless the two test DACs were running from the same clock. 'Infinitesimally' small differences in timing, when extended across several minutes would cause the two signals to 'drift' and so, in all likelihood, produce a 'difference' file with what would appear to be a large amount of content.

Yeah, that's why I don't think I can do a Null test to determine playback differences.

[Jolly Jimmy]

Originally Posted by taherbert

As far as trying to null the stereo files: Does anyone know how to record the output of a DAW into another program ITB without doing an SRC or D/A conversion on a MAC?

I've used Soundflower on a Mac to do this very kind of thing.

[Winey]

But isn't pan law on a single true stereo file irrelevant in this kind of test?

[Arksun]

If there's two of you, just blindfold yourself and get your mate to load up whichever program he wants, DP or Logic, play the recording, without saying a word of course, then load up the other program (or possible the same program) and THEN see if you can clearly make the distinction of which is which.

Do this a few times, whats your success rate?. Do you always detect which was Logic or DP playing the wave file?

It's surprising the impact visuals can have on perception of sound.

Oh and obviously make sure both DAW's are using exactly the same audio interface driver and latency settings etc.

[taherbert]

Originally Posted by Jolly Jimmy

I've used Soundflower on a Mac to do this very kind of thing.

Okay, let's do this! I want to prove there's no difference, it will be one less things to think about. I'll post the results when I get them.

[kennybro]

never mind. mis-post.

[Ethan Winer]

Originally Posted by theblue1

if you have one file and pass it through two different DACs and track them both in through the same ADC, compare the files, it would be very unlikely for them to null significantly -- unless the two test DACs were running from the same clock.

Yes, but as I read this, he's importing the exact same files into two different DAWs. So in that case the two renders should null if they're identical. Even if they're not identical, the residual might give a clue to the differences. For example, if tracks panned hard left and right are absent, but other tracks are audible in the residual, that implies a pan law difference.

--Ethan

[popmann]

Originally Posted by Winey

But isn't pan law on a single true stereo file irrelevant in this kind of test?

Should be irrelevant. Panned hard left and right, no pan law is in effect, as pan law reduces the volume as you bring the pan to center.

Feel free to correct me if I'm wrong. -3db pan law means panned up the center, the volume drops 3db. So, pick whatever one you want--for panning hard left and right should not attenuate or increase the signal.

[Karloff70]

Originally Posted by popmann

Should be irrelevant. Panned hard left and right, no pan law is in effect, as pan law reduces the volume as you bring the pan to center.

Feel free to correct me if I'm wrong. -3db pan law means panned up the center, the volume drops 3db. So, pick whatever one you want--for panning hard left and right should not attenuate or increase the signal.

Logic has three laws, +3,-3 and 0dB. Isn't it that -3 actually dips the middle by 3dB so the sweep from side to side doesn't go for a bump up in the middle and likewise but the other way round +3 the middle stays full and it sees to the sides not dropping to achieve the same result? But Logic still defaults to 0dB I think, where the middle will be louder by 3 dB, which would make sense here, as this means the sides in DP would likely be 3dB louder, alas it would seem wider, and maybe also brighter, considering the way most mixes (and the described one) has more meat in the middle and sizzle on the sides........

So that could still account for it......IF the pan law affects a stereo file like it would two mono panned out.....not sure if it does, haven't tried that out actually. Play that mix and switch that pan law and find out.

I still feel instinctively that regardless whether that might explain this particular instance that DAWs subjectively sound different and that changing the pan law doesn't change the general perception of that. Kind of like the sound changes on a desk with a stiffer power supply. Subtle but real. I bet you could null a mix off a desk with a soft supply with one off the same desk with a stiff supply yet many people on this board would feel/hear a difference.

[Muser]

any small level differences and Pan laws between the DAWs need to be totally illuminated first. I wouldn't rely on the levels being exactly the same without checking.

set both pan laws to 0 and then use something with a very precise level meter, like Nugen Audio Visualizer and work to get those levels as exact as possible.
both in the channel and in the master section.

you can use FreeG or the yellow volume bar in logic to enter high resolution gain values. you click on the yellow bar / move it then use the +/- keys and the value will move fractions of a db.

Then try to get them to sound indistinguishable.

if you use FreeG settup that plug-in preferences to zero db pan law as well and the most precise trim resolution.
then hit (apply). then check those setting have been applied.

[lame pseudonym]

If you'll excuse an interruption in all this fancy technical talk, mixes from my Audacity are very transparent and the same mixes from my Adobe Audition, in comparison, are quite muddy-sounding. I would love to find a software that works like Audition but sounds cleaner. And doesn't cost a million bucks.

[RKrizman]

Originally Posted by popmann

Should be irrelevant. Panned hard left and right, no pan law is in effect, as pan law reduces the volume as you bring the pan to center.

Absolutely correct. Hard left is hard left, and hard right is hard right, no matter what the panning law. The relationship between things in the center and things on the edges has already been established in the mix.

-R

[George Necola]

I hear differences in DAws. Even the manufacturer do (Avid claims the new pt10 sounds a lot better then 9.x).

The things I check in a DAW:
1. Phase stability
2. Alignment/delay compensation
3. Prioritizing of channels

They all sound different.

Sent from my GT-I9100 using Gearslutz.com

[Karloff70]

Originally Posted by George Necola

I hear differences in DAws. Even the manufacturer do (Avid claims the new pt10 sounds a lot better then 9.x).

The things I check in a DAW:
1. Phase stability
2. Alignment/delay compensation
3. Prioritizing of channels

They all sound different.

Sent from my GT-I9100 using Gearslutz.com

I love your no messing straight to the point-ness, George!!! Love it! Especially delivered from the mouth of your avatar.....lol.