Plz work on a good sound engine (Been DONE, and done again. and again. and again...)

[manning1]

all i want for xmas is an aston martin
with the vantage engine....ho ho.

[Till]

!ignore

[cgrafx]

Ok,

I'll try and post a reasoned answer without stirring the pot. Actually there are differences in the sound of every DAW Engine. They can range from subtle to .. well not so subtle.

If each DAW engine processed the data in the exact same way, then there would of coarse be no differences. However, each engine does not process/sum the data in the same way.

Each has its own set of math routines, with its own set of rounding errors, number limits, etc, etc.

While the differences in most cases are probably very small, there are differences, and no two DAW engines will produce the exact same result. (The precision of the math is different).

[Jae.Thomas]

Originally Posted by cgrafx

Ok,

I'll try and post a reasoned answer without stirring the pot. Actually there are differences in the sound of every DAW Engine. They can range from subtle to .. well not so subtle.

If each DAW engine processed the data in the exact same way, then there would of coarse be no differences. However, each engine does not process/sum the data in the same way.

Each has its own set of math routines, with its own set of rounding errors, number limits, etc, etc.

While the differences in most cases are probably very small, there are differences, and no two DAW engines will produce the exact same result. (The precision of the math is different).

where exactly is the reasoning/evidence in your post?

welcome btw, nice to meet you

[ieso]

Fred and Monty are back -- in force this time.

http://forum.cockos.com/showthread.php?t=19062

But who am I to argue with a guy who has a disco at his disposal?

I mean, a whole fucking disco.

[LOSER]

Originally Posted by ieso

But who am I to argue with a guy who has a disco at his disposal?

I mean, a whole fucking disco.

Hahah, man you're cracking me up!

[sceyefeye]

I want the whole disco too, still, it probably won't make me a better - actually scrap that "better" - won't make me a musician or a top notch DJ

but perhaps the loud noise will garner me some fans?

[cgrafx]

Originally Posted by Jason Brian Merrill

where exactly is the reasoning/evidence in your post?

welcome btw, nice to meet you

Thanks for the welcome

You always have to consider the precision when doing math. Are you rounding to the nearest whole number, tenth, hundredth, and so on.

You will get different results using floating point routines then when using integer routines.

When you program, you have to explicitly pick math types (single-precision, double-precision). This explicitly deals with the number of bits used in the calculations.

Keep in mind, the theoretical numbers will be the same, but the specific implementation has limits. If I use the built in math routines supplied by the hardware I will get one result. If the hardware changes, I will may get a different result.

I don't mean that 2+2 = 4 will be different, I mean that 5/7 will give a different result in the decimals (is the result 0.7143 or 0.71428571 or 0.714285714285714)

If I use the first result (0.7143) for further calculation such as another division or multiplication I will get a different number than if I use the second or third representation.

So (5/7) x (5/7) x (5/7) =

0.36445335 (using 0.7143)
0.36443148 (using 0.71428571)

This is a very simple example, but consider the number of calculations that are being done and I think it should become clear that the results will vary depending on the specific coding and math routines used.

The end result being a variation in what you hear. Subtle differences perhaps, but they will exist.

pg

[ngarjuna]

The end result being a variation in what you hear. Subtle differences perhaps, but they will exist.

You would at least agree that these differences would show up in a null test, right? Specifically, that two files which differ in rounding would not have the bit for bit accuracy to completely null?

[ieso]

now, having said all that, prove that reaper sounds different than any other daw.

when I first tried reap (coming from cubendo) I definitely noticed a difference -- then I set up the pan laws, and adjusted everything in the prefs as I was accustomed to in cubase and, ta da, same sound.

now all this silly 'sound of daw' shit can just cease until anybody can actually prove it.

[cgrafx]

Originally Posted by ieso

now, having said all that, prove that reaper sounds different than any other daw.

when I first tried reap (coming from cubendo) I definitely noticed a difference -- then I set up the pan laws, and adjusted everything in the prefs as I was accustomed to in cubase and, ta da, same sound.

now all this silly 'sound of daw' shit can just cease until anybody can actually prove it.

I wasn't talking about reaper specifically. All DAW engines sound different (not necessarily vastly different). You certainly aren't going to argue that audio software hasn't changed (I don't mean features) in sound quality over the years.

16-bit, 24-bit processing, 32-bit, 48-bit or greater mix engines. It does change the math results.

[ngarjuna]

Originally Posted by cgrafx

I wasn't talking about reaper specifically. All DAW engines sound different (not necessarily vastly different). You certainly aren't going to argue that audio software hasn't changed (I don't mean features) in sound quality over the years.

16-bit, 24-bit processing, 32-bit, 48-bit or greater mix engines. It does change the math results.

And do the differences null? If not, please post the tests that prove that the files don't null. While you're at it, be ready to serve up the actual project files so that people can confirm that no extra/different DSP occurred, that pan laws all conformed equally, etc. If you're going to take the position that there are differences in the mix bus (an old chestnut which has never, in any thread I've ever checked out, come down to anything other than user error), you better have some ammo ready.

[cgrafx]

Originally Posted by ngarjuna

You would at least agree that these differences would show up in a null test, right? Specifically, that two files which differ in rounding would not have the bit for bit accuracy to completely null?

Yes, a null test would show variation. (Not huge differences, but there will be bit level differences.

Please don't read this as a slight against Reaper, I'm not discussing the audio quality of any product here, only that differences between DAW audio engines exist. They will produce different final results. Different, not being better or worse.

Each programmer has to make choices about how thier specific engine processes audio. These choices make a difference in the digital processing chain and will produce variances from each other.

pg

[ngarjuna]

Originally Posted by cgrafx

Yes, a null test would show variation. (Not huge differences, but there will be bit level differences.

Please don't read this as a slight against Reaper, I'm not discussing the audio quality of any product here, only that differences between DAW audio engines exist. They will produce different final results. Different, not being better or worse.

Each programmer has to make choices about how thier specific engine processes audio. These choices make a difference in the digital processing chain and will produce variances from each other.

pg

Okay, as long as we're speaking the same language here...again, please post examples of this. Every actual test I've ever seen online has nulled. Entirely. 100%. The only ones that didn't had variations in DSP in the signal chain or differences in how the pan laws were set. If we're going to accept this allegation, that DAWs sum differently, there needs to be some evidence of this; we're even in agreement that the evidence should exist if your point is true. It's not enough to theorize that different programs might round differently. We need to prove this or it's just another baseless allegation.

And it's not about whether or not your allegations are denigrating REAPER or declaring it the supreme DAW sound in the universe. It's about proving an allegation that has been shown to be baseless over and over again throughout the history of teh interwebs.

[nightscope]

Originally Posted by cgrafx

Yes, a null test would show variation.

Have you ever done a null test? If so with which DAW's.

ns

[nightscope]

Originally Posted by ngarjuna

And it's not about whether or not your allegations are denigrating REAPER or declaring it the supreme DAW sound in the universe. It's about proving an allegation that has been shown to be baseless over and over again throughout the history of teh interwebs.

Exactly.

cgrafx, address the question of null tests and your empirical experience with them. Have you any?

ns

[ieso]

Originally Posted by cgrafx

I wasn't talking about reaper specifically. All DAW engines sound different (not necessarily vastly different). You certainly aren't going to argue that audio software hasn't changed (I don't mean features) in sound quality over the years.

16-bit, 24-bit processing, 32-bit, 48-bit or greater mix engines. It does change the math results.

I understand all that. Of course, reaper itself will sound vastly different if I mix at 16 bit instead of 24 or 32

Everybody already knows this. But what we're saying is prove that reap sounds different than daw x at the same sample rate, bit depth, etc.

Can't be done. They always null.

[cgrafx]

Originally Posted by nightscope

Exactly.

Okay, as long as we're speaking the same language here...again, please post examples of this. Every actual test I've ever seen online has nulled. Entirely. 100%. The only ones that didn't had variations in DSP in the signal chain or differences in how the pan laws were set. If we're going to accept this allegation, that DAWs sum differently, there needs to be some evidence of this; we're even in agreement that the evidence should exist if your point is true. It's not enough to theorize that different programs might round differently. We need to prove this or it's just another baseless allegation.

ns

I hope your not suggesting that every application that processes audio, does so in exactly the same way with exactly the same results (regardless of bit depth, headroom, or math processing).

When you say that every test you've seen online has nulled unless there were variations in the DSP signal chain is exactly the point. There ARE variations in the DSP signal chain, by virtue of the fact that they all are designed by different programmers and use different math engines.

Just exactly where those variations will show up in the final audio chain will depend on design factors that only the programmers have knowledge of.

I wasn't adding comments here to get into a battle of some sort. Only to provide some explanation for the original posters comments (well stated or not).

You can use the information I have provided or not.

pg

[schwa]

In the real world, two DAWs mixing the same sources with the same pan laws and the same resolution do null.

Somebody really needs to do a solid blog post or something on this, some persistent resource people can point to instead of just arguing the same points over and over.

[Lawrence]

Originally Posted by schwa

In the real world, two DAWs mixing the same sources with the same pan laws and the same resolution do null.

Somebody really needs to do a solid blog post or something on this, some persistent resource people can point to instead of just arguing the same points over and over.

The problem is, and I agree with you in theory, that raw summing is not real world use. Because daws null at static levels during summing doesn't mean that the more complex math being performed in a real world session will come out identical.

A better (more realistic) test would be to use two daws on the same system with identical plugins calibrated to about 0.1db and then see what happens.

It may still null, but summing tests are not the final arbiter of the final sound. We do a lot more than basic summing when mixing.

I'd like to see a simple 8 track mix, with a couple of plugs on each track and with pan laws identical and with the plugs at precise processing levels and then try to null.

Just to see what happens.

I'm pretty sure some daws handle plugin bus summing differently than others. I know Nuendo supposedly uses some proprietary "dither-like" process on it's audio so... it's not exactly a closed discussion.

I think that if there are differences they are irrelevant in real world use but I also think there actually are differences.

[Zee]

Originally Posted by cgrafx

I hope your not suggesting that every application that processes audio, does so in exactly the same way with exactly the same results (regardless of bit depth, headroom, or math processing).

pg

Sadly, very sadly, YES THEY DO !!!! (They being them and not the daws they use.)
I am amazed.

[cgrafx]

Originally Posted by Zee

Sadly, very sadly, YES THEY DO !!!! (They being them and not the daws they use.)
I am amazed.

Ok.. maybe they do...

[schwa]

Originally Posted by Lawrence

I'd like to see a simple 8 track mix, with a couple of plugs on each track and with pan laws identical and with the plugs at precise processing levels and then try to null.

Awesome, thanks for volunteering!

[Lawrence]

Originally Posted by schwa

Awesome, thanks for volunteering!

Hehe.

I actually just did a quick single track two plugin test and proved your point. I may take it farther.

Same audio file, with a RenComp and a Linear EQ inserted using the same presets in Reaper and Cubase. Imported the Reaper render into Cubase, imported the original file into Cubase and added the same two plugs with the same two presets and rendered that track to a new file and flipped phase.

Silence.

Next? Busses / groups with plugins. Then maybe a straight linear fade from identifyable peak to peak.

[ngarjuna]

I don't disagree with what you're saying, Lawrence, but that is a whole different matter. But the old chestnut that shows up again and again is that the actual raw summing is different. And no matter how many times it comes up, the people making the claim are never willing or able to provide actual non-summing examples. Hence why people call BS and these threads turn into silly discussions of naked Twister.

And yes, cragfx. That is exactly what's being suggested. That the pure mix engines of various DAWs do the same exact math. That is the entire argument in a nutshell.

Schwa is right, though: it's time to put this baby to bed. I'm going to go back to that thread from earlier this week about "Which other DAWs do you use" and see if I can enlist the help of some other REAPER users. Who knows, maybe I'll get results I'm not expecting. I'm frankly an open-minded guy and I will believe the evidence when I see it; it's just that nobody has ever ponied up anything better than "Me and my friends can totally hear the difference!!!1one!" Since the brain is intricately influenced by what it's expecting to hear and makes lots of subtle compensations (some known, some probably unknown at our current level of science/anatomy), that could possibly be the worst test on earth.

But indeed, let's see if it really nulls. And I agree, in addition to just pure mix summing, it would be appropriate while we're at it to compare the way effects are processed as well. Let me do some thinking and see if I can construct something conceptually to compare apples to apples.

[Zee]

Yeah, I know it is anything but required here but hey guys ?
How about using your ears too ?

[cgrafx]

Originally Posted by Lawrence

A better (more realistic) test would be to use two daws on the same system with identical plugins calibrated to about 0.1db and then see what happens.

The difficulty here is in the test process.

You need to test for both frequency and amplitude across at least 20hz to 20khz and from 0 to clipping.

This would need to be done for a both single and multiple channels.

And while I have no doubt variations between DAWs would be found, the end result would unfortunately not really be of great use, as the argument would still be about which one sounds better.

And as we all know, MY DAW sounds the best. Thats why I use it. (please note smiley).

Anyway.. as many will note, I have Zero credibility in this forum as these are my first posts so I hope they will be taken as the informational responses they were intended as.

For the record and by way of some minimal introduction:
------------
- I'm 48.
- I was building Analog Audio gear by age 12.
- My formal background is in Electrical Engineering.
- I worked as a live sound engineer for almost 20 years.
- I used to write custom assembly language software for 8088 and 6800 processors back in the day. I mostly deal with PHP and MySQL these days.
- Music for me is hobby, but one I take seriously and that I have at least a fair level of competence in.
- Troubleshooting for me doesn't mean swapping out boards until it works. It means finding the actual source of the problem (weather that is a blown transistor or a fubar'd line of software code)

[Xenakios]

Originally Posted by Zee

Yeah, I know it is anything but required here but hey guys ?
How about using your ears too ?

Ears are not reliable for any testing.

[bullshark]

Originally Posted by Zee

Yeah, I know it is anything but required here but hey guys ?
How about using your ears too ?

Ears are the worst test instrument on earth. Scientific fact number one: humans have extremely poor memory of sound. Number 2: How humans perceive sound is heavily influenced by their other senses, meaning if you listen to a sound in a room that smells bad, you'll think it sound bad too, or at least worst than if the room smelled nice and fresh (tip for aspiring engineer: make sure your mixing room smell nice and is painted in clean, calming colors, your clients will think you're a genius). Third, to prove or disprove such an assertion, one needs facts, not delusional dreams.

[schwa]

Originally Posted by cgrafx

You need to test for both frequency and amplitude across at least 20hz to 20khz and from 0 to clipping.

I hope we can all agree that simply observing that the actual sample data is identical (that is, it nulls) is sufficient evidence that the actual audio content is identical.

[vocalnick]

Originally Posted by Zee

Yeah, I know it is anything but required here but hey guys ?
How about using your ears too ?

If you're looking at two cows and one of them looks like a llama to you, you're not actually dealing with a defective cow.

You've just got something wrong with your head.

[vocalnick]

Originally Posted by schwa

I hope we can all agree that simply observing that the actual sample data is identical (that is, it nulls) is sufficient evidence that the actual audio content is identical.

That seems (amazingly) to actually be a contentious point here!

[Lawrence]

Hmm... I just did a test on a 3 bar section of stereo audio.

1. Straight fade from 0 from exact start to exact end.
2. Starting at dead center and panning up to the first bar, all the way back across to the second bar and then all the way back across to the end, with the automation nodes exactly on the bar lines.

Pan laws at 0 in both. Project tempos at 120.

The files were different. I didn't bother summing/nulling as I can see they're different. The Reaper render is on the bottom.



It does seem to suggest that not *all* the math is identical as the Reaper render faded much slower and much less.
Here is the Cubase screenshot...



And the Reaper shot...

[Zee]

Well, I don't know for you guys but....

If I record a track in a daw and then feel like eq'ing and adding compression and then some reverb then some high-end exciter and high shelf eq...

If then, I use the same setup to record the same riff in an other daw and then feel like ''wow, it already sounds good''...

At first I'll think I just magically got lucky
But, when it hapends for every take of every track and then I realise how easy it gets to make the mix sound alive, I'll tell myself that the second daw sounds better.... Even if I wouldn't know why...

[LOSER]

Originally Posted by cgrafx

You need to test for both frequency and amplitude across at least 20hz to 20khz and from 0 to clipping.
[...]

Good luck with that since there is virtually no possibility to clip (internally) in a 64bit floating point environment.

[Magoostus]

Originally Posted by cgrafx

Thanks for the welcome

You always have to consider the precision when doing math. Are you rounding to the nearest whole number, tenth, hundredth, and so on.

You will get different results using floating point routines then when using integer routines.

When you program, you have to explicitly pick math types (single-precision, double-precision). This explicitly deals with the number of bits used in the calculations.

Keep in mind, the theoretical numbers will be the same, but the specific implementation has limits. If I use the built in math routines supplied by the hardware I will get one result. If the hardware changes, I will may get a different result.

I don't mean that 2+2 = 4 will be different, I mean that 5/7 will give a different result in the decimals (is the result 0.7143 or 0.71428571 or 0.714285714285714)

If I use the first result (0.7143) for further calculation such as another division or multiplication I will get a different number than if I use the second or third representation.

So (5/7) x (5/7) x (5/7) =

0.36445335 (using 0.7143)
0.36443148 (using 0.71428571)

This is a very simple example, but consider the number of calculations that are being done and I think it should become clear that the results will vary depending on the specific coding and math routines used.

The end result being a variation in what you hear. Subtle differences perhaps, but they will exist.

pg

ya, ok. last time I checked, my AD converters didn't even record in fractions

[Lawrence]

My theory (which is near untestable) is that as different parts of daw code start interacting like they do in a real work mix or whatever, there will be measurable and maybe even audible differences.

It has nothing at all to do with sound 'quality' as you turn knobs until you hear what you want to hear.

[Xenakios]

Originally Posted by Lawrence

Hmm... I just did a test on a 3 bar section of stereo audio.

1. Straight fade from 0 from exact start to exact end.
2. Starting at dead center and panning up to the first bar, all the way back across to the second bar and then all the way back across to the end, with the automation nodes exactly on the bar lines.

Pan laws at 0 in both. Project tempos at 120.

The files were different. I didn't bother summing/nulling as I can see they're different. The Reaper render is on the bottom.

For some amusement, try that same test with a 1000hz sine wave sound source... edit : hmm, actually your test is maybe a bit too long in time to reveal the "interesting" properties of Cubendo's automation...edit 2: argh, it's not dependant upon that, only the ASIO buffer size...Too sleepy over here...

[bullshark]

Originally Posted by Lawrence

Hmm... I just did a test on a 3 bar section of stereo audio.

1. Straight fade from 0 from exact start to exact end.
2. Starting at dead center and panning up to the first bar, all the way back across to the second bar and then all the way back across to the end, with the automation nodes exactly on the bar lines.

Pan laws at 0 in both. Project tempos at 120.

The files were different. I didn't bother summing/nulling as I can see they're different. The Reaper render is on the bottom.



It does seem to suggest that not *all* the math is identical as the Reaper render faded much slower and much less.
Here is the Cubase screenshot...



And the Reaper shot...

Yes, but there are 6 different shape of nodes in Reaper automation, all with different algorithm...and how many are there in Cubase? You'd have to choose nod with identical algo for a test like this to be a valid test of the audio engine IMO.

[vocalnick]

Originally Posted by Zee

Well, I don't know for you guys but....

If I record a track in a daw and then feel like eq'ing and adding compression and then some reverb then some high-end exciter and high shelf eq...

If then, I use the same setup to record the same riff in an other daw and then feel like ''wow, it already sounds good''...

At first I'll think I just magically got lucky
But, when it hapends for every take of every track and then I realise how easy it gets to make the mix sound alive, I'll tell myself that the second daw sounds better.... Even if I wouldn't know why...

Yes, but if you decide to interrogate that assumption a bit further with some empirical testing (your ears and mind are far from objective) you might find that the output from the two DAWs are in fact bit-for-bit identical.

And if that occurs, and you are capable of logical reasoning at a basic level, you'd have to conclude that your ears were in fact playing tricks on you.

Even if you didn't know why