Mastering software

[Timeline]

Waitaminnit, I think I misread the original post, based on BobK's response and then re-reading some things here.

I thought you were referring to Audio Mastering, but I guess you meant CD Mastering? And in that case, DDP is of prime importance, n'est-ce pas?

My point was that Null may not be the end all be all test for sound. Bill Whittlock reiterated the tests done by Dean Jensen way back in '88 but is thinking about revisiting the subject for such things as Digital Workstations or software. I hope he looks closely at computer interactions and CPU speed and how these might also change sounds we hear in an individual application.

If Jensen was right about the spectrum of sound not representative in the tone from one to another then possibly other types of test gear can get to the bottom of why my ears, or, highly exaggerated imagination could make things appear different. I can say this though. I have three pairs of Sennheiser 280 pro cans that are NOT knockoffs and they all sound different to me yet some sound better than others and i have a favorite. So, it's analog too we need to discover about because when i plug them into my Tascam DA-3000 monitoring the same sample rate from my RME and it's headphone output the sound is different as night and day yet, both AD-DA's did the work of presenting me with the same analog feed. Yes maybe impedance but both headphone outputs are as close to DC/low impedance as one would expect.

There are so many reasons sound can change as mentioned above including buffer settings I'm sure there are better answers why mastering software could differ sonically.

I'm tech wise and not all knowing by any means but seek the experience of supper techs like Whittlock and others to tell me how they measure and what's going on with "Null". I hope some of my posts from techs help and thanks for chiming in TO ALL from what ever perspective your coming from, even those who ridicule the possibility that there actually might be other measurements that are needed to sus tone of systems, albeit subtle as they may be.

Bills followup after reading on of these posts with a few interesting experiences:

I had no idea folks were doing a digital version of “null testing”.

In the digital domain, of course, proving that what came out is identical to what went in is ridiculously easy – even to find a single non-identical bit in billions. But maybe that’s not what’s being discussed. The reply does ramble a lot and makes one error: to change the volume of a digital file, you don’t just add a number to each one in the music file, you must multiply each number in the file … and then comes round-off or truncation errors, etc. Manipulating numbers in a digital file, a.k.a. DSP, is just as prone to audible problems as the analog counter-part. The unquestioned benefit of digital is that files can be copied over and over without error (provided, of course, that some sort of on-the-fly concealment or interpolation isn’t at work … as it is in CD players, for example, when the redundancy built into the format can’t fix read errors). All that being said, I’m not sure I even know what “mastering software” does these days. However, I do remember from my years at Capitol (1981-1988), when Capitol was evaluating hardware for CD mastering we found some truly awful sounding sample-rate-converters. The digital hardware wasn’t the problem … it was the math programming that was riddled with overflow and truncation errors … along the lines of forgetting that when you divide by zero, the result is certain to be beyond full-scale … and what does the software do then? Philosophically, there’s always the problem of folks (especially the young) with a superficial knowledge of technology thinking that certain things are either simple or perfect. After you work in engineering long enough, you learn that virtually everything is a tradeoff or, put another way, there’s no free lunch.

[Shooshie]

Do not confuse analog null tests with digital ones. The null test is not nearly as useful in the analog domain, especially when dealing with amplifiers, as in the digital domain, where the sound is simply assembled from a bunch of math problems which in null tests either do or do not add up to zero when you put a minus sign on one set of them.

The criticisms of null tests when dealing with amplifiers is valid. Nevertheless, it does not apply to digital tests within a DAW.

Shooshie

[Shooshie]

Bills followup after reading on of these posts with a few interesting experiences:

I had no idea folks were doing a digital version of “null testing”.

[deleted part]

Philosophically, there’s always the problem of folks (especially the young) with a superficial knowledge of technology thinking that certain things are either simple or perfect. After you work in engineering long enough, you learn that virtually everything is a tradeoff or, put another way, there’s no free lunch.

Yes, that can be a problem, but do not assume that when someone believes that under certain conditions a null test will work (especially when it is demonstrated and does work), that said person is superficial, young, not an engineer, or doesn't understand about tradeoffs. Someone always pays for the lunch, but in certain cases the null test always tells the truth, regardless of who is paying. Understanding when you're looking at one of those cases is probably the most important part of the test. That does require knowledge!

When you're right about a null test, and YOU KNOW WHY you're right, but someone says "he's just too inexperienced to know that he's wrong," the doubters are just too full of themselves to finally come to grips with the truth. Those people need to reassess their assumptions.

You've got to know when and why a null test will yield valid information. Then, when it does, you've got to trust it and trust your own knowledge. And if you have a phalanx of doubters, who gives a flying f*ck? You know why it's accurate in that case, so you use your information to help you build a stronger studio, to increase your knowledge, and you move on. Who cares whether the rest of them do or not?

Shooshie

[bayswater]

Consider a world where a digital null test does not work: This would be where two files, identical, except one is the inverse of the other, are actually different, even when you reverse the inversion. So you could copy a file, invert it, invert it again, and it would be different. It means you could make a copy of an audio file and it would be different from the original. You could save sound from RAM to disk in a save operation, then from disk to RAM in a load operation and get something different from the sound you started with. The songs you burn to disk would be different from the songs you recorded. Each copy of your CD would be different. Each time you played a CD it would produce something different.

[Shooshie]

If my patience is wearing thin, please understand that it's not because I want to ridicule people who disagree with me. I'm genuinely growing angry at this constant implication that if we just understood more about engineering, or if we'd been there in the 60s or 80s or whenever, then we'd know how wrong we are. I say "we," meaning those of us who know that you can use null testing to show certain tracks either are or are not identical.

If you introduce any outside differences, such as analog impedances, dirty power, worn components, or different d/a or a/d codices into the picture, then those variables may cause invalid readings (but they also may not, depending on how the test is conducted).

I've spent my entire adult life studying physics, acoustics, electronics, audio, wave principles, chaos theory, fractal geometry, and digital computing principles as they became known and accessible beyond the labs in which they were invented or developed. I was knowledgeable on some of these things as far back as the 1960s. I watched the science of chaos theory and fractal geometry being developed after the publication of James Gleick's book, Chaos. If anyone is sensitive to the potential differences between two digital files, I'm certainly one of them. I have made many predictions in my life, based on my own studies, which soon became tested, published, factual experiments done by others.

I'm often wrong about things, but I quickly adjust my thinking when my errors are pointed out. That said, I think we should lay to rest this ridiculous argument about null tests.

Gary, your arguments that null tests aren't the "end all be all test for sound" are extremely misdirected, and your implications that we are naive are insulting to the integrity of people who have put countless hours —years— into the sciences behind the arts that we practice. Doing a null test on two tracks from the same DAW, especially when the actual test takes place before the digits ever become "sound," produces either a pass or a fail state. It depends a little on the ingenuity of the tester to make sure that the test is pure and reliable, but when the test comes up null, there's nothing random, accidental, or variable about it. It is NOTHING like testing two amplifiers side by side. These are two strings of numbers. If you subtract one from the other and get zero, those strings are identical. If we're talking about pure zeroes, null tests are inarguable. Without difference. If you A/B those tracks and think you hear a difference, it's the same as hearing one track over and over, and thinking it is different each time.

If you read this post again (assuming I do not edit it), the words will not change. You will read the same words over and over. Your reaction to it may change as you first assume that my intentions are hostile, and then gradually realize after several readings that I'm merely defending my own integrity from implications based on misunderstandings about what the word ZERO means. Humans are wired up in such a way that they understand more deeply each time they hear or read something, whether it is words or music. Same with art. Familiarity brings not contempt, but understanding. That's all I'm asking for: please understand the actual facts behind the words, and please stop implying that there is "more for the wise." I KNOW the things you are saying, Gary. But you are not getting what •I'M• saying. We're talking about apples and oranges. Apples and apple-pickers. Or apples and jpegs of apples.

Let's quit talking analogies and use a concrete example:

Two tracks, a brief sample:

454805132, 454805134, 454805138, 454805147, 454805162, 454805184, 454805172, 454805195, 454805200
454805132, 454805134, 454805138, 454805147, 454805162, 454805184, 454805172, 454805195, 454805200

Those are identical. If ALL other samples from those two tracks are also identical, and if you subtract one from the other, you get null. Not some of the time, not in some places but not others, but ALL the time in EVERY place. You can't play those two and get different results. They are the same.

We're not talking about machine failures.
We're not talking about amplifier fluctuations.
We're not talking about bad code written by immature audio coders.

We're talking about two tracks — two strings of numbers — which are identical. What happens to them once you play them through all your gear and effects is your business, but you cannot deny that these two tracks are identical before they reach the variables you add to them.

If they are not identical, they will not produce null results. But if they produce null results, they are identical. There is no finessing that. There is no deeper understanding of it. Your ears cannot hear differences that are not there. (remember? If there are audible differences, the null test failed.)

Could we please end this now, and stop with the posts that attempt to undermine the integrity of people who have seen null results from tests such as this? Your ears cannot hear differences between identical tracks. Period. That's why we have words such as "identical." Identical doesn't mean "close." It means there is no difference. Any difference you hear was put there between your ears. It didn't come from the tracks.

Shooshie

[Timeline]

Yes I'm a Null doubter but I'm not young and i never said "extremely misdirected ever Shooshie. Those are your words.

For me its not black and white and opinions are fine but for me I lean on experienced engineers, designers who have years of scholastic and design knowledge to test these assumptions and come up with correct answers. These are the people who seem to understand more about digital sound out there than people like me and some of you who are basically audio mixers, recording engineers, musicians etc. with hardened opinion.

I look forward to reading the next vetting progression of the Null test in the AES journal and I'm happy to keep an open mind and friends in that part of the industry who can explain it all so I can understand it. I knew the late Dean Jenson and he and I discussed sound differences in analog many times. He was working on digital about the time he blew his head off for some crazy reason, could it have been the null test controversy ? Hw coined the industry concept that digital could not become a standard over analog sonically until SR's were operating at at least 200khz. This became the goal and why we have high, HD apps and hardware today and even some at twice 192K SR..

Number one on my personal need for undrsranding list is why some mastering/DAW applications sound different from each other. Would features be the only reason these makers feel they can charge such a different range of prices for their apps and get it? Sonic Studios Amarra is $2500+, Stienbergs are up there too, Motu and others reasonable but why the extremes and varied pricing if they all do the same thing and "NULL". And If they all null, as I have seen written here and on other more technical boards discussing this, then whats the point of caring at all and why are we not just buying the cheapest app with the most features. Something else seems to be afoot here my fellow Sherlocks so lets listen and care, "Give a F_ck".

For me, I'm keeping the open mind.

[Shooshie]

Yes I'm a Null doubter but I'm not young and i never said "extremely misdirected ever Shooshie. Those are your words.

For me its not black and white and opinions are fine but for me I lean on experienced engineers, designers who have years of scholastic and design knowledge to test these assumptions and come up with correct answers. These are the people who seem to understand more about digital sound out there than people like me and some of you who are basically audio mixers, recording engineers, musicians etc. with hardened opinion.

Extremely misdirected means that you're spreading the topic around to other things. All I'm arguing is the meaning of NULL. Two identical digital tracks — two tracks which cancel each other out —  cannot have audible differences.

Anything otherwise which you might wish to say is misdirected. It means you're not understanding the word "zero." And when you bring obfuscation and opinions into the domain of fact, you're EXTREMELY misdirecting the argument.

I'm arguing one thing, Gary: Two tracks with identical numbers will nullify if one is subtracted from the other. If they nullify, they are the same. If they are the same, you cannot possibly hear differences between them.

That is the sum-total of my argument. It is undeniable. If it were up for opinion, DAWs would not even work.

If it were up for opinion, raising your volume fader might lower the volume. Saving a file might erase it. Setting your desktop color for red might yield blue one day and green the next. But those things do not happen, because computers do numbers very well. With error correction, they're pretty much exact, all the time.

CDs are a different story, but we're not necessarily talking about that. I'm JUST talking about two tracks that nullify each other. I don't care where they came from. If they produce a null result, YOU CANNOT HEAR ANY DIFFERENCE BETWEEN THEM.

To say that there are lots of opinions on this is misdirecting it, because you are not looking at the root of this argument: null is null. Two identical tracks are identical. "No difference" means that neither you, nor anyone you can name, living or dead, can or ever could hear ANY difference between two identical tracks.

That's not subject to debate. It's fact.

Shooshie

[Timeline]

OK Shooshie. I understand your you are likely a hundred times more knowledgeable than I on these Digital null experiences and could very well be right that by using the Null Test no sound differences can be heard. If i was to lock into that brain fixed stereotype than I could never be open minded enough to learn anything new again IMHO. Where was Physics when String Theory came along? Oh excuse me, another obscuration.

I plan to keep an open mind

[FMiguelez]

Where was Physics when String Theory came along? Oh excuse me, another obscuration.

As far as I know, String Theory is still mostly philosophy, since their proponents are still not even sure of exactly what it is it's telling them. I don't think it can be considered truly scientific at this point in time, since it gives no clear predictions, and most importantly, we currently have no means of testing those predictions.
But it is definitely fascinating and totally worth exploring more.
If it turns out to be correct, and we ever actually understand what it tells us, THEN it will completely fall off from philosophy and become main stream science.

I plan to keep an open mind

Keeping an opened mind might be fine and everything. Up to a point... Like I often hear in scientific circles:
Don't let it get too opened, lest your brain falls out of it.
Just saying...

I'd rather stay with the current evidence. If new findings are brought up, I'll happily change my mind, but always based on evidence, especially when something as biased as human perceptions are concerned (where's the evidence that null tests are not telling us the whole story and there's more to it ??).

Another interesting point, which is not necessarily aimed at you, is that too often, people who claim to be "open minded", are actually close minded in reality.
The following link illustrates this beautifully. Granted, it aims at different things, but THE SAME PRINCIPLE applies (simply substitute "supernatural or magic" for some of the arguments in this thread)
https://www.youtube.com/watch?v=T69TOuqaqXI

[Timeline]

FM. Appreciate your point and the brains are secure. I may have a hard time with authority too. For Digital Null tests, it seems the vetting will continue by PHD types much more certified than me who may someday explain why two different mastering systems can sound different to the ear. If Null tests are found to be settled science, great. Where this all started anyway was comparing two Mastering apps. Not the null tests itself but it all got of topic i'm afraid. Bill will write a paper about all of this soon and be at AES for those who care in November.

I look forward to his findings as he has the newest gear to test and discover. Shooshie and many of you guys may be exactly right about Null in the digital domain. I have thus posted the comments of some who design this stuff here and would love to see more of this type of engineering opinion posted on Motunation so we can learn.

Shooshie, I never said you were wrong. Again, I'm just keeping an open mind as to why two DWS or Mastering apps may vary sonically in digital. Sonic Studios possibly is running some algorithms that null may find different and if so, and Null reveals them than I'm good with that but only to a point because if a difference exists, it still comes down to "what does the difference mean to sound". Where would the digital changes affect tone and how would it relate to what we hear. It takes more than null to vet that. It seems Shooshie you were making a point about Null Test only and on the same system. I was not but enjoying seeing you freak out after your sarcastic post naming me. BTW i have been listening since the '60s as well and recording since then so we have our valuable experiences to relate to in this new time and I agree they could vary. I respect that.

[Shooshie]

OK Shooshie. I understand your you are likely a hundred times more knowledgeable than I on these Digital null experiences and could very well be right that by using the Null Test no sound differences can be heard. If i was to lock into that brain fixed stereotype than I could never be open minded enough to learn anything new again IMHO. Where was Physics when String Theory came along? Oh excuse me, another obscuration.

I plan to keep an open mind

Gary, I'm just saying one thing: two identical tracks cannot be differentiated by your ears. Period. If the samples are exactly the same, there are no differences to be heard by you or by anyone you know or ever will know or ever have known or ever will hear of or ever won't hear of. NOBODY, alive, dead, real or imaginary can hear differences in identical tracks, because they are identical.

You don't have to know much about digital audio to recognize that's true. A New York City cab driver knows it, because if two people ask him to go to 490 West End Avenue, New York, NY, 10024, then they are going to the same cab destination. One of them doesn't mean "the other 490 West End Avenue, New York, NY 10024 in New Jersey." The addresses are identical, so they'll be getting off at the same place.

If two tracks of audio have the same strings of numbers, EXACTLY, then you cannot distinguish between them with your ears.

Your personal attacks on me will not change that. Your implications that I don't know enough because I'm not old enough (geez, 60 years is not old enough?) or experienced enough (a lifetime of experience, just like you) are attempts to distract from the essential fact: identical tracks are identical.

Gary, are you actually keeping an open mind in hopes of some sort of discovery that identical tracks are different? And are you keeping an open mind in hopes that your ears can tell the difference between identical tracks? A true open mind would immediately tell you that you're merely being obstinate. An open mind does not mean you can't know anything, or that nothing is real.

What the heck? Gary? Identical doesn't mean "sorta similar." It means they are for all intent and purposes THE SAME TRACK. Do you know what the definition of IDENTICAL is?

We're not talking about identical twins. There are always differences in biological twins, even identical ones; you can tell them apart. No, we're talking about two strings of numbers that are exactly the same. You really think that you can find some difference between two identical strings of numbers? Two identical tracks will null when subtracted from each other. And if truly they pass a null test, they are identical.

Shoosh

[Shooshie]

For Digital Null tests, it seems the vetting will continue by PHD types

It requires no vetting. Identical is identical. Null is Null. You don't need a Ph.D. or a 4th grade education. Identical is identical.

Two mastering systems? Yes, they'll sound different. Unless, of course, they are set up to produce identical files. IF they produce IDENTICAL files, then they will pass the Null Test, and that will mean that the two files they produced cannot be distinguished in any way from each other.

But set up the mastering systems differently, and you'll get two tracks that fail the Null Test. Your ears will be able to tell them apart if the differences are broad enough to be in the audible range.

I'm speaking about basics here. Just definitions of words. Identical means what it says. Null means what it says. NOT identical can be anything else. NOT Null means the tracks are different, and you may be able to hear it.

By the way, there's a lot of of deviation from null that is possible before it even becomes audible. But that's a whole other story.

Shooshie

[Timeline]

OK OK, I think you've convinced me for the moment Shooshie. Anyone as passionate as you to preach Null theory over and over like this can't be all that far off. I agree to table this part of it.

So, lets discuss instead your comment on how far off numbers can be to detect audible differences and how you know what it takes in these readings to hear a change in sound. That alone fascinates me. Also, if one does get a difference in a null, then what? What do you do with the info? Do you give up? Maybe this is where more measurement comes in as mentioned before. This is where i wanted the conversation to go many posts ago.

Since nobody reading this channel has bothered to do a null including me or you, if Sonic S systems is differing in these ways from DSP-Q and even DP, possibly they are forcing out information in their engine they have decided is not ear palatable in masters. Could they be right based on their industry success? Maybe just doing a throughput test they will null. If they don't null then how in the differences of these numbers would you sus out what they are doing for mastering sound?

Their entire advertising is based on their sound being better than all other apps. Since digital audio is commonly so close, then they must be doing something special. The fact is, the mastering industry uses this companies products like God wrote their applications and hense the price.

More on this needs to be learned and measured because i swear it sounds different to me. As i mentioned, a null would do me no good because what i heard included their EQ but if someone else can do a proper Null with their demo even just in throughput, no EQ, I'm certainly interested. Shooshie i think you should do it because you seem to have the most experience with null. We both use DSP-Q but I hope your passion like mine to discover sound not limited just to null differences only but the whys of digital sonics still drives something inside you.

[bayswater]

So, lets discuss instead your comment on how far off numbers can be to detect audible differences and how you know what it takes in these readings to hear a change in sound.

As mentioned above, differences were heard in a blind test in Radiogal's blind test. When I tried a null test using general test files on two of the apps involved, DP and Logic, I found differences around -132db. So you don't need much.

[Timeline]

So, lets discuss instead your comment on how far off numbers can be to detect audible differences and how you know what it takes in these readings to hear a change in sound.

As mentioned above, differences were heard in a blind test in Radiogal's blind test. When I tried a null test using general test files on two of the apps involved, DP and Logic, I found differences around -132db. So you don't need much.

Wow. -132. That's interesting BW.