24bit vs 16bit and Hz

[Lt. Bob]

we can measure anything but that doesn't mean we DO measure everything or that we even know all the things that might be useful to measure.

And Ethan, while very knowledgable, is also very pendantic and has a closed mind to anything he doesn't agree with.

 

[Beck]

Where I disagree though is your apparent disregard of structured listening tests. These are the ONLY way to be sure of what you're hearing.

Well it's not a disregard or ignorance of them, but a flat out rejection of them. There are a few problems when it comes to the tests I'm familiar with, one of which, Meyer-Moran, which I was one of the first, if not the first, to thoroughly debunk based on some simple oversights they made regarding the test setup. The biggest problem is the whole idea. That is, the idea that you can grab a room full of "Expert listeners" in a "Controlled setting" who will decide better than countless engineers in real-world settings. Next to that the biggest problem is missing some little details in the methodology, often something in the equipment used that skews the results and makes them unreliable.

Many people are driven to find an answer to all of life's questions; so driven that they will find one whether its accurate or not. It's more about human behavior than hard science. IMO it's more about religion or to the point of OCD (Obsessive Compulsive Disorder) that leads to a person's inability to accept that there are things we don't know yet and can't measure yet and thus they will have to remain a mystery until we have a way. But anyway, Meyer-Moran is really more academic at this point because we didn't move on to hi-rez end-user formats like SACD or DVD Audio. Instead we moved down to inferior compressed formats like mp3... and the unwashed masses seem to happy with that.

On the other hand, certain types of listening tests have been used to establish principles in Psychoacoustics, such as the Fletcher-Munson Equal Loudness Contour (Everyone should be familiar with). It's important to know that there are many things that must be done by ear because there's no way to lend evidence to something like Fletcher-Munson without the involvement of human hearing. Many principles in audio are based on human perception rather than something a scope or meter says because a scope or meter can't measure it.

 

[ecc83]

Whilst there is obviously quite a divide here "philosophically" and some doubts as to the validity of listening "panels" there is no questioning the results of null tests.

For those that may not know:
An amplifier (say) is put in a rig that allows the difference in its output (at rated power) and the ORIGINAL SIGNAL to be summed in anti-phase.
This is a bit tricky, there being phase shift to trim out but is not that bad with modern* DC coupled sstate amps.

The two outputs are then monitored and the signal switched between the thru' path and the nulled amplifier output. The result is silence for a very good amplifier!
That result is not open to "perception", interpretation or the edujukashun of the listeners! The distortion artifacts of such an amplifier are at or below its noise floor and that should be a good 100dB below 100watts (e.g.) .

This work and these results led Peter Walker of Quad fame to concluded more than 20 years ago that it was possible to make a "subjectively perfect" power amplifiier, which is what he did.
*Modern only in the sense of the last 20yrs or so.
Dave.

 

[Bobbsy]

I wish you hadn't mentioned QUAD, ecc83.

In a fit of madness before moving to Australia I sold my old QUAD 405 and my LS3/5A speakers.

Worst move I ever made. I miss them all worse than some of my family.

 

[Ethan Winer]

I don't think the sole metric in choosing sample rate/bit depth is whether a wave file encoded at a lower rate/depth would "null out" one at a higher rater/depth.

My point about nulling was to prove to people who believe in magic that two signals are identical, even if they seem to sound different. Once you know they're the same, then we can explore why they seem to sound different even when we know they're not. I don't know if a 44.1 KHz file can null completely against the same content recorded at 96 KHz, but that wasn't my point.

I simply don't see how manipulation of sound through application of effects and other processing prior to mastering is going to yield identical results regardless of sample rate and bit depth and, indeed, I can readily hear the difference with some kinds of processing.

I honestly don't know what you're saying.

--Ethan

 

[Ethan Winer]

Sometimes we don't measure all the right things when doing the basics

Sure, though a null test reveals all differences including those you're not even looking for.

Measurements can't always determine what sounds "good" to individuals.

Yes, that's a completely different issue. There is no metric for "good" because it's different for each person. Of course there are common features that most people agree on. I don't think many people would disagree that Dark Side of the Moon sounds excellent.

--Ethan

 

[Ethan Winer]

My mixes are done 90% by ear (and where I involve the numbers it's things like avoiding clipping

Exactly, and I mix by ear too. The only time I look at tech details is, for example, when soloing a track to better hear if it has LF content that should be filtered.

a lot of things lend themselves to sorting out by the numbers and double blind testing (bit depth, samples rates, that $300 power cord)

Yes again. The only reason I rail against audio myths is to save people money. You don't have to spend $2,000 per channel to get a good mic pre or A/D/A converter, and that knowledge helps countless people. Sample rates and bit depths are similar, because there's a "cost" to wasting CPU and disk space by recording files that are three times larger than needed.

--Ethan

 

[Ethan Winer]

Ethan, while very knowledgable, is also very pendantic and has a closed mind to anything he doesn't agree with.

That's simply not true. All the time I explain that I'm glad to change my opinion as soon as someone shows me proof. As a perfect example, right now there's a thread going on in Lynn Fuston's forum about (yawn) the value of higher sample rates. One of the fellows there claimed that phase shift from equalizers is an audible problem, so I asked him to post an example Wave file proving the point. I'm still waiting. The same has happened dozens of times in other forums and threads. Someone will claim [whatever] and I'll ask them to prove it. They never do and, worse, they just repeat the same nonsense a week later. When you push them they claim they're too busy to do such a test, or "I don't need to prove anything to you," or any number of other lames excuses.

I don't claim to always be right about everything, and I've learned a lot over the years and changed my opinion more than once. I have a very open mind. I just don't accept sighted anecdotal evidence as proof of anything. If you believe I'm wrong about any specific aspect of audio, please tell me what it is and we'll discuss it.

--Ethan

 

[Ethan Winer]

There are a few problems when it comes to the tests I'm familiar with, one of which, Meyer-Moran, which I was one of the first, if not the first, to thoroughly debunk based on some simple oversights they made regarding the test setup. The biggest problem is the whole idea. That is, the idea that you can grab a room full of "Expert listeners" in a "Controlled setting" who will decide better than countless engineers in real-world settings.

You completely miss the point. The value of Meyer/Moran isn't that they used experienced listeners rather than professional recording engineers! The value is that the tests were blind, and highly controlled, and they tabulated the results to determine what people can and cannot hear with statistical significance. I've seen a few minor criticisms of Meyer/Moran that are valid, but nobody has "thoroughly debunked" the basic premise of what they proved.

--Ethan

 

[Lt. Bob]

lol @ many pendantic posts!

Winer's one of the guys I keep on ignore because while he IS very knowledgable about this stuff he always comes across as superior to any of us mere mortals as if he's somehow important.

I don't like being talked down to and neither does anyone else which is why he's so disliked so often.

 

[Ethan Winer]

lol @ many pendantic posts!

You can't refute what I say, so all that's left is insults? Nice.

--Ethan

 

[PTravel]

I honestly don't know what you're saying.

--Ethan

I'm saying that a single track played at 44.1 KHz/16-bits may, to the human ear, sound the same as the identical track played at 96 Khz/24-bits. However, I think that, when that track is subjected to processing, e.g. stretching, shriking, pitch shifting, EQ, there's going to be an audible difference between applying these effects to a 44.1/16 and a 96/24 track and, I think, that difference would express itself as unwanted harmonics, distortion and increased noise. Now, to be clear, I'm talking strictly, "it seems to me" -- I'm not remotely close to a professional at this and I'm very open to someone explaining why I'm wrong.

 

[Ethan Winer]

^^^ Okay, got it. Yes, some processes are more accurate when done at a higher sample rate. But plug-ins that will benefit generally up-sample within their own code. So there's no reason to record your track files at a higher sample rate.

It's not difficult to test this stuff for yourself. I've done lots of such tests using the basic tools in Sound Forge and SONAR. To test this specifically, extract a music track off a CD that you think sounds very good, then make a copy and convert it to 88.2 KHz. Then you can load both files into separate tracks in a DAW and apply (for example) EQ plug-ins with the same settings that make a small change such as a slight treble shelf boost or cut. Then play one after the other and see if you can hear a difference. Better, have someone else switch while you listen blind at least ten times in a row. If you can identify one versus the other all ten times, then the difference is probably real. You could also reverse the polarity of one track in the DAW to see how well they null. If the null residue is 60 dB down or softer, then you know both files are sure to sound the same.

Edit: Most DAWs won't load tracks at different sample rates, so I guess you'd have to process each file, convert the 88.2 file back down to 44.1 after processing, and then compare them.

--Ethan

 

[Bobbsy]

lol @ many pendantic posts!

If Ethan were to flood the forum with "pendantic" posts would that be a pandemic of pendantic?

(Sorry Ethan--I'm basically with you on most of the issues but I can't pass up a good pun for anyone!)

 

[gecko zzed]

I don't like being talked down to and neither does anyone else which is why he's so disliked so often.

Actually, this observation is a clue to the 24bit vs 16bit debate, or the analog or digital debate, or to most of the debates around here. It highlights the significance and influence of one's perspective.

I've been following this thread, and I never gained the impression of Ethan talking down to anyone, so Lt Bob's observation came as a surprise. It may have come as a surprise to others as well. However, I daresay there will just as many for whom it is no surprise.

 

[Greg_L]

What a hideous thread.

 

[Ethan Winer]

I never gained the impression of Ethan talking down to anyone

It certainly is not my intent to be condescending. I try hard to stick to the facts and remain calm, even as others are insulting.

--Ethan

 

[Beck]

You completely miss the point. The value of Meyer/Moran isn't that they used experienced listeners rather than professional recording engineers! The value is that the tests were blind, and highly controlled, and they tabulated the results to determine what people can and cannot hear with statistical significance. I've seen a few minor criticisms of Meyer/Moran that are valid, but nobody has "thoroughly debunked" the basic premise of what they proved.

--Ethan

Perhaps you refuse to see it, Ethan. I think so because I highlighted "Expert listeners" and "Controlled setting." I don't believe they provided either of those.

Years ago on Anderton's forum I presented the death blow to Meyer-Moran based on information concerning the design and function of the HHB CDR-850 CD recorder, which they used... and I owned at the time, and still do. After speaking at length with HHB they confirmed my suspicions about the CDR-850 converters and how the unit handles signal once it is converted.

In short, Meyer and Moran did not have the 16-bit conversion they thought they did in using the CDR-850. Thus, following the dictates of sound scientific procedure, the study has no validity. Such sweeping claims as those from Meyer-Moran cannot have some validity. It's either valid or its not. Such a gross error in the signal chain is no small issue. It puts them back to square one in fact.

 

[Ethan Winer]

Years ago on Anderton's forum I presented the death blow to Meyer-Moran

I'm sure you think you did.

After speaking at length with HHB they confirmed my suspicions about the CDR-850 converters and how the unit handles signal once it is converted.

More details please. Link?

Meyer and Moran did not have the 16-bit conversion they thought they did in using the CDR-850.

So what did they have?

Even if M&M's tests were completely wrong, which I really doubt, this much earlier test came to the same conclusion using antique 1984 conversion:

Boston Audio Society - ABX Testing article

Even with that "old" digital stuff, a self-professed golden ear couldn't tell which was which.

--Ethan

 

[Beck]

I'm sure you think you did.



More details please. Link?



So what did they have?

Well, I know I did... when you see people referring to the problem with the so-called 16-bit converters I was the one who brought that to everyone's attention back then. It's old news. If you're comparing 16-bit to 24-bit and you later discover your 16-bit converters were not 16-bit, I think that pretty much debunks the "Study" don't you? It wasn't that hard to do.

I had all the links, the CDR-850 signal flow diagram and the pertinent email exchanges with HHB posted on Anderton's forum back then and you were there too debating the same topic. You couldn't accept it then and you can't accept it now because you have too much invested in it. You are however helping us all see very clearly how one's preconceptions help to contaminate pure science, as Meyer-Moran did as well. It's more about social engineering than engineering in this case.

For those who really want to know, the Multi-Level Delta Sigma converters in the HHB CDR-850 are equivalent to 20-bit.

And again, these focused listening tests smack of control freakism. You might just as well conduct a controlled taste test and then tell people what they like to eat, despite their protests that they don't like what you say they like and they in fact taste a difference.