Maximum Sound Quality Settings?

Ethan Winer has designed a simple bit depth comparison that perhaps some may enjoy:

http://www.ethanwiner.com/BitsTest.html

Joe

AJT, you nail it right on the head. It's about being satisfied with the results you get and I'm the same on this. Someone can tell me it's all worng, but if I like it that's all that matters. From one sound sytem to the next there are going to be differences so this is why I am not too concerned with comparing in that way or using it as a basis for saying it would not be an accurate or fair test. I would even go as far to say 2 identical systems set up exactly the same may sound a bit different because one was made on a Monday morning and the other on Friday (and just before happy hour!)

I would simply be interested in hearing some samples to see if I can hear a difference. I'm very into the sound aspect and if there's anything I can do to make an improvement with what I have to work with, I will certainly at least give it a try.

Who wants to throw a few samples up for a listen? I would, but I don't own any 24 bit sets.

Marc

Here's another quick and dirty test. I started with a five second sine wave in 64 bit floating format and sloped the volume to fall off to silence. Unlike organ samples, there is absolutely no wind or other noise to mask quantization noise in these files. Next, a 24 bit copy was made and and 16 bit copy was made. The 16 bit copy was resampled to 24 bits so that the files will appear identical in length or if inspected by a sound editor. Resampling was done with a high quality dither. I then made between one and nine copies of the 24 bit file and between one and nine copies of the 16 bit file and assigned each an arbitrary name. Only I know which files are 24 or 16 bit. If you wish to try to identify them, send me a list of names that you think are the 16 bit versions by private message. I promise to tell you honestly how accurate your answers are.

https://sites.google.com/site/vtheatreorgan/Home/files/Bill.wav
https://sites.google.com/site/vtheatreorgan/Home/files/Bob.wav
https://sites.google.com/site/vtheatreorgan/Home/files/Ed.wav
https://sites.google.com/site/vtheatreorgan/Home/files/Fred.wav
https://sites.google.com/site/vtheatreorgan/Home/files/Jim.wav
https://sites.google.com/site/vtheatreorgan/Home/files/Larry.wav
https://sites.google.com/site/vtheatreorgan/Home/files/Paul.wav
https://sites.google.com/site/vtheatreorgan/Home/files/Ron.wav
https://sites.google.com/site/vtheatreorgan/Home/files/Sam.wav
https://sites.google.com/site/vtheatreorgan/Home/files/Tom.wav

This is more for entertainment than pushing the boundaries of scientific investigation.

Joe

Thanks Joe. I will certainly take a listen and let you know what I think but tonight I have some practicing to do for the upcoming Easter events. No, luckily they don't have to listen to me play the organ but they do have to listen to me sing!

Marc

I've been following this thread with interest, although I doubt my ears are good enough to discern minor differences in sounds caused by 16 or 24 bit depths. I watched Ethan Winer's youtube video linked by Joe and that confirmed my suspicions about my hearing. Winer has an interesting article (http://ethanwiner.com/believe.html) on his website that suggests any comparisons would pretty much need to be done with headphones if the listener is to hear the samples without a chance of the environment altering perception, since even a little head movement will change the sound from one sample to the next using speakers.
Craig

Hello Clive, haven't heard from you in awhile, I hope all is well there with you.

Yes, fine, thanks. I've been reading new posts here occasionally, but without adding any of my own.

This is getting way more complicated than what I had in mind. [...] I'm simply suggesting to have some recordings posted, have people listen to them and have them tell us if they think the recording is 16 bit, 24 bit or if they can't tell the difference [...] What it ends up proving? I'm not too worried about that.

As far as the outcome or throwing statistical numbers at it to draw some sort of conclusion, I'm really not interested in that either as we all know too well the debate will continue anyways. I just simply want to see if I or any others CAN honestly say they hear a difference and we can then all draw our own conclusions from there. [...]

Fair enough. I commented because, at first, I thought you were talking about going down the path that Joe suggested.

One thing, though: however it's done (and especially if there aren't many files to listen to), it would be best not to routinely offer the same number of 16-bit and 24-bit files - eg two of each or three of each - because if listeners know how many there are, either that makes their 16/24 decision for the final recording just a consequence of what they've already decided, or that extra information might influence their decision (eg going back and altering their earlier assessments in order to get the right number of 16s & 24s).

Of course, that's not to say having an equal number of 16s & 24s ought to be deliberately avoided - that could be just about as bad (eg if listeners knew there were three of one and one of the other).

Ideally, the person posting the (four, or whatever) recordings would determine randomly how many there would be of each kind, and not tell the listeners how many.

Oh, and that isn't for the sake of any follow-up anaylsis - just a way to avoid influencing people's perception by giving them too much info.

I'm no statistician, but I think the Chi square test would apply to judge the validity of results of listening to samples and deciding if they were 16 or 24 bit. You can go to the web site: http://math.hws.edu/javamath/ryan/ChiSquare.html, and play with the first applet. It appears that one would need at least 12 samples to make even one error and still have statistical validity at the 5% level, and that of those 12, at least 3 of one type, with probability of a valid result the highest with an even distribution. I have absolutely no idea why this is so, I am only able to plug in numbers and let the program do it's thing.
Craig

Q. If you have multiple eyewitnesses to an event and you ask them all for their story,
how do you tell which ones are lying?

A. The ones who's stories MATCH exactly.

The point is that human perception is such a subjective variable that it is hard to use 'witness' to prove anything but I think the 24bit sounds better. I think I have heard the same thing AJT has and my ears are OLD


Pax,

Thomas

Yes, AJT is one who said he was hearing differences between certain stops. Amateurorganist is saying he also hears a difference. If anything I would like to hear just 2 examples, one in 16 and one in 24 to see if I notice anything. AJT seemed to not be able to say exactly what the difference was but said he heard something that sounded better none the less. It would be interesting to have people listen to the 2 examples I suggest and just tell us what they heard, or what was different and see how many stories do match up.

Marc

The 16 bit copy was resampled to 24 bits so that the files will appear identical in length or if inspected by a sound editor. Resampling was done with a high quality dither.

Changing from 16 to 24 bits does not require dither; in fact by dithering at this point you have added that little extra noise that could make the difference for someone distinguishing two files.

Paul

<pedant>Changing the sample depth is not "resampling", which term applies to changing the sample rate</pedant>

There is really just one key point that I had hoped to respectfully make: knowing what one is listening to inevitably influences one's perception. If we believe A will sound better than B because it is more expensive, because it is praised by others, has more impressive numbers, or any other reason, this knowledge will affect our perceptions whether we want them to or not. If there is a real and significant audible difference between two sounds, then logically we should be able to reliably identify that difference without knowing which sound is which.

Joe

The 16 bit copy was resampled to 24 bits so that the files will appear identical in length or if inspected by a sound editor. Resampling was done with a high quality dither.

Changing from 16 to 24 bits does not require dither; in fact by dithering at this point you have added that little extra noise that could make the difference for someone distinguishing two files.

Paul,

As usual, you are quite correct. Dithering is a type of noise applied when reducing bit depth to randomize the quantization noise. This is opposed to merely truncating the extra bits. This does not occur when upsampling to higher bit depth. I could have been more precise, but I was actually referring to the downsampling from 64 bits.

Joe

Listened several times through the samples Joe prepared.
I could hear NO difference between any two or more of them.

O.K., I listened to a number of Joe's examples here, but not all. These 48 year old (and still pretty decent even though my wife says otherwise) ears can't pick up any difference but the samples are very short consisting of a single note played once. I would have to hear a succession of notes or an actual HW audio sample in both bit rates to be able to judge further. If there is (or isn't) a difference I would think a longer sample would settle it once and for all for me.

Marc

I always try each configuration, for at least two weeks,

if I am then "dissatisfied",

then I try a new setup,

or go back to a prior one that I did enjoy.

I had my pipe speakers aiming at the walls behind my console,
and became dissatisfied, I then turned them back to face the end walls, and now I like this sound better.
Time will complete this story, "or repeat it" !