The 24-bit challenge

Ethan Winer · #1 03-12-2002

Is 24-bit recording really better than 16? Can people actually hear the difference between dithering to 16 bits versus truncating? Can you tell the difference? How about only 11 bits? Take the 24-bit challenge at www.ethanwiner.com/BitsTest.html.

camn · #2 03-13-2002

Ill do it. Will email.. though I really dont think a simple volume up/volume down is enough error accumlation to really take advantage of the superiority of 24-bit.

xoxo

camn · #3 03-13-2002

Hold up.

Do you really understand bit depth? I ask merely for information, you understand. I dont claim to know much about it, really.. but I can do math, and I have a working knowledge......

so let me get this str8:

You recorded a sound... as close to unity as you could get. 24/441.

Then you copied it.

Then you dithered this UNPROCESSED sound....(copy A)

Then on a copy.. you TRUNCATED this UNPROCESSESSED sound.(copy B)

then you did a couple volume processes on a 16-bit sound??(c, D, and E)??

I must say, this is a completely non-scientific and useless test, and Ill tell you why.

A and B will be virtually identical. you arent processing the sound.. thus there are NO rounding errors... you basically are testing the difference in sound between truncating and dithering here.... which will be minimal . . . and subjective.

Now, for c, d, and e...
Where do you get the idea that 6db = 1 bit?? If I drop volume be 18db.. Im NOT chopping 3 bits off... Im NOT dithering.. Im NOT truncating... basically Im filling those last 3 bits with ZEROs. Which is MUCH different then chopping the bits.. because when I bring the volume back up... the sample can use those bits again!! they are STILL THERE!!! so your volume changes seem extremely minimal to me... maybe you introduce a little rounding error.. but its only two moves, dude.

maybe Ill do my own test, now that Im so annoyed at people's even-more-shallow-than-my-own understanding of this matter. But not anytime before April

xoox

Ethan Winer · #4 03-13-2002

camn,

> I must say, this is a completely non-scientific and useless test ... A and B will be virtually identical. ... you basically are testing the difference in sound between truncating and dithering here. ... which will be minimal . . . and subjective. <

Actually, this is quite scientific. According to a lot of people, the whole point of 24-bit recording in a 16-bit CD world is that after dithering, the result is better than just using 16 bits in the first place.

> Where do you get the idea that 6db = 1 bit?? <

And where did you get the idea that 6 dB. is NOT one bit?

> If I drop volume be 18db ... Im filling those last 3 bits with ZEROs. Which is MUCH different then chopping the bits. <

No, it's exactly the same. Since audio programs won't play a 13-bit file, the only recourse is to leave the file at 16 bits but zero out the lower bits. Which is just what I did.

> because when I bring the volume back up... the sample can use those bits again!! they are STILL THERE!!! <

Yes, they are still there, but they're all zeros! To take this to the extreme: Suppose I drop the volume by whatever it takes so only one bit is active, and then raise it back up. What do you think it will sound like? Do you see the point now?

--Ethan

Dolemite · #5 03-13-2002

I believe that the methodology used to produce the .wav files may be flawed. I do understand the db level = resolution thing, and theoretically this should work, but you really don't know what's going on in the software.

When you lower the volume, SoundForge may be saving those extra bits almost like a "level of undo" and simply replacing them when you then raise the volume. What assurances do you have that you're actually losing resolution? What was the procedure in between the lowering/boosting db level?

Also, I find that a highly dynamic voice, and in some cases, live drums are the most revealing sources for such a test. Acoustic guitar passages without much dynamics don't do much for me.

guhlenn · #6 03-13-2002

Please let me participate

camn,

> I must say, this is a completely non-scientific and useless test ... A and B will be virtually identical. ... you basically are testing the difference in sound between truncating and dithering here. ... which will be minimal . . . and subjective. <

>Actually, this is quite scientific. According to a lot of people, the whole point of 24-bit recording in a 16-bit CD world is that after dithering, the result is better than just using 16 bits in the first place.

NO, the processing is done at 24 bit -> the processing uses the extra bits... hence the quality difference... and then you go to 16 bit.

> Where do you get the idea that 6db = 1 bit?? <

>And where did you get the idea that 6 dB. is NOT one bit?

You CANNOT say that it's "quite scientific" when you try to dismiss a serious question by a lame joke.

> If I drop volume be 18db ... Im filling those last 3 bits with ZEROs. Which is MUCH different then chopping the bits. <

>No, it's exactly the same. Since audio programs won't play a 13-bit file, the only recourse is to leave the file at 16 bits but zero out the lower bits. Which is just what I did.

> because when I bring the volume back up... the sample can use those bits again!! they are STILL THERE!!! <

>Yes, they are still there, but they're all zeros! To take this to the extreme: Suppose I drop the volume by whatever it takes so only one bit is active, and then raise it back up. What do you think it will sound like? Do you see the point now?

--Ethan

guhlenn

Lt. Bob · #7 03-13-2002

Let me comment on this. I don't know the scientific jargon as well as some of you. But I am sort of an audiophile and I can tell you flat out that I can tell the difference between 16 bit A/D's and 24 bit A/D's Is this the same thing? I have several audio burners; some of which have 16 bit A/D's and some have 24 bit A/D's and there is definately a difference and in fact I hear differences between A/D's of different manufacturers even when the bits are the same. I guess I'm gonna find out that this is totally different that recording resolution;

but I wanted to make the observation and if it is different; will someone please explain to me about A/D and D/A bit rates as opposed to recording word length bit rates.

Ethan Winer · #8 03-13-2002

Dolemite,

> When you lower the volume, SoundForge may be saving those extra bits almost like a "level of undo" and simply replacing them when you then raise the volume. <

Excellent point. But in this case, SoundForge is not doing that. I specifically faded out the very end of each Wave file to hide the noise that became more and more obvious as the bit depth was reduced. So you can't hear the change in that part of the Wave files, but I sure could. Which convinced me that SoundForge was in fact doing what I expected it to.

> Also, I find that a highly dynamic voice, and in some cases, live drums are the most revealing sources for such a test. Acoustic guitar passages without much dynamics don't do much for me. <

Okay, I'll use those for the next test.

--Ethan

Ethan Winer · #9 03-13-2002

guhlenn,

> NO, the processing is done at 24 bit -> the processing uses the extra bits... hence the quality difference... and then you go to 16 bit. <

I'm not sure what you are saying here. All I did was reduce the bit depth of the files. There was no other processing.

> You CANNOT say that it's "quite scientific" when you try to dismiss a serious question by a lame joke. <

Not a lame joke at all. It is a fact that each bit represents 6 dB. of dynamic range. How is the test unscientific? The point is to see how many people can hear the difference between various bit depths in a totally blind test. Does this test not do that? Please listen to the files yourself and send me an email with your answers.

--Ethan

Ethan Winer · #10 03-13-2002

Bob,

> I can tell the difference between 16 bit A/D's and 24 bit A/D's Is this the same thing? <

Can you tell the difference using only the 24-bit A/D when one file is recorded at 24 bits and another is at 16 bits? If not, then the real issue is the quality of the two different A/D convertors, not the bit depth itself.

--Ethan

Slackmaster2K · #11 03-13-2002

Ethan,

I've found that in wavelab when I process dynamic changes, it maintains the lost resolution while you're working, such that if you crank the level back up, it sounds identical to the source. I verified this by cutting a 16bit file by 90db, "applying" the processing (e.g. creating a new temp wave file), and then increasing the gain by 90db and applying the processing (creating another new temp file). It should have sounded like "ckejrkadslfkjadsflkajfdfasdrer" but it sounded like the original. The key was to actually *save* the file after each application. I'm not sure how soundforge works, but I figured I'd bring this to your attention just in case. Sorry to interrupt!

Slackmaster 2000

mcr · #12 03-13-2002

I think what Camn was pointing at is that you have to do the recording (i.e. the AD conversion) using 16bit and 24bit. Whatever sound file you get after dithering or truncating the 24bit file has to be compared to the file recorded at 16bit ...

I don't know poo about this topic but it's quite interesting to get into. If my above paragraph is nonsense, feel free to call me an idiot.

Slackmaster2K · #13 03-13-2002

Another point I'd like to make. The actual usable range of a 16bit system will not be a perfect 96db. However, you can (or you better be able to) get a perfect 96db when recording at 16bit using 24bit converters, because the converters are still working at 24bit. So comparing this test to testing seperate 16 and 24bit converters isn't exactly correct.......HOWEVER, I agree with Ethan that you can't compare two different sets of converters, because you suddenly introduce too many uncontrollable variables.

Slackmaster 2000

VOXVENDOR · #14 03-13-2002

This is a very interesting thread and for the most part I have enjoyed reading it.... BUT...

Can you all quote each other with the proper method..

All im seeing is sentences starting with this ">"...

Fucking Annoying...

I don't know who is saying what..... OK well I do, but It's given me a headache trying to seperate each person..

atomictoyz · #15 03-13-2002

I have a side question here, it just popped into my head.

What does a 24 bit soundcard do when it plays a 16 bit CD or a .wav file? Do the cards have dual converters and automatically sort the incoming digital infomation to the proper converter? Or does the 24 bit card resample the 16 bit data into the 24 bit format? When your monitoring your PC during mixdown are you listening to a 24 bit mix or a 16 bit mix? Or does the cards allow you to hear a 16 bit dithered signal whatyouhearwhatyouget mix so that the CD will playback exactly what you mixed?

Peace,
Dennis

TexRoadkill · #16 03-13-2002

Why is this even an issue? The advantages to 24bit and the increase in SNR is pretty obvious to anyone who uses 24bit. This isnt an esoteric theoretical issue. Just use your ears.

Why wouldnt you use 24bit? Harddrives are dirt cheap now so storage isnt that big of an issue. Why even worry about it?

If you want a real test listen to reverb created in 16 vs 24bit. THEN you will really hear the difference.

Ethan Winer · #17 03-13-2002

Slack,

> I've found that in wavelab when I process dynamic changes, it maintains the lost resolution while you're working <

SoundForge doesn't seem to do that, because the lower resolution was very audible as the last chord faded out. Of course, I faded the files on my web site before that point!

> The actual usable range of a 16bit system will not be a perfect 96db. <

Agreed, but how many studios have an ambient noise floor even close to that?

--Ethan

Ethan Winer · #18 03-13-2002

VOX,

> Can you all quote each other with the proper method.. <

I agree. I go out of my way to address each person by name. In newsgroups this is less of a problem because you can see the thread order. But here, all the replies are stacked one after the other in the order they were posted, not the order of conversation.

--Ethan

Ethan Winer · #19 03-13-2002

Tex,

> This isnt an esoteric theoretical issue. Just use your ears. <

Okay, so please do that and let me know which file is which!

> Why wouldnt you use 24bit? Harddrives are dirt cheap now so storage isnt that big of an issue. Why even worry about it? <

Hard drives are indeed cheap, but throughput is still not unlimited. If someone can get only 12 tracks at 24 bits they could probably get 18 tracks at 16 bits.

> If you want a real test listen to reverb created in 16 vs 24bit. THEN you will really hear the difference. <

Yes, others have suggested this too. That test will be next.

--Ethan

Dolemite · #20 03-13-2002

Quote:

Originally posted by Ethan Winer
I specifically faded out the very end of each Wave file to hide the noise that became more and more obvious as the bit depth was reduced. So you can't hear the change in that part of the Wave files, but I sure could. Which convinced me that SoundForge was in fact doing what I expected it to.

Quote:

SoundForge doesn't seem to do that, because the lower resolution was very audible as the last chord faded out. Of course, I faded the files on my web site before that point!

Well, then you're already masking the difference! What good is resolution without dynamics?

Quote:

> Also, I find that a highly dynamic voice, and in some cases, live drums are the most revealing sources for such a test. Acoustic guitar passages without much dynamics don't do much for me. <

Okay, I'll use those for the next test.

I look forward to hearing it! I think the differences would become more apparent.

I found this test, done be Keyboard Magazine, to be interesting at least:

http://archive.keyboardonline.com/fe.../bitwars.shtml

I surprised myself and identified the original bitrate of 4/5 of the clips.

windowman · #21 03-13-2002

Quote:

Originally posted by TexRoadkill
The advantages to 24bit and the increase in SNR is pretty obvious to anyone who uses 24bit.

Even an SB Live has a signal to noise ratio of around 93db at 16/48. Man I'm sorry but the difference between 93 and 100 in the SNR department is so infinitesimal as to be nothing. I mean sure you could hear that tiny bit of hiss if you really crank the volume but good grief man! How gay would a dude have to be to get that persnickety over anything in life? It's just incredible to me that something like that would be an issue with anyone. I remember that not long ago we used to think that if you owned a Studer or Otari machine and you ran your tape at 30-ips that you'd get a SNR of over 70-db and you didn't have to use noise reduction because that was perfectly acceptable. And you're crying like damn girl over 93?!! I mean this is what bugs me the most about the people who would actually cry over having to use a 16 bit card. First, I don't believe a darn one of them could possibly hear the difference. Second, it's just so freaking GAY! I mean be a man for Pete's sake. Who the hell cares? I'd be perfectly happy recording on a Fostex B16 with a SNR of 72 and and a frequency response of 40 to 18k. A lot of great records were made on those little machines. They sound perfectly fine. An SB Live or any other 16 bit card will sound even better but I don't give a damn. I mean how freaking lame do you have to be to cry over something like that? It's like refusing to go outside because it's 68 degrees and not a perfect 70. I mean, I expect Richard Simmons to act that pampered but frankly I wouldn't buy his album if he made one. Somebody like that wouldn't have enough charcater in him to make worthwhile music. He'd be in the MP3 clinic saying, "Listen to my latest remix man! Ain't it cool?" Yeah, that's what I want to hear, your damn remix....

Ethan, I appreciate what you're trying to do but believe me, a sissy will always find an excuse to be a sissy. No matter what you bring out to the testing floor they'll cry "unfair" to. It'll always be one lame excuse after another. I once put several people to the test with sound files like that. I made a small 16/44.1 wave file and then copied it, then dropped the copy to a 128k MP3, then back to a 16/44.1 wave, then dithered to a 16/48, then back to a 128k MP3 copy again. Then back to a 16/48 wave. Then dithered again to 16/44.1 That's a hell of a lot of signal degradtion that these golden-eared people should have been able to hear according to them. Most people guessed the wrong file was the original. They actually thought the file that I put through all the changes/resaves was the better sounding file! But do you think they'd admit their shortcomings in the hearing department? Not on your life. It was one childish excuse after another. This test was flawed for this reason or that. Right....

Do yourself a favor, get off the internet and go to the gym and hang out with some men. And watch out for the guys who bring a towel to wipe down the equipment before they'll touch it. This is sickening!

As to the listening challenge I'd give em this order of best to worst:

file 1 best
file 5 second best
file 2 third best
file 3 fourth best
file 4 worst

To be honest though 2, 3 and 4 sounded identical to me. 1 and 5 were pretty much identical too except they had a bit more treble than the others.

mcr · #22 03-13-2002

finally a couple of recordings in 24bit and 16bit ... now, I can't fucking believe the big fuss everybody's been making about the huge quality difference. The difference is a big, fat 1MBit JOKE.

I got three right and two wrong ... which is best explained by luck. Statistically, significance is usually set at 5% or 10% levels ... hence, in order to show that your decisisons are more than mere guesses you should be able to have all FIVE right. Getting four right out of a sample of five is not enough and would not be considered as proof by any scientist.

I'm sure there're some people who are gifted with exceptional ears. However, I dare to suggest that most people who ramble on about 24bit's superior quality in comparison to 16bit are probably just babbling ...

MAYBE, the quality difference is more audible after recording several audio tracks (e.g. 12). That'd be a something I'd like to have put before me. I guess that's what many 24bit defenders would say ... and they might be right. What do I know. But as far as my ears are concerned (and I grew those myself): 24bit = 16bit ... if you hear a difference, then you're really great, deserve a medal and I'll call you my hero ...

Lt. Bob · #23 03-13-2002

Hey guys, if you put these sounds to Mp3 and put them on the net to listen, then they have no relevence to how they would sound in the original format because Mp3 is such an inferior format. The differences between formats and bit rates can be pretty subtle which is why some people either can't hear them or don't care about them. So if we're talking subtle differences--it seems to me that Mp3's could make them inaudible because of Mp3's lower resolution.

windowman · #24 03-13-2002

Quote:

Originally posted by Lt. Bob
Hey guys, if you put these sounds to Mp3 and put them on the net to listen, then they have no relevence to how they would sound in the original format because Mp3 is such an inferior format.

Hee, hee! I just knneeewwwwww somebody was going to say that!

Lt. Bob · #25 03-13-2002

Well-- why not? As I said when I edited my post--if the differences are subtle, then putting them in a format that has lesser resolution could certainly mask subtle differences. I'm not saying the differences are night and day. But for those who record say...mostly delicate acoustic stuff; even a subtle difference might matter.