The 24-bit challenge

[Ethan Winer]

Re: Re: point made

Shailat,

> Did you ever consider the fact that some of us are on a old modem? <

I feel your pain.

> If Ethan would mail me a cd I would gladly participate. <

I can't mail everyone a CD. But even at 56k, 11 MB takes only 15 to 20 minutes. And you can do it at the same time you're reading the posts here!

--Ethan

 

[Basslord1124]

Granted some may have a 56K modem but not everyone can connect at 56K. I'll admit, we run on a 56K modem too If the file was 11MB and with the average speed we connect at (which is around 36000) I'd say it'd take close to an hour or more to download that (and this is assuming AOL doesn't boot me...LMAO, I'd like to see that). Now I could go to school where it's all on T3 lines....but regardless, I don't think they'd apppreciate hogging the bandwidth there. I always laugh at the idiots trying to file share at school using Morpheus or whatever. Granted I do file sharing but I would never do it at school...I think the fear of getting my account taken away is why.

 

[dosdan1]

camn,

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

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

From mathematics: decibels are a ratio of powers with each bit increasing the resolution by a power of 2 (a bit can be ON or OFF = 2 possible states). The formula for comparing two amplitudes (or voltages) is 20 log(V2/V1). When we are just considering the effect of the addition of 1 bit's extra resolution it's 20 log 2 = 6.02 db per bit.

Why "20" log 2?
because
1. decibels are 1/10 of a bel = "10"
2. Voltage or Amplitude ratios are not the same as Power ratios but are squared to get the Power (remember P=V^2/R). When dealing with a log, to square it, we multipily it by 2.

So 10 * 2 = 20

Why "add" 6 dB per bit? Each bit multiplies the resolution (number of quantisation step) by 2 i.e. 2^14(bits)=16364 steps, 2^15(bits)=32768 steps. When dealing with logs, to multiply a number by two, we add the value of log 2 which, with decibels, was adding 6.02 dB per bit, to the current log value.

 

[ff123]

Ethan,

I submitted my results to you by email just a short while ago. I can statistically analyze the results for you if you wish. Just assign scores as described below:

There were five samples, so assign values from 1.0 to 5.0, where 5.0 is the best-sounding, and 1.0 is the worst sounding. In the case of a tie, average the values. For example, say that someone ranked the files File3=worst, File2=next worst, File5=best, File1=File4=next best. Then the scores would be:

File1: 3.5
File2: 2.0
FIle3: 1.0
File4: 3.5
File5: 5.0

Arrange the scores in a table of listener vs. samples, where there are 5 columns of samples, and X rows of listeners.

I will run my resampling program (http://ff123.net/bootstrap/) on it to look for differences which are significant at the 0.05 level.

ff123

 

[camn]

From mathematics: decibels are a ratio of powers with each bit increasing the resolution by a power of 2 (a bit can be ON or OFF = 2 possible states). The formula for comparing two amplitudes (or voltages) is 20 log(V2/V1). When we are just considering the effect of the addition of 1 bit's extra resolution it's 20 log 2 = 6.02 db per bit.

Why "20" log 2?
because
1. decibels are 1/10 of a bel = "10"
2. Voltage or Amplitude ratios are not the same as Power ratios but are squared to get the Power (remember P=V^2/R). When dealing with a log, to square it, we multipily it by 2.

So 10 * 2 = 20

Why "add" 6 dB per bit? Each bit multiplies the resolution (number of quantisation step) by 2 i.e. 2^14(bits)=16364 steps, 2^15(bits)=32768 steps. When dealing with logs, to multiply a number by two, we add the value of log 2 which, with decibels, was adding 6.02 dB per bit, to the current log value.

Ok, ok... Ive been gone.. and this thread has exploded so much that Im just gonna have to jump in at the end here.

1 bit does not equal 6.02 db, when you're talking about bit-depth.

Now.. TRY and understand what Im saying here... just as an intellectual exercise....

if I take a 16 bit sample, and I dither it down to 14 bits.. Im NOT attenuating by 12 db. I may be losing 12db of HEADROOM.. but if Im not up at clip, then im not actually losing VOLUME. I know this makes sense to you.

for the same reasons.. if I lower VOLUME.. Im not losing bit depth! Im not losing headroom just because Im not using it, you know?

oh, and Ill have my 24-bit challenge out to youall.. one that actually is signifigant... around september

xoox

 

[dosdan]

1 bit does not equal 6.02 db, when you're talking about bit-depth.

if I take a 16 bit sample, and I dither it down to 14 bits.. Im NOT attenuating by 12 db. I may be losing 12db of HEADROOM.. but if Im not up at clip, then im not actually losing VOLUME. I know this makes sense to you.

for the same reasons.. if I lower VOLUME.. Im not losing bit depth! Im not losing headroom just because Im not using it, you know?

xoox

I don't know what you mean by the term "bit-depth"? Do you mean S/N?

I've only seen a reduction of bit-rate done with the FS (Full-Scale) values being the same in both cases. So a -3 dBFS 16-bit encoded value was still -3 dbFS after converting it to 14-bit. There is no change in headroom (the difference between the current value and FS) or volume. There is a decrease in dynamic range. The electronic noise floor is increased in the 14-bit version compared to the 16-bit one. That is also an acoustic noise floor based on the background ambient noise level in the studio but this is not affected by redithering.

If you lower Volume | Amplitude then you decrease the S/N ratio (assuming that you calculate the S/N using the current value rather than the FS value) because your current value is closer to the noise floor. Whether or not this decrease is noticeable depends on whether the noise floor is below the ambient noise level in the playback environnment. Of course if you boost a -60 dbFS 16-bit signal and a -60 db 14-bit signal by +60db of gain the 14-bit would sound much noiser that the 16-bit one (assuming that the signal in question is a electronically generated signal rather that a signal recored acoustically.)

The only way that headroom comes into the discussion is that you might tend to record at a higher level on a 14-bit system and thus increase the likelyhood of clipping to keep away from the higher noise floor on a 14-bit system.

But that has nothing to do with converting a 16-bit sample to 14-bit that you used as an example. That is not "attentuation" of the sample by 12 dB but is is an increase of approx 12 db in the electronic noise floor (depending on the linearity and monotonicity of the encoder).

 

[Pio2001]

Let's not forget that we are comparing 16 vs 16 bits files, here.

It would be interesting to exend the test to 24 bits files : the original vs a 16 bits version converted back to 24 bits.
It would give an idea of what the DVD-A is promising (a true 96 kHz 24 bits sample vs a 44.1 / 16 converted to 96 /24 would be perfect).

Of course, only people with 24/96 soundcards could participate.

 

[Slackmaster2K]

Camn, you are correct on your first point, but wrong on your second.

You are losing bit resolution as you decrease volume. 6db represents a *doubling* of voltage amplitude, which is what makes it an interesting figure. (recall that the db scale is logarithmic, not linear)

In a 16bit system, -96db to -90db is represented by one bit, or two discrete values. -96 to -84db is represented by 2 bits (4 values). -96 to -72db is represented by 3 bits (8 values). And so on and so forth. Hence you'll often see 16bit converters advertised as having 96db of dynamic range, which is 16*6. In theory that's correct. A 24bit system has 144db to work with in theory.

So technically, as you decrease volume you are also decreasing your resolution. Hence if you decrease the volume of a file by say 72db and commit it to disk, there is absolutely no way to regain that lost resolution, and if you crank the gain back up by 72db, you're simply cranking up a truncated 3bit waveform.

dosdan,

Headroom comes into play because you are simply offered more resolution at higher volumes in a 24 bit system, plus a lower noise floor, hence in theory you can track at lower volumes, which for all intents and purposes means that you have more headroom.

Slackmaster 2000

 

[Slackmaster2K]

Pio,

Converting a 16bit file to a 24bit file will in no way improve the sound. DVD audio is promising because audio can be rendered at 24bit and remain at 24bit.

Consider a simple case:

Original sample: 1111 1111 1111 1111

24bit: 1111 1111 1111 1111 0000 0000

Converting from 16 to 24 doesn't do anything for you unless you will be applying DSP, which isn't going to happen in this case.

Slackmaster 2000

 

[Basslord1124]

Go figure at a time when I complain about this stuff we are doing in my college calculus class (logs, linear, etc) that it's actually used in an area that I have interest in (well I enjoy recording but this does help educate on how a computer records your music). Yeah, in our class we already talked about logarithms, linear, exponential and all that good stuff. Talking about derivatives now. Of course I was already aware this stuff has uses...it's just when you're 21 you most often want to b*tch about it rather than think it's useful.

 

[windowman]

Basslord;

Right there with ya pal. My only electronics knowledge comes by way of a one-year certificate course in computer maintenance back in the late 80's. We didn't really delve that far into digital theory though because the main function of a computer repair tech was already that of a board swapper by then and it hasn't changed since. So actually most of the theory I have is old component-based analog technology that they still taught in the electronics classes. I can do an okay job of swapping out capacitors in a tank circuit or adjusting the playback speed of your tape deck and so on but that's about it. And the physics I know is just stuff I picked up from reading Hawkings, Einstein and Scientific American etc. So really this conversation is a little beyond my realm of knowledge at this time. I'm still an analog man trying to keep up in a digital world.

PIO;

Even most of the people who do record at 24 bit will still use a 44.1 sample rate. I think most would agree that going to 96 isn't going to change much. However, if you want to go higher try 88.2 instead. You'll find it's much easier on your machine because it's a multiple of 44.1, therefore your saves/changes etc, will take much less time and be less of a burden on your processor. Just a suggestion.

Slack;

"Converting from 16 to 24 doesn't do anything for you unless you will be applying DSP, which isn't going to happen in this case."

I agree of course. Just like taking a cassette tape and transferring to a wave file won't make it sound better and in fact may help bring out more undesirable noise artifacts. I suppose though that the exception would be that if you have a recording that you want to make a gob of changes to, then taking it up to 24/88.2 may be a good idea just so you'll have more bits and samples to work with even though it may be redundant information. People claim that you can do more edits like this without degradation of sound. I wonder about the validity of that claim though. Just not sure on that one....

 

[Pio2001]

Converting a 16bit file to a 24bit file will in no way improve the sound. DVD audio is promising because audio can be rendered at 24bit and remain at 24bit.

Of course, I know that

I was talking about trying to sort out this "false" 24 bits wav made from a 16 bits one, from the "true" 24 full bits original that was recorded in the first time.

I wonder if the difference is audible... it must depend on the DAC behaviour, I guess. I don't have acces to any 24 bits source.

 

[Hippo]

Just my take on things

The big advantage to me of 24 bits is the ability to record without compression by lowering levels without a vast increase of noise.

Makes for a natural recording that is useable rather than artificially faked one using compression.

Arguably the human ear can go from a jet engine roar at 1 meter from a pin drop in a quiet room which is about 115db dynamic range in a young person 24 bit can do this 16 bit can not, analogue got close at the very top endof the range. (using the very best of the old systems)
High quality large PA also come close.

(24 bits lowers noise floor down, given the restraints of the room and equipment used, it cant increase digital max)

Now how much is required for actual music recording for domestic reproduction is a different debate!!

Hippo

 

[windowman]

"The big advantage to me of 24 bits is the ability to record without compression by lowering levels without a vast increase of noise.

Makes for a natural recording that is useable rather than artificially faked one using compression."

Everyone under the sun that's been recording for any length of time knows enough to use compression when laying tracks, especially acoustic instrument tracks. You'll never in a million years get a decent acoustic sound without one. It makes everything sound bigger and fuller. And contrary to what you say it makes things sound MORE natural, not less. You won't show me a CD made anywhere by a top acoustic musician that's not using some real nice compression. It does two main functions:

1. It brings low level signals up. Just listen to any of the old recordings of solo classical guitar by Segovia that were done without compression and you'll quickly see why everybody uses it now and has for 30 years. (Yes, it also brings high signal spikes down but that's not as big a concern as bringing up the low ones.)

2. It gives acoustic instuments a wonderful airy feel to them that you'll never get without one. (And personally I wouldn't get caught dead without one in my electric guitar setup although I don't use it all the time. Probably 75% though.)

 

[j]

i like the compression on 'hellhound' by robert johnson.

 

[Pio2001]

Unfortunately, since 10 years, compression has been used mostly to butcher the sound of FM, clubs, and sadly, CDs, for the sake of playing louder than the neighbour.

Interesting comments here :
http://www.musicplayer.com/ubb/ultimatebb.php?ubb=get_topic;f=3;t=002464

 

[windowman]

"i like the compression on 'hellhound' by robert johnson."

Got anybody you can name that actually made a record "after" compression invented?

"Unfortunately, since 10 years, compression has been used mostly to butcher the sound of FM, clubs, and sadly, CDs, for the sake of playing louder than the neighbour."

Yes it's used to extremes by radio station programmers. That's common knowledge and we all yell about it. And occassionally albums (one only need listen to a Boston or Van Halen album.) But none of us would condone "over" compression. What's that got to do with compression that's done "well" in order to bring those low level signals up (so the listener doesn't constantly have to grab the volume control every 2 minutes as we all do on our favorite "live" bottleg cassettes that were recorded by an audience mic without compression) and to make acoustic instruments sound fuller and more "natural"?

Show me a recording by a top acoustic guitarist made between 1975 and 2002 that doen't have some compression. I'm not saying you should do it because everyone else does but rather because you'll never get a decent acoustic recording without it. Believe me, I tried when I was younger. It's impossible. And I didn't find that link interesting at all. Just your typical internet, teenage morons who couldn't make a friend in real life talking about something of which they haven't a clue. That's why they call it the weird wide web and the world's biggest lonely hearts club.

 

[Pio2001]

And I didn't find that link interesting at all. Just your typical internet, teenage morons who couldn't make a friend in real life talking about something of which they haven't a clue. That's why they call it the weird wide web and the world's biggest lonely hearts club. [/B]

Maybe you didn't find it interesting, there are no infos about compressing, just moaning against nowadays habits, but some of them are not exactly " teenage morons", since it's a professional recording forum, and Ted NightShade, for example, is sound engeener, if I read properly.

 

[windowman]

Hey I'm not complaining about that forum. I seldom post but go by just to read now and then. And you're right in that there are actually quite a few well known engineers who posts on occassion. What turns me off about all forums though is the "interaction" between questioners and the expereinced engineers. It's the same problem most of us have with National Public Radio programs that have listener call-in segments. They don't screen their calls as well as they could and half those callers have nothing to say worth saying. They just don't have anything better to do than sit around calling into radio shows. It's really infuriating to have to listen to that amist what would be an otherwise well-done show.

BTW, you know, I might actually agree with you on the "no compression" thing if everybody had a great listening room with a killer sound system they used exclusively for music. But the fact is that most people (myself included) are listening in the car, or through a headphone RIO type unit while jogging etc. The low level signals get completely lost amid the traffic noise and so on. I have a copy of The Glass Harp's "Live at Carnege Hall" that has a 30-minute jam on it which was recorded with an audience mic and sounds like it. Everything's very distant and burried in tape hiss. And there was no compression or noise reduction used during the recording. (It was done in 72.) If you try to imploy compression afterwards to a recording that has a lot of tape hiss it'll just make it a lot worse so they were smart enough to leave it alone. But I have to grab for the volume knob constantly because of the volume dynamics. It's really frustrating to have to do that while you're trying to drive. Same with a lot of my favorite classical CD's. I actually prefer tlistening to the classical radio station in town than to my classical CD's (a lot of classical music isn't compressed at all) while in the car because they're compressing the station signal and bringing those low level signals up for me so I don't have to strugle with the volume control so much.

What's unfortunate is that staions often go overboard with this compression stuff. Another thing that's common for them to do (which I detest) is to speed up the recordings by a hair. Supposedly that'll make them sound more "zippy" and lively or something. Sure doesn't work for me. Oh well.

 

[Pio2001]

There, I completely agree.

I'm always confused by a friend who says she always have to play with the volume because the records she listen to are lifeless and boring, but at the first dynamics peak she screams "what a noise, damn them !".

I think she's got a problem with her tweeters (BC Acoustique Araxe speakers), that are too loud.