24 Bit more about resolution, not dynamics...

Posted: Thu Aug 31, 2006 10:47 am

I just started posting so my cred may not be high, but I have read ALL the posts here and probably anywhere pertaining to the subject. Again Cucco and I speak in concert; I disagree with anyone who says 24 bit is "not more resolute". This is why I started this thread. Total misunderstanding on this topic.
Again, anyone who says 24 bit is for capturing a wider MUSICAL dynamic range is missing the boat. It IS for capturing more detail at the lower voltages, which happen hundreds of times a second, no matter what the perceived musical dynamics are.
So literally hundreds of times a second, 24 is capturing and encoding information that 16 bit is incapable of.

Quote:

The digital geeks like Nika and Bob.K and others ... including the Digi boys themselves ( Dave LB )
have often said that resolution in the way you have used it to describe PCM encoding is incorrect.

they say that 24 bit is not more resolute than 16 bit
it does provide for more dynamic range if required
if the signal is of low dynamic range then the detail of the sound at the given sample rate will remain the same

- Kev

Bob Katz (Bob K?) does NOT say this Kev Surprised

You are misunderstanding bro. If you found someone who did, they are misunderstanding.

Like Cucco said, the burden of proof is on "them" Laughing

Dont believe what people say, believe what you can prove. The Earth is ROUND peeps!!! Represent!!!

Peace,
Nate

Posted: Thu Aug 31, 2006 5:48 pm

I'm no one to talk, but I can tell from a measly 2 years involved in recording, that my 24 bit converter sounds miles better than my old 16 bit converter. Mainly happens on things like cymbals (the high end sounds smoother or "nicer"..it was harsh in 16 bit), and vocals (more detail, more "reality" to the sounds). SO! I don't know...just felt like I had to stick my nose in and relate my limited experience to the topic I guess! ANDY

Posted: Thu Aug 31, 2006 7:04 pm

Cucco wrote:

Yep, and I've gone on record several times disagreeing with these folks. My feeling is, the burden of proof is on them.

No offense to either Dan or Nika, as I respect them greatly, but their arguments border on the insane. To me, it's apparent that they are attempting to put science before logic. Neither can exist without the other.

Sorry...Rant over.

yep
and I know you have been in many of those and over a good portion of the time I suggested above.

The occasional rant is fine ... Thumbs Up

I know what you maen about the nature of their argument at times.
This stuff is harder to explain and it does require the heavy maths.

I've been think about this again ... over night
I promised myself I wouldn't and as I said before I don't want to start yet another endless thread
BUT
I had a thought ... Shocked

... and even though it is not technically right it might help to bring a direction or point of perspective ... devils advocate if you like.

Vision / Video
1 bit video gives the two state situation. BLACK or WHITE
8 bit gives the 256 grey scale
but that scale falls between the same two extremes of the BLACK and the WHITE.
There is more information within the range.
... then comes the 16 bit and true colour stuff ... but still falls into the between BLACK and WHITE.
BUT cameras with CCD blocks can see into the Infra Red ... beyond the above range
???

PCM Audio
both 16 and 24 bit have the upper limit of 0dBFS ( Full Scale )
the largest 16 bit word send you down to the 16 bit noise floor
and the 24 bit work can send you LOWER.

so part of the expanded 16 to 24 bit is to expand the area we are working with.

wait for it

Does the expanded 16 to 24 bit ALSO provide for more detail within the OLD 16 bit area.

Rolling Eyes

get my drift ?

as I said above , this analogy is not strickly correct so don't all go juming on me ... just trying to present it in a deferent sort of way.

if that does sit ok then I can present the idea that a given signal captured into a 16 bit system can be identical to the same signal captured into a 24 bit system.

Posted: Thu Aug 31, 2006 7:19 pm

CrackBuddha wrote:

Bob Katz (Bob K?) does NOT say this Kev Surprised

You are misunderstanding bro. If you found someone who did, they are misunderstanding.

nup
I have a very clear understanding and Bob has at times used and chosen not to used the word Resolution
Currently he does use the phrase more resolved
http://www.tnt-audio.com/intervis/digidoe.html

Quote:

In fact, with a little less lazy engineering, 44.1K/24 can be extremely good. (16 bit is out of the running; 20 and 24 bit at either sample rate is always superior, wider, warmer, deeper, more resolved). It's a lot easier (with the relaxed filter requirements) to make a 96/24 (or 88.2K/24) recording sound good.

is the filtering the dominant factor in what is being heard ? - a Kev comment only
... BUT in the past in some thread here ( I think) and at PSW has has tended to agree with Nika in that Resolution was less than a correct descriptor.

This is a very old subject and one must go back to the original PCM maths and we should use the words in the original context of PCM theory.

I think Lavry's white paper is a good presentation
http://www.lavryengineering.com/documents/Sampling_Theory.pdf
but it tends to focus on rates
... it would be nice to see an equally depth look at Bit depth and amplitude

Quote:

Like Cucco said, the burden of proof is on "them"

and I totally agree
and feel that to date they have not done so well enough.
I fear that it will end up in a detailed look at converts and a chart full of mV values and digital words.

and meaningless to those that say ...

Quote:

but I can tell from a measly 2 years involved in recording, that my 24 bit converter sounds miles better than my old 16 bit converter.

too often the white papers on bit depth end up being detailed on noisefloor and then introduce Dithering Theory.
It would be nice to see some detail on signals recorded in both 16 and 24 well above noise floor and below FS and see how the same signal compares in the two formats.

Buddha,
if you go back through some of those ancient threads you will see on many occasions I have banged heads with Nika and even though I do respect him greatly ... I often did not agree with many of Nika's presentations and interpretations. You may even find I have presented the exact agruments that you may chose to present here. I bring only part of an alternate point of view.

I think that when either side of a argument takes a point to an extreme, it can lose some credability unless it is based in an exact and restricted frame of reference.
Hence the use of Math and exact definitions that often seem to have no bearing on the types of signal we want to record.

The bottom line and beyond much of the math and theory is the fact that people can and reliably hear differences. Norrowing this down to just BIT depth as against the influances of Dithering and Error Correction and/or other algorithins contained within the hardware and chipsets is so very hard.

Posted: Thu Aug 31, 2006 8:31 pm

curious that you bring up dither at this point as IMO thats what makes the argument for the resolution camp... all that little detail in the lsb end of things once truncated needs to be synthesized to make it sound right again... why would that be except that through poor gain staging brought on by the erronious belief that some perceived increase in available dynamic is going to save you from overs... when in fact you've simply been lulled into not useing the detail it affords you...

Posted: Fri Sep 01, 2006 12:10 am

dementedchord wrote:

curious that you bring up dither at this point ...

and bringing up dither is exactly what I don't want to do but nearly every white paper on the subject or 16 and 24 bit PCM audio , that I can find at the moment, does end up looking at noise floor and dithering

as I said above, it would be nice to see a detailed presentation of a limited dynamic range signal recorded in both 16 and 24 to see how they differ in accuracy to the original signal
and the same signal recorded at diferent levels with respect to FS in both formats and see which one is closer to the original and why/how

it can be difficult to present someone else's argument when you tend not to agree with their premise

I tend to like the way Roger Nichols presents a concept.
Even though this paper doesn't talk about the act of sampling and is more about subsequent processing and mixing,
it is worth a read
http://www.rogernichols.com/EQ/EQ_2001_03.html

Quote:

If you get your input level up above -144 you will start to record some information. With no noise shaping or dithering your recording will be very distorted, but you can easily tell what is being recorded. At that level you are basically recording with a one-bit converter.

then it should follow that if you record at the other end and near to FS the detail should be at it's best
... I think this is dementedchord's line of thinking and to be honest I find it hard to side with Nika.

One of Roger's articles ends with the line

Quote:

Now you will have to excuse me, I have to make room on my desk for my new 48bit/ 192kHz Digital Audio Workstation.

and it's a cool point
why don't we just make the leap now and be done with it or will it be a case of diminishing returns
and those golden ears will just continue to complain that ... "it's still not quite right"

roll on DSD ?

Posted: Fri Sep 01, 2006 1:04 pm

boy i know what you mean about taking the otherside... i only hope that you (like i) find that it often helps us reevaluate our own beliefs... walk amile in my shoes kinda thing... i have always held rather strongly that any position worth defnding HAS to be worth reexaming... i appreciate you brother...

Posted: Fri Sep 01, 2006 3:56 pm

dementedchord wrote:

... i have always held rather strongly that any position worth defnding HAS to be worth reexaming... i appreciate you brother...

thanks mate

not easy and I've racked my brain over the last couple of days to find a new way to present the idea but I just can find a way that doesn't end up falling ito the same lsb and hard math sort of thing.
AND that math is not my strong point and although can follow I'm not the one to try to drive a point of view.

it is actually a harder subject to investigate than the more simple sample rate
even then the high sample rates 0f 96 and 192 has it's detractors
but
it is a much easier concept to see that 192 has more sample points in each second and therefore there is more data as the sampling takes place.
more = more resolution

The vertical or amplitude axis is clouded by the log scale and noise floor ... and dither ... and algorithms

An A to D for the sole purpose of tracking say temperature between a highest and a lowest defined point if much easier than tracking an audio signal.

I have a feeling that a presentation of PCM should be based around the zero crossing points
rather than the usual peak to peak presentations as found at the horrible ol' Wikipedia . Wink

http://en.wikipedia.org/wiki/Pulse-code_modulation

Perhaps it is just easier to hope and accept that next years ADDA is cheaper and better sounding than the one I am currently using and in turn that was better than my old one.

I admit that I have been living that way since I traded my 888 for the newer 888/24 .... waaay back at around PT4.3
and now on HD with 96I/O's and yet I still record at 44.1/24.

What will the future bring ?
will it be higher sample rate AND an increase in bit depth ... 32 or 48 with an even bigger increase for the mix engines
or
will we jump from PCM and find a new method (DSD) ??

Can anyone remember the first time they were presented with serious digital audio ?
For me it was 81 or 80 and was the 16/44.1 destined to become CDaudio and I still have fond memories of the first Fairlight 16 I ever saw and heard (it was a CMI I think) .

Posted: Fri Sep 01, 2006 4:44 pm

Kev wrote:

it is a much easier concerpt to see that 192 has more sample points in each second and therefore there is more data as the sampling takes place.
more = more resolution

i've always found it curious why it can't be held similarly... 2 to the 16th power versus 2 tothe 24th (no calculator sorry) if seen as discrete slices of the maximum 24 has 256Xs the # of possible representations....

The vertical or amplitude axis is clouded by the log scale

and perhaps thats where it breaks down for me... while i found math easy in school i chose not to take much... had to practice... and it's not too hard to imagine that some of that 256 is eatten up pretty quickly by a decade or two...

What will the future bring ?
will it be higher sample rate AND an increase in bit depth ... 32 or 48 with an even bigger increase for the mix engine

it's already a lovely bit of masturbation... because we can ??? there's better things to research... i might fancy leaving this rock for instance... then there's oh don't get me started...

will we jump from PCM and find a new method (DSD) ??

doesn't that qualifiy as everything old is new kinda thing??? not sure here but as i understand it it's just based on what sony tried to do with those mix down decks built on video transports... hey there's an idea...lol but the sony decks were pretty serious nakamichi made one that i had for a while... predictive delta modulation i think they called it i think lexicon also played with it some if memory serves...

Can anyone remember the first time they were presented with serious digital audio ?

well i never got my hands on it but saw a very early denon maschine at CES in chicago once... whoa... also sold hi-fi for years and carried the first philips cd players...

I still have fond memories of the first Fairlight 16 I ever saw and heard (it was a CMI I think) .

i know what you mean somewhere around here i have a copy of the original Synclavier brochure....
BTW sorry if my comments show inside your's afraid i'm rather inept at this ...

see ya round the playground....

Posted: Sun Sep 03, 2006 7:05 pm

Kev wrote:

Vision / Video
1 bit video gives the two state situation. BLACK or WHITE
8 bit gives the 256 grey scale
but that scale falls between the same two extremes of the BLACK and the WHITE.
There is more information within the range.
... then comes the 16 bit and true colour stuff ... but still falls into the between BLACK and WHITE.
BUT cameras with CCD blocks can see into the Infra Red ... beyond the above range
???

Well...a couple things here. First, 8 bit is not entirely relegated to B&W or Grayscale. In fact, you can have 8 bit color.

Also, Infrared is not solely available to digital cameras with CCDs. In fact, any camera can see InfraRed. However, bear in mind, this is not to be thought of as "beyond the range." In fact, much like digital audio, digital video/photo is broken into 2 types of incoming signal - in audio, it's frequency and amplitude, in video/photo, it's luminence and chrominence. Infrared qualifies as a luminence (or light) value, not a color value

Kev wrote:

PCM Audio
both 16 and 24 bit have the upper limit of 0dBFS ( Full Scale )
the largest 16 bit word send you down to the 16 bit noise floor
and the 24 bit work can send you LOWER.

so part of the expanded 16 to 24 bit is to expand the area we are working with.

wait for it

Does the expanded 16 to 24 bit ALSO provide for more detail within the OLD 16 bit area.

Rolling Eyes

get my drift ?

as I said above , this analogy is not strickly correct so don't all go juming on me ... just trying to present it in a deferent sort of way.

if that does sit ok then I can present the idea that a given signal captured into a 16 bit system can be identical to the same signal captured into a 24 bit system.

Well, clearly you're playing devil's advocate here... Wink

But, of course, we know that, as more values (voltage) are able to be represented on an exponential basis in binary, then obviously the increase from 2 to the 16th versus 2 to the 24th obviously represents a broader range than merely 1.25x increase in SNR. Clearly, we are getting more values within ALL ranges (as is analogous to color/imaging/video)

J

Posted: Mon Sep 04, 2006 5:11 pm

Cucco wrote:

First, 8 bit is not entirely relegated to B&W or Grayscale. In fact, you can have 8 bit color.

yep
I was just trying to keep things narrow just to illustrate a point and as I said vision is not a great way to explain audio but it can help to highlight an error that may also be in audio with a given technique if we go looking for it.
Still there really is no parallel with audio and vision in sampling and compression techniques.

Cucco wrote:

Also, Infrared is not solely available to digital cameras with CCDs. In fact, any camera ... Infrared qualifies as a luminence (or light) value, not a color value

yep
I only mentioned the CCD as it was something that many could relate to here as this type of camera is common place now.

I wonder if I could get an IR image out of one of our old tube cameras

Cucco wrote:

... we know that, as more values (voltage) are able to be represented on an exponential basis in binary, then obviously the increase from 2 to the 16th versus 2 to the 24th obviously represents a broader range than merely 1.25x increase in SNR. Clearly, we are getting more values within ALL ranges (as is analogous to color/imaging/video)

and here is I think the point at which the two camps seems to get stuck

I think if you rescale the peak of the signal being sampled to the maximum value then you do have more bits/values to describe the original.
Hence the original premise of the thread.

But if you stay at -14dBFS or -20dBFS = 0dBu ... for example
then there may be little difference between the two samples.

Even when this re-scaling is brought up the Digital Boffins still continue the argument that resolution is incorrect.
It is here where I fall in a heap trying to present their argument cos I can help but fall in with Cucco's

Quote:

... we are getting more values within ALL ranges ...

as I said above ... the grey scale idea was the best I could come up with. The boundaries of white and black are fixed and the 254 variations in between.
In audio one boundary of FS is fixed and the other boundary is moved
... SN increases ... and the scale is logarithmic

I haven't got a pictorial/graphical way to present this and an Excel Spreadsheet of values is too big and hard to visualise.
Rolling Eyes

so I fail at being devils advocate.

sometimes these things are not worth worrying about and I tend leave a little margin of say -6 to a max of -12dBFS on any take and even if it falls outside of that but is a RIPPER then I keep the take and move on. The performance is probably more important than anything.
Once you have chosen your gear and the sample rate and bit depth ... and the Mic and Mic-pre then all this geek stuff should be pushed aside and just record.
Rolling Eyes

Posted: Thu Sep 07, 2006 10:18 pm

not to be an ignorant asshole but which is the best 24 bit, or 48? has it been ruled out that 16 is unfit for natural selection?

do you believe that 192kHz is useless? not only that but truly decimation?

i honestly wanted to believe that analogue tape and albums and what not display beyond 20-20khz. but he does make a point that if it goes to cd whats the point.

where can you even find a 48/192?

Posted: Fri Sep 08, 2006 2:23 am

mmm
... is someone stirring here ?

liquidstudios wrote:

do you believe that 192kHz is useless? not only that but truly decimation?

not useless
one simple useful fact for the high sample rate is

for a given function that may take say, 10 samples
at 192 khz it takes 1/4 of the time to get that job done than at 48khz
so latency is shorter

that could be a good thing for mix engines and live work

Quote:

... but he does make a point that if it goes to cd whats the point.

because the calculation should be at better acuracy than the result
... calculators go to many decimal places internaly and yet you may choose to display only 2 decimal places

most stuff is being distributed at low MP3 qualities
so
we should record at lower quality

I'm happy with 44.1 and 24 for the time being

Quote:

where can you even find a 48/192?

as a mix engine .. or as a recorded file ?
there is a few mix engines and I've hear of some prototype stuff but not for retail
and I don't think there is an AES or EBU standard yet
but then I'm not that interested and may have been asleep

Posted: Fri Sep 08, 2006 7:51 am

liquidstudios wrote:

not to be an ignorant asshole but which is the best 24 bit, or 48? has it been ruled out that 16 is unfit for natural selection?

do you believe that 192kHz is useless? not only that but truly decimation?

i honestly wanted to believe that analogue tape and albums and what not display beyond 20-20khz. but he does make a point that if it goes to cd whats the point.

where can you even find a 48/192?

Nope, not coming across as an ignorant a-hole, just an inquiring mind.

48 bit is superior to 24 bit. However, there are no converters or engines capable of recoring at 48 bit currently. (Even those DAWs which claim to operate at 64 bit don't record anything at 64 bit, the converters simply don't put that many bits out.)

48 bit is usually reserved for processing on digital devices. For example, Dan Weiss's gear operates at 48 bit fixed point as does some Z-sys and Waves stuff. Then, these devices have to dither back down to 24 bit. The good news about this is that much of the advantage of the 48 bit processing is retained when dithering back down to 24 bit. (The advantages being a cleaner treatment of the processed audio due to the fact that any artifacts are well outside the audible range amplitude-wise).

Personally, I believe that 192 IS a benefit despite what Dan Lavry states. As Kev mentions, there is less overall latency, but also less in-band distortion.

As far as analog tapes/albums going past 20kHz - it's entirely possible, but the limitations are that the hardware (analog) capable of doing this is quite expensive. Normal Sony or Pioneer tape decks or JVC turntables just aint gonna do it.

J.

Posted: Fri Sep 08, 2006 11:46 am

Quote:

we should record at lower quality

I'm happy with 44.1 and 24 for the time being

That's high quality!

Quote:

Once you have chosen your gear and the sample rate and bit depth ... and the Mic and Mic-pre then all this geek stuff should be pushed aside and just record.

Yes sir!

also - higher sample rates allow an alias filter much farther above the frequencies of interest, thus skewing them less. Then, when you are processing, you are processing a better representation of the original. Eventually the rate has to come down, but higher sample rates allow you to forgo that damage until you are done doing your creative damage!

Having said that, do I think it real