Regarding digital recording formats, what is the bit depth/sampling rate beyond which the human ear/body can no longer detect any improvement in quality?

Is there any reason to increase bit depth from 24-bit, and if so then how far? And is there any reason to increase sampling rate from 192kHz, and if so then how far?

Sorry this question is not exactly mastering-related! I put it here because I thought mastering engineers would be the ones most likely to know the answer. I am just curious, that is all.

24-Bits (144dB) is more than enough.

As for sample rates, a few studies have shown that listeners can discern differences between 96 and 44.1 and even 44.1 and 96 downsampled to 44.1. Mainly to do with the perception of pre-echoes.

This paper is a bit old and converter design has come a long way but still an interesting read:

http://www.cirrus.com/en/pubs/whitePaper/DS668WP1.pdf

There's similar research if you delve into AES journals.

ken

--------------------
I'm All Ears.

It somewhat depends on the music, arguably 16 bit as a final bit depth for a format is perfectly ok for low dynamic range music, hiphop/dance/pop/rock etc. when all preceding stages have been at 24bit capture/mixing/mastering.

I am doubtful if there is anything audibly different at more than 24 bit resolution even in the quietest of rooms. It brings into question the definition of the word "quality" as well, which is a factor at many stages more important than bit depth IMO.(with maybe the exception of A/D conversion)

cheers

--------------------
Mastering online mastering

Thanks, I wondered about this for some time. So it seems existing digital recording technology is more than capable of doing the job, and (production aside) any quality issues are due to format of transmission/playback...

Well that is quite surprising (to me)!

Quote Richard Vine:

---

Thanks, I wondered about this for some time. So it seems existing digital recording technology is more than capable of doing the job, and (production aside) any quality issues are due to format of transmission/playback...

---

I think it depends on what your definition of "doing the job" is.

Can you record an analog 1/2 inch master tape to the "best" digital format, though the best "a to d" converters and still be able to tell the difference between the source and digital recording blind?

IME, yes, you can tell a difference, so is there still room for improvement in conversion quality?

--------------------
PM for Affordable Rates - High Quality Analog Mastering
www.waltzmastering.com www.waltzmastering@gmail.com

Quote Richard Vine:

---

So it seems existing digital recording technology is more than capable of doing the job, and (production aside) any quality issues are due to format of transmission/playback...

---

That's not what was said. Current sample rates and bit depths are more than adequate for capturing audio, but that doesn't mean there aren't improvements to be made in digital recording technology. The sound chip on my motherboard and a £1500 audio interface can both record at 44.1 kHz and 16 bit but that doesn't mean they'll produce identical recordings.

Quote Waltz Mastering:

---

IME, yes, you can tell a difference, so is there still room for improvement in conversion quality?

---

Playing devil's advocate for a momeont... I'll say no!

There is no doubt that you can tell the difference between a music track recorded onto 1/2-inch and to digits (via the best converters available). But then you';d expect that surely because they are entirely different formats -- just as you'd expect 16/44.1 to sound different to 24/96 given wide band source test material.

There are two important questions to my mind:

Which one is the more accurate (ie introduces the least distortions)?

And which sounds the most pleasant for musical enjoyment?

I don't think many would argue anything other than highest the quality digits for the first.

The second is entirely subjective of course...

There are some occasions where the charactearistics of tape enhance the sound character in a pleasing way... and there are others where accurate fidelity is important.

Hugh

--------------------
Technical Editor, Sound On Sound

hi,

just to 'amplify' the above responses, for your reference, there are three simple formulae to keep in mind:

1) Nyquist sampling limit - the well known "maximum frequency captured by pulse code modulation sampling is one half the sampling frequency". as with any technology, the difference between theory and implementation can be large, notably in the anti-aliasing filters used in the converters. that said, few people can hear anything above 15 kHz.

2) the dynamic range is determined by the number of bits in the sample. the formula is each bit gives an additional 6 dB of dynamic range. that's why 16 bit sampling used in CDs yields a theoretical 96 dB dynamic range.

3) each D/A conversion costs approximately 3 dB of dynamic range. that's why most CD players quote as 93 dB dynamic range.

so: 44.1 kHz sampling, 16 bit samples: 22.05 kHz max frequency, 93 dB max dynamic range.

96 kHz sampling, 24 bit samples: 48 kHz max freq, 141 dB dynamic range.

some symphonic music can approach this dynamic range limit but most doesn't. btw, 'brick wall' limiting dramatically reduces dynamic range needed (that is why "make everything louder" requests drive good mastering engineers to contemplate murder a/o suicide depending on their mood a/o how much they need the job to pay the rent).

btw2, there are many variations of PCM sampling in use that change these numbers a little but basically the above math holds.

hope this helps,

d

--------------------
"Patsy had the drug tolerance of Keith Richards and the moral rectitude of Brian Jones." - Dr. Walter Bishop, "Fringe"

The only people who REALLY care about super-high resolution numbers work in the marketing department of equipment manufacturers. Everyone else knows that there are hundreds of factors more important than this in making a good recording.

Quote dubbmann:

---

...the difference between theory and implementation can be large, notably in the anti-aliasing filters used in the converters. that said, few people can hear anything above 15 kHz.

---

This is (simplistically) why 96kHz sampling has greater fidelity in some practical applications than 44.1kHz; why a 60kHz sample rate would have been a much better idea all round in the first place; and why 44.1kHz is an acceptably pragmatic solution for run-of-the-mill domestic applications.

Quote:

---

the formula is each bit gives an additional 6 dB of dynamic range. that's why 16 bit sampling used in CDs yields a theoretical 96 dB dynamic range.

---

That's a close approximation to the signal-noise ratio of an undithered quantiser. For a dithered quantiser (which is the only kind of any use in quality audio applications) the rough forumla is SN ratio =6n-3dB, where n is the number of bits in the wordlength. (The dither noise accounting for half a quantising level's worth of noise!)

Hence a practical 16 bit audio system gives a signal-noise ratio of 93dB. A theoretically perfect 24 bit system would be 141dB, but typical implemenations are currently around 19, 20 or 21 bit resolution giving SN ratios of 111, 117 or 123dB.

Quote:

---

3) each D/A conversion costs approximately 3 dB of dynamic range.

---

This is true, because the noise floor is incoherent for each conversion stage and thus the noise floor rises by 3dB for each A-D-A conversion.

Quote:

---

that's why most CD players quote as 93 dB dynamic range.

---

No it's not... see above!

Quote:

---

96 kHz sampling, 24 bit samples: 48 kHz max freq, 141 dB dynamic range.

---

In your dreams

Quote:

---

btw, 'brick wall' limiting dramatically reduces dynamic range needed

---

Only if you don't mind the original source material being distorted...

Hugh

--------------------
Technical Editor, Sound On Sound

Quote Exalted Wombat:

---

The only people who REALLY care about super-high resolution numbers work in the marketing department of equipment manufacturers.

---

It looks like a lot of people posting on some other well known pro-audio forums care about 'super high resolution numbers' too. And some of those buying the stuff must do so because they believe it, so it's more than just the marketing people!

hugh

--------------------
Technical Editor, Sound On Sound

Quote Hugh Robjohns:

---

Hence a practical 16 bit audio system gives a signal-noise ratio of 93dB. A theoretically perfect 24 bit system would be 141dB, but typical implemenations are currently around 19, 20 or 21 bit resolution giving SN ratios of 111, 117 or 123dB.

---

24 bit converters however have any explicit dither buried under many bits of self noise so arguably the numbers would be straight 6 times effective word length (Obviously that is coming from the wrong direction as in reality it is the noise performance that is specified, not the number of bits).

In practise I don't think anyone explicitly dithers 24 bit conversion, the thermal noise inherent to the electronics is more then good enough.

Quote:

---

Quote:

---

96 kHz sampling, 24 bit samples: 48 kHz max freq, 141 dB dynamic range.

---

In your dreams

---

One interesting observation is that SNR is specified over a stated bandwidth, so you COULD run a 16 bit converter at way above 48K and hide all the dither energy up above 20Khz, giving you a 16 bit converter with a noise floor in the audio band better then 16 bits....
This is in fact how most modern converters operate (but most are WAY shorter then 16 bits at the input stage).

One other thing I spotted up thread was an all to common misstatement of the sampling theorem.
The actual condition is NOT that the sample rate must be twice the bandwidth to be recorded, but that it must be GREATER then twice the bandwidth to be recorded.
I think this misunderstanding gives rise to a lot of confusion when considering say a signal at 22.05Khz with a sample rate of 44.1Khz (which actually fails the test at least if the intended bandwidth extends to DC).

Regards, Dan.

--------------------
Audiophiles use phono leads because they are unbalanced people!

Quote GaryM:

---

Quote Richard Vine:

---

So it seems existing digital recording technology is more than capable of doing the job, and (production aside) any quality issues are due to format of transmission/playback...

---

That's not what was said. Current sample rates and bit depths are more than adequate for capturing audio, but that doesn't mean there aren't improvements to be made in digital recording technology. The sound chip on my motherboard and a £1500 audio interface can both record at 44.1 kHz and 16 bit but that doesn't mean they'll produce identical recordings.

---

Oops that was indeed a slack comment, lol. I only meant the bit depth and sample rate, and assume that over time continual improvements will still be made to the machines/processes.

Another way to ask the question would be whether we might be using 32-bit 384 kHz in 100 years? Except it sounds like there would be no point. I should have said "So it seems the existing digital recording format is more than capable of doing the job, and (production/technology aside) any quality issues are due to format of transmission/playback..."

Quote dubbmann:

---

...that said, few people can hear anything above 15 kHz.

---

I read/heard something really interesting recently, it might have been here but am not sure - I will check when time permits as it is bugging me now.

The speculation was that people do not merely hear music, they feel it too. In other words music causes vibrations in your body (cavities) which may affect you even though you cannot hear the frequencies responsible. If I remember correctly the specific example I am looking for concern a well-known producer/engineer (?) who had a problem with a certain channel on an expensive desk. At first no one could find the fault, but it turned out to be a high-frequency emission that was annoying the chap even though technically he was unable to hear it. I will watch this video again, although if anyone can find/remember where these comments were, please let me know. I meant to Google it before I forgot the bloke's name!

Anyway, assuming there is something in this, then it occurred to me that we might yet require a higher than expected resolution to cope with inaudible frequencies too?

Hooray, found it at 1:07:35 in the link posted above! One source of this allegation is here, about halfway down the page (scan for Geoff Emerick). This bloke Neve says that a sample rate of 192 kHz is required, and it is because of frequencies we cannot hear!

Scary stuff

Whilst higher sampling frequencies and greater bit depths might be advantageous in some applications, a significant limitation of digital audio data has always been the bit-depth-dither-screw (BDDS) where bits become twisted out of polarity from the norm and slip unprocessed through the umbilical filters. It's a subtle effect that is only audible through the left ear (unless you are sitting to the left of a centre). Balanced phono leads can lift this out of the electromagnetic spectrum but the effect is never entirely eliminated.

It is of course well known that the use of recursive compression in the bit-phase domain will lead to a minimisation of the maximum polar wobble and mask, psycho-acoustically, the effect of bit twisting (sometimes mistakenly referred to as bit-scribbling – an entirely different phenomena) as can, quite by chance, the orientation of a virtual valve-pod placed within a hundred meters of the amplifiers current bucket. However, even the most determined cannot mask sub-bit elastic phasing which can dynamically alter the effective height of each bit such that analogue signals flow back into the DAC causing a reversal of the bit clock and a negative shift in the phase locked loop. This can result in a total vacuum.

Hope that all makes sense.

--------------------
twas brillig and the slithy toves did run the country..

Quote dmills:

---

In practise I don't think anyone explicitly dithers 24 bit conversion, the thermal noise inherent to the electronics is more then good enough.

---

Quite so -- although I suspect it's actually the case that the inherent thermal and electronic noise of the analogue circuitry swamps any conceptual dither noise.

Quote:

---

One interesting observation is that SNR is specified over a stated bandwidth, so you COULD run a 16 bit converter at way above 48K and hide all the dither energy up above 20Khz, giving you a 16 bit converter with a noise floor in the audio band better then 16 bits....

---

Yep.. noise shaping is a wonderful thing!

Quote:

---

One other thing I spotted up thread was an all to common misstatement of the sampling theorem.
The actual condition is NOT that the sample rate must be twice the bandwidth to be recorded, but that it must be GREATER then twice the bandwidth to be recorded.

---

Absolutely.

Hugh

--------------------
Technical Editor, Sound On Sound

I find it interesting that everyone is concerned about the HF part of the spectrum but an analogue tape machine rolls off way down compared with digital, to that end they are both imperfect systems.

I suppose arguably, typical playback systems are potentially less limited in response up top than down below.

As soon as analogue equipment is in the chain the noise floor is often above dither noise floor levels(and it is not even noise shaped) and this is not usually deemed a problem. So discussing the limits of 24bit systems when your typical analogue floor is -85/-90dBFS is somewhat a limiting factor in itself.

It is rare a digital system is worth discussing in isolation as there has almost always been an analogue aspect involved.

Quoting from Manley Vari Mu manual :

Noise Floor -80 dB : 20Hz to 80Khz: -90 dB typical

P.S. In the name of not confusing a casual browser it might make sense if the twaddle from Mark G above was deleted, (amusing though it was)

--------------------
Mastering online mastering

Quote Mark G:

---

Whilst higher sampling frequencies and greater bit depths might be advantageous in some applications, a significant limitation of digital audio data has always been the bit-depth-dither-screw (BDDS) where bits become twisted out of polarity from the norm and slip unprocessed through the umbilical filters. It's a subtle effect that is only audible through the left ear (unless you are sitting to the left of a centre). Balanced phono leads can lift this out of the electromagnetic spectrum but the effect is never entirely eliminated.

It is of course well known that the use of recursive compression in the bit-phase domain will lead to a minimisation of the maximum polar wobble and mask, psycho-acoustically, the effect of bit twisting (sometimes mistakenly referred to as bit-scribbling – an entirely different phenomena) as can, quite by chance, the orientation of a virtual valve-pod placed within a hundred meters of the amplifiers current bucket. However, even the most determined cannot mask sub-bit elastic phasing which can dynamically alter the effective height of each bit such that analogue signals flow back into the DAC causing a reversal of the bit clock and a negative shift in the phase locked loop. This can result in a total vacuum.

Hope that all makes sense.

---

Private Eye would just love that one... I bow to your knowledge, but it's mostly gobbledegook to a lowly muso like me.

In the end, even with 96 (or 93, to be accurate to Hugh's specs) db of dynamic range available at 16 bit, most modern records fail to achieve even 5db of dynamic range, due to the loudness war, and most punters are listening to our stuff on pretty [ ****** ] systems.

Yes, record as best you can, but more and more, it's the basic musical elements of what you're recording (the tune, the beats, the performances) that require the most attention, not the bit rate or sampling frequency.

A [ ****** ] tune will always sound [ ****** ], regardless of the recording quality - a good tune always comes through (even if sung by 2 kids of dubious talent in the back of the car all the way back from Cornwall, as I had to endure recently. Their renditions of Penny Lane, Barbara-Ann, and other ancient ditties, often with new words and sometimes morphing from one tune to another in the middle of a verse, just proved the point. Somehow the genius of the songwriters still held up)

That's what sells music - good tunes, not good recordings.

--------------------
http://anotherfineday.bandcamp.com/ http://anotherfineday.co.uk http://apollomusic.co.uk

Quote SafeandSound123:

---

I find it interesting that everyone is concerned about the HF part of the spectrum but an analogue tape machine rolls off way down compared with digital, to that end they are both imperfect systems.

---

They are both imperfect -- but in different ways. The tape roll off tends to be gentler, and so with a more 'natural' phase response. Digital tends to be much more abrupt, and usually with a linear phase response which results in unnatural pre-echoes. The latter are significantly reduced with higher sample rates, which might be another reason (along with the removal of other filter artefacts) why some people prefer the sound of higher sample rates.

hugh

--------------------
Technical Editor, Sound On Sound

I love how these threads on SR always end up advising the OPs that there are many other factors to consider when recording than the converters.

If one calmly re-reads the OP, one can surmise that he has asked a question about the theory and science behind sampling. Had he had asked "will my recordings get better using higher a higher SR?" then we could probably advise on all other factors. As it stands, he's asked a theoretical question about the science involved.

ken

--------------------
I'm All Ears.

You're quite right -- and in answer to his first question I would say about 20 bits wordlength and 60kHz sample rate.

Hugh

--------------------
Technical Editor, Sound On Sound

Quote ken long:

---

I love how these threads on SR always end up advising the OPs that there are many other factors to consider when recording than the converters.

If one calmly re-reads the OP, one can surmise that he has asked a question about the theory and science behind sampling. Had he had asked "will my recordings get better using higher a higher SR?" then we could probably advise on all other factors. As it stands, he's asked a theoretical question about the science involved.

---

But if you went to your doctor, you'd be grateful for his holistic approach rather than wanting "a straight answer to a straight question". That's how online forums work - determinedly holistic :-)

(Funny how practitioners who CLAIM to be holistic often reject 90% of possible therapies - usually the tried-and-tested ones that actually work. But that's another topic.)

If I went to a car forum and asked how using a different oil may improve the performance of my engine (and the scientific reasons why), I wouldn't want some smart ass telling me that changing the way I drive will have more of an impact on performance than changing the oil. Yes, it might be true - but it wouldn't be what I asked for.

I think OP was specific, not only in his question, but also where he chose to post: the Mastering Forum - Not a place much concerned with Recording Techniques. There's a forum for that.

ken

--------------------
I'm All Ears.

Yeah - we just WON'T do exactly what we're told, will we! You must find it very annoying! :-)

Yes!...............especially when i don't understand all the big words mummy.

--------------------
My head hurts!

Mark, that was totally classic.

Ref the original question:
There is a difference between the needs of production and the needs of a final distribution medium, mostly related to the lack of need for headroom in the distribution case, and while I personally consider 20 bits to be slightly overkill for a distribution medium (it asks a LOT of the listening room and repro chain if the bottom few bits are to have any meaning and most of them just are not anything like that good), I would basically agree that 20 bits and 60Khz is more then sufficient as a distribution medium (And is in all likelihood sufficient as a production medium for all but the most critical of jobs).

This is not to say that there is no way to improve the AD/DA parts, but basic word length and sample rates are no longer the limiting factors.

Regards, Dan.

--------------------
Audiophiles use phono leads because they are unbalanced people!

Quote ken long:

---

I love how these threads on SR always end up advising the OPs that there are many other factors to consider when recording than the converters.

If one calmly re-reads the OP, one can surmise that he has asked a question about the theory and science behind sampling. Had he had asked "will my recordings get better using higher a higher SR?" then we could probably advise on all other factors. As it stands, he's asked a theoretical question about the science involved.

---

Yup it was a simple question, perhaps too simple, but the answers are nonetheless interesting - certainly much more to this than meets the eye. Mark G's post had me suckered right up until the words "total vacuum".

If you think that post was bizarre check out the comments Rupert Neve made regarding this subject. He says 24-bit is fine except 192 kHz is required as the frequency range has to be free (of distortion/noise) up to about 75 kHz. Not doing this is very dangerous. Why? Because the Japanese showed (bad) distortion/noise above 20 kHz makes the brain produce 'electric radiations' associated with (brain) discomfort, frustration, and even anger. Therefore Neve wonders whether we cannot blame the CD for some of our social problems?

So, is it true that the frequency range ought to be considered up to 75 kHz? Is it possible that mp3s could cause bit-rage? Is this bloke actually on to something, or is he off his swede?

That Japanese paper is a red herring as far as I can see, even if the result is valid, it assumes the repro chain produces significant noise above 20K and that is far from clear, as is the relationship between experiments done with ultrasonic noise generators and the need to RECORD the energy up there (Lots of SMPSU switching noise I should imagine).
Showing that the presence of a high level of ultrasonic noise has a physiological effect is interesting (and possibly not all that surprising), but it is a big leap to go from that to a psychological effect never mind needing to record and reproduce it.

How many speakers are capable of meaningful reproduction at 50Khz anyway?

Rupert is basically an analogue console audio guy, and in that context pushing the passband out to 75K might make a kind of sense, mostly to keep the in band phase shifts small, it is one thing to design a 75K bandwidth in for example a mixer or such, in order to keep the in band phase shifts under control, it is another to claim a need to record that bandwidth.

This is the same reason the input caps on good gear are often many times what a simple calculation would predict (If everything was -3db at 20Hz then an entire record chain would accumulate a huge low frequency phase shift due to all those cascaded filters). The same basic idea applies at the high end (Where if you have many cascaded low order filters you do NOT want them going over anywhere near 20Khz....

However, the rules change somewhat when you have convenient ways to build filters that are essentially flat (in both frequency and phase) out to just past the audio band then drop like a rock, all that phase shift occurs very suddenly outside the audio band, so it doesn't matter.
Digital gives us the means to (if we are clever about it) have near DC to 20Khz with reasonably sharp filter either end and without tending to accumulate phase errors at the inputs to every bit of gear in the chain (There is normally less gear to start with which helps).

I still say there are bigger issues in digital audio then sample rate and word length.

Regards, Dan.

--------------------
Audiophiles use phono leads because they are unbalanced people!

"beyond which the human ear/body "

I cannot remember the unwavering technical specifications of a standard human ear/brain combination

The fact that hearing is variable between all of us means there is no scientific absolute and the question become less direct with less chance of a direct answer which I think has already been covered.

--------------------
Mastering online mastering

Quote SafeandSound123:

---

"beyond which the human ear/body "

I cannot remember the unwavering technical specifications of a standard human ear/brain combination

The fact that hearing is variable between all of us means there is no scientific absolute and the question become less direct with less chance of a direct answer which I think has already been covered.

---

Are you suggesting there is NO cut-off point beyond which you can say "humans can't hear this"? Some manufacturers would love to be able to suggest an ultra-wide range just MIGHT have audible effect!

On a loosely related note, did anyone else suffer Radio 4 this morning? (Was it Womans' Hour, or the program before it?) A "researcher" has decided that when Mum goes out to work it can actually benefit her children. No new evidence, just old data re-evaluated. A classic example of "tell me the conclusion you want, I'll make the research prove it" I think! Next question "why will staying at work until you drop be enormously beneficial to all concerned?"

Quote Richard Vine:

---

...the Japanese showed (bad) distortion/noise above 20 kHz makes the brain produce 'electric radiations' associated with (brain) discomfort, frustration, and even anger. Therefore Neve wonders whether we cannot blame the CD for some of our social problems?

---

Two things: I believe that paper has been discredited, and show me a conventional hi-fi systemn that can reproduce significant amounts of energy above 20kHz.

If ultrasonics really do cause behavioural problems, then best stay well away from everyone who uses a laptop, everyone with an LCD TV, everyone who has any equipment with a switched mode power supply, every percussionist, every brass player, anyone who uses a dog whistle, and anyone who has bats flying around in their garden at night!

Quote:

---

Is this bloke actually on to something, or is he off his swede?

---

Definitely fallen off the turnip

Hugh

--------------------
Technical Editor, Sound On Sound

Quote SafeandSound123:

---

I cannot remember the unwavering technical specifications of a standard human ear/brain combination

---

You're not an audiologist then. As I understand it, vibrations are converted into nerve impulses by tiny hairs in the cochlea in the inner ear with different sets of hairs for different frequencies. The highest frequencies that these hairs can detect seems well known and extends to around 22kHz in young people but it gradually reduces with age as some of the finer hairs flatten and no longer do their job.

James.

--------------------
JRP Music - Audio Mastering and Restoration.
http://www.jrpmusic.net

Quote Richard Vine:

---

Because the Japanese showed (bad) distortion/noise above 20 kHz makes the brain produce 'electric radiations' associated with (brain) discomfort, frustration, and even anger. Therefore Neve wonders whether we cannot blame the CD for some of our social problems?

---

You are talking about a paper by Oohashi et al which, as others say, has been discredited on at least 2 counts. Firstly, no-one else has been able to repeat his observations and, secondly, his observations could simply be down to listener bias as (as I understand it) the two options were always presented in the same order so the listener was always more familiar with the material when listening to the second sample.

James.

--------------------
JRP Music - Audio Mastering and Restoration.
http://www.jrpmusic.net

Quote:

---

Are you suggesting there is NO cut-off point beyond which you can say "humans can't hear this"? Some manufacturers would love to be able to suggest an ultra-wide range just MIGHT have audible effect!

---

No, we all know it's around 20kHz under best conditions.

What I am saying is that you cannot second guess a humans hearing response as everyone hears a little differently, the shape of your ear/face/ear canal dimensions alters the response. Move 2 cms from your current head position and your in room hearing response has just altered.

Quote:

---

You're not an audiologist then.

---

No. I understand the bio mechanism of hearing, but it does not give an absolute where you can judge everyones hearing, sure take an average but averages don't correlate to straight Q's/A's which someone was requesting.

--------------------
Mastering online mastering