Quote kris:

---

I'm talking about low-level signals where you DON'T have a lot of sampling resolution (although this distortion occurs at all bit-depths, it's level is inversely proportional to bit depth).

---

Er... hoist on your own petard perhaps.

In a correctly dithered quantiation system, there is no quantisation distortion. Only in a crude undithered system is the level of distortion inversely proportional to the bit depth.

The whole point of dithering is to linearise the transfer function, to achieve nominally zero distortion at the expense of a fixed but very small amount of pure noise (whose flavour depends onthe dithering method chosen).

Resolution absolutely does not diminish with amplitude, only the signal to noise ratio. In a correctly dithered system, it is perfectly possible to encode undistorted signals of a smaller amplitude than the smallest quantisation level.

Hugh

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

Quote Lighthouse_Mastering:

---

Off the top of my head, Burr Brown make a 24bit high sample rate capable multibit DAC, and I know that there are others.

---

I guess you are thinking of chips like the PCM1704 and others in that family that all use sign and magnitude conversion... but with 16x(44.1) oversampling filters. Okay, so not quite as heavy handed as true delta-sigma... but then no one uses true (single bit) delta-sigma in pro audio applications anyway. Even these 'multibit' converters employ all the same oversampling techniques, digital filters, noise shaping and so on. There really is no getting away from it.

Quote:

---

I agree that digital is better today than it was 20yrs ago, but just not as good as it could be. Good to see you are an optimist.

---

Always the optimist! I fear we may be converging on agreement again here, Dave

hugh

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

Below are my views based on my limited knowledge. I don't claim to be an expert on this subject so I might be making some misstakes. Please anyone correct me if I am wrong.

Quote kris:

---

So, let's say you sample at 10hz and you input a 5hz sine-wave (remember, in this example we're using a 1-bit system).

---

This is the first error. Nyquist states that you need a sample rate that EXCEEDS twice the highest frequency of the signal you want to sample.

Quote:

---

Now try a 4hz sine-wave. We have a problem, 101 is a 5hz sine-wave (after reconstruction),

---

This is the next problem. A sine wave (sine (x)) doesn't just start out of the blue and end abruptly. Imagine a curve (or imaginery sine wave for lack of a better term) that starts at the point in time of the first of the three samples and stops after the third. You would have angles at the beginning and end of the curve where it joins the flat lines on either side. These angles represent frequencies way beyond 5Hz and thus do not fit withing Nyquist's sampling theory.

In other words, you are describing a wave form that is akin to a true square or saw wave that has infinite bandwidth and can not exist in the analogue world. Things that don't exist can't be sampled.

Quote:

---

If it can only see a 1 followed by a 0, how can it possibly know the input wasn't simply a 5hz sine-wave?

---

(Ignoring the 5Hz for a sec). Luckily we never have this situation in real life. CD audio has 44100 samples per second. We humans can certainly not recognise the pitch based on a sound that only lasts 1/22050 of a second.

It would sound as a click wich effectively is a sound with alot of high frequency content way beyond the actual frequency of the wave we were sampling. (That is if anyone does 1/22050 second recordings in the first place. So this fits with what I say above about infinite (or at the least extend) bandwidth.

Quote:

---

Now take a 100hz sampling rate with the same input. 5hz could be 10x1s, followed by 10x0s. 4hz could be 12x1s followed by 13x0s.

Despite the fact that ALL freqencies are below the Nyquist limit in both cases, the higher sampling rate gives us a better approximation of the same source.

---

As I have explained above, if the wave you describe could be created in real world, it would have frequency content way beyond Nyquist.

Btw Barish, unlike Hugh, with the exception of giving some nice links to documents about sampling theory and alluding to maths, you have explained ABSOLUTELY NOTHING! (I checked every single one of your posts on this thread). You didn't even seem to pick-up on some errors I posted here that obviously were quite a bit beyond the "joining the dots" missunderstanding for which you have no patience.

On the contrary! You have been insulting, denigrating, aggressive and, contrary to what some have posted, totaly lacking patience. So get off of your high horse. You might have knowledge to share but you are certainly not doing it in this thread.

UnderTow
Edited for blatant typos and other 2:30am brain farts

Edited by UnderTow (02/02/06 01:30 AM)

Quote kris:

---

by low bit-depths, I'm talking about frequencies that modulate only the last few bits in a digital system (i.e. at a very low level).

---

If you're going to talk technical, can we keep the terminolgy accurate to avoid further confusion, please. I think you mean you're are talking about signal levels (not frequencies) that modulate the last few bits..

Quote:

---

For example, there could be frequencies that only modulate the lowest bit (of course there's dither, but I'm not discussing that here).

---

Er... how can you talk sensibly about the problems of digitisation in modern converter design without taking dither into account? This is ludicrous. It would be like talking about tape recordings without AC bias, and conveniently managing to ignore the fact that it would sound terrible as a result!

Quote:

---

Now try a 4hz sine-wave. We have a problem, 101 is a 5hz sine-wave (after reconstruction), but 110011 would be a 2.5hz sine-wave. It's therefore IMPOSSIBLE to accurately sample this 4hz sine waves, despite the fact it falls below the nyquist limit.

---

Oh dear... this is childishly naieve and doing you no favours at all. I strongly suggest you go away and read a few good books on the subject by reputable authors. Dr John Watkinson is as good a place to start as any: The Art of Digital Audio on Focal Press.

Quote:

---

Now Hugh makes a good point about the reconstruction filter taking care of this, but I'm afraid it's NOT applicable here because the filter doesn't have sufficient input resolution to produce anything like a valid output!! If it can only see a 1 followed by a 0, how can it possibly know the input wasn't simply a 5hz sine-wave?

---

Because the little digital goblins have told it so. They listen in on what you send in and make the binary magic happen.... Makes about as much sense as the stuff you are spouting!

It is very simple to feed a high frequency, low level tone in a digital converter, and then decode the output and prove that what comes out is the same frequency tone as you sent in, but obviously well down into the noise -- just as you would expect. Try it... you might surprise yourself.

Hugh

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

Quote kris:

---

Barish yes I've studied calculus, but I really don't see the need for your hostile tone here, I'm not being agressive at all.

---

I appreciate you lack of hostility here, and I apologise for Barish's tone -- I'm sure it is born out of sheer frustration rather than malice.

Quote:

---

Anyway, a 1-bit waveform has no slope, it moves instantly from 0 to 1, so why in this case would derivatives help?

---

You are choosing to ignore the fundamentally crucial role of dither once again, but I agree with you... we have gone as far as humanly possible on this. Let's all go away and make happy music

hugh

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

Hugh is the new avatar cos you've finally come out of the mic cupboard about being the real Chumley-Walmer??

For those not in the know, what Hugh is trying to say is this..

The dashed clever boffins down at the laboratories have come up with this totally top hole new way of recording things. You talk into here, points at the microphone, and magick little number pixies convert it to a a collection of smaller bits and then recreates it as a big bit again through your speakerphones.

The larger the number of magick number pixies working on the sound the better they can re build it and the more mellifluous it is on the ear.


At present most people can only afford 44 or 96 pixies but we at SoS towers have the services of 192 of the little chappies so they can rebuild the bits far more accurately than 44.

Simple really chaps...

--------------------
nibbled to death by an Okapi http://www.soundclick.com/tubilahdog

Waw, 192 pixies....and what do you feed them? Be carefull not to feed too much of Babyshamble's music through them, or they will develop a drug habit and start dating Kate Moss, and you don't want her messing up your pixies!!

BTW, I totally stayed out of this thread because it is way over my head, and I admire people like Hugh that know what they are talking about and take the time to explain it. But the Hitler reference by mister Barish was totally uncalled for, no matter how big his knowledge on the subject.

--------------------
"Do not fear mistakes. There are none."
Miles Davis

Waow. Rereading what I wrote last night, I see that I didn't state what I was pointing at: Although you only need two sample points to represent any frequency below 1/2 Nyquist, it has to be in a stream of bits longer than just two or three samples. At least that is how I understand it.

UnderTow

Quote Dunewar:

---

But the Hitler reference by mister Barish was totally uncalled for, no matter how big his knowledge on the subject.

---

Or was he simply invoking an old Usenet tradition.....?

But I guess that we've all been through thought processes similar to Kris's back in the days when trying to get to grips with the basics of digital audio. Thankfully Hugh still has the patience to expalain how it really works. Maybe there ought to be a digital audio basics page somewhere on the SOS website (if there isn't already) that we can point newcomers to whenever these kinds of discussions arise.

Cheers

James.

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

Quote UnderTow:

---

Btw Barish, unlike Hugh, with the exception of giving some nice links to documents about sampling theory and alluding to maths, you have explained ABSOLUTELY NOTHING! (I checked every single one of your posts on this thread). You didn't even seem to pick-up on some errors I posted here that obviously were quite a bit beyond the "joining the dots" missunderstanding for which you have no patience.

On the contrary! You have been insulting, denigrating, aggressive and, contrary to what some have posted, totaly lacking patience. So get off of your high horse. You might have knowledge to share but you are certainly not doing it in this thread.

---

Why? Is it because I didn't rephrase here the same concepts someone else had perfectly put in another location, but rather gave the links instead?

If someone doesn't know the fact that mathematical functions can be derived and integrated and what that means in terms of AD/DA Digital Analog conversion shouldn't be so smugly discussing the fine tunings of it in the first place.

And if they don't know what that is then they can open up the book and study it. I spent three years studying that [ ****** ] and it got in my head even years after that what those theorems meant in audio, so I'm not going to try to explain it all in a forum thread which will disappear in the depths of this forum in two weeks.

I've just shown the way. Those who are REALLY interested in designing one, can go and study it.

Suffice to say, to put it in simple words, complex mathematical functions sometimes yield two different y results at some points on the x axis, which is called alias. That's what exactly happens in the DA audio conversion. Increasing the bandwidth does not solve this problem. It is anti-aliasing that sorts out that problem. And the more time that you give the filter to process, the more accurate the filtering is. If you increase the sampling speed to increase bandwidth, you steal from the time anti-aliasing filters could use. That's why unnecessary expansion of bandwidth has a negative impact on the optimum quality an AD/DA convertor can achieve. It actually compromises the accuracy. You may like what sounds through it, but you can't call it more accurate.

In most cases you'll find that other people's silliness makes you sound arrogant. Some people don't know the most crucial concepts and theorems that are used in designing a converter, yet they are trying to establish a logic by themselves to which they expect the converters to operate to. It doesn't work that way.

Can anyone show me a white paper from another manufacturer that claims the contrary of Lavry paper? I haven't seen any yet. If there is one, please let me know.

Thank you.

B.

Edited by kris (02/02/06 03:21 PM)

Deleted. I did say I'd try not to participate in this anymore, and I'm sorry my weakness got the better of me!!

Edited by kris (02/02/06 03:22 PM)

Food for thought.

Does High Sampling Frequency Improve Perceptual Time-Axis Resolution of Digital Audio Signal?

http://www.aes.org/e-lib/browse.cfm?elib=7217

Inaudible High-Frequency Sounds Affect Brain Activity: Hypersonic Effect

http://jn.physiology.org/cgi/content/full/83/6/3548#B29

Quote Barish:

---

Why? Is it because I didn't rephrase here the same concepts someone else had perfectly put in another location, but rather gave the links instead?

---

There is nothing wrong with you not having enough time or patience to repeat what others have allready clearly explained but there is absolutely no need for rudeness and agression. Just refrain from posting.

Quote:

---

Suffice to say, to put it in simple words, complex mathematical functions sometimes yield two different y results at some points on the x axis, which is called alias. That's what exactly happens in the DA audio conversion. Increasing the bandwidth does not solve this problem. It is anti-aliasing that sorts out that problem. And the more time that you give the filter to process, the more accurate the filtering is. If you increase the sampling speed to increase bandwidth, you steal from the time anti-aliasing filters could use. That's why unnecessary expansion of bandwidth has a negative impact on the optimum quality an AD/DA convertor can achieve. It actually compromises the accuracy. You may like what sounds through it, but you can't call it more accurate.

---

Thanks for sharing your insights.

Quote:

---

In most cases you'll find that other people's silliness makes you sound arrogant.

---

Usually it is the tone that makes you sound arrogant. Not the content.

Quote:

---

Can anyone show me a white paper from another manufacturer that claims the contrary of Lavry paper? I haven't seen any yet. If there is one, please let me know.

---

I don't think there is one. That doesn't mean _too_ much on itself though. That having been said I think that Dan Lavry is correct in as far as my knowledge goes.

UnderTow

Sorry, I can't help myself - I have to jump in again..

Barish has at least made some effort here to direct people to the correct information, but if they have chosen to ignore it, or have read it but don't fully understand it, then that's really not his fault is it.

If people then start claiming they know exactly what's going on, yet show in their explanations they clearly don't, you can understand why his tone may have risen *slightly* above y=0....

Unless you're equally qualified to argue the t*ss, you really should avoid doing so.

Edited by Stoney (02/02/06 04:58 PM)

Quote UnderTow:

---

I don't think there is one. That doesn't mean _too_ much on itself though. That having been said I think that Dan Lavry is correct in as far as my knowledge goes.

UnderTow

---

How wonderful of you to validate Dan's work ! I'm sure he'll be overjoyed to know you have given him the nod!

Can we end this thread now?

--------------------
"Nothing in the world can take the place of persistence. Talent will not; nothing is more common than unsuccessful people with talent."

Hypothetical , humorous conversation thing....

DL. "there's no point in making 192KHz converters.
\
DCS. "Yes there is...."
/
DL. "well you would say that, what with being someone who makes them..... you're obviously biased aren't you? "
/
DCS, " strange we thought the same about your position, what with , being someone who doesn't make them......"
\
PRM. "pair of wannabees".




Kris 1 bit 10Hz converters.

really mate, you are simply not , for a change, right.,

--------------------
if you don't know who i am, i aint gonna tell you.

Quote DCompton:

---

Food for thought.

Does High Sampling Frequency Improve Perceptual Time-Axis Resolution of Digital Audio Signal?

http://www.aes.org/e-lib/browse.cfm?elib=7217

---

As I understand it from reading the abstract, this simply shows that different filters can sound different. Chances are, they would have obtained similar results by using different designs of filters at the same frequency.

Quote DCompton:

---

Inaudible High-Frequency Sounds Affect Brain Activity: Hypersonic Effect

http://jn.physiology.org/cgi/content/full/83/6/3548#B29

---

This has already been discussed earlier in this thread.

Cheers

James.

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

Quote DCompton:

---

Food for thought.

Inaudible High-Frequency Sounds Affect Brain Activity: Hypersonic Effect

http://jn.physiology.org/cgi/content/full/83/6/3548#B29

---

What part of "this has been discredited" don't you get mate? You've been pushing this the second time in this thread, even though it has already been replied to. do we have to go through things over and over again?

And then they accuse me of arrogance.

Quote:

---

Does High Sampling Frequency Improve Perceptual Time-Axis Resolution of Digital Audio Signal?

http://www.aes.org/e-lib/browse.cfm?elib=7217

---

I'll check what that is.

B

I like you Barish
don´t get me wrong i also like girls

--------------------
I could´nt sleep till I put it down

Personally, I don't like those frequencies that 96khz gives my recordings. The guitar booms horribly in the low end and my vocalsss beam in the top. I can't eq it until it is right either... there is just too much information. Nearly all my best-sounding recordings, I've found out lately, are done in 24bit/44khz.

Edited by Kaw-Liga (02/02/06 06:37 PM)

Quote Barish:

---

Quote:

---

Does High Sampling Frequency Improve Perceptual Time-Axis Resolution of Digital Audio Signal?

http://www.aes.org/e-lib/browse.cfm?elib=7217

---

I'll check what that is.

B

---

I think Mike Storey at dCS published an AES paper on this topic too, and I recall it seemed to make sense when reading it, but I know John Watkinson (amongst others) spent some time trying to discredit it. Can't remember the details now though. There was a lot of debate inthe pages of Studio Sound magazine in the early-mid part of 1998 too if memory serves. I'll try to dig them out.

Hugh

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

M Story: "A Suggested Explanation For (Some Of The) Audible Differences Between High Sample Rate and Conventional Sample Rate Audio Material"

You can find the paper in the Technical Papers section of the dCS website on the link I gave way back up this thread somewhere. It sticks with some simple ideas and avoids going into the maths but it makes some fairly convincing reading as it stands. Obviously though, with this kind of thing, the devil is in the detail and I'd've preferred to have seen his "workings". I took part in some listening tests related to this and another paper on the dCS site and I have to say that whatever the actual maths involved, empirically, it does stack up alongside what was heard.

Enlightenment strikes!

The point of 192kHz, to answer the original question, is to provoke heated and often incomprehensible debate, thereby diverting attention from music.

Any other function is imaginary.

I rest my case, m'Lud.

--------------------
Dynamite with a laser beam...

Quote Hugh Robjohns:

---

I guess you are thinking of chips like the PCM1704 and others in that family that all use sign and magnitude conversion... but with 16x(44.1) oversampling filters. Okay, so not quite as heavy handed as true delta-sigma... but then no one uses true (single bit) delta-sigma in pro audio applications anyway. Even these 'multibit' converters employ all the same oversampling techniques, digital filters, noise shaping and so on. There really is no getting away from it.

---

Yes, I think you are right, there is no escape. All the current implementations, have everything built into the single chip. This was not the case a few years ago. This has been done in order to reduce costs.

Quote:

---

Quote:

---

I agree that digital is better today than it was 20yrs ago, but just not as good as it could be. Good to see you are an optimist.

---

Always the optimist! I fear we may be converging on agreement again here, Dave

hugh

---

I too try to be optomistic, but I am also a perfectionist. I really do believe that we have the capability to have much better performance with the current technology.

Cheers Dave

--------------------
Lighthouse Mastering

I know your post was in humour but just one thing:

Quote Max The Mac:

---

DCS, " strange we thought the same about your position, what with , being someone who doesn't make them......"

---

The Lavry black actually can run at 192Khz. It just doesn't have anything on the front panel (or marketing) to show that it does.

UnderTow

Because I'm not an AES member I couldn't access to AES paper for a discounted price and frankly speaking, I didn't want to pay $20 for an article just to prove a point on an internet forum.

But I have read both the abstract at AES website and 0VU's link. The abstract didn't give me any clue about the details of the test (well, obviously, $20 stuck in there to get it) but dCS guy has a case, but even in his diagrams and explanations, as soon as you go over 60kHz sampling rate, the FIR ringing becomes no issue as the ringing is already pushed out of known audible range. So it is already covered by 88.2 kHz sampling rate, never mind 96. For the pychoacoustic effect of the energy above there is still no consensus. One side say they experimented and proved something the other party fails to replicate.

I'm looking forward to reading Hugh's findings from his archive.

B.

Musical Instrument spectra:
http://www.cco.caltech.edu/~boyk/spectra/spectra.htm

Coding high quality digital audio:
http://www.meridian-audio.com/ara/coding2.pdf

Both links are fairly informative without all the messy math.

BTW Barish, to truly analyze digital audio, one does not just take derivatives and integrals, one must also do a Fourier transform (more precisely a Discret Fourier Transform) to get out of the time domain into the frequency domain to do proper frequency analysis. So get off your high horse and quit beating up on people and trying to show them off with triple derivatives and triple integrals (most of which are never used in DFT's).

Quote Grimm Reaper Sound:

---

BTW Barish, to truly analyze digital audio, one does not just take derivatives and integrals, one must also do a Fourier transform (more precisely a Discret Fourier Transform) to get out of the time domain into the frequency domain to do proper frequency analysis. So get off your high horse and quit beating up on people and trying to show them off with triple derivatives and triple integrals (most of which are never used in DFT's).

---

The Nyquist theorem is reached to and proven through all derivatives and integrals:

http://en.wikipedia.org/wiki/Nyquist-Shannon_sampling_theorem

http://www.digital-recordings.com/publ/pubneq.html

Thank you.

B.

Given that a many instruments (or random sounds) have content above 22khz, and that human hearing doesn't "officially" extend to this range, perhaps we should divide everyone involved in this debate into two camps -

a) You are happy with the fact your digital audio system discards that information because ultimately it's not going to be heard by the end listener, and with a quality recording and playback system, the frequencies they *do* hear are properly reproduced.

b) You'd rather that whatever system you used, it should capture all the available information from the source, even though human hearing limits the audible frequencies, and most consumer formats discard them anyway.

which one are you?

which leads me to my next question....
What's the point of 96khz?

sorry....
Bob

First a word to Hugh -- you are great in not losing your temper and continuing to try to convince. Please keep it up.

Just to give discussion even a bit more fuel, here is a rather nice page on the subject matter:
http://en.wikipedia.org/wiki/Nyquist-Shannon_sampling_theorem

What you might notice is that there is a difference between a continous signal (say a sustained note) and a changing signal (say the attack of a drum hit). The simplified methods work for continuos signals, but you need a bit more advanced mathematics to handle changing signals. Rest assured though that they totally agree and the results are the same as far as the mathematics goes. If you want to do the "connect-the-dots" exercise, please at least use sinc-functions.

In the real world you have to conted with reality, and then things seldom are quite as easy. I still think that ears are made for listening though, so go on and listen. If you happen to hear a difference between sampling at 44.1 and 192 it is NOT because of Nyquist sampling theorem (which proves that there is no hearable difference), but because of the practical imperfections of real-world equipment.

And in the real world, different converters behave differently. My converters might actually sound best at 44.1, yours at 192, who knows?

Gunnar

Quote Grimm Reaper Sound:

---

BTW Barish, to truly analyze digital audio, one does not just take derivatives and integrals, one must also do a Fourier transform (more precisely a Discret Fourier Transform) to get out of the time domain into the frequency domain to do proper frequency analysis. So get off your high horse and quit beating up on people and trying to show them off with triple derivatives and triple integrals (most of which are never used in DFT's).

---

The *triple* derivatives and integrals certainly had me confused (and I'm an engineer in a DSP company, so I should know, or something )... but fundamental Fourier math does involve the common occurance of integrals.

The continuous-Fast Fourier Transform for e.g. is a single integral of a product (multiplied by 1/2pi). Fundamentally though, DFT which you mentioned, is just a sum-of-products.

Barish the point of these articles was not to prove you wrong or other wise but simply present another point of view on the subject.
nothing else

Cheers

Quote Marky:

---

The *triple* derivatives and integrals certainly had me confused (and I'm an engineer in a DSP company, so I should know, or something )... but fundamental Fourier math does involve the common occurance of integrals.

---

My point was to point out that you can derive a high level function a couple of steps down, and then integrate them to reach to the original function healthily, just to support the fact that sound waves can be similarly restored from digital form with no loss by a similar integration. But you are right, I should have made it more clear in my phrasing.

Thanks for pointing out.

B.

Quote DCompton:

---

Barish the point of these articles was not to prove you wrong or other wise but simply present another point of view on the subject.
nothing else

Cheers

---

Agreed.

B.

The maths involved here are way over my head, and I respect everyone that has a far better knowledge of them than I do.
But I have a practical question. Since none of us can listen to 0's and 1's directly, or to voltages, we all need monitors to translate that stream of signals into moving air.
I'm no expert on the thing, but what is the upper limit of a monitor's frequency response? Surely, any sound above that is not present on its output, and can't influence the signal (acoustical or psycho-acoustical or whatever).

Might this just be a case of 'a chain is as strong as its weakest link'? I don't understand the need for DA converters that give you pristine signals up to 96khz (192/2) if your monitors can't translate that into real sound.

Of course, I am making major abstractions of any advantages on the AD side (do microphones record that high?), and of any other technical advantages that may or may not be present.

And by all means, feel free to slaughter me on account of my humble knowledge

--------------------
"Do not fear mistakes. There are none."
Miles Davis

Quote Dunewar:

---

I'm no expert on the thing, but what is the upper limit of a monitor's frequency response?

---

Most speakers are falling off pretty steeply above about 22kHz, but some now have 'supertweeters' which claim to extend the output to 40kHz or thereabouts (Tannoy are a big fan of this approach, for example).

Quote:

---

Surely, any sound above that is not present on its output, and can't influence the signal (acoustical or psycho-acoustical or whatever).

---

Any energy fed into a transducer will cause some effect, even if it is only local heating -- but these effects can have secondary effects on more audible parts of the frequency range. Beats and intermodulations, for example. Also, ultrasonic signals can cause intermodulation and non-linaarity issues with some electronics, again resulting in audible side effects. None of this is ever entirely straightforward because of the real world issues of practical implementations.

Quote:

---

Might this just be a case of 'a chain is as strong as its weakest link'?

---

Absolutely, yes. Which is why I alsways argue that the first need is to sort out the room, and then to buy the best monitors you can afford. Most people choose to ignore the room and buy the cheapest monitors they can, and spend all their money on flash computers and boxes with flashing lights instead! Even though they won't be able to hear the differences.

Quote:

---

I don't understand the need for DA converters that give you pristine signals up to 96khz (192/2) if your monitors can't translate that into real sound.

---

I agree, there is little point if that is the benefit. I don't think it is, though. There are obvious technical benefits from using a higher sample rate for acquisition and post-production, even if the final output is destined for 44.1. These include more accurate EQ curves, more accurate metering, and more accurate dynamic control. There is also the obvious benefit of being able to accommodate less than perfect anti-alias and reconstruction filters without impinging the audible frequency range. The down side is more data storage, higher straing on DSP processess, and the need for far more stringent clocking regimes. ON the whole, though, I think the advantages outweigh the disadvantages. Moving up to 192 or 384, to my mind, introduces more disadvantages with the current level of technology, and offers no further advantages at all.

Quote:

---

Of course, I am making major abstractions of any advantages on the AD side (do microphones record that high?), and of any other technical advantages that may or may not be present.

---

Very few mics have a response that remains anything like flat above 16kHz. Most are already begining to fall gy then. The better small diaphgram mics will be flat to 20kHz or thereabouts, and then fall fairly quickly again. A very few have been specially designed to extend flatish to 30 or 40kHz, but there are significant side effects in pushing transducers to do that -- mainly involving higher broadband noise and distortion artefacts.

hugh

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

Quote:

---

Any energy fed into a transducer will cause some effect, even if it is only local heating -- but these effects can have secondary effects on more audible parts of the frequency range. Beats and intermodulations, for example. Also, ultrasonic signals can cause intermodulation and non-linaarity issues with some electronics, again resulting in audible side effects. None of this is ever entirely straightforward because of the real world issues of practical implementations.

---

Aha, there's the flaw in my reasoning. Thanx Hugh, now it makes sense. I thought that effects like beats and intermodulations took place in the airwaves of sound, but now I understand that it happens in the electronics before the tweeters start flapping about...

But the "weakest link" approach is (according to me) the best approach, and is an approach that makes perfectly clear why sample rates like 192 khz and beyond are useless for people like me. My room is far from perfect, my monitors are cheap, and my recording methods are far from ideal. So moving my converters up to 192khz would be like putting a porsche engine into a volkswagen... So for me this is a theoretical problem, i've got bigger problems to sort out in my recording chain before that. And I bet it is for a lot of people.

Of course, when you are talking about high-class recording facilities, with excellent sounding rooms and 10K+ microphones, then i'll shut up, because I don't know squad about that. That's like a hobbiest car-fixer telling a formula 1 engineer what to do.........

--------------------
"Do not fear mistakes. There are none."
Miles Davis

Quote Barish:

---

I'm looking forward to reading Hugh's findings from his archive.

---

Had a quick rummage through the archives. Studio Sound 1998. Feb edition, John Watkinson argued against the need for sampling at 96kHz and above on the basis that theory says you don't need to. Any apparent sonic differences must be because the 48kHz designs are imperfect.

April Edition, Mike Story write about the perceived benefits of 96kHz in relation to greater spatial accuracy because of improved filter responses (shorter impulses).

June edition, John Watkinson defends his corner in typical fashion. He suggests that the dCS experiments were flawed because they contain no proof that only the sample rate changed between tests -- suggesting that other (presumably filter-related) aspects may have changed too.

He goes on to argue that the dCS arguments rely on the assumption of linear phase filters, but that any practical decimation filter implementation involving noise shaping will inherently involve recursion which is not necessarily phase linear.

He suggests that the resulting group delays from non phase linear filters may indeed be audible, and that a doubling of sample rate would halve this group delay, which would certainly be an audible change. He therefore claims that in switching from 48 to 96kHz, the test listeners were simply detecting the improved phase linearity in the audio band. This would certainly have the effect of improving the spatial resolution noted in the dCS experiments, but not for the reasons Mike Story claimed.

He went on to ask for measurements of the phase linearity of the dCS converters at 48 and 96kHZ to see if it changed. As far as I can see, no response was ever provided to that challenge.

In the August edition another writer claimed that perhaps the audible benefit of 96kHz (and higher) was because of the inherently much shorter filter impulse responses.

He claimed that the ear can react to the ultrasonic pre-ringing of linear phase filters. Pre-ringing is something we never hear from analogue filters, or in normal real life, of course.

Higher sampling rates reduced the period of pre-ringing, and it was argued stimulates the ear less. He also suggests that filter designs that don't have pre-ringing characteristics (i.e more like traditional analogue filters) would sound better -- and this is something dCS also claims.

It's all a minefield, and it's all been debated at length before by people with far better brains than any of us!

Hugh

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

Yes, this is very much like arguing over which religion is better. No one wins and the life goes on.

Thanks for summing this up with a fashionly manner, Hugh. I hope we all got something out of this. I for one realized after all this debate that I don't need 192kHz converters, but I think I could use a bit of those anger management courses.

Regards,

B.