Hi,
I have just bought a Presonus FP10 as my first home studio recording equipment (up until now I have just been using a laptop sound card).

Should I record at 96khz now that I have the capability? What benifit does this bring to me?

My recordings sound ok at 44khz so im assuming the actual quality of the sound wont change?

Thanks,
AM.

Debatable. If you can, do so, but you'll find that it chews up resources far faster than when working with 44.1. Bitrate (24bit vs 16bit) would be my preference in a pinch, as it gives more flexibility as well as (Debatably) better sound quality. Actually there's loads of articles that'll be far more useful than people's opinions. This: http://www.soundonsound.com/sos/sep07/articles/digitalmyths.htm is a good starting point, carry on from there

--------------------
Stuff what I done

Stick to 44.1kHz - you'll be wasting your time & hard disc space running 97kHz

--------------------
Mac Pro, Logic Pro, lots of software and 17 hard drives!

I dont know about that

I can hear a huge difference between 96 and 97khz

--------------------
Matt
www.krcollective.org

i have worked with "big name producers" that record at 44.1

though this is all pretty much rock guitar bandy stuff.

--------------------
Live Sound and Studio Engineer

44, unless you have a really specific reason for 96, and your room acoustics and recording chain are top rate, and you have the available resources to "waste" (double the CPU load, double the disk space, etc).

what about 48khz?

--------------------
www.myspace.com/shambolic-charm

You always want to maintain the highest sample rate as long as possible. However just because you have the capability to record at 96 does not mean that you should. There are a lot of factors to consider and I won't go into detail as I am no expert but space is one of them, power is another, internal components of your AD is another. Even if you have the capability that doesn't mean that your gear operates best at 96. It really is quite confusing and there are a million previous threads debating such questions, one or two of them mine.

Bottom line after all the reading and research I feel I know less than when I started, or else I know more which only makes it more confusing thus it feels like I know less

I can hear a difference between 96 and 48, however what that difference is, is debatable and once I listen to 48 for a while it sounds just fine.

At the end of all of that I would say unless you are recording Classical music I would stick to 48/24, or 44.1/24 and forget about trying to understand any of it.

48k is for film and broadcast - I usually export "up" to 48k from 44.1 when handing in broadcast stuff.

--------------------
Stuff what I done

Some say certain plugins sound better at higher sample rates. And I have seen convincing claims that lower end converters perform better at higher sample rates as well, this seems counterintuitive I know. But space and computing power will most likely be the deciding factor for most home studios. Dedicated DSP cards like those in the UAD line and HD Pro Tools systems up sample to as high as 192khz for their internal processing I believe, which negates the need to record at higher sample rates.

I could be wrong, this is just what I have gathered from my untrained research!

Quote Jocoserious:

---

You always want to maintain the highest sample rate as long as possible.

---

Hmm... bit of a blanket statement that.

Why?

Quote passerrby3141:

---

Some say certain plugins sound better at higher sample rates.

---

Yes. There are technical reasons for this, but it can be true that, especially for aliasing reasons, some virtual instruments benefit from going up that high. But this is a completely different issue to audio recording.

Quote passerrby3141:

---

Dedicated DSP cards like those in the UAD line and HD Pro Tools systems up sample to as high as 192khz for their internal processing I believe, which negates the need to record at higher sample rates.

---

It's nothing to do with DSP cards per se, it's the DSP algorithms used. Some algorithms and processes benefit from upsampling, and some native plugins do this as well where it makes sense. However, upsampling does tend to make the plugin much heavier on the CPU, so plugin designers try to avoid high-cost algorithms unless it makes sense, or is necessary to use them.

Basically.....

Human hearing goes up to about 20 kHz.... but you need to double that to eliminate something called aliasing, which is something like a "rounding error" that causes mucho distortion and strange frequencies.

However, I regularly argue that 80 kHz or higher is necessary to get a good representation of the frequencies around 20 kHz.

If you imagine a sine wave at 20 kHz... and then you sample that twice per period (which is a 40 kHz sampling rate), you don't happen to get a great representation of the wave... For example, if you're unlucky and you happen to sample the sine wave at the two zeros each period, you'll get nothing.

in fact, you're guaranteed to get a "lower volume" reproduction of the wave. Thus, I suggest at least 4 samples per period of the highest frequency you're interested in reproducing accurately.

192 kHz means you're getting 8 samples per period of a 24 kHz wave... that ought to reproduce those high frequencies much better.

Anyhow - that'd be the difference! Much music is "low end heavy" anyway.

Like others have said, I'd stick with 44.1 unless you're doing something very delicate in a very nice room. I used to do everything at 96. It was a pain to mix large sessions because my computer really couldn't handle all of that data. Now my classical recordings are done at 96/24 while pretty much everything else is at 44.1/24. A friend of mine uses 48 to get away from a few of the side-effects of 44.1. He even uses 48 for classical work and there's no problem in the sound. Unless you can do 192k or... SACD rates (2822k etc...) I'd stick with lower rates. 192 and sacd are definitly far superior, but is it worth the strain on your computer?

--------------------
Winding Road Studios
Speaking Audio Blog

Ok, i made a rough approximation... because I'm hitting the same problems at 48 versus 96...

Basically, I roughly computed that if you wanted 48 to sound more like 96, if you put an EQ that raised 20 kHz by 1 dB, 10kHz by 0 db, and made a smooth line between the two (eh, I have a hand drawn eq)... it'd be pretty close.

So, then I tested it with some songs in my collection... basically, if you've already boosted the high end on cymbals and similar, it doesn't really matter too much.. just puts a small punch on the cymbals.

Whelp, so. I find this stuff fun to think about.

Well I think 96k sounds better...the difference gets smaller as converters get better but I think it's clearly a sonic advantage IF you're actually recording stuff and not just using loads of samples.

However it is really inconvenient because it chews resources, makes backups take ages, and introduces a boring extra conversion at the end of the process.

My advice would be: Anything that's really important, try it at 96 because it's certainly a bonus. Do your demos etc at 44.1 because it's convenient and sounds very good.

J

--------------------
www.jackruston.com

This has been covered (extensively) before, but just to add my 2 penneth...

recording at 96k doubles your PC overhead - for an improvement that is hardly perceivable in real-life use. (don't quote your maths at me)

It ain't worth it.

Just remember to record 24 bit, 'cause that makes one hell of a difference.

--------------------
www.thediplomatz.com

Check out Paul Lehrmans artice in Mix Magazine (available on the web) - The Emperors New Sampling Rate.

The title says it all really. Even Ethan Winer, of Real Traps who I respect greatly when it comes to sound (his ears are much more golden than mine) suggests that differences heard (and I'm not saying that people who hear a difference are lying) can be due to minute changes in head position between changing the material, affecting the phase of the sound. Mr Lehrmans article refers to double blind tests where there was no statistical difference and apparently 'educated' ears were used. Of course, he has been flamed within an inch of his life in some places but many people, in particular manufacturers and those who have been saying for years that they can hear a difference have a vested interest in 88.2 and up.

Now, I'm not one to go against the word of many much more experienced and sucessful engineers on this but in my (admittedly very limited and unscientifc) tests I can't hear a difference in my little studio. And as I can run a few more plug ins (not as many as you would think, I run Guitar Rig in HQ mode which does internal upsampling and use UAD plugs which also upsample where required) and many more tracks I keep using 44.1 or 48.

Here's the link. Interesting article, thanks for posting that:

http://mixonline.com/recording/mixing/audio_emperors_new_sampling/

Quote JohnnyT:

---

Mr Lehrmans article refers to double blind tests where there was no statistical difference and apparently 'educated' ears were used.

---

Double blind tests, in a big enough sample of people, are about the only thing that actually proves anything. All the rest can be easily explained by the placebo effect.

I believe any hesitation in shouting about the results is due to the size of the sample, if I recall. I didn't think to post the link - my, how 20th Century!

I correct that, I just re-read the article. The sample size was 'hundreds' of people. Perhaps not definitive but pretty persuasive, certainly enough to stop and think about it and hope for more studies. In the meantime, maybe stick to 44.1 :-)?

Once again, sorry. Hundreds of trials.

Quote desmond:

---

Here's the link. Interesting article, thanks for posting that:

http://mixonline.com/recording/mixing/audio_emperors_new_sampling/

---

Interesting article. I enjoyed reading that

Quote JohnnyT:

---

Once again, sorry. Hundreds of trials.

---

I can not find anything wrong with the study at all. I suppose the only thing you could do to improve it would be having some kind of frame to lock the listener head in the exact same position every time. Something like they had in clockwork orange maybe? lol

I know I can not wear any differences in 96KHz audio as opposed to 44.1KHz, but then again my hearing is not great after a few years playing gigs so Id on't really think this proves much.

It reminds of a test I saw in some audiophile forum once where they did a blind test between themselves in order to prove they could hear a difference between 16/44.1 Khz and 320 Kbps MP3. They only guessed correctly about 40% of the time, and conclude that deaf people could probably do as good a job as they did or better, since anyone has 50% probability of guessing right by picking a random answer.

Anyway, I blame all this on gearslutz.com. lol

Quote Interweaved:

---

If you imagine a sine wave at 20 kHz... and then you sample that twice per period (which is a 40 kHz sampling rate), you don't happen to get a great representation of the wave... For example, if you're unlucky and you happen to sample the sine wave at the two zeros each period, you'll get nothing.

in fact, you're guaranteed to get a "lower volume" reproduction of the wave. Thus, I suggest at least 4 samples per period of the highest frequency you're interested in reproducing accurately.

---

This is why you need a sampling rate greater than instead of equal to twice the highest frequency you are sampling.

It's only when the frequency of the sampled signal is exactly half the sampling frequency that this problem occurs, as there are an infinite number of sine waves of frequency fs/2 with different phases and amplitudes that could all give rise to the same sampled data.

If however the sampling rate is even sightly greater than twice the frequency of the sampled sine wave then there is only a single sine wave with a frequency less than fs/2 that will fit through all the sample points, and thats your original signal.

Mark has already tackled this, but it is a common fallacy and deserves the emphasis:

Quote Interweaved:

---

Basically.....

Human hearing goes up to about 20 kHz.... but you need to double that to eliminate something called aliasing, which is something like a "rounding error" that causes mucho distortion and strange frequencies.

---

The Shannon/Nyquist sampling theorem states that you need to sample at a rate which is more than twice the bandwidth of the required signal. Sampling is a modulation process, and the 'more than twice' element is to ensure that the source and it's modulated sidebands remain separate and separable.

Interweaved Quote:

---

However, I regularly argue that 80 kHz or higher is necessary to get a good representation of the frequencies around 20 kHz.

---

It is one way to do it, but it is not 'necessary.' Given a wanted bandwidth of 20kHz, how much higher than 40kHz the sample rate needs to be is determined by the efficacy of the anti-alias and reconstruction filtering.

To be honest, 44.1 was too low when the standard was set -- 60kHz would have been a far better choice for a host of reasons. Equally, 96kHz is arguably wastefully high. However, technology hasn't stood still for the last 25 years, and most decent converters can now provide very satisfactory performance sampling at 44.1kHz.

Twenty plus years ago, you needed the very best of converters to achieve adequate performance sampling at 44.1. Thankfully, that is (arguably) no longer the case, and there are some extremely competent mid-price converters out there now.

Interweaved Quote:

---

If you imagine a sine wave at 20 kHz... and then you sample that twice per period (which is a 40 kHz sampling rate), you don't happen to get a great representation of the wave... For example, if you're unlucky and you happen to sample the sine wave at the two zeros each period, you'll get nothing.

---

This is true, but hardly relevant since the theorem requires the sample rate to be more than twice the bandwidth of interest. In the case you cite, the lower modulation sideband of the source signal will lie precisely on top of it, and what you hear will be the summation of the two, which -- if they are perfectly aligned because the sample rate is exactly twice the source frequency -- could be nothing at all, as you say!

Interweaved Quote:

---

Thus, I suggest at least 4 samples per period of the highest frequency you're interested in reproducing accurately.

---

Yes, this will obviously work, but it is a bull in a china shop approach, which is wasteful of the information capacity provided by sampling at that high a rate.

The theorem is 100% accurate in its claims. The only issue is in the practicalities of implementing it -- specifically in designing the filters to do what they are required to do without damaging the wanted signal in the process.

There was a time when working at 96kHz (or higher) provided a clear sonic advantage to working at 44.1kHz, simply because the filter artefacts where removed above the human hearing range. However, as filter design and clever techniques like delta-sigma converters have evolved, the difference has become far less pronounced -- in some cases to the extent that the difference is barely audible at all.

Countering that, computer processing power and data storage capacity has increased (and continues to increase) enormously, and while working at 96kHz involved serious overheads for most systems a few years ago, that is less the case now.

Personally, I generally work at 24/96 unless specifically requested not to, because all of my equipment can handle that rate doing the kind of work I do without a significant performance reduction. But on the odd occasions that I am required to work at 44.1 or 48kHz, I don't hang my head in shame at the poor quality -- the difference is often undetectable.

But as a working principle, it makes sense to record source material at the highest possible resolution and quality, because the quality can only suffer after that.

Quote Interweaved:

---

Basically, I roughly computed that if you wanted 48 to sound more like 96, if you put an EQ that raised 20 kHz by 1 dB, 10kHz by 0 db, and made a smooth line between the two (eh, I have a hand drawn eq)... it'd be pretty close.

---

I think what you are suggesting here is compensation for the transition area of the 'brick wall filter(s)' since some (not all) exhibit a mild roll off as they approach the turnover frequency. But of course, amplitude is not the only aspect of a filter to consider -- phase is as important (if not more so) and by introduceing your HF lift, you will also introduce phase shifts that will change the character of the sound in some (indeterminate) way.

Hugh

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

Edited by Hugh Robjohns (05/07/08 01:37 PM)

Not much to add to the above, save to say that if you have (or aspire to) a Focusrite Liquid Mix, you can run 32 instances at 44/48k as opposed to 8 instances at 88/96k.

That's how much extra DSP power the higher sampling rate requires.

192k is definitely for the birds (or anyone else with genuinely ultrasonic hearing).

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


Edited by Steve Hill (05/07/08 01:36 PM)

Quote Interweaved:

---

Basically.....

Human hearing goes up to about 20 kHz.... but you need to double that to eliminate something called aliasing, which is something like a "rounding error" that causes mucho distortion and strange frequencies.

However, I regularly argue that 80 kHz or higher is necessary to get a good representation of the frequencies around 20 kHz.

If you imagine a sine wave at 20 kHz... and then you sample that twice per period (which is a 40 kHz sampling rate), you don't happen to get a great representation of the wave... For example, if you're unlucky and you happen to sample the sine wave at the two zeros each period, you'll get nothing.

in fact, you're guaranteed to get a "lower volume" reproduction of the wave. Thus, I suggest at least 4 samples per period of the highest frequency you're interested in reproducing accurately.

192 kHz means you're getting 8 samples per period of a 24 kHz wave... that ought to reproduce those high frequencies much better.

Anyhow - that'd be the difference! Much music is "low end heavy" anyway.

---

That's a nice thought provoking explanation.

I keep thinking of the maths and mechanics of how the highs are dealt with (and lows) because this is where the audio begins to suffer in standard home recordings.

I believe software and components hold the key to how the sound is dealt with.
I mean it's easy for a soundcard to state audio clarity,but add a dodgy plugin(which you don't know about)and it will mess your audio signal up.
It's a minefield but I love digital.

Quote Music Manic:

---

but add a dodgy plugin(which you don't know about)and it will mess your audio signal up.

---

This is just as relevant to hardware and analog processing too. At the end of the day, you are putting your signal into a "black box", which is closed and does some undefined process on that signal, the design process and choices of which you weren't part of, and then you get the output.

Without knowledge of what's happening inside the box, you have no idea what's going on, apart from what you can deduce from the output signal.

It doesn't matter whether the box is a software plugin designed by a 12-year old with a poor understanding of audio signal processing, a plugin written by some DSP maths genius, a hardware box an the entry level with cheap poorly designed electronics or a boutique piece of expensive analog gear.

At the end of the day, you have to understand and be familiar with your gear, educate yourself and make informed choices about how to process, why, and what to use, and this will give you better results than just inserting some random magical black box on a signal and hoping it's going to make your song sound better, somehow...

I found:
www.lavryengineering.com/documents/Sampling_Theory.pdf
Very interesting.

I have been doing 88.2 projects for a couple of years now, and I have to say that I think the quality of the sound I am getting is amazing. It doubles the system resources required however, and that can put a kink in things when I am working with large projects (Core2Duo + 3 UAD-1 cards).

It has been so long since I have done a 44.1 project that it is hard to remember the difference, HOWEVER, I have been recently taking a trip down memory lane in my car going through a bunch of old sessions that I have recorded in my studio and not listened to in a long time. Bands that have long since broken up and moved on, etc.

Just from listening to the CDs, I could not tell you which ones were recorded at 44.1 and which ones were recorded at 88.2, and trust me, I tried as hard as I could to see if I could tell the difference!

Maybe it is my old, overused ears...

--------------------
James
"Never interrupt your enemy when he is making a mistake" ~Napoleon Bonaparte~

Quote Hugh Robjohns:

---

The Shannon/Nyquist sampling theorem states that you need to sample at a rate which is more than twice the bandwidth of the required signal. Sampling is a modulation process, and the 'more than twice' element is to ensure that the source and it's modulated sidebands remain separate and separable.

---

Let me repeat my argument again... Assume the Shannon/Nyquist theorem only eliminate aliasing, so your sidebands are eliminated. But it does not guarantee anything about your sample being a great reproduction.

Now, suppose we're sampling at 40 kHz, and there's a 20 kHz sine wave, like thus:



Now... say your sampler samples at the blue points. Your final sample will just be zeros. Say i samples at the yellow points. Your final sample will be something like a 20 kHz sine wave, but reduced in volume for the original. So, on average, 20 kHz waves will be somewhat lower in volume when sampled in 40 kHz than your initial sample.

However, if you sampled at 80 or 160 kHz... you'd get a much closer representation of the original wave.

Quote Interweaved:

---

Now, suppose we're sampling at 40 kHz, and there's a 20 kHz sine wave, like thus:

---

But we are not, we are sampling at 44.1KHz, the minimum defined "CD quality" standard for the full range of the human hearing.

This is "more than" 2x the sample rate, rather than "exactly" 2x the sample rate.

Quote Interweaved:

---

However, if you sampled at 80 or 160 kHz... you'd get a much closer representation of the original wave.

---

Have you actually been *reading* this thread..?

Sampling at that frequency is redundant and not necessary. 44 is fine. Remember, sampling at higher sample rates does not give a "more accurate" reconstruction of the signal than 44. More dots does not equal more accuracy.

Okay now I may be way off base here and probably will not explain it well enough to be understood but here goes.

Even if someone cannot hear the difference between say 44.1 and 96 isn't there still some benefit to be had from using a higher sample rate??

There a million variables in all of this and this may not be of significance for anyone to actually hear but what about the degradation caused by plug-ins, calculations etc.?? If you are using 96khz shouldn't any degradation caused by such calculations be spread over a larger frequency range?? Thus when you convert down it should actually eliminate more of that degradation?? This makes sense to me I just don't know if I am explaining it correctly.

Not really, it's the bit depth that carries the accuracy of the maths involved, not the sample rate.

The sample rate determines which samples are used to reconstruct the wave - whether the wave is made from 44100 samples per second, or 96000 samples per second, the resultant reconstructed waveform is exactly the same, as long as your wanted audio is under the Nyquist limits for your sample rate...

For me, there is really only one tangible, fairly obvious benefit, in that some virtual instruments have a better sound at 96KHz, due to the aliasing calculations being done (some synths have an internal upsampling mode as well).

Your milage may vary, etc...

Sorry that doesn't work for me....Or rather that does not do anything to change my opinion on my previous post.

Yes I understand bit rate and all that, that really wasn't what I was getting at. Maybe my explanation is too convoluted but it still makes sense to me.

If you are using 44.1 does that not mean that the range is from 0-44.1? So when you apply something like a plug in it is contained within 0-44.1. When you convert it down it is now gone from 0-44.1 to say 0-20 thus losing 24.1 of range. Would a plug in not cover the whole spectrum? Thus any degradation it causes would be over that whole spectrum. Therefore at 44.1 you lose 24.1 yet if you are using 96 and you convert down you would be losing 76khz of range?? Wouldn't this allow the plug in degradation to be spread over a larger spectrum thus eliminating more of it when converting down???

This seems really simple to me yet oh so confusing.

Now what about calculations? I understand that the bit rate is the determining factor in the calulation but the degradation it causes would still be included in the range right? So wouldn't a higher sample rate allow more degradation to be spread out and thus when converted down you lose more of it??

Sorry if my terminology is un professor like, hopefully someone can shed some light on this not often talked about issue....

Quote Jocoserious:

---

If you are using 44.1 does that not mean that the range is from 0-44.1? So when you apply something like a plug in it is contained within 0-44.1. When you convert it down it is now gone from 0-44.1 to say 0-20 thus losing 24.1 of range. Would a plug in not cover the whole spectrum? Thus any degradation it causes would be over that whole spectrum. Therefore at 44.1 you lose 24.1 yet if you are using 96 and you convert down you would be losing 76khz of range?? Wouldn't this allow the plug in degradation to be spread over a larger spectrum thus eliminating more of it when converting down???

---

Depends on the effect. Imagine a delay/echo effect... it really doesn't care about your sampling frequency at all. Now, imagine a pitch shifter... ever notice how they sound better pitch shifting up instead of down? Now... flanger's do "pitch shifting" to some degree, so it likely has an effect there. Chorus's do some... but not a whole lot... so, maybe not so much there.

Quote Jocoserious:

---

If you are using 44.1 does that not mean that the range is from 0-44.1?

---

What "range"? 44.1 refers to the fact that 44100 sample are taken per second on a single mono stream. The actual amplitude values depend on the bit depth.

There is no 0-44.1 "range" of which you speak.

Quote Jocoserious:

---

So when you apply something like a plug in it is contained within 0-44.1. When you convert it down it is now gone from 0-44.1 to say 0-20 thus losing 24.1 of range. Would a plug in not cover the whole spectrum? Thus any degradation it causes would be over that whole spectrum. Therefore at 44.1 you lose 24.1 yet if you are using 96 and you convert down you would be losing 76khz of range?? Wouldn't this allow the plug in degradation to be spread over a larger spectrum thus eliminating more of it when converting down???

This seems really simple to me yet oh so confusing.

---

Unfortunately it seems that you really don't seem to have a grasp of how sampling and digital systems work, so any intuitive feelings you have, or conclusions you reach, are based on a flawed understanding are are therefore not going to be valid.

A bit like asking me to advise on, say, choreography or flower arranging.

I don't really like you, or rather your posts, not very helpful.

Like I said before I knew I was being convoluted. No need to call me an idiot, I can do that myself.

Yes I do understand pal, I was trying to explain something in more of a Lehmens term however I think I may have been too confusing. Essentially we are saying the same thing, either way a higher sample rate is going to provide a wider frequency range because having more samples per second allows you to capture higher frequencies because those travel faster, or something like that. I am not referring to dynamic range I am referring to frequency range. On a spectral frequency it shows from 0 up to half of the sample rate, at least that is what shows in the program I am using. Therefor 44.1 will not contain as much info as 96. Since we can't hear above 20khz anything we apply over a 96khz platform will move more into the inaudible region...

How to explain this properly is better left to someone much more qualified. Someone like Katz perhaps. "Thus there is considerable advantage of doing all our processing at higher sample rates, which moves the distortion products into the inaudible spectrum above 20khz".

Not sure who he is or how knowledgeable he really is, I am sure you are more qualified then he is. So sorry for not being able to simply communicate this very easy to understand topic. I mean there are only a million threads, articles and books on sample rate simply because it is so easy to understand and everyone enjoys debating known facts for the simple reason of debating.

All I know is that 96khz will contain more info than 44.1khz will regardless of whether we can hear it or not and regardless of exactly how and why it does. To me saying 0-44.1 or 0-96 is the same as saying samples per second sine wav this and that yada yada yada. To argue over semantics is a waste of time.

Cheers!

Look I don't really want to get this into another long winded sample rate debate where people get their panties in a bunch. I was merely being sarcastic and my last post should be taken in jest, because I felt you were a little rude Desmond.

Please feel free to lock this baby down or whatever needs to be done so this doesn't spin out of control way away from the OP's original question.

Quote Jocoserious:

---

because I felt you were a little rude Desmond.

---

If you took it that way, I apologise, but it really wasn't meant to insult you.

I felt that, from how you were describing things, that you really didn't have a grasp on the mechanics at work here. Perhaps you weren't expressing yourself very well, or perhaps I interpreted wrongly, but that's how it came across to me - if you took that comment personally, once again, it wasn't intended to be insulting at all.

After all, not understanding something is not equivalent to being an idiot, and I certainly didn't use that term or anything close (nor would I.)

Quote Jocoserious:

---

Essentially we are saying the same thing, either way a higher sample rate is going to provide a wider frequency range because having more samples per second allows you to capture higher frequencies because those travel faster, or something like that. I am not referring to dynamic range I am referring to frequency range. On a spectral frequency it shows from 0 up to half of the sample rate, at least that is what shows in the program I am using. Therefor 44.1 will not contain as much info as 96. Since we can't hear above 20khz anything we apply over a 96khz platform will move more into the inaudible region...

---

Yeah. How much actual energy is in those frequencies though is very small - I don't know if tests have been done to determine how much energy is up in those frequencies, and certainly the jury is out on whether we can perceive them (although we're fairly sure we can't "hear" them, at least as we understand conventional hearing.)

Quote Jocoserious:

---

Not sure who he is or how knowledgeable he really is, I am sure you are more qualified then he is.

---

I can assure you I am by far less qualified than Bob Katz, don't worry...

Quote Interweaved:

---

Now, suppose we're sampling at 40 kHz, and there's a 20 kHz sine wave, like thus:



Now... say your sampler samples at the blue points. Your final sample will just be zeros. Say i samples at the yellow points. Your final sample will be something like a 20 kHz sine wave, but reduced in volume for the original. So, on average, 20 kHz waves will be somewhat lower in volume when sampled in 40 kHz than your initial sample.

However, if you sampled at 80 or 160 kHz... you'd get a much closer representation of the original wave.

---

lol

This must be a joke.

Quote Jocoserious:

---

If you are using 44.1 does that not mean that the range is from 0-44.1?

---

No. It means that provided your signal was properly band width limited you will be able to perfectly reconstruct any sine waves below the nyquist frequency. That's all it means. You don't need any more points to reconstruct the sine wave, just like you don't need any more than two points to draw a straight line.

Quote:

---

No. It means that provided your signal was properly band width limited you will be able to perfectly reconstruct any sine waves below the nyquist frequency. That's all it means. You don't need any more points to reconstruct the sine wave, just like you don't need any more than two points to draw a straight line.

---

Yeah taking it out of context sure...The only thing I was trying to get at, a point which not many people seem to bring up, a point which most people seem to ignore and look past is the point of using a higher sample rate to push most of the distortion into an inaudible region.

I'm not trying to get into the more points argument, the more accurate argument, the whatever else argument. The above point about degradation and distortion was the only one I was trying to communicate and open for discussion.

If you would like I am sure you can go back to my posts and pull a lot of other one liners out that on their own are totally useless. Put them within the whole story and still you get a hard to follow muddled kind of explanation but it was all meant to try and explain ONE point, nothing else.

I am not saying that what you say is incorrect or anything of the sort, I am only saying that what you are correcting me on was not what I was talking about.

If someone has something to say regarding using higher sample rates to push distortion into inaudible ranges then I would love to hear it and will take any kind of tongue lashing and name calling and corrections you want to throw at me cause I am most definitely no expert, but please don't try correcting me on things I was not even talking about. That only shows you are either not reading the thread and my posts, skimming until you find something you can interject with your own wit and wisdom to make yourself look like the scholar, or else I make absolutely zero sense to anyone trying to follow which brings about the simpleton explanations of rudimentary nature.

Sorry for the long wind of hard to follow banter. I shall revert to the sidelines, I promise to never, ever get involved in another sample rate discussion.

haha although they do always spur on a lot of discussion lol

Quote Interweaved:

---

Let me repeat my argument again...

---

With the greatest respect, there is no point in repeating a totally flawed argument that doesn't adhere to the fundamental principle of the underlying maths as laid down in Nyquist's and/or Shannon's sampling theorum.

What you are claiming, while accurate, is irrelevant because sampling doesn't work if you use a sample rate which is exactly double the frequency of the signal of interest. The sample rate HAS to be MORE than twice the bandwidth of the source material. If it isn't, then it won't work, as you have shown. That's what the theorum explains, and that's why everything samples at a rate of more than twice the bandwidth being sampled. End of story.

Quote:

---

Assume the Shannon/Nyquist theorem only eliminate aliasing, so your sidebands are eliminated. But it does not guarantee anything about your sample being a great reproduction.

---

Oh dear... It appears that you don't understand the Nyqusit theorem. Can I respectfully suggest you read this article which will go some way to explaining the process. Or the Lavry article cited above.

Quote:

---

Now, suppose we're sampling at 40 kHz, and there's a 20 kHz sine wave, like thus:

---

There would be no point as this example contravenes the Nyquist theorum, and it won't work. So no system will try to operate this way. It's about as useful as saying, see how this petrol engine won't run on diesel fuel, and therefore the whole internal combustion theory is crap!

Quote:

---

However, if you sampled at 80 or 160 kHz... you'd get a much closer representation of the original wave.

---

It wouldn't just be 'closer', it would be totally accurate -- and the reason is that by sampling at a rate of more than twice the bandwidth of the source signal, you are now adhereing to the Nyquist rules and, surprise, surprise, it all works perfectly! Magic! But it is very wasteful and inefficient if you aren't particularly interested in the upper portion of the audio bandwidth.


Hugh

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

Quote Jocoserious:

---

If you are using 44.1 does that not mean that the range is from 0-44.1? So when you apply something like a plug in it is contained within 0-44.1. When you convert it down it is now gone from 0-44.1 to say 0-20 thus losing 24.1 of range.

---

No. sampling at 44.1kHz allows the system to know preciselty what is going on up to about 20kHz (or whereveer the anti-alias filter rolls off). That 20kHz audio is 'reflected' above and below the 44.1kHz sample rate as 'images.' Again, check out that digital myths article in February's SOS to see what I mean.

The processing is only applied to the region defined by the sampling, between 20Hz and 20kHz (or whatever the filters allow).

There are some advantages to processing at higher rates, for sure. HF EQ can be made to work in a way which replicates analogue EQs much better. Dynamic processing is more accurate, as is metering...

But in plug-ins and systems that use oversampling internally while ostensibly working at 44.1kHz, the original audio bandwidth remains unchanged -- it is still
20Hz to 20kHz.

In systems that operate from the outset at 96kHz (or whatever), the original audio bandwidth will be 20Hz to 40kHz (or thereabouts) -- although whether there is actually any useful content above 20kHz depends on the quality and design of the microphones and other sources.

Hugh

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

Quote Jocoserious:

---

Like I said before I knew I was being convoluted. No need to call me an idiot, I can do that myself.

Yes I do understand pal

---

Calm down, please. He hasn't called you an idiot, but it does appear that you don't really understand it -- and there is no shame in that because this is a complicated topic that very, very few really understand completely. And it's not helped by the enormous about of misinformation that so many people pass around as solid facts!

Quote:

---

Essentially we are saying the same thing, either way a higher sample rate is going to provide a wider frequency range because having more samples per second allows you to capture higher frequencies because those travel faster, or something like that.

---

See... you're not really sure are you. You are right, though, that a higher sampling rate allows a greater audio bandwidth to be described. So while a 44.1kHz sample rate caters for an audio bandwidth from, say, 20Hz to 20kHz, a 96kHz sample rate would cater for an audio bandwidth from 20Hz to 40kHz -- in broad terms.

Quote:

---

Since we can't hear above 20khz anything we apply over a 96khz platform will move more into the inaudible region...

---

Sampling at a 96kHz rate doesn't move anything. It just allows whatever is in the region between 20kHz and 40kHz to be captured.

Quote:

---

"Thus there is considerable advantage of doing all our processing at higher sample rates, which moves the distortion products into the inaudible spectrum above 20khz".

---

The first part of the sentence is correct. The second is confusing and misleading, and really needs more context to justify the claim. Bob Katz knows his stuff, but I fear this quote isn't doing him justice.

Quote:

---

So sorry for not being able to simply communicate this very easy to understand topic. I mean there are only a million threads, articles and books on sample rate simply because it is so easy to understand and everyone enjoys debating known facts for the simple reason of debating.

---

Don't apologise -- it is a hard topic to understand and discuss -- and of those millions of threads and articles, 90% are wrong or grossly misleading, sadly.

Quote:

---

All I know is that 96khz will contain more info than 44.1khz will regardless of whether we can hear it or not and regardless of exactly how and why it does.

---

Correct.

Quote:

---

To me saying 0-44.1 or 0-96 is the same as saying samples per second sine wav this and that yada yada yada. To argue over semantics is a waste of time.

---

No, this is not correct. These are not the same things, and the differences are important to a correct and reliable understanding of the subject. We are talking about a very clearly defined, but immensely complex process. Using confused terminology and inappropriate analogies doesn't help anyone.

Hugh

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

Quote desmond:

---

Yeah. How much actual energy is in those frequencies though is very small - I don't know if tests have been done to determine how much energy is up in those frequencies, and certainly the jury is out on whether we can perceive them (although we're fairly sure we can't "hear" them, at least as we understand conventional hearing.)

---

Yes, tests have been done. The amount of energy various, as you would probably expect, with the source. Things like trumpets and cymbals generate a huge amount of ultrasonic audio energy. So do things like switched mode (universal) power supplies and the inverters that drive the backlights in LCD screens. And if the microphones have a wide enough bandwidth (and many do, these days) all of this ultrasonic garbage that is inaudible to us naturally is being captured in pristine quality if you operate at 96kHz!

Worse... if there is some non-linearity in some of your digital processing the resulting intermodulation products may well render these ultrasonic signals to a lower frequency range where they will become audible

Hugh

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

Quote Jocoserious:

---

Yeah taking it out of context sure...The only thing I was trying to get at, a point which not many people seem to bring up, a point which most people seem to ignore and look past is the point of using a higher sample rate to push most of the distortion into an inaudible region.

---

What distortion? There is no distortion if the system is sampling properly according to the Nyquist theorum (and ignoring any quantising artefacts for the moment). Sampling is a perfect process, provided the filtering complies with the requirements -- which they do in all well designed systems. It doesn't add distortion, and sampling at 96kHz doesn't change that. It doesn't 'push' disortion products anywhere, either.

Sampling at 96kHz allows more accurate signal processing in some ways than working at 44.1kHz, and if you are using badly designed anti-alias and reconstruction filters, the audibility of their functioning will be reduced since the turnover is an octave higher. But that's it.

You are right though, threads on any aspect of digital audio are always busy and tempers tend to fray as those who do understand the topic get frustrated with those who don't, and those who don't get upset when their misunderstandings are exposed.

But as I've said, it is a very complex subject that is inherently hard to understand, and very few explanations in colleges, on the web, and even in some books get it right. To really get to grips with sampling theory, you also need a solid undertsanding of modulation theory and information theory -- both highly technical subjects in themselves.

Hugh

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

Hi Hugh,

I think the question has got way too complex. Surely the simple answer for 99.9% of people recording at home in a semi-pro - or even pro setup would be to suggest that 44.1kHz is far more appropriate than 96kHz.

Simple as that.

I would only suggest there is any point in recording at significantly higher frequencies if very high end gear was being used throughout for the purposes of audio research or very very high end (classical) recordings. The simply truth is that most other systems have other factors that affect the recording quality to a much larger degree than the sampling frequency ie. Jitter, D-A inaccuracies and distortion in both analogue and digital domains. Not to mention 'distortion' in the mixing process! And given that, again, 99.9% of people listen to music on CD or mp3 players surely recording at 96kHz becomes a bit of a nonsense.

BTW - Huge fan of SOS - bought the very first issue!

--------------------
Mac Pro, Logic Pro, lots of software and 17 hard drives!

Quote Jonnypopisical:

---

I think the question has got way too complex.

---

I don't think questions can get too complex, only the answers! But this is a complex topic and there is still a huge amount of misunderstanding about it, as has been shown here. Hopefully, threads like this will help, in some small way, to reduce this kind of misunderstanding over time. That's why I perservere with it, anyway

Quote:

---

Surely the simple answer for 99.9% of people recording at home in a semi-pro - or even pro setup would be to suggest that 44.1kHz is far more appropriate than 96kHz.

---

I'd say more practical, rather than more appropriate.

Quote:

---

The simply truth is that most other systems have other factors that affect the recording quality to a much larger degree than the sampling frequency

---

You're certainly right there. Moving a microphone a few inches one way or the other would make a far more audible difference than switching between 44.1 and 96kHz sample rates.

Quote:

---

And given that, again, 99.9% of people listen to music on CD or mp3 players surely recording at 96kHz becomes a bit of a nonsense.

---

Not a nonsense. The source recordings should always be of the highest possible quality because anything that is then done to the material can only reduce quality. That's why high speed and wide format tapes were used in the studios back in the day when the consumer had manky 7-inch singles and ceramaic cartridges, or cassette players, or AM radio.

With modern digital equipment, the processing losses are far less than they were with analogue gear, but the principle remains the same. More worryingly, it is now quite common for domestic users to have replay equipment which is beginning to rival that of professional studios. DVD-A can accommodate 24/192 material, DTS discs can accommodate 24/96 material, and SACD is roughly equivalent to 24/96. MP3 is a convenience at the moment, like cassette was twenty years ago, but it may well be superceded by something far better in the not too distant future.

Quote:

---

BTW - Huge fan of SOS - bought the very first issue!

---

Thanks muchly

Hugh

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

I've been interested in this arguement for a while. Over on another thread, we have been discussing the article...

mixonline.com/recording/mixing/audio_emperors_new_sampling/

...where Paul Lehrman (one of Mr Robjohns' peers in the audio journalism world I believe) talks about a study that points to higher sampling rates not really making a noticable difference. Of course, the study refers to stereo recordings and it may still be better to record multitrack at a higher rate before downsampling but I wonder if it does make a difference. Before the days of oversampling plug-ins which seem to improve distortion and eq algorithms (as I've found out through using Guitar Rig, PSP products with and without the 'Fat' selected and my UAD plugs) the argument for 96K may have been stronger but I'd like to see a study testing the difference with using higher rates throughout the production process.
When I say I wonder, I'm not just using a rhetorical device as I do indeed wonder. I imagine people like Mr Robjohns have much more opportunity to test these things than I do but after reading the article above I'm going to stick to 44.1 and use oversampling plugs like the UAD 1073 when I can and stay happy.
As has been said previously in this thread - if there are differences that humans can hear, they are pretty darn small if we are still arguing about them and other factors should have much more effect. I'll stick with the extra tracks and CPU thanks.

Oh, yes. While we are handing out compliments, I have Hugh Robjohns (and Paul White, despite the 'buss' travesty) from SOS on my favourite authors list on Myspace. I've been reading you guys for coming up to 20 years now. Keep it up, Mr Robjohns, I get enormous pleasure in reading your articles every month. Seeing HR, PW or Martin Walker on the byline makes me feel like I'm listening to old friends.

Quote JohnnyT:

---

...where Paul Lehrman (one of Mr Robjohns' peers in the audio journalism world I believe) talks about a study that points to higher sampling rates not really making a noticable difference.

---

In theory, it shouldn't make any difference. If you accept that we humans really can't hear anything above about 20kHz (and the majority of evidence supports that premise), then any sampling system that captures and reproduces audio up to 20kHz should be perfectly adequate, and sampling at a higher rate to capture a wider bandwidth shouldn't sound any different.

However, the real world is a tad more complex as there are a lot of variables to consider. It was certainly the case a decade or two ago that the anti-aliasing and reconstruction filters were imperfect and had an audible effect. Sampling at a higher rate removed the audibility of the filters, and hence many claimed higher rate sampling was better. It was -- but because of filter imperfections, not higher sample rates per se.

There is also the issue of JF intermodulation to consider. There is a lot of ultrasonic noise in our modern world that we humans are not aware of because we can't hear it -- yet if the electronic equipment (analogue or digital) suffers non-linearities intermodulation can result and new audible signals can be created. Again, a decade or two ago, higher sample rate systems tended to suffer less from this than lower sample rate systems, but technology has moved on and it is much less of an issue now. So again, there is negligable sonic difference between standard and higher rate sampling systems.

Another aspect was the clocking precision and jitter. To work at all, higher sample rate systems have to have a very precise clock and a while back these tended to be better than systems that only operated at standard rates. But as technology and design experience has matured, clocking is no longer the challenge it once was, and so the differnece between standard and high sample rate systems has diminished.

But myths and ideas persist for decades in the audio industry -- people are very slow to change their ideas and update their beliefs, which is why these old chestnuts keep coming around and around.

Quote:

---

I imagine people like Mr Robjohns have much more opportunity to test these things than I do but after reading the article above I'm going to stick to 44.1 and use oversampling plugs like the UAD 1073 when I can and stay happy.

---

I think that is a very sensible and pragmatic approach to take -- assuming you have reasonably modern and well designed converters. Personally, I will continue to make the majority of my source recordings at 96kHz simply because most are stereo and very few involve more than eight tracks, the equipment I have makes it easy to do without penalty, and it makes sense from a maximal quality archive point of view. But I don't break down at the thought of recording projects at 44.1 or 48 -- and I do both frequently, and if a project is going to involve a lot of source tracks, then 44.1 or 48 is way more mangeable and uber-quality is generally less of an issue because it would normally mean I was into close miking and DIs and all that.

Hugh

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

Quote Hugh Robjohns:

---

However, the real world is a tad more complex as there are a lot of variables to consider. It was certainly the case a decade or two ago that the anti-aliasing and reconstruction filters were imperfect and had an audible effect. Sampling at a higher rate removed the audibility of the filters, and hence many claimed higher rate sampling was better.

---

This is still sort of the case with the vast majority of oversampled ADCs as the decimation filters are more often then not half band designs which are only -3db at 0.5FS. They fall off quickly, but that 0.05FS window is sufficient to make ultrasonic rich sources sound unnatural. This is aggravated by the outputs of many electronic instruments containing a lot of ultrasonic hash from the noise shaping used in the DACs (Building competent post filters costs money).

The fix is easy, run the ADC output at 96Khz, add an analogue filter going over at 25 or 30Khz, and resample the output of the ADC with a correctly designed decimating filter in the software....

Of course doing this is a trade off as it adds noise (the extra electronics), adds latency (the additional FIR resampler) and adds CPU and IO load. You could of course put a small cpu (or modest gate array) in the ADC box itself to do this final downsample correctly there at which point the apparent output will be at 44.1/48 without the problem, but nobody that I know in the prosumer card market does this.

Regards, Dan.

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

I really like the last two posts. I'm not very convinced by subjective statements but the mention of JT inteference, the explanations of ultra-sonic complications within electrical systems and the non-human hearing based ideas make sense to me. Of course, all things need to be equal and it makes sense that equipment designed to work well at the higher rates should logically work better when not pushed as hard. The fact that many plug-ins do not oversample internally may also mean a better result when working with them at higher rates. If I seem a little anti high rates, then I'm sorry. I do think we need to find out why there are many reports from trusted sources saying that high rates are better, and if it's not the recorded sound itself then what it is that results in the better quality. I'm actually very interested in using 96K soon (I'm about to do some location sound for video and I'm going to use my M-Audio recorder at 96 as an experiment) but I've been burned with new concepts before - I have yet to be asked to mix or record anything in 5.1, despite spending a couple of grand on a monitoring setup and making sure everything I have is 5.1 capable. I realise this is a different type of issue but it has made me sceptical! As I said, I will continue to use 44.1 for multitrack until such time as I have excess CPU cycles and hard-disk space and the rest of my audio path is good enough. Or until someone can show me, either in a good demonstration or via a good study that it is worth it for me.

I would tend to concur with your scepticism, as I suspect that the cases where 96K sounds better are really papering over the cracks in poorly implemented hardware or software (And there is an awful lot of it out there).

It is not that hard to do right, but it will always cost more (extra electronics for hardware, more CPU load for plugins), and the headline SNR specs for hardware will be slightly worse then they would have been with it done wrong.

Unfortunately the benefits are not always obvious, especially on a spec sheet as the frequency response and SNR are worse in the box done right then they are in the one without the extra filters, and most marketing runs on specs, not real world performance.

Regards, Dan.

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

The topic is so complicated and really the only thing which matters is that you either use 44.1/48 or 88.2/96 and you hopefully try and record whatever you are trying to record well. Guitars Amps, Bass Cabs, Singers, Drums.... if not recorded well... who cares if it was recorded at 192 because it ends up being 44.1 anyway and is still crap

I think if you record at the source well, and capture what you want then regardless if it was recorded at 44.1 or 88.1 it is going to sound good.

I just feel sorry for Hugh's fingers... his keyboard must be dented from all that typing

--------------------
There is nothing Grim about this Reaper
We Fell From The Sky

This thread should be a sticky, along with a thread about multi track recording connections and equipment.

I record at 96khz lately, mostly for the lower latency I can get with my M Audio AP192 used as monitor with VST effects path digitally connected to my EMU 1212M through SPDIF connections.

Quote Henry-S:

---

The topic is so complicated and really the only thing which matters is that you either use 44.1/48 or 88.2/96 and you hopefully try and record whatever you are trying to record well.

---

Well, sort of....

But it has always seemed to me that if you want to be able to consider yourself an engineer as opposed to a technician, then you really need to understand how your tools work and the hairy edge cases where your tools stop behaving ideally and collide with the real world (And how to fix it when it happens)!
This is not really all that complex, compared to say grade 8 music theory!

There is a place for both skill sets of course, and I am well aware that I am far better at system design and implementation then I am at using that system to a creative end, my stuff does not hum, and you can use GSM phones around it with impunity, but my mixing skills (for studio based music) are still a work in progress (Live is different, mostly because good enough now is better then perfect in 3 weeks).

Regards, Dan.

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

Quote Hugh Robjohns:

---

Oh dear... It appears that you don't understand the Nyqusit theorem. Can I respectfully suggest you read <a href="/sos/feb08/articles/digitalaudio.htm" target="_blank">this article</a> which will go some way to explaining the process. Or the Lavry article cited above.

---

Eh, I know well what the theorem states... I'm stating that these guys are over-stating the results of the theorem. Believe me, I sat there at the same university Nyquist got his PhD at arguing with some well established physics professor, who was a great hobbyist with accoustics and these things, the same point I'm arguing with you.

See, if F is the frequency you're interested in, 2F is necessary in order to avoid aliasing. 2F won't give you a 100% accurate reproduction of the original wave though.

Say you're interested in recording from 0 to 22 kHz...
A) If you record at 44kHz, you won't get aliasing. Aliasing makes extra noises where they don't belong.
B) If you record at 44kHz, you won't get a 100% perfect reproduction of all signals in the 0 to 22 kHz range.

People who think that A) implies a "perfect reproduction" (of the 0-22kHz range) aren't thinking this through and don't seem to understand that sometimes results get overstated.

I could actually prove what I'm saying with my equipment, but I'm sure you'd still say I was wrong because you believed the overstatement of the theorem.

Quote Interweaved:

---

See, if F is the frequency you're interested in, 2F is necessary in order to avoid aliasing.

---

And there you just proved that you completely clueless about the whole matter.

It's sad that people that pick up a few key words from some forum come here and self proclaim themselves as some kind of authority, even referencing supposed academic credentials and then just go on to regurgitate a load of crap completely messing themselves up in the procvess and possibly confusing a lot of people.

I hope most people reading this forum understand that you don't know what you are talking about.

Interweaved - are you trying to argue that Nyquist and Shannon were wrong? Or are you saying that current equipment falls short of perfection?

And don't forget that it is > rather than =. And while we're at it, don't ignore the filters.

Cheers

James.

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

Quote James Perrett:

---

Interweaved - are you trying to argue that Nyquist and Shannon were wrong? Or are you saying that current equipment falls short of perfection?

And don't forget that it is > rather than =. And while we're at it, don't ignore the filters.

Cheers

James.

---

My argument is that people mis-interpret what the theorem does provide and what it doesn't.

Quote Interweaved:

---

Eh, I know well what the theorem states... I'm stating that these guys are over-stating the results of the theorem.

---

Eh? The theorum defines the correct mathematical operation needed to achieve perfect sampling, the correct application of which allows the source signal to be recreated perfectly. How can that be overstated?

Quote:

---

See, if F is the frequency you're interested in, 2F is necessary in order to avoid aliasing.

---

No, that is not the case. For a start, the sample rate is related to the bandwidth of the source signal, not just it's frequency... but ignoring that nicety for the moment, if F is the frequency you are interested in, the sampling rate must be 2F+n, where n is sufficiently large to enable the real-world filters to achieve the necessary attenuation of the unwanted image(s).

Quote:

---

2F won't give you a 100% accurate reproduction of the original wave though.

---

Absolutely. Which is why no practical audio sampling system samples at 2F. So why are you harping on about something that isn't done, isn't a problem, and isn't relevant?

Quote:

---

Say you're interested in recording from 0 to 22 kHz...
A) If you record at 44kHz, you won't get aliasing. Aliasing makes extra noises where they don't belong.

---

That's debateable. The bottom of the lower sideband will fall precisely on the top of the wanted audio, and thus any wanted signal components at 22kHz won't be perfectly retrievable because of its superimposed alias.

Quote:

---

B) If you record at 44kHz, you won't get a 100% perfect reproduction of all signals in the 0 to 22 kHz range.

---

Obviously, because the sample rate isn't high enough and doesn't conform to the Nyquist requirements.

Quote:

---

I could actually prove what I'm saying with my equipment, but I'm sure you'd still say I was wrong because you believed the overstatement of the theorem.

---

Not at all. I could prove what you are saying with my equipment too. But the point is that what you are arguing is NOT what the theorum states. Nyquist and Shannon both state that the sample rate must be MORE THAN twice the bandwidth of the wanted signals. Not equal to twice.

Nothing is being overstated, but there does still seem to be some misunderstanding or confusion... Hopefully you can now see where that error lies.

Hugh

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

Quote Interweaved:

---

My argument is that people mis-interpret what the theorem does provide and what it doesn't.

---

They do indeed... and it is categorically your good self that has misinterpreted it. Sad but true.

Hugh

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

Quote Hugh Robjohns:

---

Nyquist and Shannon both state that the sample rate must be MORE THAN twice the bandwidth of the wanted signals. Not equal to twice.

---

May of us on this thread have stated this multiple times, and yet still Interweaved goes "yeah, but when it's not more than twice the bandwidth, it doesn't work".

We know. This is what the theory says. I don't understand what you are debating...

Quote desmond:

---

Quote Hugh Robjohns:

---

Nyquist and Shannon both state that the sample rate must be MORE THAN twice the bandwidth of the wanted signals. Not equal to twice.

---

May of us on this thread have stated this multiple times, and yet still Interweaved goes "yeah, but when it's not more than twice the bandwidth, it doesn't work".

We know. This is what the theory says. I don't understand what you are debating...

---

I think this is what is usually called a troll?

Quote Hugh Robjohns:

---

Quote Interweaved:

---

My argument is that people mis-interpret what the theorem does provide and what it doesn't.

---

They do indeed... and it is categorically your good self that has misinterpreted it. Sad but true.

Hugh

---

Ouch !

To answer the original question the as is 44

--------------------
Mac Pro, Logic Pro, lots of software and 17 hard drives!

Quote Hugh Robjohns:

---

Not at all. I could prove what you are saying with my equipment too. But the point is that what you are arguing is NOT what the theorum states. Nyquist and Shannon both state that the sample rate must be MORE THAN twice the bandwidth of the wanted signals. Not equal to twice.

Nothing is being overstated, but there does still seem to be some misunderstanding or confusion... Hopefully you can now see where that error lies.

Hugh

---

That doesn't matter so much. MORE THAN yo mamma.

But, eh, you can record at your frequency, I'll record at mine. (I actually am using 44, though, because there really isn't much content in the upper frequencies in most music I'm doing).

There's a bloke down the pub..

--------------------
Stuff what I done

my homemade test

16 bits 44 khz
http://motthieu.free.fr/freq/with-EQ-1644.aif

24 bits 96 khz converted to 16 bits 44 khz
http://motthieu.free.fr/freq/with-EQ-2496%20converted%20to%201644.aif< br />
with my soundcard, Echo Audiofire 8, you'd better record at 96 khz. do the test with your soundcard

I like your answers...You are free to put me to shame any day Hugh, I always learn something when you do.

I should really learn to stay out of these ones as I can never compete with a lot of you guys and just end up looking stupid....

That being said I still think I understand...somewhat....I just am absolutely terrible at trying to take what is in my brain and put it into words that all you really smart people can understand....Not only that like you say it is a very complicated subject and I apologize if I appeared to be ranting or angry. I try and keep my place as much as possible in here, which is at the bottom of the food chain.I really was only trying to make a point and unfortunately sometimes you need to push a few buttons to do so...

p.s. For the record I hate to quote anybody directly out of context, and try and refrain from doing so. I used that small quote from Katz because really, I would have had to use 5 pages to make it make any sense

Ok, here's a wave file of 18, 19, 20, and 21 kHz sine waves at 24 bit 192 kHz... you should be able to play this back at different rates yourself and hear the difference.

If you do not turn down your volume before listening to this file, I will laugh when you come back here crying about it.

http://www.soundupload.com/audio/davrr8p5utguaj7z

Quote Jocoserious:

---

I like your answers...You are free to put me to shame any day Hugh, I always learn something when you do.

I should really learn to stay out of these ones as I can never compete with a lot of you guys and just end up looking stupid....

That being said I still think I understand...somewhat....I just am absolutely terrible at trying to take what is in my brain and put it into words that all you really smart people can understand....Not only that like you say it is a very complicated subject and I apologize if I appeared to be ranting or angry. I try and keep my place as much as possible in here, which is at the bottom of the food chain.I really was only trying to make a point and unfortunately sometimes you need to push a few buttons to do so...

p.s. For the record I hate to quote anybody directly out of context, and try and refrain from doing so. I used that small quote from Katz because really, I would have had to use 5 pages to make it make any sense

---

Hey that sounds familiar to someone I know..

People who aren't afraid to make mistakes learn the quickest.

Hugh's only chink in his armour is that he doesn't know how us stupid people think.

Aren't we getting bogged down into talking about a part of a whole thing?

If Hugh says that 96k can help emulate analogue EQs in a way that deals with the high frequencies, is this then a mathematical side to the processing argument as opposed to the "sampling" process of capturing an audio signal, which is mechanical?

Quote Interweaved:

---

Ok, here's a wave file of 18, 19, 20, and 21 kHz sine waves at 24 bit 192 kHz... you should be able to play this back at different rates yourself and hear the difference.

If you do not turn down your volume before listening to this file, I will laugh when you come back here crying about it.

http://www.soundupload.com/audio/davrr8p5utguaj7z

---

First impressions......

Sounded very clear and crisp (over laptop) speakers.

Would be good if you processed some cymbals with EQ to give as an idea of differences.

I don't know what you used to generate that file, but the spectrum is all over the shop (and extends well past 22K (In fact it has as much energy at 80K as it does at 20 odd K).

Regards, Dan.

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

Quote dmills:

---

I don't know what you used to generate that file, but the spectrum is all over the shop (and extends well past 22K (In fact it has as much energy at 80K as it does at 20 odd K).

Regards, Dan.

---

Well, I think that'll happen.... as a consequence of the sampling. That stuff doesn't "matter" except if you want to piss off your neighbors poodle cuz it dumped in your yard. Also, I'm wondering about the analyzer algorithms once we're getting this precise... I don't think they all react quite 100% the same.

My intention was that ya'll could take that sound and play it back at different rates to try and hear any differences (I've convinced myself some tones in that frequency range do come out sounding strange at 44.1 kHz). I was hoping it'd sound different to you if you played it back at 44.1 or 48 versus 192.

Anyhow, I used this thing to make them (it supposedly is possible to compile it for windows... but, anyhow, it should be relatively easy to compile for macs):

http://tph.tuwien.ac.at/~oemer/wavetools.html

Quote Music Manic:

---

Aren't we getting bogged down into talking about a part of a whole thing?

If Hugh says that 96k can help emulate analogue EQs in a way that deals with the high frequencies, is this then a mathematical side to the processing argument as opposed to the "sampling" process of capturing an audio signal, which is mechanical?

---

Both are fundamentally mathematical transforms, and as for higher SR being better for eq, it might make the filters less exacting at the top end, but it increases numerical precision required at the low end (especially true for IIR stages where 24 bit mantissa can cause problems at 44.1K, never mind faster).
For FIR filters, you get into O(N^2) processor load behaviour as the filter impulse becomes longer to maintain the same temporal length and you are pushing more data into it.

This is in any case no reason to sample at high frequencies, when any competent software developer is quite capable of resampling as needed to best fit whatever the requirements of the algorithm are (8 or 16 times in some cases).

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

Quote Nuno_:

---

I think this is what is usually called a troll?

---

Yup!

--------------------
Cubase, guitars.

Sure starts to have that feel about it.

Regards, Dan.

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

Quote Jocoserious:

---

I like your answers...You are free to put me to shame any day Hugh, I always learn something when you do.

---

I hope you don't really feel I'm shaming you -- that certainly isn't my intention. The only stupid question is the one that isn't asked.

Quote:

---

I should really learn to stay out of these ones as I can never compete with a lot of you guys and just end up looking stupid....

---

Not at all. If you learn from the answers, there's nothing stupid about asking the question. It's only stupid when you ask a question but don't listen to the answer

hugh

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

Quote Interweaved:

---

Quote dmills:

---

I don't know what you used to generate that file, but the spectrum is all over the shop (and extends well past 22K (In fact it has as much energy at 80K as it does at 20 odd K).

Regards, Dan.

---

Well, I think that'll happen.... as a consequence of the sampling.

---

Eh? I'm sorry... perhaps I'm being dim and I'm missing something important here... but I think you are talking complete and utter nonsense.

If you had sampled "18, 19, 20, and 21 kHz sine waves at 24 bit 192 kHz", then what we should be hearing (and seeing on a spectrograph display) is that sequence of tones and nothing else.

But what you have actually produced has spectral spikes all over the place, indicating a harmonically rich source -- certainly not a sine wave -- and more aliasing than a Soundblaster soundcard would generate on a really bad day.

This test file is not as you have described... which would indicate that you don't know what you are doing.

Quote:

---

you should be able to play this back at different rates yourself and hear the difference.

---

What difference? The sonic character of different sample rate converters? Why? Or should we be listening to a junk file replayed erroneously at the incorrect sample rate to her different tones at different frequencies?

What exactly are you trying to demonstrate?

Quote:

---

I was hoping it'd sound different to you if you played it back at 44.1 or 48 versus 192.

---

Do you mean play it back through a sample rate converter to change the sample rate to 44.1 or 48? If so, then the different filter algorithms employed by different SRCs will definitely produce slightly different sonic characters because of the very complex spectrum you have created interacting with the different SRC filter amplitude and phase responses.

On the other hand, if you just mean playing the same file back but with a different sample rate, then yes, it will sound different again because the spectral content will replay at different frequencies.

Either way, it's entirely pointless.

I'm quite bemused by all this.... Must be feeling generous tonight

Hugh

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

Quote:

---

I hope you don't really feel I'm shaming you -- that certainly isn't my intention. The only stupid question is the one that isn't asked.

---

Yes you are! BUT it is a good thing I wouldn't want it any other way. If I was unable to accept that I am wrong or am unwilling to listen to others who are more knowledgeable than me, how can I ever expect to be as knowledgeable as them?

Quote:

---

Not at all. If you learn from the answers, there's nothing stupid about asking the question.

---

No there is not, however how that question is asked or communicated can be. Sometimes quick fire typing from the top of the brain with no time for reflection leads to stupid sounding questions, or answers for that matter, especially on a forum containing so many incredibly knowledgeable people. Not only that, when posing questions or answers to better educated people, it is usually smart to not offend them, or belittle them. I sometimes forget that probably 99% of the people on this forum are more knowledgeable than me, and just because the majority of it is anonymous doesn't give me any right to be a jacka$$

Just because I think I know something doesn't mean that everyone else doesn't know it better

Quote Hugh Robjohns:

---

Eh? I'm sorry... perhaps I'm being dim and I'm missing something important here... but I think you are talking complete and utter nonsense.

---

Then record at 44.1 kHz your whole life and live happy, man.

Or make a file like I did which doesn't have those artifacts.

Put your wave files where your mouth is, redcoat.

It should be pretty simple to take the file I uploaded and plot the actual graph it creates. One way or another, you should be able to do something more than sit around misspelling the word color in this conversation.

Edited by Interweaved (07/07/08 03:42 AM)

Quote dmills:

---

Sure starts to have that feel about it.

Regards, Dan.

---

Seems we're in troll-city at the moment.

The OP is the same guy who thinks that Cubase (which seems to be a cracked version he's using) is for Mastering. He's been flooding the forum with nonsense questions and insists on posting his newbie questions in the MRT forum.

--------------------
An Eagle for an Emperor, A Kestrel for a Knave.

Quote Interweaved:

---

Or make a file like I did which doesn't have those artifacts. (snip) It should be pretty simple to take the file I uploaded and plot the actual graph it creates.

---

Fair enough. Here is a spectral display of your 'misrepresented' file:



This is a screen grab using Izotope RX in spectral display mode. Time is along the horixontal axis, and frequency is the vertical axis, rising to about 96kHz at the top (to show the full audio bandwidth of a 192kHz sampled signal).

Clearly, these are not sine wave tones -- there are spectral components all over the place, almost certainly due to gross aliasing!

The screen resolution and jpeg-ing probably makes it hard to see, but looking at the four tones in your sequence, the lowest spectral components are at 3kHz, 1.5kHz, 6kHz and 1.5kHz again. Assuming the source signals really were sine waves at 18, 19, 20 and 21kHz, these sub-fundamental components would support the suggestion that there is gross aliasing going on.

However, closer inspection of the frequency region between 17 and 22kHz (see below) reveals that none of the four tones in this sequence has any frequency component at 19kHz or 20kHz (although there are frequncy components at 18kHz in all four, 21kHz in the first, second and third, and 19.5kHz in the second and fourth).

This irrefutably indicates that the source files are not as you described at all. An accidental error, or complete incompetence? I'll leave it to the forum members to make up their own minds.

Quote:

---

Put your wave files where your mouth is, redcoat.

---

Okay. Here is a spectral display of a file which I created to do what you said yours was supposed to. Created in Adobe Audition 3, sampled at 192kHz, containing a sequence of true sine waves at 18, 19, 20 and 21kHz, each lasting about 3 seconds and all at a level of -18dBFS (to reduce the risk of frying tweeters).



Again, time is on the horizontal axis, frequency is vertical up to 96kHz, and as you can see, there are just the four spectral lines showing true sine wave tones at 18, 19, 20 and 21kHz.

Should you want to try the wav file yourself, it is here for download (it is 8MB)

If you process this file with any sample rate converter to render a 44.1 or 48kHz file, you'll find all four tones preserved, intact and (hopefully) completely undistorted. I've tried it with SRCs in Wavelab 6, Izotope RX and AA3, all without any problem or significant degradations.

Typically most SRCs will impose a slight (less than 0.5dB) attenuation of the 21kHz tone -- it all depends on the way the filter is designed around the turnover point.

Quote:

---

One way or another, you should be able to do something more than sit around misspelling the word color in this conversation.

---

How witty.

It would appear, from this and numerous of your other posts, that you are far less experienced and knowledgeable than you profess to be and/or may enjoy a degree of 'trolling.' While I (and others here) will be happy to help you redress the former, the latter won't be tolerated.

Hugh

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

Edited by Hugh Robjohns (07/07/08 11:58 PM)

Quote Interweaved:

---

Assume the Shannon/Nyquist theorem only eliminate aliasing, so your sidebands are eliminated. But it does not guarantee anything about your sample being a great reproduction.

---

Erm. Yes it does.

--------------------
Nik
Godin, Axon, Tonelab, Repeater & the skin of my teeth!

this is just a random idea, not based on any scientific analysis, but maybe a better test of 44 vs 96, rather than blind A-B tests, would be a sort of "sustained listening" test ?

ie, is listening to stuff at 44.1 more fatiguing than 96, over a period of hours ? do your ears (or your brain) tire from having to reconstruct those missing frequencies?

maybe those ultrasonic frequencies which we can't actually hear (and which are better captured/reproduced @ 96kHz) have a more long-term or cumulative effect?

a straight A-B comparison doesn't really reflect real-world listening behaviour - for example, i can't imagine most casual listeners quibbling over whether they can hear more detail in a cymbal in two different recordings, however i would expect them to maybe switch a CD off after awhile due to a subconscious "ear fatigue" effect - something that i feel has increased since the advent of CD / MP3 etc vs LP / cassette (although maybe it's just as i've got older i can't handle listening to music the same way ... )

Quote Shingles:

---

Quote Interweaved:

---

Assume the Shannon/Nyquist theorem only eliminate aliasing, so your sidebands are eliminated. But it does not guarantee anything about your sample being a great reproduction.

---

Erm. Yes it does.

---

Have just read the article Hugh linked to earlier on - sorted my head out on a number of issues. One of these was the idea that increasing the sample rate is a step up in 'resolution' akin to moving from DV to HD. I now (feel, at least!) as if I understand how the reconstructed analogue sound that comes out of my speakers is reconstructed identically from either a 44.1 recording OR a 96 recording. (disregarding any 'non-ideal DA' stuff).

I understand the little I do by reference to thinking of the sample rate as storing a 'carrier signal', rather than a 'pixelisation' of the sound. Secondly, from the fact that a dithering process creates a randomly noisy linear signal, in place of a distorted one.

A couple of questions on it, though:

1. The 'dithering process' in question here - in order for the result to be a noisy analogue signal, where the dither noise is what is used to join up the steps in reconstructing the waveform, I am assuming that all DA converters add an *analogue* dither noise after the distorted/stepped signal has been created. Is this right?

2. I had only thought of dither in terms of word length conversion, as a noise that was added to force information regarding low amplitude detail further up the word (?! sorry...) so that it is not just chopped off. Are these two distinct uses of dithering, or am I missing the connection between the two?

Thanks,

Hector.

--------------------
Anto mo Ninja, Watashi mo Ninja

Quote BubbleButt:

---

this is just a random idea, not based on any scientific analysis, but maybe a better test of 44 vs 96, rather than blind A-B tests, would be a sort of "sustained listening" test ?

---

Certainly listening fatigue is a useful indicator of audio quality -- and it is one area where mp3s and other data reduced formats tend to fall down.

Quote:

---

do your ears (or your brain) tire from having to reconstruct those missing frequencies?

---

In the case of many data reduction systems,the asnwer is yes. However, many decades of using reduced bandwidth systems like 78rpm records, AM radio, cassette and others would seem to suggest that los bandwidth isn't such an issue -- the fatigue problem tends to manifest when small spectral chunks are missig from within the bandwidth (which is what heppens in most data reduction algorithms).

Quote:

---

maybe those ultrasonic frequencies which we can't actually hear (and which are better captured/reproduced @ 96kHz) have a more long-term or cumulative effect?

---

Maybe. Personally, I think perceived differences are more to do with the small phase differences caused by the anti-alias and reconstruction filtering that occur within the 20Hz-20kHz region for base rate and higher rate sampled systems.

Hugh

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

Quote molecular:

---

Have just read the article Hugh linked to earlier on - sorted my head out on a number of issues.

---

Glad it helped.

Quote:

---

1. The 'dithering process' in question here - in order for the result to be a noisy analogue signal, where the dither noise is what is used to join up the steps in reconstructing the waveform, I am assuming that all DA converters add an *analogue* dither noise after the distorted/stepped signal has been created. Is this right?

---

No, dither is not applied in the D-A process. It is applied when a signal is first quantised in the A-D, or re-quantised to reduce its wordlength. In effect, the dither signal is added (mixed with) to the input audio signal, and the whole thing is then quantised. As Dan explained earlier, the noise forces the signal to jump between adjacent quantising levels in a way that, statistically, represents its true amplitude.

Quote:

---

2. Are these two distinct uses of dithering, or am I missing the connection between the two?

---

They are separate applications of dither, and the technique used differes slightly between the two processes -- but the underlying concept is the same: dithering linearises the quantising transfer function, providing a (theoretically) perfectly linear system with a defined noise floor.

In an A-D converter, the dither signal is am independent noise source of some form. In requantising to reduce word length, the dither signal is usually derived from the lower order bits that would other wise be discarded in the truncation process.

Hugh

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

Hi,
I am the original poster of the question, asking if I should use 44 or 96.

I have read this thread and decided to use 24bit/44.1.

I would however like to give a massive thanks to everyone who has contributed, I really appreciate the efforts everyone has gone to.

Thanks,
AM.

... and relax...

Quote Hugh Robjohns:

---

Quote Interweaved:

---

Or make a file like I did which doesn't have those artifacts. (snip) It should be pretty simple to take the file I uploaded and plot the actual graph it creates.

---

Fair enough. Here is a spectral display of your 'misrepresented' file:

---

---

Well.. your image didn't load for me. But I'll take your word on it that it's distorted, really. Don't know why it did that (I just downloaded some source code that says it generated a sine wave).

Truth is, I have this example that convinces me using another technique, but I don't think recording it will work nor convince anyone, because it adds another "sampling" layer when recording it.

So.... anyhow, I'm not trolling. As for credentials, I really didn't say much about them.

I'm done with screwing around with the question, because, like I said, not much I'm doing has much energy above 15kHz anyhow.

If you can generate high frequency (17kHz plus) sine waves at 192 kHz/24 bit sampling... then try and play them back at different sampling rates, I think you should hear a difference. Up to you, though...

In fact, I recommend playing up a major scale, first from 8 kHz, then from 16 kHz.. with sine waves... created/sampled at 192/24... then playing it back at different sampling frequencies.

Edited by Interweaved (07/07/08 11:48 PM)

Quote Ken I Cummin:

---

... and relax...

---

hugh

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

thanks hugh for these answers.

I realise I'm drifting ever further away from the posters topic, but he seems to have come a wise decision - certainly the same one I have come to, as I am running a laptop and don't want to be the one to say 'no you can't double track that harmony... I'm sampling at 96khz for reasons I can't hear, and I'm maxed out!'

I want to run my understanding about averaging/dithering at AD stage past the forum...

So, I have a sine wave at 100hz, of an amplitude of the first single step of my wordlength (i'll call it '1'...), and I'm sampling at 44.1 - this means that the first quarter cycle, which rises linearly from 0 to 1, is covered by 1100 and a half samples.

What I end up with in the digital domain (once dithered and sampled) is that the first few of these samples are most likely to be zeroes, but as we move through the samples, they become more and more likely to be 1s, following a graph of 'probability of being 1' against 'time' which looks very similiar to my original sine wave...after sample 1101, the likelihood of a sample being registered as a 1 decreases again, until the cycle moves into its negative phase.

Without the dither, I would get ALL zeroes, followed by ALL ones, followed by ALL zeroes in the first half of the cycle.

Is this right? apologies for the gobbledegook explanation, but it is 1am!

Ta,

Hector.

--------------------
Anto mo Ninja, Watashi mo Ninja

Quote Interweaved:

---

Well.. your image didn't load for me.

---

Apologies. It was referenced from an in-house file store that may not be visible to ordinary forum users. I have now re-referenced the file to an external web host, so it should be properly visible to everyone. Sorry about that.

Quote:

---

If you can generate high frequency (17kHz plus) sine waves at 192 kHz/24 bit sampling... then try and play them back at different sampling rates, I think you should hear a difference. Up to you, though...

---

I can, as I have demonstrated -- and so too can most people with a half decent DAW -- but if you play a 192kHz file back at a lower sample rate without sample rate conversion -- yes, some DAWs and wave file editors will let you do that -- you will simply lower the frequency of the source tone in a simple 'varispeed' manner.

So what you really need to do is sample rate convert the source file (I'm sure you meant that) -- but then the results will depend entirely on the design and efficacy of the SRC's anti-alias filtering -- and pretty much every SRC does it slightly differently. Inherently, the nearer the source tone is to the notional turnover frequency (which is typically going to be between 20 and 22kHz) the more its amplitude is likely to be reduced, and the more likely it is that some harmonic aliasing might occur (depending here on the harmonic precision of the source audio).

As I explained in my post above, I did actually SRC my own test file described above using the SRCs including within AA3, Wavelab 6 and Izotope RX pro (this last with several different filter options).

In some cases I couldn't hear or measure any differences at all, but in others I could -- but I'm not surprised by that. No one has denied that practical anti-alias and reconstruction filters can, and often do, have a sonic impact on HF signals. You would expect them to because of the inherent amplitude and phase response wobbles as you approach the turnover frequency.

As I explained several pages ago, that is the precise reason that operating a poorly designed converter at a 96kHz sampling rate can often sound subjectively better than when using it at 44.1. It ain't rocket science.

But of course, none of this comes remotely close to justifying your earlier statements and claims relating to futile 2F sample rates and your apparent ignorance or misunderstanding of the Nyquist theorum.

Quote:

---

In fact, I recommend playing up a major scale, first from 8 kHz, then from 16 kHz.. with sine waves... created/sampled at 192/24... then playing it back at different sampling frequencies.

---

Easily done, but time consuming and I'm out of the office all day tomorrow. Why not redeem yourself and impress us all with your technical skills in generating such a potentially interesting test file that we can then all enjoy experimenting with?

Hugh

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

Quote molecular:

---

What I end up with in the digital domain (once dithered and sampled) is that the first few of these samples are most likely to be zeroes, but as we move through the samples, they become more and more likely to be 1s, following a graph of 'probability of being 1' against 'time' which looks very similiar to my original sine wave...after sample 1101, the likelihood of a sample being registered as a 1 decreases again, until the cycle moves into its negative phase.

---

Yes, essentially. The statistical probability is as you have described, and the degree of 'randomness' between adjacent quantising levels of adjacent samples is the element we hear (and measure) as the noise floor.

Quote:

---

Without the dither, I would get ALL zeroes, followed by ALL ones, followed by ALL zeroes in the first half of the cycle.

---

Yes. Precisely. There is no noise floor -- the system is perfectly silent -- but the wanted signal ceases to be audible until it reaches a level when the quantiser can switch between levels -- and then it is horrendously distorted because it's amplitude envelope is essentially being clipped into square waves.

You can hear this effect very clearly in the 3-bit examples from my february 2008 digital myths article.

I took the analogue output from a CD player reproducing some piano music and passed it through an A-D converter modified to operate with just three bits and no dithering.

truncated - 3 bits

What you hear when the piano is played loudly is gross quantising distortion, rendering the audio very unpleasant and barely recognisable. As the piano tones die away on sustained notes, they break up and then mute, as the quantiser is forced to generate all zeros.

However, when correctly dithered, the piano is distortion free, even during the die away of sustained notes -- but the noise floor is ridiculously high, as you would expect in a three bit system.

dithered - 3 bits

The third file uses noise-shaped dither to move much of the spectral elements of the dither signal to the extreme HF, where is is subjectively less objectionable and intrusive.
Noise shaped dither - 3 bits

The result is a surpsingly listenable version of the piano track, distortion free, with the tail end of sustained notes fading gracefully under the noise floor.

So yes, you are right and I hope these examples cement the ideas for you.

Hugh

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

Quote Hugh Robjohns:

---

So what you really need to do is sample rate convert the source file (I'm sure you meant that) -- but then the results will depend entirely on the design and efficacy of the SRC's anti-alias filtering -- and pretty much every SRC does it slightly differently. Inherently, the nearer the source tone is to the notional turnover frequency (which is typically going to be between 20 and 22kHz) the more its amplitude is likely to be reduced, and the more likely it is that some harmonic aliasing might occur (depending here on the harmonic precision of the source audio).

---

Uh, this I think was the "distortion" I was complaining about... amplitude loss and frequency inaccuracies at high frequencies.

Anyhow, as per doing something more with that, perhaps I'll come back to it in a few weeks. I mess with this stuff in my spare time, I have some other priorities.



Edited by Interweaved (08/07/08 03:25 AM)

Quote Interweaved:

---

Uh, this I think was the "distortion" I was complaining about... amplitude loss and frequency inaccuracies at high frequencies.

---

A small reduction in amplitude is certainly possible -- but in most cases we are talking about fractions of a dB. Not only can most people not perceive that small a level difference, but most people won't be able to hear a 21kHz sine tone anyway.

I'm not sure what you mean by 'frequency inaccuracies' The pitch of the source tone can't change through sample rate conversion, and if the source tones were pure sinewaves there will be no harmonics to cauase aliasing distortions. So if you can hear additional frequency components, that would suggest an imperfect source signal combined with inadequate anti-alias filtering. Both are possible.

So if they are hearing some significant and repeatable difference, the liklihood is that they are hearing some kind of non-linear distortion products or aliasing -- which inherently means that the source signals weren't processed properly with an accurate anti-alias filter before digital conversion (or sample rate conversion), and/or the replay system is inadequately engineered to reproduce HF signals without adding significant distortion products of its own.

Neither of these things is inherent in digital recording/processing per se, but are possible through poor equipment designs. An important distinction to make.

Quote:

---

Anyhow, as per doing something more with that, perhaps I'll come back to it in a few weeks. I mess with this stuff in my spare time, I have some other priorities.

---

As do we all. I hope your understanding of this complex topic has improved through our dialogue

Hugh

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