The No.1 Website for Pro Audio
 Search This Thread  Search This Forum  Search Reviews  Search Gear Database  Search Gear for sale  Search Gearslutz Go Advanced
if higher sample rate doesnt matter then why .... Effects Pedals, Units & Accessories
Old 8th September 2011
  #31
Lives for gear
 
doug hazelrigg's Avatar
Samplers/convertors are physical machines with real-world physical limits determined by physical law -- they're not magic boxes. A 768fs is NOT going to be "more accurate" by ANY definition of that phrase, because after a certain point, as a physical machine operates faster and faster, the measurements it takes will be less and less accurate. It's FAR better to have 44,100 samples that are all reasonably accurate measurements than having 768,000 woefully inaccurate ones. Don't believe the hype!

Also, "resolution" is only tangentially related to sample rate.

Errors in computing peak values DO decrease as one increases sample rate, so 48kHz is better than 44.1, and 88.2 or 96 is better still. A higher fs also minimizes any passband issues -- a steep filter between 20kHz and 22.5kHz will have a more pronounced effect on the audible range than a more gentle filter between 20kHz and 88.2/2 or 96/2 kHz. But the truth is that around 60kHz these improvements become so minimal as to be negligible. That's why I use 88.2kHz*


*I don't use 88.2khz because it divides more "roundly" or "evenly" when downsampling to 44.1 -- that's yet another digital sampling "urban myth"
Old 8th September 2011
  #32
Lives for gear
 
TurboJets's Avatar
Quote:
Originally Posted by psycho_monkey View Post
The effect of the moon is measurable, calculable, predictable, visibile and provable - and probably lots of other words ending in "ibble" (random Red Dwarf quote there).

I'm not sure if that's the same for supersonic frequencies...
Well that's just it though, isn't it?

You're "not sure".

No one is really "sure".

That's one of the beauties of science for those who can appreciate it...there is still so much we don't know.

Doesn't matter really. I just thought it was an interesting point. Just because you can't see a thing doesn't mean it doesn't exist or have a physical effect.

However, it can be agreed that 768kHz does exist as a measurable frequency. As a measurable cyclic rate. When it's vibration is introduced to the human body, who knows what the effect is? As all things vibrate, surely it stands to reason there must be some influence or effect on things in its proximity or environment whether animate or inanimate.

Who would have thought 1,000 years ago that sound (or ultrasound) could help break up a kidney stone? Right?

We may not be able to hear ultrasound but it does affect the human body.
Old 8th September 2011
  #33
Lives for gear
 

Quote:
Originally Posted by doug hazelrigg View Post
Samplers/convertors are physical machines with real-world physical limits determined by physical law -- they're not magic boxes. A 768fs is NOT going to be "more accurate" by ANY definition of that phrase, because after a certain point, as a physical machine operates faster and faster, the measurements it takes will be less and less accurate. It's FAR better to have 44,100 samples that are all reasonably accurate measurements than having 768,000 woefully inaccurate ones. Don't believe the hype!

Also, "resolution" is only tangentially related to sample rate.

Errors in computing peak values DO decrease as one increases sample rate, so 48kHz is better than 44.1, and 88.2 or 96 is better still. A higher fs also minimizes any passband issues -- a steep filter between 20kHz and 22.5kHz will have a more pronounced effect on the audible range than a more gentle filter between 20kHz and 88.2/2 or 96/2 kHz. But the truth is that around 60kHz these improvements become so minimal as to be negligible. That's why I use 88.2kHz*


*I don't use 88.2khz because it divides more "roundly" or "evenly" when downsampling to 44.1 -- that's yet another digital sampling "urban myth"
i mentioned the physical limitations.
but those limitations depend on the quality of the design and the build. just being bigger does not mean it has lost quality (yet).

88.2 to 44.1 is NOT a myth
you can divide by any power of 2 with no extra artifacts at all
only lose the natural resolution by lowering it

48 or 96 to 44.1 will cause noticeable problems
192 to 44.1 needs golden ears to hear the artifacts

you can disbelieve the AES results that showed 192 was better than 88/96 was better than 44.1/16 was better than going to 24bit. i will use the highest rate that the specific component's clock jitter allows to work better.
Old 8th September 2011
  #34
Lives for gear
 

Quote:
Originally Posted by ArnieInTheSky View Post
768KHz sample rate, that'll take a snap shot of 384KHz. What note is that? Does it sound to my ears as infrared looks to my eyes? I don't see it... but it is there...

WTF

looks like its between f# and g 11va

only if you are sampling the infrared and recreating it
Old 8th September 2011
  #35
Lives for gear
 

Quote:
Originally Posted by psycho_monkey View Post
I'm not sure that anyone has said higher sample rates aren't better, or can provide a noticeable difference. I think it's just been debated how much difference that makes. Is it the "night and day" that some claim, or is it just a small improvement, that could be offset by the workflow implications?

At any rate, the above listing is pure marketing BS. Human hearing dynamic range doesn't reach 144dB (the dynamic range of a perfect 24bit converter). To put it another way, if you turned something up until your ears were bleeding at full scale deflection, you wouldn't be able to hear the quantisation noise in the reverb tails at the end of the song. Since a perfect 24bit converter doesn't exist, we're looking more at 21, 22bit conversion in practice - which is still more than good enough.

This advert is aimed at people who think they do, but don't actually understand, digital audio.

FWIW I've seen you go on about "intersample peaks" before, I'm not truly convinced you know what you're talking about or why you think it's relevant. Intersample peaks happen because there's no guarantee that a particular sampling interval will be at the peak of a waveform. Given the sampling frequencies used (from 44.1 upwards), it's enough to make sure you never limit above -0.1dBFs, and you won't have any issues. Nothing to do with a ridiculously specced, BS hyped converter - the only thing which it has going for it is the ability to separate a fool from his money.
nobody is using that DR
so it is irrelevant

irrelevant even if they did use it

even with a lower DR the higher conversion rate leads to more accurate reproduction that can be heard

AES demonstration proved it empirically up to 192k
golden eared stereophiles using green magic markers on their cds can even do better

yes there is marketing BS
and this may be an example

but the FACT is that higher sample rates, within the limits of resolution, clock jitter, noise, yada yada are better than lower rates.

we need some good gear and triple blind tests at higher rates to see where psychoacoustically there is no benefit

i suspect that golden eared stereophiles will always equate more cost with better sound.

i DO KNOW that intersample peaks are non existent.
what people call intersample peaks are the confusion between analog and digital domains and not understanding ALL of nyquist.

one a/d/a vendor has a white paper that shows it is the d/a design that causes problems not some inherent peak problem waiting to byte your @$$

logic and intellectual honesty along with a valid analysis confirms their statement (and mine). there is NO INTERSAMPLE PEAK PROBLEM.

you can go to =-0.1 if you want. you can clip, limit, and otehrwise mangle your signal. just dont complain about intersample peaks.

bob katz and others have noted , quite correctly, that you should stay away from 0dB with a significant cushion in addition to headroom.

it is possible to make your signal as loud as you want without distortion, so there is no need to use such methods as peaking at -0.1, limiting, etc.

if you want distortion then you dont need to worry about peaks of any kind.
Old 8th September 2011
  #36
Lives for gear
 

Quote:
Originally Posted by psycho_monkey View Post
The effect of the moon is measurable, calculable, predictable, visibile and provable - and probably lots of other words ending in "ibble" (random Red Dwarf quote there).

I'm not sure if that's the same for supersonic frequencies. The 96k argument is valid because it allows for less steep aliasing filters, which could quite easily have an audible effect below 20k. Unlikely for there to be a difference once the slope of the filter is out of the way above 20k. Still, if anyone wants to take a blind test on this...I'm all ears.

huh ???
steep aliasign filters should just make the signal more nyquist compliant and should have less effect than less sharp cutoffs.
Old 8th September 2011
  #37
Quote:
Originally Posted by oldeanalogueguy View Post
nobody is using that DR
so it is irrelevant

irrelevant even if they did use it

even with a lower DR the higher conversion rate leads to more accurate reproduction that can be heard

AES demonstration proved it empirically up to 192k
golden eared stereophiles using green magic markers on their cds can even do better

yes there is marketing BS
and this may be an example

but the FACT is that higher sample rates, within the limits of resolution, clock jitter, noise, yada yada are better than lower rates.

we need some good gear and triple blind tests at higher rates to see where psychoacoustically there is no benefit

i suspect that golden eared stereophiles will always equate more cost with better sound.

i DO KNOW that intersample peaks are non existent.
what people call intersample peaks are the confusion between analog and digital domains and not understanding ALL of nyquist.

one a/d/a vendor has a white paper that shows it is the d/a design that causes problems not some inherent peak problem waiting to byte your @$$

logic and intellectual honesty along with a valid analysis confirms their statement (and mine). there is NO INTERSAMPLE PEAK PROBLEM.

you can go to =-0.1 if you want. you can clip, limit, and otehrwise mangle your signal. just dont complain about intersample peaks.

bob katz and others have noted , quite correctly, that you should stay away from 0dB with a significant cushion in addition to headroom.

it is possible to make your signal as loud as you want without distortion, so there is no need to use such methods as peaking at -0.1, limiting, etc.

if you want distortion then you dont need to worry about peaks of any kind.

huh ???
steep aliasign filters should just make the signal more nyquist compliant and should have less effect than less sharp cutoffs.
I don't mean any offense, but what in the hell are you talking about?!? I'm generally pretty on top of my reading comprehension, but there's some weird stuff in your last few posts. Maybe I'm missing something...

Is there a link to this AES paper?

Intersample peaks are fact; they most certainly are NOT "non existent".

"It is possible to make your signal as loud as you want without distortion"? If you've found a way, you might wanna pursue a career in mastering, cuz you'll make a lot of loudness war types very happy.

And the audible effect of steep anti-aliasing filters is well-demonstrated and easily Googled.

Last, but certainly not least, I don't think that anyone is arguing that higher SRs might sound better. For whatever it's worth, it is my opinion that they do (I typically work at 96k). But there's a tradeoff in workflow. Things like track counts, plugin availability, processing power, and the amount of time you feel like sitting around to back up a full record's worth of 100-track tunes at 96/24. For those of us working in the real world, these are considerations that need to be weighed on a project-to-project basis. Unfortunately, it's not always JUST about the sound quality. I don't think you'll find anyone vehemently opposed to working at higher SRs. Certainly, you won't find anyone opposed to other people working at higher SRs.

Maybe there are some typos in your posts? I dunno...This is all very strange.
Old 8th September 2011
  #38
Lives for gear
 
TurboJets's Avatar
I can't help but try to inject a little more reality into this thread. A little more "whoa big boy, slow down; where are you going with this" kind of thing.

But, these guys are audiophiles/stereophiles and the product being discussed isn't portrayed as something that should be inserted into a recording/monitoring chain; or to be even more specific, a mixing chain. Yet that seems to be how most people are discussing it.

This is a playback device intended for listening pleasure.

Pleasure...not work.

So typical, so funny. It really is.

People have hobbies, people have things they do just for fun. Sometimes they may seem unreasonable, but it's their pleasure so who gives a hoot. And before anyone brings up the $$$ issue and how much $$$ people spend on stuff like this, consider that it's their money not yours and some people have the opportunity in this world to spend what may seem like an unreasonable amount of money on things that may seem unreasonable or unnecessary to others. So try and get a grip.

Back to life, back to reality.
Old 8th September 2011
  #39
Lives for gear
 
popmann's Avatar
Quote:
Originally Posted by frans View Post
A lot of folks don't understand digital.

a popular error:

-"my interface sounds soo much better at samplerate "X" versus samplerate "Y" - EVERY OTHER INTERFACE has to be like that.
Meaning that if people would just buy Burl/Apogee/Larvy/Prism/IZ level converters thy could save themselves some drive space and relatively cheap CPU cycles?

Is that really what you're suggesting?
Old 8th September 2011
  #40
Lives for gear
 

Quote:
Originally Posted by bgrotto View Post
I don't mean any offense, but what in the hell are you talking about?!? I'm generally pretty on top of my reading comprehension, but there's some weird stuff in your last few posts. Maybe I'm missing something...

Is there a link to this AES paper?

Intersample peaks are fact; they most certainly are NOT "non existent".

"It is possible to make your signal as loud as you want without distortion"? If you've found a way, you might wanna pursue a career in mastering, cuz you'll make a lot of loudness war types very happy.

And the audible effect of steep anti-aliasing filters is well-demonstrated and easily Googled.

Last, but certainly not least, I don't think that anyone is arguing that higher SRs might sound better. For whatever it's worth, it is my opinion that they do (I typically work at 96k). But there's a tradeoff in workflow. Things like track counts, plugin availability, processing power, and the amount of time you feel like sitting around to back up a full record's worth of 100-track tunes at 96/24. For those of us working in the real world, these are considerations that need to be weighed on a project-to-project basis. Unfortunately, it's not always JUST about the sound quality. I don't think you'll find anyone vehemently opposed to working at higher SRs. Certainly, you won't find anyone opposed to other people working at higher SRs.

Maybe there are some typos in your posts? I dunno...This is all very strange.
i am talking about the people that dont understand nyquist
and most dont know fourier either

they have a problem
allege that it is some intersample peak issue
which is total bullbleep

if you think there is a problem and wont look at the facts then i wont bother trying to explain it to you again

i found the paper with google
you could too

there are NO INTERSAMPLE PEAKS!!!!!!!!!!!!!!!!!!!!!
THAT IS JUST NORMAL NYQUIST IN ACTION
Any problem is due to your lack of knowledge or a badly designed d/a converter.

i am planning to patent the method and perhaps offer a service. too old to futz with setting up a corp and selling products.

violate nyquist and the filter could byte your @$$
apparently 99.99% of teh intersample peak believers dont know the full nyquist theorem nor how to use it

there are many people on both sides of the SR issue
and plenty of them who claim 2xBW is all you need
and some of them do deride those who use higher rates

to me storage is not an issue so i use highest quality

3TB for $129 usb3.0 on sale this week !!
with autobackup software

no excuse for not having quality recordings
nor backing up

but most of those arguing that intersample peaks are real are destroying their "music" with extreme processing to make it louder and so get more distortion and artifacts etc. so they can just use 44.1/16 initially and save disk space.
Old 8th September 2011
  #41
Quote:
Originally Posted by oldeanalogueguy View Post

if you think there is a problem and wont look at the facts then i wont bother trying to explain it to you again
My point is this: I don't think there is a problem. I'd be happy to look at facts if you would present them; as far as I can tell, all you've done is made vague (and possibly misinformed or erroneous) references to Nyquist, and insisted that everyone on GS hates high sampling rates and blames intersample peaks for the destruction of music. None of that appears to be the case from where I'm sitting.

Please, if you have a point beyond what I just mentioned, please state it clearly, and offer up some evidence that supports the claim. I'm happy to learn a thing or two.

Quote:
Originally Posted by oldeanalogueguy View Post
violate nyquist and the filter could byte your @$$
apparently 99.99% of teh intersample peak believers dont know the full nyquist theorem nor how to use it


How does an end user "violate Nyquist"? The only way that seems possible is if a company manufactures a converter without low-pass filtering and/or oversampling, and the end user uses that product. But is that really "violating Nyquist"? You make it sound like people are choosing to sample audio at low SRs without a LP filter, then crying about intersample peaks. To be frank, in the many years I've spent killing time on GS, I don't think I've ever come across this complaint. I can't even recall any recent posts bemoaning aliasing, which seems a much more likely issue with the kinds of situations you're describing.

And how would the filter "byte your ass" in this instance?

Along the same lines, how does one "use Nyquist"? We're not exactly designing and building our own converters here, so I fail to see your point.

Like I said before, this is all very strange...
Old 9th September 2011
  #42
Lives for gear
 

Quote:
Originally Posted by bgrotto View Post
My point is this: I don't think there is a problem. I'd be happy to look at facts if you would present them; as far as I can tell, all you've done is made vague (and possibly misinformed or erroneous) references to Nyquist, and insisted that everyone on GS hates high sampling rates and blames intersample peaks for the destruction of music. None of that appears to be the case from where I'm sitting.

Please, if you have a point beyond what I just mentioned, please state it clearly, and offer up some evidence that supports the claim. I'm happy to learn a thing or two.





How does an end user "violate Nyquist"? The only way that seems possible is if a company manufactures a converter without low-pass filtering and/or oversampling, and the end user uses that product. But is that really "violating Nyquist"? You make it sound like people are choosing to sample audio at low SRs without a LP filter, then crying about intersample peaks. To be frank, in the many years I've spent killing time on GS, I don't think I've ever come across this complaint. I can't even recall any recent posts bemoaning aliasing, which seems a much more likely issue with the kinds of situations you're describing.

And how would the filter "byte your ass" in this instance?

Along the same lines, how does one "use Nyquist"? We're not exactly designing and building our own converters here, so I fail to see your point.

Like I said before, this is all very strange...
i have explained it several times

at least one maker of d/a gears agrees
so read their white paper
i put the reference on wikipedia
but some idiot didnt like the truth and rewrote the whole article to make it wrong again and deleted the reference
taht showed there is no intersample peak problem

never said everybody hates higher SR
but a lot of GSs do

if you really know nyquist and fourier
then this is a simple exercise for the student
hint: reread the theorems completely and carefully

not going to rehash it again
Old 9th September 2011
  #43
Lives for gear
 

Quote:
Originally Posted by oldeanalogueguy View Post
huh ???
steep aliasign filters should just make the signal more nyquist compliant and should have less effect than less sharp cutoffs.

The argument is that distortions at certain frequencies affect the sound of other frequencies above and below it. So having a cut off filter at 22k can create distortions within hearing range. A higher sample rate moves the filter and it's associated distortions further away from the purported human hearing range making it less noticeable. That's how the story goes anyhow.

One thing that I think people have a hard time grasping is that, while Nyquist, the theorem stands and is not really open to debate, no one has been able to implement it. That is, a perfectly band limited signal can be perfectly reconstructed... Yeah well, we can not perfectly band limit a signal.
Old 9th September 2011
  #44
Gear Head
 

Quote:
Originally Posted by oldeanalogueguy View Post
nope
they draw the original digital samples back on the analog image
and say: there! the analog is higher than the digital
which is total nonsense
Nobody is doing this. Please show us all here that this is not just in your head.

Quote:
Originally Posted by oldeanalogueguy
Worse they assume that the d/a is designed to duplicate the levels in the a/d. the d/a circuit is not a mirror of the a/d and can be designed to be eg 60 db higher or lower than the original signal if the designer wanted to do that. now that would be a peak problem if it were designed to be +77dbu peak out for max digital bits for example.
Of course the D/A converter is designed to reproduce the voltages denoted by the digital signal. Why would it be any different? If it was designed to be different there would be distortion of the intended audio. The anti-aliasing filter just removes the square wave harmonics that would be reproduced if it wasn't there, along with frequencies represented by the digital wave that would "foldback" in to the audible frequency range. After the signal has been recreated the analogue components can then amplify the signal to whatever level is desired. This part has nothing to do with the absolute voltages represented by the digital signal and converted in to AC voltage by the D/A converter.

Quote:
Originally Posted by oldeanalogueguy
clearly they are confusing the digital and analog domains if they claim there are intersample peaks that are a problem because that is just nyquist working correctly.
It doesn't have anything to do with nyquist.

Quote:
Originally Posted by oldeanalogueguy
no valid nyquist signal that is run back through d/a has any peaks at all. it is just the original signal which was sampled and the samples have to be below the analog peaks in almost all cases of real signals.
There is no such thing as a "valid nyquist signal". It just depends how you configure your reconstruction filter as to how the audio will sound. For foldback distortion to occur at the output stage of your converter, the cut-off frequency of the low-pass filters would have to be higher than 50% of the sampling rate of the digital audio being sent to your converter!


Quote:
Originally Posted by oldeanalogueguy View Post
you are confused
bandwidth has nothing to do with cpu use
only the number of bits you are crunching
and infinite bandwidth does not make more bits
just a different arrangement of their values
What are you talking about? There is no "infinite bandwidth" in the digital domain. I don't care whether your signals are square waves or whether you "merrily clipped and compressed them with non-linear processing", your bandwidth is limited by the number of samples per second you use to represent your audio. Even in the box, this is not infinite - plugs still operate at particular rates, even with oversampling. I will say it again: if your sampling rate was infinite you would have no CPU power left, so there's your proof that it doesn't happen. It is impossible.

Please don't deflect this to bit depth - it has nothing to do with resolution.

You are good at making strawmen though.

Quote:
Originally Posted by oldeanalogueguy View Post
88.2 to 44.1 is NOT a myth
you can divide by any power of 2 with no extra artifacts at all
only lose the natural resolution by lowering it

48 or 96 to 44.1 will cause noticeable problems
192 to 44.1 needs golden ears to hear the artifacts
Check out the maths behind sample rate conversion.

Quote:
Originally Posted by oldeanalogueguy View Post
but the FACT is that higher sample rates, within the limits of resolution, clock jitter, noise, yada yada are better than lower rates.
But that's the point isn't it - clock jitter etc. does matter.

Quote:
Originally Posted by oldeanalogueguy
i DO KNOW that intersample peaks are non existent.
what people call intersample peaks are the confusion between analog and digital domains and not understanding ALL of nyquist.
It has nothing to do with nyquist. Nobody is confusing analogue and digital domains. Intersample peaks are a real phenomenon, testable and provable. They have to do with the difference between adjacent samples when their interpolated output exceeds full-scale. If there is enough headroom in the analogue pathway, the peaks will pass un-clipped. If they exceed the headroom they will clip the analogue circuit because there is not enough voltage to represent those peaks. The peaks are very real when translated out of the digital domain in to real-world voltages. They can be predicted mathematically by metering plugins.

Quote:
Originally Posted by oldeanalogueguy
one a/d/a vendor has a white paper that shows it is the d/a design that causes problems not some inherent peak problem waiting to byte your @$$
Intersample peaks are not the fault of shoddy manufacturing. Please understand this. Every analogue circuit has a certain amount of headroom. You may as well say that a Neve has shoddy design because it only has +26dbu headroom. The analogue part of a D/A will enevitably crap out at some voltage level, that's just the way it is.

Quote:
Originally Posted by oldeanalogueguy View Post
there are NO INTERSAMPLE PEAKS!!!!!!!!!!!!!!!!!!!!!
THAT IS JUST NORMAL NYQUIST IN ACTION
Any problem is due to your lack of knowledge or a badly designed d/a converter.

i am planning to patent the method and perhaps offer a service. too old to futz with setting up a corp and selling products.
You keep going on about "it's just nyquist in action". Do you really know what are talking about when you say that? I don't think you do. Intersample peaks have nothing to do with foldback distortion caused by exceeding the 'nyquist frequency'. Intersample peaks are a real and provable phenomenon.

Quote:
Quote:
Originally Posted by oldeanalogueguy
violate nyquist and the filter could byte your @$$
apparently 99.99% of teh intersample peak believers dont know the full nyquist theorem nor how to use it
Quote:
Originally Posted by oldeanalogueguy
i am talking about the people that dont understand nyquist
and most dont know fourier either

they have a problem
allege that it is some intersample peak issue
which is total bullbleep
No, it seems YOU don't understand nyquist theorum. Most others seem to know what they are talking about. I don't think you do.

Quote:
Originally Posted by oldeanalogueguy
but most of those arguing that intersample peaks are real are destroying their "music" with extreme processing to make it louder and so get more distortion and artifacts etc. so they can just use 44.1/16 initially and save disk space.
Again, intersample peaks have nothing to do with choice of sample rate and bit depth, nor do they have anything to do with an engineer's choice of bus compression or limiting.

Last edited by child of Gaia; 9th September 2011 at 02:13 AM.. Reason: clarification
Old 9th September 2011
  #45
Quote:
Originally Posted by oldeanalogueguy View Post
88.2 to 44.1 is NOT a myth
you can divide by any power of 2 with no extra artifacts at all
only lose the natural resolution by lowering it
I'm starting to think that maybe you are just some advanced troller model or something, the T2011.
Old 9th September 2011
  #46
Lives for gear
 

Quote:
Originally Posted by Dean Roddey View Post
I'm starting to think that maybe you are just some advanced troller model or something, the T2011.
you been sniffing the glue in your green magic marker ?

there is NO problem going from 88.2 to 44.1
you lose resolution
but there are NO other artifacts or issues

if you believe that internet wizdumb then it is time for you to start thinking for yourself
Old 9th September 2011
  #47
Gear Maniac
 
ArnieInTheSky's Avatar
 

Quote:
Originally Posted by oldeanalogueguy View Post
you been sniffing the glue in your green magic marker ?

there is NO problem going from 88.2 to 44.1
you lose resolution
but there are NO other artifacts or issues

if you believe that internet wizdumb then it is time for you to start thinking for yourself
You're losing half your information when you dither it down. Look! You went from 88.2 to 44.1. See... half.

If you sample at 768 and dither down to 44.1, you're only seeing about one twentieth of what it was you recorded. You're right, it's the D to A conversion that doesn't allow us to hear it but why buy this thing if there's no way to convert it? There's no medium to listen to it on. Besides, most people are looking at getting AIFF's for their iPods or listen to their CD's in the car with 80db of road noise howling at their windows. What super sonic sounds am I gonna hear listening on those formats?

If I'm gonna sit and really listen to tunes, I flip up the dust cover on my record player and throw back some Innis and Gunn. Bliss.

But hey man.... your money. Go buy/make whatever gear you want that inspires you to make the best music/sounds you can. I don't have a sig but if I did it would read:

"I'll use a Neve, or a Peavy, or ****ing Ham-sandwich if it works!"
Old 9th September 2011
  #48
Lives for gear
 

Quote:
Originally Posted by child of Gaia View Post
Nobody is doing this. Please show us all here that this is not just in your head.

many people have done this. see other threads attackign me where people who do not understand nyquist make that very noob mistake.



Of course the D/A converter is designed to reproduce the voltages denoted by the digital signal. Why would it be any different? If it was designed to be different there would be distortion of the intended audio.

FALSE. the levels can be higher/lower with zero distortion.
And the d/a design sets the levels not the digital signal.




The anti-aliasing filter just removes the square wave harmonics that would be reproduced if it wasn't there, along with frequencies represented by the digital wave that would "foldback" in to the audible frequency range. After the signal has been recreated the analogue components can then amplify the signal to whatever level is desired. This part has nothing to do with the absolute voltages represented by the digital signal and converted in to AC voltage by the D/A converter.

the analog anti aliasing filter cant remove something that has already been folded back in the digital domain

It doesn't have anything to do with nyquist.

it has everything to do with CORRECTLY USING NYQUIST
and NOT VIOLATING the requirements of the theorem


There is no such thing as a "valid nyquist signal".

what can i say. if you dont think nyquist theorem requires a valid nyquist signal that complies with the theorem then there is no point trying to explain anything to you



It just depends how you configure your reconstruction filter as to how the audio will sound.

exactly. and i can configure it so there is never any alleged intersample peaks occuring


For foldback distortion to occur at the output stage of your converter, the cut-off frequency of the low-pass filters would have to be higher than 50% of the sampling rate of the digital audio being sent to your converter!

you ignore other issues such as the input stage where the digital is already folded.

What are you talking about? There is no "infinite bandwidth" in the digital domain.

if you clip your signal to look like a square wave and it does have infinite bandwidth and when you do the d/a you will see it.

I don't care whether your signals are square waves or whether you "merrily clipped and compressed them with non-linear processing", your bandwidth is limited by the number of samples per second

that is just total bull. you do NOT KNOW SQUAT ABOUT NYQUIST OR FOURIER.

you use to represent your audio. Even in the box, this is not infinite - plugs still operate at particular rates, even with oversampling. I will say it again: if your sampling rate was infinite you would have no CPU power left, so there's your proof that it doesn't happen. It is impossible.

i have no idea what you are talking about. and neither do you. nobody has an infinite sample rate. nobody said anybody was using that rate.

Please don't deflect this to bit depth - it has nothing to do with resolution.

deflect what to bit depth? bit depth is irrlevant to the problem at hand. you get better quality with more bits. but fewer or more bit depth is irrelevant to the original issue.

You are good at making strawmen though.

i am good at analysis and do not parrot internet whizdumb.

Check out the maths behind sample rate conversion.

been there done that. proved the theorems. solved the problems. all in my graduate classes.


But that's the point isn't it - clock jitter etc. does matter.

i always said that clock jitter was key to being able to use higher sample rates. but that has nothing to do with the bogus claims of intersample peak problems.

It has nothing to do with nyquist. Nobody is confusing analogue and digital domains.

it has everything to do with understanding and correctly using nyquist. everybody claiming there are intersample peaks confuses the two domains.

Intersample peaks are a real phenomenon, testable and provable.

not for the reasons people claim.
there is NO INHERENT INTERSAMPLE PEAK PROBLEM.
The only problem is badly designed d/a
OR people who violate nyquist and then wonder why they get results they dont expect.

They have to do with the difference between adjacent samples when their interpolated output exceeds full-scale.

nonsense. dont interpolate if that causes a problem.

If there is enough headroom in the analogue pathway, the peaks will pass un-clipped.

a valid nyquist signal in digital domain will always recreate without clipping in the analog domain unless the d/a was badly designed.

If they exceed the headroom they will clip the analogue circuit because there is not enough voltage to represent those peaks.

wtf. you are clueless wrt proper design then.

The peaks are very real when translated out of the digital domain in to real-world voltages.

of course there are peaks. that is nyquist in action. if you sample an analog signal most samples will be below the peak. so when you go back you get exactly the origianl signal AND YOU GET IT WITHOUT CLIPPING.

They can be predicted mathematically by metering plugins.

rotflamo. using erroneous tools to prove an erroneous claim.

Intersample peaks are not the fault of shoddy manufacturing. Please understand this. Every analogue circuit has a certain amount of headroom. You may as well say that a Neve has shoddy design because it only has +26dbu headroom. The analogue part of a D/A will enevitably crap out at some voltage level, that's just the way it is.

not shoddy manufacturing. ONLY IF THE DESIGN WAS ERRONEOUS or you violate nyquist in the digital domain.
a properly designed d/a will never clip in the analog domain IF THE DIGITAL SIGNAL WAS NYQUIST COMPLIANT.

You keep going on about "it's just nyquist in action". Do you really know what are talking about when you say that?

absolutely. but the folks makign erroneous claims about intersampl peaks do not fully understand ALL of the theorem.

I don't think you do. Intersample peaks have nothing to do with foldback distortion caused by exceeding the 'nyquist frequency'. Intersample peaks are a real and provable phenomenon.

LET ME REPEAT IT AGAIN
THERE ARE WAYS FOR PEAKS TO OCCUR AND THEY ARE:
BAD D/A DESIGN
INVALID NYQUIST SIGNAL IN DIGITAL DOMAIN

THERE IS *NO* INHERENT PEAK PROBLEM THAT HAS TO OCCUR.

No, it seems YOU don't understand nyquist theorum. Most others seem to know what they are talking about. I don't think you do.

take it to the bank. i know what i am talking about.
unless they changed nyquist and fourier since i did my graduate work.

Again, intersample peaks have nothing to do with choice of sample rate and bit depth, nor do they have anything to do with an engineer's choice of bus compression or limiting.

never said it did. dont know where you keep coming up with these things that nobody ever mentioned as if they were important to the discussion.

however compression and limiting CAN CAUSE PEAK PROBLEMS IN THE ANALOG DOMAIN.
not going to tell you how it can be avoided.
that is an exercise for the student.

see bold face above
Old 9th September 2011
  #49
Lives for gear
 
doug hazelrigg's Avatar
Quote:
Originally Posted by oldeanalogueguy View Post
i mentioned the physical limitations.
but those limitations depend on the quality of the design and the build. just being bigger does not mean it has lost quality (yet).
True, quality implementation is important. But again, after a certain point, it's not a matter of design or the current state of technology, it's a matter of the laws of physics. One example is that as a sampler operates faster and faster, it generates more and more heat, which is noise. It's a physical limitation that can't be surmounted, ever.


Quote:
Originally Posted by oldeanalogueguy View Post

88.2 to 44.1 is NOT a myth
you can divide by any power of 2 with no extra artifacts at all
only lose the natural resolution by lowering it

48 or 96 to 44.1 will cause noticeable problems
192 to 44.1 needs golden ears to hear the artifacts
It's a myth, my friend. There are no artifacts introduced. Don't go by my word, go by actual convertor designers like Dan Lavry and Nikka Aldrich.

And again, sample rate is only tangentially related to resolution.

Quote:
you can disbelieve the AES results that showed 192 was better than 88/96 was better than 44.1/16 was better than going to 24bit. i will use the highest rate that the specific component's clock jitter allows to work better.
I've read all manner of papers... the one that interests me is the one from last year's AES, which tested 88.2, 88.2 downsampled to 44.1, and 44.1, and found that overall they COULD perceive the difference between files recorded at 88.2 and the other two. I don't dispute this and never did (maybe you missed where I said I use 88.2). Somewhere on this forum I posted some stats showing that the potential error in computing peak values at various freq's DOES get very small up at a fs of 192. I don't use that rate due to storage space issues, but I wouldn't criticize anyone that does use it.
Old 9th September 2011
  #50
Lives for gear
 

Quote:
Originally Posted by ArnieInTheSky View Post
You're losing half your information when you dither it down. Look! You went from 88.2 to 44.1. See... half.

you do not dither when you reduce sample rates.
you should dither when you decrease resolution.
of course you lose resolution going from 88.2 to 441
BUT you do not cause any other problems.

If you sample at 768 and dither down to 44.1, you're only seeing about one twentieth of what it was you recorded.

that is not dithering
that is downrezzing.

You're right, it's the D to A conversion that doesn't allow us to hear it but why buy this thing if there's no way to convert it? There's no medium to listen to it on. Besides, most people are looking at getting AIFF's for their iPods or listen to their CD's in the car with 80db of road noise howling at their windows. What super sonic sounds am I gonna hear listening on those formats?

you do not hear supesonic sounds.
you do hear audio sounds with more fidelity .


If I'm gonna sit and really listen to tunes, I flip up the dust cover on my record player and throw back some Innis and Gunn. Bliss.

But hey man.... your money.

i am not going to buy it
my point was that higher sampling rates = better quality
while many people here claim that 2x is plenty.


Go buy/make whatever gear you want that inspires you to make the best music/sounds you can. I don't have a sig but if I did it would read:

"I'll use a Neve, or a Peavy, or ****ing Ham-sandwich if it works!"

and many people use things when it doesnt work.
and then blame it on some bogus problem like
the alleged intersample peak problem.
see bold text
Old 9th September 2011
  #51
Lives for gear
 
doug hazelrigg's Avatar
Quote:
Originally Posted by oldeanalogueguy View Post
huh ???
steep aliasign filters should just make the signal more nyquist compliant and should have less effect than less sharp cutoffs.
100% wrong, sorry
Old 9th September 2011
  #52
Lives for gear
 

Quote:
Originally Posted by kats View Post
The argument is that distortions at certain frequencies affect the sound of other frequencies above and below it. So having a cut off filter at 22k can create distortions within hearing range. A higher sample rate moves the filter and it's associated distortions further away from the purported human hearing range making it less noticeable. That's how the story goes anyhow.

now we are back in the world of real implementation
versus the theory
but filters are not causing the problems people are complaining about. so assume perfect filters for this discussion. and perfect resolution and linearity and jitter free clocks.

One thing that I think people have a hard time grasping is that, while Nyquist, the theorem stands and is not really open to debate, no one has been able to implement it.

absolutely !!
what we are discussing is a NON problem when the perfect a/d/a is performed, and which normal reality considerations like rez, jitter, linearity, etc. are not causing either.


That is, a perfectly band limited signal can be perfectly reconstructed... Yeah well, we can not perfectly band limit a signal.

nothign is perfect. engineering works on approximatiosn.
some things are close enough for govt work. some things do matter.

however none of these things cause the alleged intersample peak problem

that peak problem is caused only by either
bad d/a design
invalid nyquist signals in the digital domain that are sent to a d/a circuit whilst expecting nyquist type results.

see text in bold
Old 9th September 2011
  #53
Gear Maniac
 
ArnieInTheSky's Avatar
 

Quote:
Originally Posted by oldeanalogueguy View Post
see bold text

Last edited by ArnieInTheSky; 9th September 2011 at 05:19 AM.. Reason: This is more like my response.
Old 9th September 2011
  #54
Lives for gear
 

Quote:
Originally Posted by doug hazelrigg View Post
100% wrong, sorry

sorry but 100000000000000000000% right.

that is nyquist theorem - perfect filters and perfect results.

dont confuse that with reality and engineering approximations.
Old 9th September 2011
  #55
Lives for gear
 
doug hazelrigg's Avatar
Quote:
Originally Posted by oldeanalogueguy View Post
sorry but 100000000000000000000% right.

that is nyquist theorem - perfect filters and perfect results.

dont confuse that with reality and engineering approximations.
Er, right... I think

There are many differences between Nyquist-Shannon theory and the actual design and construction of an ADC, and filter design is perhaps the biggest. My reading of the post I responded to is that YOU were essentially saying that a steep filter actually is preferable in an ADC, and THAT assertion is 100% totally and indisputably wrong wrong wrong. And quite basic, too, I might add. 50% of the argument for using higher sampling rates is about using gentle filter slopes in the passband!
Old 9th September 2011
  #56
Lives for gear
 
filipv's Avatar
Quote:
Originally Posted by oldeanalogueguy View Post
you been sniffing the glue in your green magic marker ?

there is NO problem going from 88.2 to 44.1
you lose resolution
but there are NO other artifacts or issues

if you believe that internet wizdumb then it is time for you to start thinking for yourself
No matter what sample rates you want to convert from-to, you can always upsample to the lowest common denominator, right?
Old 9th September 2011
  #57
oldanalogueguy - half of what you post doesn't make sense (either grammatically or from a technical viewpoint). If you'd like to communicate your thoughts, please do so, but at the moment it's just random references to technical words.

The rest of what you post is factually incorrect. Intersample peaks, as bgrotto points out, ARE fact, they ARE what happens in digital audio, they're NOT necessarily a problem, and they ARE totally irrelevant here!

Likewise, your whole spiel about Nyquist....er, what?! I didn't get one modicum of sense out of your last 5 posts....and my degree was partly in this subject, so don't trot out the " if you don't understand, I can't be bothered to explain" argument - I understand this sh1t, I (like many others it seems) just don't have a clue what you're trying to say!

The more you post, the more it just seems to be that you don't understand digital audio AT ALL! Stick to analogue maybe?! At least talking about it...

Plus the whole 88.2 to 44.1 "simple division" - much cleverer people than either of us have repeatedly stated that this is a myth and with modern day SRC it doesn't matter. IF you have proof to the contrary, please provide it!

Actually, I wonder...there was a guy interviewed in Resolution Magazine ybe April 2010 that swore his method of SRC was more transparent...where to convert 48 to 44.1 he'd upsample to 88.2 then use a bitsplitter...Tony Faulkner was the guy's name...wasn't you was it?! You're obviously not American....
Old 9th September 2011
  #58
Quote:
Originally Posted by TurboJets View Post
Well that's just it though, isn't it?

You're "not sure".

No one is really "sure".

That's one of the beauties of science for those who can appreciate it...there is still so much we don't know.

Doesn't matter really. I just thought it was an interesting point. Just because you can't see a thing doesn't mean it doesn't exist or have a physical effect.

However, it can be agreed that 768kHz does exist as a measurable frequency. As a measurable cyclic rate. When it's vibration is introduced to the human body, who knows what the effect is? As all things vibrate, surely it stands to reason there must be some influence or effect on things in its proximity or environment whether animate or inanimate.

Who would have thought 1,000 years ago that sound (or ultrasound) could help break up a kidney stone? Right?

We may not be able to hear ultrasound but it does affect the human body.
No I agree I'm not sure.

Thing is, neither is anyone else. Neither has anyone else been able to provide any sort of supporting evidence why it should be the case.

The moon isn't a great example for this, since you can't exactly stop it's motion and thus test the results, but you can do gravitational experiments on a smaller scale that demonstrate the same effects.

With all this crazy conversion talk, you can sit down and do blind tests to PROVE that things make a difference. Thing is, none of the believers ever will - because they can "hear the difference - it's like night and day man!".

Well, some things are, some things aren't. And if it's so obvious, a blind test should be easy to prove it right? but rarely does anyone do this, and when they do, they fail it. So excuse the scepticism - I'm open when it comes to science, not understanding everything in the world, etc - but I'm not open to wasting time and money if there's not an audible effect, just because there "might" be something not being captured! if it can't be discerned in a blind test, it's not there as far as our ears are concerned.

I also understand my brain never works quite as I think it should....something that again audiophiles will never want to believe.
Old 9th September 2011
  #59
Lives for gear
 

Quote:
Originally Posted by psycho_monkey View Post
I (like many others it seems) just don't have a clue what you're trying to say!
That's a good summary!
Old 9th September 2011
  #60
Lives for gear
 

And yet it's the processing of plugins at higher sample rates that yeilds better audio results.

I can't hear past 16khz but I can hear the improved quality of eq's and other plugins at 88/96khz over 44khz.

Could I tell you if a .wav was recorded at source either at 44khz or 96khz probably not... but add some plugins and the difference becomes clearer.
Post Reply

Welcome to the Gearslutz Pro Audio Community!

Registration benefits include:
  • The ability to reply to and create new discussions
  • Access to members-only giveaways & competitions
  • Interact with VIP industry experts in our guest Q&As
  • Access to members-only sub forum discussions
  • Access to members-only Chat Room
  • Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD $20/year
  • Promote your eBay auctions and Reverb.com listings for free
  • Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.


 
 
Slide to join now Processing…
Thread Tools
Search this Thread
Search this Thread:

Advanced Search
Forum Jump
Forum Jump