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 12th October 2011
  #871
Lives for gear
 

Quote:
Originally Posted by swim View Post
I don't really think this requires explanation, as it is pretty much common knowledge.

You gain about 6dB for each 4x increase, and that translates to effectively 1 bit.
1 bit of what and where?

Since you don't answer my questions I can only draw the conclusions that you don't understand this.

I don't need an explanation but to move the discussion forward I need YOU to give an explanation. Simply to show where you are wrong.


Quote:
You are simply wrong about that, and, with all due respect, I do not have a duty or a desire to go 'round and 'round with you on something that you are apparently unwilling to accept, or incapable of understanding.
Oh my! Since you don't offer any substance or arguments what so ever but seem to feed on rudeness and personal attacks I will not spend much more time on you either.


Quote:
Just as an aside, I believe that it would be helpful to your ability to communicate on these forums if you would try to understand the difference between plural and singular nouns and verbs. I'm not trying to be snarky, but you are trying to combatively argue technical topics in an English speaking forum on a frequent basis, so some effort in that area should probably not seem too much trouble for you.
As you seem to understand english is not my native language but I feel I have no problems at all communicating technical issues. It's part of my work and no one complains there. During the ten years or slightly more that I have been participating in various forums I have never had a problem. You are the first that ever complains so maybe this speaks more about you than me?

Something makes me feel that you won't last long under your new moniker, Admins are informed about your trolling, personal attacks and non-constructive posting.


Quote:
As a threshold issue, with regard to your statement, note that the ratio of noise to signal is better for larger signals.
But of course it is.. but that does not matter becasue the noise in modern converters is inaudible. Most listners doesn't even have playback systems or program material that makes the noise audible in well produced 16bit audio.

Quote:
The noise does not set the limit of accuracy or of precision.
Yes, in modern well designed converters the analog noise IS the factor that makes the signal deviates most from "perfect". Lowering quantization noise makes no difference when it is already 20-30dB below the analog noise and the analog noise is well below the threshold of audibility.

Quote:
Frankly, I think you are rather unclear on this entire subject. But, again, I am not the guy who is going to spend page after page explaining things to you just because you have nothing else to do.
It's very obvious just by reading this thread that you need to take a look at yourself. It's there for everyone to see. I have a lot to do actually and will not try to help you understand much more than I already have.


Quote:
That's an absurdity.
Explain to me what is absured about random noise with power well above the power of the quantization error effectivelly shifts harmonically related spectral components into noise. That is what dithering is all about. The lower the level of the noise in relation to the quantization error the more important the spectrum and type of noise.

Get into the game. You're all attitude and no arguments or substance.


Quote:
Sure, man. Whether you realize it or not [you probably don't], what you, and, less glaringly, some others, are actually arguing in favor of is about 12 bit sampling [for contemporary stuff anyway].
No, I'm not arguing for 12bit sampling. Since I never done that I don't see how you can make up such lies and put words in my mouth again.


Quote:
Again with all due respect, I don't really care about your stance or opinion on these matters.
So it's all about the heat and rudenes? You don't care about the actual discussion and moving forward? Since you still spend time "discussing" with me obviously you are a troll.


Quote:
I don't know if it would be meaningful for you. I think it would be meaningful for you to learn more about this topic so that you would not be spreading confusion.
I'm all for learning but how can I learn when you offer no explanations or constructive arguments?

You see in contrast to you, learning and sharing knowledge is why I'm here.

Quote:
Second of all, just as something for you to think about, if you like:

An MRI machine can have the effect of rearranging all of the atoms in a person's body, yet that person may not consciously perceive it while it is happening.
And?


Quote:
IMO, I honestly don't think you are in a position to be changing textbooks.
First of all you should understand that I talk about "we" "us" as in the scientifical and audio engineering community. If someone with golden ears sit in their studio and find out something that is missing in our global understanding of physcis and human perception, then I think "we" should bring it out so everybody can learn from it.

I'm sure you'll find a way twisting that (and most of what I wrote) to something negative.


/Peter
Old 12th October 2011
  #872
Gear Maniac
 

So, I´m kind of some pages late and hadn´t read all the thread, but it strikes me as odd that an "Old Analogue Guy" has studyed with professors 40-50 years ago when there were more professors without PhD and at that time they had lab proof that 192kHz is better than 96kHz....
Old 12th October 2011
  #873
Lives for gear
 
jupiter8's Avatar
 

Quote:
Originally Posted by swim View Post
I don't really think this requires explanation, as it is pretty much common knowledge.

You gain about 6dB for each 4x increase, and that translates to effectively 1 bit.
Yes but AFAIK ever AD converter samples at a fixed rate and then downsamples to the target rate. Does this not mean that 96 kHz is more accurate according to you and thereby contradicting yourself ?
Old 12th October 2011
  #874
Lives for gear
 
doug hazelrigg's Avatar
Quote:
Originally Posted by oldeanalogueguy View Post
as i noted
it is the ENGINEERING APPLICATION of the theorem

nonsense
the math theorem says nothing about practical realisation using circuits
Agreed.

Quote:
within the error limits of voltage resolution
and clock jitter then higher bit depth and higher sample rates do improve the a/d/a in actual practice
And I agreed with this, just not for the reason you were citing. Leaving aside these issues, a band-limited signal can be reproduced just as accurately using 2B, 4B, 8B, etc.

I think we're actually on the same page here if you think about it a bit
Old 12th October 2011
  #875
Lives for gear
 
doug hazelrigg's Avatar
Quote:
Originally Posted by Audiop View Post
But the theroem says that a bandwith limited signal will be reproduced.

It does not say that it will be reprodcued with squashed peaks that happened to fall between two samples.

Intersample peaks anyone?


/Peter
I'm not referring to intersample peaks at all. A signal can be accurately reproduced in terms of frequency, but, depending on the sample rate, there will be varying degrees of potential error in calculating peak values. Here's the math (attribution: forum member pj geerlings):



A -0.5dB step is approximately 0.9441 ( from "1.0" ) -so- the minimum sample rate of a sine wave which guarantees no greater than a 0.5dB error is given by the following equation ( values rounded ) :

360 / (2 * acos(0.9441)) = 9.3483

Briefly:

1) acos(0.9441) means "at what angle (in degrees) is the value of the cosine effectively 0.5 dB down from full scale?"

2) multiply that angle by 2 because the cosine is symmetric around "0 degrees"

3) there are 360 degrees in a complete cycle

so at 44K1 we have 4717.75Hz
and at 48K we have 5134.64Hz

-----------------

The implications for frequency limits are fairly straightforward then; frequencies above those given above will be subject to continually worse peak value estimates (that is to say, as frequency increases the naively computed peak value will be more and more imprecise)

But maybe 0.5 dB is little too restrictive - if we relax the error to 1dB then we have:

360 / (2 * acos(0.8913)) = 6.6743

@44K1 = 6607.44
@48K = 7191.77

Hey, not much of an improvement - so finally, if we relax the allowed error even further - to 3dB - then we have:

360 / (2 * acos(0.7079)) = 4.0060

@44K1 = 11008.49
@48K = 11982.03

So, for example, at a sample rate of 48kHz, there is a (roughly) 3dB potential error when calculating the peak value of a 12kHz signal. That is probably audible, especially when doing some dynamics processing, etc.
Old 12th October 2011
  #876
Lives for gear
 
TurboJets's Avatar
Quote:
Originally Posted by swim View Post
Even if you could prove the absurd theory that a certain method's impact could never engender, or contribute to, an "audible" result [which you could never truthfully prove], it remains that it is simply not necessary for something to be audible for it to desirable, or to be of benefit.
Just wanted to cut in to agree with this and add that's it's really interesting how many of these arguments eventually come down to admitting they can see errors or differences in screen shots, or that yes, cakewalk handles anti-aliasing differently than cool edit or pro tools, etc. And even though brick-wall filters are constantly inserted into these arguments, we all know there's really no such thing as a "brick wall" filter and we all know they're not even something we want in the circuit because of phase issues.

Anyway, so when it's finally admitted there are verifiable errors and differences, the inevitable phrase that follows is always "but it's doubtful that it's audible". What a crock. A fella says he hears a difference, eventually the scientifical believer's admit there's a calculable difference and then cry "but you can't hear it". What?

Plus, the word "audible" bothers me as well. Listening is not just hearing, it's an experience. the human body physically feels sound as well and that adds to the listening experience. Sound at movies, concerts, live acoustic performances, etc. are a total human physical experience, not just an "audible" experience. Frequencies well below and well above the audible spectrum are felt by the body and the body translates those energies into information that adds to the overall experience.

Again, it's just so difficult to converse with any believer about things like this, whether you're discussing converters, sample rates, bit depth, or religion...believers just won't look beyond what's in "The Book".
Old 12th October 2011
  #877
Lives for gear
 

Quote:
Originally Posted by TurboJets View Post
And even though brick-wall filters are constantly inserted into these arguments, we all know there's really no such thing as a "brick wall" filter and we all know they're not even something we want in the circuit because of phase issues.
Therefore use linear phase filters and place them above the audible spectrum. While most people won't hear well implemented 44.1kS/s if we go with 48kS/s or say 50-100kS/s we are home. Phase relation of inaudible frequencies does not matter.

Quote:
Anyway, so when it's finally admitted there are verifiable errors and differences, the inevitable phrase that follows is always "but it's doubtful that it's audible". What a crock. A fella says he hears a difference, eventually the scientifical believer's admit there's a calculable difference and then cry "but you can't hear it". What?
But if it's verified by listening tests it is verified. I don't get you here. If it's verified it's verified. Measurments are a guide and a tool but obviously you need to test humans to learn what humans hear. Pretty basic I'd say. :-)

If there are audible effects/differences it is interesting to learn why.. By using measurements we can learn what parameters and thresholds which are important.

Quote:
Plus, the word "audible" bothers me as well. Listening is not just hearing, it's an experience. the human body physically feels sound as well and that adds to the listening experience. Sound at movies, concerts, live acoustic performances, etc. are a total human physical experience, not just an "audible" experience. Frequencies well below and well above the audible spectrum are felt by the body and the body translates those energies into information that adds to the overall experience.
Yes of course we feel sound and obviously this are also "measured" by the human performing the listening test.. unless using headphones or mediocre speakers with limited bandwith and dynamic range.

We do not feel ultra sound from music though. Infra, yes, ultra no!

Quote:
Again, it's just so difficult to converse with any believer about things like this, whether you're discussing converters, sample rates, bit depth, or religion...believers just won't look beyond what's in "The Book".
Now who's the believer and who's not? :-)

The believer is the one that believes. After performing some controlled tests we learn and know.


/Peter
Old 12th October 2011
  #878
Lives for gear
 
jupiter8's Avatar
 

Quote:
Originally Posted by swim View Post
I don't understand your question. Maybe try to rephrase it?

The fact that almost all modern converters used in audio are sigma delta oversampling is often ignored in these types of discussions, and it does render a lot of the argument rather irrelevant, or moot.

However, the eventual system rate [i.e. what you call "target rate"] is what it is, and the concept of 6dB "improvement" per 4x increase [over base rate] just has to do with how many samples per second are ultimately representing the waveform for reconstruction. A 192kHz system rate is still 4x 48kHz, even if it was arrived at by decimation of a still higher rate.

I don't know if that was responsive to your question or not, but hopefully so.
The initial sampling is done at a very high fixed rate with just a few bits. You gain bits by decimating. Thereby 96 kHz should be more accurate than 192 since it is one more decimating step, hence you contradict yourself when you say 192 kHz is more accurate.
Old 13th October 2011
  #879
Lives for gear
 

Quote:
Originally Posted by Audiop View Post
But the theroem says that a bandwith limited signal will be reproduced.

It does not say that it will be reprodcued with squashed peaks that happened to fall between two samples.

Intersample peaks anyone?


/Peter
there are NO INTERSAMPLE PEAKS ANYWHERE !!!!!!!!!!!

in the digital domain there is NOTHING BETWEEN SAMPLES
hence no intersample peaks or anything but nothingness

in the analog domain there are NO SAMPLES AT ALL hence no intersample peaks there either

one paper has stated it better as analog values between two samples that are higher than the reproduced values of the two samples are

and this is NEVER A PROBLEM as that is exactly what nyquist said would happen. most samples are below the peak analog signal values so there has to be analog values WHEN RECREATED PROPERLY that will be higher than many of the digital samples recreated analog values woudl be.
Old 13th October 2011
  #880
Lives for gear
 

oh come on ... we've already been though this ... let it go
Old 13th October 2011
  #881
Lives for gear
 
doug hazelrigg's Avatar
Quote:
Originally Posted by doug hazelrigg View Post

Leaving aside these issues, a band-limited signal can be reproduced just as accurately using 2B, 4B, 8B, etc.
Quote:
Originally Posted by swim View Post
No. There will be less quantization error in the audio band for the higher sample rate. The power of the quantization error is the same for all sample rates, but it is spread over a larger bandwidth in the faster sample rate.
Right -- note I said "leaving aside these other issues" which was specifically referring to quantization error. We're on the same page here (in fact, it appears you and I are the only two who agree on this point)
Old 13th October 2011
  #882
Lives for gear
 

Quote:
Originally Posted by doug hazelrigg View Post
Right -- note I said "leaving aside these other issues" which was specifically referring to quantization error. We're on the same page here (in fact, it appears you and I are the only two who agree on this point)
you cant leave aside the other issues with real gear
that is the real issue

the mathematics is proven
its real life that needs analysis

and the truth is that within voltage resolution and clock jitter limitations higher sampling rates do give better a/d/a results
Old 13th October 2011
  #883
Lives for gear
 

Quote:
Originally Posted by oldeanalogueguy View Post
there are NO INTERSAMPLE PEAKS ANYWHERE !!!!!!!!!!!
Yes there are, if you only accept the terminology used by the majority.

Quote:
one paper has stated it better as analog values between two samples that are higher than the reproduced values of the two samples are
Now you contradict yourself but I think you'll have to accept that most people find it easier to just write... (here it comes).... intersample peaks.

Quote:
and this is NEVER A PROBLEM as that is exactly what nyquist said would happen. most samples are below the peak analog signal values so there has to be analog values WHEN RECREATED PROPERLY that will be higher than many of the digital samples recreated analog values woudl be.
Don't tell that to me, another poster brought up what he think is a problem with correct representation of analog peak levels and I questioned it and reminded him about the theorem.

Yelling will not make people stop using accepted terminology. :-)


/Peter
Old 13th October 2011
  #884
Gear Nut
 

Quote:
Originally Posted by Audiop View Post
Yes there are, if you only accept the terminology used by the majority.

Now you contradict yourself but I think you'll have to accept that most people find it easier to just write... (here it comes).... intersample peaks.

Don't tell that to me, another poster brought up what he think is a problem with correct representation of analog peak levels and I questioned it and reminded him about the theorem.

Yelling will not make people stop using accepted terminology. :-)


/Peter
Now this is some jiggy nerdiness coming down the interspunk.

Go on..
Old 13th October 2011
  #885
Lives for gear
 

Quote:
Originally Posted by doug hazelrigg View Post
I'm not referring to intersample peaks at all. A signal can be accurately reproduced in terms of frequency, but, depending on the sample rate, there will be varying degrees of potential error in calculating peak values.
Hi!

You are contrdicting yourself. If a sine is "reproduced" with 3dB cuts of peaks that frequency is not reproduced anymore but heavily processed. This is gross distortion and would be easily heard and measured.

Your math friend is not correct and if he was I would expect to see him in Stockholm soon.


/Peter
Old 13th October 2011
  #886
Lives for gear
 
doug hazelrigg's Avatar
Quote:
Originally Posted by Audiop View Post
Hi!

You are contrdicting yourself. If a sine is "reproduced" with 3dB cuts of peaks that frequency is not reproduced anymore but heavily processed. This is gross distortion and would be easily heard and measured.

Your math friend is not correct and if he was I would expect to see him in Stockholm soon.


/Peter
Not sure what you mean... a 12kHz signal is STILL a 12kHz signal whether the digital data represents a +/- 3dB error or not (and it's a potential error, not an absolute error). The math is solid because it's the same math that Nyquist-Shannon uses to assert that fs>2B will allow perfect (in theory) reconstruction of a sine wave. (FWIW, the same rubric can be applied to RMS values, too)
Old 13th October 2011
  #887
Lives for gear
 

Doug, if you check you math/source you'll see that it's false for a signal that is bandwith limited. Of course you'll get errors with square waves which goes without saying.

Also it was you who brought up:
Quote:
so at 44K1 we have 4717.75Hz
and at 48K we have 5134.64Hz
Which I read as sines.

So yes, square waves with energy above the Nyquist frequency will be better reproduced with higher sample rate. Sines/bandwith limited signals below Nyquist=total absence of the errors you talk about.


/Peter
Old 13th October 2011
  #888
Lives for gear
 
TurboJets's Avatar
Quote:
Originally Posted by oldeanalogueguy View Post
you cant leave aside the other issues with real gear
that is the real issue

the mathematics is proven
its real life that needs analysis

and the truth is that within voltage resolution and clock jitter limitations higher sampling rates do give better a/d/a results

Whether or not people want to hear this, OAG is correct here. However, I don't necessarily agree that the mathematics is proven. The PhD's in the thread are going to freak out about that statement, but, whatever.

But, it's Nyquist that's being cited, pursued by "Shannon mathematically proved Nyquist was right". But Shannon used Boole's math to prove Nyquist was right and quite a number of people (yes including PhD's of course) feel Boole's was wrong. Boolean logic seems to fall apart in the face of quantum reality.
As far as Nyquist needing analysis in real life, well, that's already been done and to me it seems very clear that higher sample rates (like 96kHz) compensates for some errors that occur when sampling at 44.1.

So while 44.1 is adequate (and of course it IS adequate), capture and reconstruction is improved when capture is performed at 96kHz. I just don't see any way around that.

OAG is obviously right about voltage resolution as well.

Anyway...

One PhD says "yes", another says "no"; and they both say (as Norcoman said) "I'm a doctor, trust me". Very funny, also quite boring really.

I was mixing a new song yesterday and thinking how much more fun it is to be in the studio recording and mixing than it is to have your nose buried in some book or research paper. But then again, to each their own, right? To each their own.
Old 13th October 2011
  #889
Quote:
Originally Posted by swim View Post
Track count is currently the obvious main reason why 192kHz is not used too much. Today's productions usually exceed the track count available with 192kHz, unless you're willing to sync more than one rig together, which is impractical for most.

96kHz is much more common.
I don't know that anyone works at 192kHz in the popular music world. Maybe in the audiophile world. Anything you've ever heard on the radio was 96k or below. I've never seen anyone work at 192k, apart from one archiving session perhaps.
Old 13th October 2011
  #890
Lives for gear
 
dcollins's Avatar
Quote:
Originally Posted by TurboJets View Post
I was mixing a new song yesterday and thinking how much more fun it is to be in the studio recording and mixing than it is to have your nose buried in some book or research paper. But then again, to each their own, right? To each their own.
True, but remember that if it wasn't for those pointy-headed beard-scratching Ph.D's you would have nothing in the way of high-tech.

Or even low-tech.


DC
Old 13th October 2011
  #891
Lives for gear
 
dcollins's Avatar
Quote:
Originally Posted by Audiop View Post
Yelling will not make people stop using accepted terminology. :-)
Or things that are trivial to show actually exist. Like inter-sample peaks.



DC
Old 13th October 2011
  #892
Lives for gear
 

Quote:
Originally Posted by oldeanalogueguy View Post

and the truth is that within voltage resolution and clock jitter limitations higher sampling rates do give better a/d/a results
How do you define better in this case?


/Peter
Old 13th October 2011
  #893
Lives for gear
 
TurboJets's Avatar
Quote:
Originally Posted by Audiop View Post
How do you define better in this case?


/Peter

Maybe he means something like a more accurate representation of the content of the original real-world signal? In simple terms.
Old 13th October 2011
  #894
Lives for gear
 
DaveUK's Avatar
Not better just different ....
Old 14th October 2011
  #895
Lives for gear
 

Quote:
Originally Posted by Audiop View Post
Yes there are, if you only accept the terminology used by the majority.

Now you contradict yourself but I think you'll have to accept that most people find it easier to just write... (here it comes).... intersample peaks.

Don't tell that to me, another poster brought up what he think is a problem with correct representation of analog peak levels and I questioned it and reminded him about the theorem.

Yelling will not make people stop using accepted terminology. :-)
/Peter
i do not accept anything that is wrong, irrational, andor illogical no matter how many idiots misuse or abuse internet whizdumb
Old 14th October 2011
  #896
Moderator
 
narcoman's Avatar
 

Quote:
Originally Posted by oldeanalogueguy View Post
i do not accept anything that is wrong, irrational, andor illogical no matter how many idiots misuse or abuse internet whizdumb
<response left intentionally blank>
Old 14th October 2011
  #897
Lives for gear
 

Quote:
Originally Posted by Audiop View Post
How do you define better in this case?


/Peter
minimum least squared error
when running a signal through the a/d/a chain
and comparing with the original
Old 14th October 2011
  #898
Lives for gear
 

Quote:
Originally Posted by dcollins View Post
Or things that are trivial to show actually exist. Like inter-sample peaks.
DC
the only way intersample peaks exist
is if people used some bad weed left over from the clinton administration or are totally mentally challenged and can only parrot internet whizdumb and not think for themselves
Old 14th October 2011
  #899
Lives for gear
 

Quote:
Originally Posted by TurboJets View Post
Whether or not people want to hear this, OAG is correct here. However, I don't necessarily agree that the mathematics is proven. The PhD's in the thread are going to freak out about that statement, but, whatever.

But, it's Nyquist that's being cited, pursued by "Shannon mathematically proved Nyquist was right". But Shannon used Boole's math to prove Nyquist was right and quite a number of people (yes including PhD's of course) feel Boole's was wrong. Boolean logic seems to fall apart in the face of quantum reality.
As far as Nyquist needing analysis in real life, well, that's already been done and to me it seems very clear that higher sample rates (like 96kHz) compensates for some errors that occur when sampling at 44.1.

So while 44.1 is adequate (and of course it IS adequate), capture and reconstruction is improved when capture is performed at 96kHz. I just don't see any way around that.

OAG is obviously right about voltage resolution as well.

Anyway...

One PhD says "yes", another says "no"; and they both say (as Norcoman said) "I'm a doctor, trust me". Very funny, also quite boring really.

I was mixing a new song yesterday and thinking how much more fun it is to be in the studio recording and mixing than it is to have your nose buried in some book or research paper. But then again, to each their own, right? To each their own.
the math is proven
our math textbook did it
students did it
our engineering textbook (grad class) proved it
the students proved it

the theorem for nyquist and sampling is a proven fact

the real issue that we need to discuss are the
engineerng practicalities in approximating that theorem
with real world stuff that is "good enough"
and then see what we can do to make it better

the higher sample rate allows real components to provide a better a/d/a chain in the real world

i only know shannon from commuications theory and bandwidth snr channel capacity etc. relationships

our book only had nyquist for samplign theory (and the theorems that support it)

the a/d has problems with getting exact samples
there is voltage resolution/noise vertically
and clock jitter horizontally
so you get little boxes of where the signal might be
not clean dots where it absolutely is
as long as the boxes do not overlap
increasing bit depth and sample rate does improve the a/d/a

dont recall boole having anything to do with quantum physics
Old 14th October 2011
  #900
Moderator
 
narcoman's Avatar
 

I see the pickles are still here..... want one?
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