Benchmark talked about headroom for intersample peaks in DAC, does it really matter?

[cloggins]

If all the information is between the sampling points, then that would include the peaks of the sine waves would it not?

Um, yes. ALL non-clipped peaks of ALL waveforms within the Nyquist range will be reconstructed accurately, with no exceptions. None. There is no "lost information between samples" in a properly functioning Nyquist space.

 

[gregorio]

Um, yes.

Ah, now we're getting somewhere! The only person talking about lost information is you. It's because the information is reconstructed accurately that we have intersample peaks! That information about the sine wave peak which is between the sample points.

G

 

[cloggins]

sine wave peak which is between the sample points.

In a Nyquist space, there is no unexpected or un-measurable or un-reconstructable clip "between the sample points" (unless the original PCM is clipped). You don't understand Nyquist theory. What you're inferring is impossible in a proper Nyquist space.

This is true for program -1dBFS or -0.1dBFS, or whatever.

There was a graph on page 2 of this thread that shows two consecutive FS samples, with a big sine wave between the samples. This is a PCM clip, nothing more.

 

[gregorio]

[1] In a Nyquist space, there is no unexpected or un-measurable or un-reconstructable clip "between the sample points" (unless the original PCM is clipped).

My, you are having a bad (read: dense) day aren't you. Having a bad day is OK, we all have them BUT insulting others because you are being dense is NOT acceptable and additionally, it just makes you look like a fool. That's not really what you want is it?

Again, I've already told you a couple of times and others have already told you, we are NOT (that's bold, underlined and all caps!) talking about un-reconstuctible signals, we are talking about entirely reconstructible, legal signals!!!! Let's use simple logic and your OWN words: All the information is between the sample points, right? If all the information is between the sample points, that must include the peaks of the sine waves, right? The peak has the highest amplitude value of any part of the sine wave, that's why it's called the "peak", right? If the peak is between the sample points then it must have a higher amplitude value than the sample points! Getting it yet or is even this simplest of logic beyond you?

G

 

[cloggins]

If the peak is between the sample points then it must have a higher amplitude value than the sample points! Getting it yet or is even this simplest of logic beyond you?

Two consecutive FS MSB PCM samples is defined as a clip, and is beyond our Nyquist space. If your two sample points are both FS MSB, then you have a program clip feeding your DAC.

All non-clipped waveform peaks in a valid Nyquist space are represented correctly, with no "missed intersample overs."

ALL peaks, not some peaks, not 99.999% of peaks, ALL peaks. You'll have to show me a case where this isn't true. Please, use a graph or math, not hand-waving.

 

[castleofargh]

thank you for that. my attempt at making up a physical case with voltage limit, didn't help. but this is as explicit as it gets.

or not.

Incorrect, my friend.

ALL (not "most") of the information is "between the sampling points" and ALL of that information (in theory) can be perfectly reconstructed. There is no "special case" (in valid Nyquist space) where information is "lost" or "indeterminate" between sample points. You need to go back and drill this into your head. Say it over and over: In a properly designed Nyquist space, there is NO information (none, nada, zilch) that cannot be PRECISELY reconstructed in analog space. Go ahead, say it over and over until it sinks in. It's OK, you're among friends.

my first reply to you was saying that you were fighting the wrong fight. since you haven't tried to understand us a single time. we say bananas aren't round and you reply, "of course apples can be round", then look at us like we're idiots for talking about bananas in a banana topic. you're not wrong because you misunderstand Nyquist, you're wrong because you refuse(or fail) to consider the very specific situations we're talking about.
you could get a clue with how we call it intersample clipping and not reconstruction distortion, or loss of data, or whatever it is you think you're discussing. clipping is a clear problem with clear cause, we want to get our signal to an amplitude but the maximum amplitude available is lower. if we increase the gain too much, the signal get stuck to 0dB in the digital domain or at max amplitude in the analog domain. yes I'm explaining clipping to you, yes I know you know what clipping is, but apparently that's where we're at right now, getting you to look at the actual discussion.

from the start we're saying that it's basically a gain setting issue(like any clipping issue). it's a guy pushing the gain too high when normalizing the signal or whatever, but not seeing it's too high because he used sample peaks to visualize the signal peaks. we could sum it all up as a reading mistake leading to a small area where we have no digital loss of signal and yet can end up with a clipped signal.
so stop your nonsense about defending digital audio's honor against nobody, and start reading this topic as a topic about clipping a signal. or if you really can't do that, maybe stop posting because you're off topic.

 

[gregorio]

Two consecutive FS MSB PCM samples is defined as a clip, and is beyond our Nyquist space. If your two sample points are both FS MSB, then you have a program clip feeding your DAC. ... All non-clipped waveform peaks in a valid Nyquist space are represented correctly, with no "missed intersample overs." ... ALL peaks, not some peaks, not 99.999% of peaks, ALL peaks. You'll have to show me a case where this isn't true. Please, use a graph or math, not hand-waving.

So the answer to my question then is "no" you're not getting even that simplest of logic ... Oh dear, must be a really bad day!!! What was it that you didn't understand? You own words or what the peak of a sine wave is? If you don't understand your own words or what the peak of a sine wave means, you've got a lot more homework than I guessed!

One last try: Lets say we have a sine wave one octave below the Nyquist point (half the Nyquist freq, IE. A 11.025kHz sine wave). At 44.1kS/s that sine wave will have 4 sample points, let's say; one on the falling side of the trough "A" and one on the rising side "B", one on the rising side of the peak "C" and one on the falling side "D". Let's say that the value of C is -8dBFS and the value of D is say -9dBFS. The peak, which according to you is between all the sampling points (in this case between C and D) must be higher than C or D, let's say 2dB higher. No problem, when reconstructed our DAC can handle the reconstructed peak which is at -6dBFS. What happens though if we increase the amplitude of the sine wave (all the sample points) by 7dB? Sample point C is now a perfectly legal -1dBFS and sample point D is -2dBFS, also perfectly legal. When reconstructed, what's the value of the actual peak of this sine wave (which occurs between C and D)?

G

 

[cloggins]

What happens if we increase the amplitude of the sine wave (all the sample points) by 7dB?

What is the original peak value of the 11kHz sine wave?

 

[cloggins]

so stop your nonsense about defending digital audio's honor against nobody, and start reading this topic as a topic about clipping a signal.

When someone tells me that Nyquist, under rare circumstances, misses certain information "between the samples" -- they are simply wrong, and spreading really strange ideas. Where is Monte Montgomery when you need him?

Yes, this is about clipping.

No, the clipping has nothing to do with "missing information between samples."

 

[gregorio]

What is the original peak value of the 11kHz sine wave?

It doesn't matter, the peak value is always going to be higher than the sample values because the peak is between the samples, as you stated. The original value of the waveform could have been -6dBFS but that's irrelevant as virtually never is the recorded level of the waveform the level at which it's released, after mixing and mastering.

G

 

[cloggins]

the peak value is always going to be higher than the sample values

My friend, you don't understand sampling theory. There is no peak (none, nada) in valid Nyquist space that is "missed between samples" Yes, of course, there is often a maximum level higher than either of two adjacent samples, but that maximum level (peak) remains defined perfectly in a valid Nyquist reconstructional space (unless the original PCM signal is clipped). This is true 100% of the time (not 99.999%) for all Nyquist-band frequencies and levels. You have this strange idea that Nyquist sometimes "misses" certain peak information within its space. I assure you, It doesn't (unless the original PCM signal is clipped). Where did you pick up this idea? I'd like to see the book you're reading! Sounds like Internet folklore.

 

[gregorio]

My friend, you don't understand sampling theory. There is no peak (none, nada) in valid Nyquist space that is "missed between samples."

OK, you're clearly insane as you're arguing with yourself. You're arguing with yourself because you are the only one talking about "missed between the samples". As far as intersample peaks are concerned, you're obviously right and the AES, ITU, EBU and ATSC are all nuts for specifying something which does not exist. The problem with insanity is that the insane person thinks they are sane and everyone else is insane. I've tried but as you're obviously unable to follow simple logic based on your own words, there's really nowhere else to go. So I'll leave you to argue with yourself ....

G

 

[cloggins]

As far as intersample peaks are concerned, you're obviously right and the AES, ITU, EBU and ATSC are all nuts for specifying something which does not exist.

Intersample peaks are a real phenomenon, but they occur for reasons vastly different than what you have been selling here. There is also the issue of devices (digital audio meters, etc.) which look only at discrete samples, not at reconstructed analog waveforms. Yes, those devices can miss absolute peaks, yet those peaks remain perfectly reconstructed in Nyquist space. This is why mastering engineers must be careful and not simply use digital metering to assess headroom. Nyquist is never invalidated, but our tools often fail us. The main distinction here is that any consecutive FS MSB PCM data is defined as a clip. EDIT: Any original unclipped signal remains 100% reconstructable.

My last sentence was incorrect (edited) and I apologize. It got me thinking about Nyquist math, which shows elements of both power law and voltage law. I believe we can draw a general rule that "any two contiguous Nyquist-valid samples under -3dBFS assure non-clipped original program." But I can envision synthetically-generated bit patterns designed to confuse normal filters, resulting in clips with pre-post samples under -3dBFS.

The practical takeaway here (at least for me) is that audio producers using sample-point metering are playing with fire. Any non-reconstructive digital metering system can give wrong data, and should never be trusted when working anywhere near FS. Another takeaway which I hope has been drilled into Grigorio's head is that Nyquist data is always 100% reconstructable. If there is a clip in the original signal, it will clip in reconstructed Nyquist space. If there is no clip in the original signal, there will be no clip in a valid reconstructed Nyquist space.

 

[gregorio]

When someone tells me that Nyquist, under rare circumstances, misses certain information "between the samples" -- they are simply wrong, and spreading really strange ideas.

You've steadfastly refused to answer my previous questions but answer this one: Who is that "someone"? It's certainly not me or anyone else who has responded to you. That "someone" is a figment of your own imagination, that's why I stated you are arguing with yourself!

Intersample peaks are a real phenomenon, but they occur for reasons vastly different than what you have been selling here.

"Peak meters in digital audio systems often register “peak-sample” rather than “true-peak". ... The problem occurs because the actual peak values of a sampled signal usually occur between the samples rather than precisely at a sampling instant, and as such are not correctly registered by the peak-sample meter. ... Since sampled signals may contain overloads even when they have no samples at, or even close to, digital full scale, overload indication by a peak-sample meter is unreliable. Overloads may cause clipping in subsequent processes, such as within particular D/A converters or during sample-rate conversion, even though they were not previously registered by the peak-sample meter (and were even inaudible when monitored at that point). "

That is taken from The International Telecommunications Union's publication: ITU-R BS. 1770 (2006), which forms the basis of technical specifications published by the EBU, ATSC and AES. What "vastly different reasons" have you got in mind, reasons which differ from all the world's main international audio bodies?

G

 

[castleofargh]

the insane part is really not necessary. he clearly got caught up in his own interpretation of some of the first posts, and then admittedly had a real hard time forgetting about that. it can happen to anybody with a little self confidence. TBH it happens to me rather often. I read something I find silly, start responding and by the end of my post I'm on fire. I go seek the guy's quote to really rub his nose in it, except there is no quote to be found because I misread his sentence.