Understanding Sample Rate

[adamdea]

19 hours ago, beerandmusic said:

As to not disrupt the "best bang for the buck thread", any more, I am starting a new thread.

 

I am conceding that I "MAY" possibly have a misconception of sample rate, but after reading below, i "believe" my conception is still the same.

 

Can we agree on this as the "basics" for any debate?

 

https://techterms.com/definition/sample_rate

 

 

Really good news. Jim Lesurf has made his "information and Measurement", an excellent undergraduate text, available for free

http://jcgl.orpheusweb.co.uk/InformationAndMeasurement_PDF_Book.pdf

If you really want to learn something read this.

 

If you don't understand why a book on information and measurement is relevant, you really need to read this.

 

If (with very few exceptions) you think you don't need to read this, you really really need to read this. 

[adamdea]

1 minute ago, mansr said:

I've read other books that probably cover most of the same topics. Is that a good enough excuse?

You might be one of the exceptions.

[adamdea]

My view is that at some point one either says I don't actually care about how this all works, or one tries to work out how it all works. Of all the things that get boring on the internet, fairly high up the list is -people pretending to want to understand something without being prepared to admit that what they think they know might be wrong.

[adamdea]

Just now, beerandmusic said:

 

for me, i had a genuine curiosity initially, but i see it is more complicated and don't want to devote the time to understanding.

If it was something that could be spoon fed where people wouldn't get upset, that would have been great, but seeing that is not the cae, and i doubt I will get on the same page, i personally have no desire to explore it deeper...but my belief has not changed that SACD can sound better and can be more accurate, has not changed.

 

 

I see. Well learning and questioning your own beliefs can be tiring and it might get in the way of all the attention-seeking. Onto a new thread?

[adamdea]

17 minutes ago, mansr said:

A Dirac pulse is a signal whose value is non-zero at one single instant and zero for infinite duration before and after. It contains all frequencies from zero to infinity throughout the infinitely long interval.

This is I think why Fokus reminded me that when you apply a Dirac to a linear phase filter, the ringing is caused by what the filter takes away, not what it adds. 

[adamdea]

1 hour ago, jabbr said:

 

That’s fine. A certain degree of over engineering is reasonable. So let’s say record at 24/96 or 24/192 ... but recording beyond that is really really hard pressed to justify any benefit (unless you were doing a ton of post-processing — even then)

 

On the other hand, upsampling for the purpose of improving the DAC is totally reasonable and an excellent technique — in fact the argument that you don’t need it would be the harder one to justify.

Hmm alot of people go along with recording up to 96khz (but probably still distribution at 44.1) but quite a few people argue that 192 is actually too much. The problem is that you are actually letting in a whole lot of spuriae which might be better filtered out and in any event not all the editing and production tools work at that rate anyway. 

There was a case for the possibility that a bit more wriggle room over 44.1 was needed. No one was really arguing for 192 because whilst its possible to speculate that 44.1fs filters might have audible problems, the same is difficult even for a fertile imagination with fs 96 khz filters because there is so little information at the transition band and hearing 40 odd khz tones beggars belief.

The problem is that people who tended to claim that 96khz fs actually sounded better then had a habit of thinking that 192khz fs sounded better (no surprise) and so on 

As for upsampling well you need that for a digital filter whether you call it a feature (as some dacs do) or just get on with it (as most of them do). 

[adamdea]

30 minutes ago, beerandmusic said:

proven for band limited signal

So unfortunately that must mean that dsd doesn't work either.

[adamdea]

2 minutes ago, firedog said:

He apparently has an obsession with his idea of resolution and accuracy and has almost dropped it, and then goes looking on the net for material he doesn't understand so that he can come back and say what we've been telling him is wrong. 

 

It's pretty much time to give up responding to him. 

 

And yes, beerandmusic, there are conditions for the Shannon -Nyquist theorem that are never actually met in real life. That means nothing as we can get so close to the conditions that the gap is meaningless.

If you don't believe that, I suggest you stop using every bit of technology you own, and never get in a car or on a plane or train. It's all based on "almost" getting there in calculations.

NASA got to the moon using calculations made by a slide rule, because it was "accurate enough" that in real/practical terms you could call it 100% accurate.

It seems to me that one way of looking at the impossibility of perfectly bandlimiting is to compare with pi- you can;t actually write down all the digits but you can compute them to whatever degree of precision you require in an application. 

Rob Watts is on a mission to keep bringing out a new dac or two every year with more filter taps. It could go on forever

[adamdea]

3 minutes ago, firedog said:

But at some point he may decide that he’s gotten to the point that adding more taps doesn’t actually improve the sound in practice. 

errm maybe. I think that if we were going to get there we would have got there. 

[adamdea]

7 hours ago, crenca said:

As far as sampling theory, your trying to tinker in one of the few areas where there is actually a substantial body of scientific, mathematical certainty and proof in audio.  The differences in sound you experience and rightly question, the cause is not because of sampling method (i.e. the math and digital construction of the waveform), but rather because of the implementation of the equipment, room, music, and person.  

 

Divert to these other areas for fruitful and interesting things!

Whilst I mainly agree with you, I would say that there is far more established knowledge in audio than most people seem to think or draw on (see for example the dreadful farrago of a discussion of auditory localisation on another thread); it's not quite so mathematical as the sampling theorem (being amongst other things experimentally derived, and in more developing fields)  but sill it's solid scientific understanding, fairly easily accessible  and frankly no one seems to give a damn. 

Most people insist on starting from what they think they know and only cherry-picking technical stuff so far as it seems to support that. I find it extraordinary that we still have discussions about the sampling theorem; I would have assumed that everyone involved would either have understood it by now, given up hope or decided that they didn't care. I would also have hoped that people would stop posturing and pretending either to understand it or to be interested in trying to understand it. Why bother?

Our intuitive understanding of every process involved (maths, engineering, psychoacoustics no doubt more else) is muddled and about as helpful as the intuition that thunder must be caused by the gods moving chairs around. . As long as that intuitive understanding is valued above any actual knowledge we get the wrong answer. Where much of it is debatable that may not make much difference. Where it isn't debatable is just leads to a very long very boring river of sludge, enlivened by the occasional interesting anecdote, charming generous and warm-hearted character or flash of wit.

[adamdea]

1 hour ago, Spacehound said:

'Digital', audio or not,  is fully understood  and has been for a very long time. We made it  from scratch to work as we intended it to, we didn't 'discover' it. At first it was entirely 'mechanical' but the principles have not  changed at all.

 

The 'small area' where it changes from digital to analog (if reguired) is only fully understood in its digital part.

 

The only 'digital' advance needed (or possible) are in computing speed, which is not important in audio as it already vastly faster than audio will ever need. .   

Yes but you still can't get answers out without (consciously or unconsciously) applying a psychacoustic model. If we were cats, we might need a higher sampling rate than 44.1 kHz.

That is not because of any limitation in the maths, it's just that our specification of what we need need from the system comes from something outside. 

I have been dying to get my hands on someone who really understands current thinking on psychacoustics because the information about time resolution of the perceptual system is a bit difficult to get hold of. I have got as far as understanding that there seems to be some current thinking about the system integrating over a 6ms window. I have a hunch that this might be important to certain current controversies.

[adamdea]

Just now, Spacehound said:

Cats and psychoacoustics have zero to do with computing or the  sampling theorem.

As I previously said it even works fine on bus timetables.  And as mansr said it will work  100% on  things not yet discovered.

 

The rest is 'vague (not your fault). It's like  'logic machines' such as computers. Be very  exact in  what you wish for  

Of course not, but they do have a lot to do with the question what is the minimum sample rate required to capture all the information you can hear. You need to understand the sampling theorem and know something about human hearing to answer that question. Like durrr

[adamdea]

Just now, Spacehound said:

He appears to point blank refuse to understand anything he doesn't understand already. Unless perhaps  our sample rate is too low.

 

The "something  about" is the maximum frequency we can hear.  All the 'time' and 'impulse' stuff follows from that and most people's 'understanding' of it is nonsense because they don't understand that fundamental point.

Yes. That much became apparent over 10 pages ago. Having thought about it and already worked out the answer to your earlier question (why does OP bother asking a question if he doesn't care about the answer?) I answered in my own mind as "to get attention". So I think it's time to follow the logic of my own conclusion and engage the ignore button. 

[adamdea]

5 minutes ago, mansr said:

You can also ask the question, what is the minimum sample rate require to capture all the sound there is? The answer is roughly 500 kHz. Recordings done at 352.8 kHz are now readily available. If you examine one, you will see that they have precisely zero signal content above 100 kHz or so and very, very little above 50 kHz. This means that a sample rate of ~200 kHz is enough to capture sound in actual music, whether we are cats or humans.

True and I agree about the physics of sound and the spectrum of real musical sounds. 

Either way, the knowledge required comes from outside the sampling theorem itself. Of course for a whale or a creature on another planet the specs might be different.  

Similar points arise in relation to bit depth etc.

 

Ultimately you have to have a model of what matters. One of the problems of this hobby is that even those people who seem to take an orthodox approach to the sampling theorem like Rob Watts can take a pretty weird view of  what is audible -hence the 1 million tap filter and the claim somewhere to be able to hear things at -200dB . In fact that refusal to accept any limits to the spec drives a lot of arguments. I think the technical engineering side of it can become a bit of a red herring. You can show differences between the noise floor of dacs maybe 30dB below the 16 bit noise floor. You can show lowr phase noise in the clock. You can show objectively shinier paint on the router.

[adamdea]

1 minute ago, Spacehound said:

 

In audio 'spheres' there are more things that don't matter than do. 

What saddens me is that all the noise about things that don't matter tends to drive out consideration of what does matter. The real science seems to be driven more by gaming than music. If ambisonics had caught on way back we could have a catalogue of proper surround recordings available to be played on whatever configuration you chose and basically future proof. Even if it had been at 13/32 it would be more use than any dsd stereo recording. 

[adamdea]

7 minutes ago, mansr said:

The difference is that this outside knowledge is easier to acquire and validate than that concerning our ability to perceive sounds.

Possibly as regards bandwidth- but just as easy to ignore if one takes the fluctuations of one's listening experience and one's personal interpretation of the meaning of those experiences as the only critical data, alongside a general faith in progress. At one point it looked as though DXD might become a de facto "this far and no further" standard. But forget it. Limits are offensive and if you play someone a dxd file and a 32/ 768 file pretty much the same group of people who can hear the difference between 16/44 and 24/96 will claim to hear the improvement in the 32/768. 

 

When first came to this hobby I assumed that the 16/44 spec must have been cobbled together as a result of the limits of technology in the early 80s and that it made sense that it must be easy to improve on it now. After all I remembered what computers were like in 1982 and assumed that analogies with video were closely applicable. It took me a long time to "unlearn" that.

 

It's useful to learn about the sampling theorem (it being beautifully elegant and worth knowing)  but I think that learning about perception provides what is in many cases the real answer to the eternal question- why do I hear a difference when....

 

And amongst the really clever technically minded people I have met on the internet forums are some who I feel just confuse the issue by finding brilliant and subtle technical explanations as to why there might be a peculiarity of the kit or set up leading to perceived result. (I don't mean you)

 

[adamdea]

22 minutes ago, mansr said:

The way I see it, the goal of audio engineering should be to accurately reproduce the stimulus a listener is subjected to, i.e. the sound waves entering his or her ears. The study of perception belongs to the domain of neuroscience (or something similar). The only way it relates to the engineering is when its findings allow engineers to take shortcuts. A prime example would be the discovery of two-channel stereo and its ability to create the illusion of arbitrary left/right positioning without using a speaker for every possible direction.

I think there are a number of rational approached which could be taken to this. I think quite  a lot of engineers would say that the goal was to reproduce everything you can hear. Others might say that it is to maximise the verisimilitude of the recording

 

Either way I am surprised at the extent to which people get worked up at any shortcomings of the 16/44 channel compared with the shortcomings of only having two channels, the need to reproduce the lowest audible octave,  and the problems caused by rooms.

 

Taking your goal of  "accurately reproduce the stimulus a listener is subjected to, i.e. the sound waves entering his or her ears" I'm not 100% sure you can dispense with any psychoacoustics. I assume there are two ways of reproducing the sound waves in the ear: 

One characterised by ambisonics would be to measure the 3d vector and then try to produce that in the listening room; the other is to pre-treat the sound with the hrtf and pipe it to the ear

 

I'm not convinced that you could ever perfectly achieve the reproduction part of the ambisonic soundfield , so I suspect you would still need a theory of what really matters. I also suspect that perceptual science would still have something to say about the limitations even of a "perfect" reproduction though. Because you aren't in the recorded acoustic space  it's difficult to "learn" it.

Similarly with the hrtf it is psychoacoustics that tells us of the need to have head tracking isn't it? And in the absence of perfect hrtf measurements (possibly matching the clothes I am wearing today) you need a theory of what really matters.

 

[adamdea]

23 minutes ago, mansr said:

The process leading to the specific design of the CD is interesting. The rough requirements were determined by the already well known limits of human hearing, both in frequency and dynamic range. Sony and Philips argued a bit over whether to use 14 or 16 bits, mostly because one of them had a 14-bit DAC chip and the other had a 16-bit one. The choice of precisely 44.1 kHz as sample rate was made because existing equipment could easily be made to handle it. The physical design of the disc (laser wavelength, track spacing, etc.) was mostly based on what could be manufactured at a reasonable price. As for the size of the disc, one story has it that it was chosen such that the longest known recording of Beethoven's 9th symphony would fit.

Yes very interesting and well covered here https://www.amazon.co.uk/Perfecting-Sound-Forever-Story-Recorded/dp/1847081401

[adamdea]

16 hours ago, fas42 said:

 

I find it quite amusing that so many need a terribly, terribly technical reason for what is needed to get high quality sound - part of the male game to be able to hold up the Big Breakthrough over one's head in triumph, I guess.

 

What really matters is sufficient quality in the replay chain - because the inner workings of the ear/brain, and  a lifetime of learning to interpret the meaning of subtleties in the sound then does the rest. Using various multi-channel gimmicks, and fancy room setups all can help the mind grok what is intended to be heard - but it's talking the hard road; far better ROI is to fix the audible flaws; and it also means that the whole century plus of recordings history we have becomes fully accessible.

My formulation

I find it quite amusing that so many need a terribly, terribly technical reason for what is needed to get high quality sound - part of the male game which involves being over fixated with gadgets

What really matters is sufficient quality in the replay chain - because the inner workings of the ear/brain, and  a lifetime of learning to interpret the meaning of subtleties in the sound then does the rest. 

Fortunately that sufficient quality is not too high. All you need to do is to let go of the notion that ever increased kit complexity and/or expenditure is required for enjoying music. What makes the difference between one listening experience and another has far less to do with kit than you think. That's why most music lovers don;t care much about hifi. 

A whole century plus of recordings history is already accessible. It's OCD which is getting in the way.

[adamdea]

5 hours ago, Spacehound said:

Sandyk,

I just picked your above post at random so I could 'reply' to you. Don't read anything into it.

Thought you might find this of interest:

One of 'our' latest computers. It may look like something out of Babbage's workshop but it isn't.  It  has a very strong potential to render all other computers as obsolete as the slide rule.  And 'we'  have already increased its power tenfold since September 2017.

 

 

 

Off topic, Mr Hound, but as a matter of interest do you think there is any chance it might  undermine blockchain as well? 

[adamdea]

1 minute ago, pkane2001 said:

 

All encryption-based technology that relies on the difficulty of computation can become useless should quantum computing become a reality.

Yes. agreed.  I'm not sure whether blockchain avoids this by having a distributed register. But I'm a bit hazy

[adamdea]

42 minutes ago, Don Hills said:

Aha, I misread the original post. What JJ actually said was "1/ ( 2 pi bandwidth number_of_levels)", not "sample rate". 110 ps, not 55. How embarrassing, I've had it wrong the last 3 years. So he was describing the limit case (22.05 kHz for 16/44.1) whereas your (Mans') derivation in the "MQA is Vaporware" thread is the general derivation.

 

 

 

Don, I too share your concern becasue I have assumed JJ got it right. But he definitely said it was 1 

"1/ (2 pi fs nlevels)

for CD that is 1/( 2 pi 44100 65536)."

https://hydrogenaud.io/index.php/topic,91126.0.html

Fs isn't bandwidth. Are we reading the same thing. Sorry if I am being obtuse.

[adamdea]

Just now, mansr said:

That's definitely wrong.

Just to be clear I assuming that you are saying that what he said was wrong not that I have misquoted him.

[adamdea]

Great. Can you just talk me through your correction from 220 ps to 110ps

[adamdea]

5 minutes ago, mansr said:

Scroll down a little on that page. As esldude points out, a level above ½ LSB will be rounded up to 1.

Ah so the minimum time difference is equal to the difference which will cause a 1/2 lsb amplitude change in a 22.05Khz wave at the sampling instant.