Sample Rate Question.

[KevWind]

Originally Posted by Psalad

This is sensible, lots of what you write is sensible, but given how difficult it is to successfully hear the difference between higher sample rate audio and 44.1k, I think it's pretty likely they got it right.

If being able to hear the difference requires one to be in a perfect room, with some perfectly designed tests, then maybe that itself is good enough for me. That seems like a pretty logical conclusion to me. How would you argue? Many tests have been done, by both individuals and experts.

The AES test method I believe is listed here: http://www.bostonaudiosociety.org/explanation.htm

Here is a link to the paper itself which can be purchased: https://secure.aes.org/forum/pubs/journal/?ID=2

You again might be right there is something different in music compared to tones that might make things different when it comes to audio listening. Sure, why not, seems possible right? That might very well be possible. Next step would be a scientist would take that thought and test it... knowing the burden of proof is with existing science, he would then create a test for his hypothesis. He wouldn't ask the other side to prove tones are good enough. The existing science is the baseline.



This was meant to read as we (you and I) disagree, but I see how you might take it otherwise, I didn't word it well. Please assume I'm NOT being snarky, as I hope it's perfectly clear there is a complete absence of snark in this conversation from my side, OK?



I have no problem with your thought process, or Lavry's for that matter. You both make very good logical arguments for the potential need for higher sample rates. But the proof for me still comes out in the testing. The reasons why it is so important to test are:

- Expectation bias is HUGE when it comes to audio..I'm sure you're aware of how the human bias makes objective testing very difficult
- Humans can only "remember" audio qualities for a matter of seconds (making comparison difficult)
- Moving your head just a little bit, as we've all experienced, changes the sound in a significant way (making comparison difficult)

So in the absence of testing, you have... just faith, and magical thinking. I too was one of the people who really thought 96k made a "night and day" difference. The openness! The imaging! Finally I can really hear cymbals again. But then I tested myself, and I believe it to be my own expectation bias.

You've made some clear arguments about the limitations in every test that you have seen, and that is fine, but I would love to see you design a better test... and actually why not create the e2e test yourself? I'm not at all being snarky, just to be clear. I would enjoy seeing a better test, even participating in it. You might find a way to prove there really is a difference!

But in the meantime we have a lot of evidence showing the lack of audible difference in blind testing, which goes to show one thing for certain... IF there is a difference, one that has been so difficult to prove, the difference cannot be heard in casual listening. It can't be heard in a/b testing so far. So I conclude if there is a difference, that difference is tiny and irrelevant. I think that's a pretty logical conclusion, and it drives my thoughts as to why 44.1 is the right choice.

Oh and one thing we haven't discussed is how much of existing modern gear might have non linearities at frequencies above 20k, which is another reason why at a very minimum one should pay attention to specs in the gear one uses if they use higher sample rates.

I did not think you were being snarky, I actually thought you throwing in more humor with an interesting twist on the first person, but I understand now that you meant you and I disagree.

I did see the Meyer-Moran paper shortly after it first came out .
And as I said in another post while it makes a excellent argument for the Idea of recording in high resolution to obtain the best possible results when down sampled to CD standard, given that the results were that people with trained ears were only able pic the high res. SACD, over down sampled from SACD to the Standard CD version 52% of the time.

And I completely agree with expectation bias being "HUGE" and objectively there is no reason to believe it does not work equally both ways in this issue, for example it is bubbling over in the Mix Online article from the obviously biased title "The The Emperor's New Sampling Rate" to the overly generalized claim of design intention.

Because Meyer-Moran only tested material recorded in SACD high resolution then looped played back through a CD recorder that down sampled high res. to be the example of standard CD resolution .
To make the claim as the author of the Mix article did "It was designed to show whether real people, with good ears, can hear any differences between “high-resolution” audio and the 44.1kHz/16-bit CD standard" is either specious and misleading, or is as you say merely a "guess" or "Faith" or "Magic thinking". As is offering similar single source methods of testing, as some kind of established baseline science.

I already laid out what my criteria would be for viable test in a previous post.

"Which would mean using a single performance, that is split and simultaneously recorded in both 44.1 and say 88 or 96. and then played back in 44.1 and 88 or 96. respectively and then perhaps another test played back again in only a single reduced resolution .

It seems logical to me that the only way would be taking the signal from the same front end pre amp outputs and splitting it:
Send it to two of the same brand and model of AD/DA converter (as I don't think one unit can handle two different sample rates at the same time)
One at 44 and one at say 88. and into two sessions ( one in 44 and one in 88 ) in the same brand DAW, on two of the exact same brand and config. of computers.
Then take the outputs from the computers back to the two AD/DA's and back through the same playback system.
Then see if people can determine which is which.

[rick-slo]

KevWind, you might enjoy reading this book:
http://www.amazon.com/Digital-Audio-...audio+engineer

[Trevor B.]

I'd love to know more about the "Guitar(s) by Derek Coombs" but I guess that's a topic for another thread.

[KevWind]

Originally Posted by Trevor B.

I'd love to know more about the "Guitar(s) by Derek Coombs" but I guess that's a topic for another thread.

Trevor in answer to your basic questions about what sample rate and why. For years I recorded in 44.1 but now tend to use 88.2 or 96 (still have not made up my mind)
A quick history:
My journey in home recording began specifically to be able record to produce a CD.
With this goal I started and stayed for many years with 44.1k believing that #1. it covered the entire range of human hearing and #2. if you were going to end up there you might as well start there.

What started to change my thinking was as I stated earlier while being involved some blind critical listening testing. I was not surprised when my upper range hearing toped out between 14.7k and 15.3k on the prototypical single freq. sign wave test. BUT was quite literally dumbfounded when my score for being able to detect both cuts and boosts to pink noise and music at 16k was something like accurate 74% or 76% of the time.( my age kicking in again )

Humm ? What does that mean? The only logical conclusion myself and the people I was involved with could reach was (and yes this is only a guess) , . Even though I could not hear 16 k as single sign wave frequency, that perhaps I could detect some kind of interaction the effect of a 16k cut or boost was having on frequencies I could hear.

Now back to your question about "air " or presence, or 3D perception (as opposed say flatter 2D and less perception of presence) No matter the disagreement on sample rate, I think it safe to say 24 bit is fairly unanimous.
And I think everyone agrees there are other factors that will have a much greater and more noticeable effect on presence.
Two things that I think make a noticeable improvement in "presence' is using Pre Delay in Reverb ( anywhere from 30 to 80 ms) and the judicious use of Compression.

[KevWind]

Originally Posted by rick-slo

KevWind, you might enjoy reading this book:
http://www.amazon.com/Digital-Audio-...audio+engineer

Thanks I will put it on my list

[Joseph Hanna]

Originally Posted by rick-slo

KevWind, you might enjoy reading this book:
http://www.amazon.com/Digital-Audio-...audio+engineer

I worked with Nika for some years. His job at the time put him in a unique position to collaborate with both audio engineers and audio conversion manufacturers and his book was a culmination of the raging debates we used to have back in the embryonic days of digital audio. We used to discuss this topic almost endlessly and Nika being a pretty focused kinda guy took the opportunity to at least get the cards out on the table.

In the end all the consternation proved to be of little enduring sonic value and by looking at the content of this thread the arguments remain nearly identical to this day. For me the fuel ran out of my sample rate furnace years ago. I've heard all the spec combinations sound good and all sound bad.

It's ironic to note however that Nika has not been in the audio world for years now and last I heard he was practicing law in New York City. I'm guessin' the fuel ran out on his sample rate furnace as well

[sam.spoons]

I usually record at 44.1 24 bit. If you don't have a good condenser microphone, decent preamps and, above all, a nice sounding recording space there is little point in recording at a higher sample rate (nearly true, this article explained why you might in some circumstances choose a higher sample rate). 24 bit, as has been said, is a no brainer. But, there has been a bit of misinformation..... 24 bit does not make an intrinsically better quality recording than 16 bit it simply increases the available headroom (i.e. the number of dBs between the noise floor and dBFS) from a theoretical maximum of 96dB for 16 bit to potentially, 144dB for a 24 bit recording. Since the rest of your signal path will introduce significantly more noise anyway the advantage of 24 bit is that you can record at a lower level (as much as 20dB lower in practice) without increasing noise and, thus, be safer from overloading your system into clipping (and digital clipping is an absolute no-no).

[rick-slo]

Originally Posted by Joseph Hanna

I worked with Nika for some years. His job at the time put him in a unique position to collaborate with both audio engineers and audio conversion manufacturers and his book was a culmination of the raging debates we used to have back in the embryonic days of digital audio. We used to discuss this topic almost endlessly and Nika being a pretty focused kinda guy took the opportunity to at least get the cards out on the table.

In the end all the consternation proved to be of little enduring sonic value and by looking at the content of this thread the arguments remain nearly identical to this day. For me the fuel ran out of my sample rate furnace years ago. I've heard all the spec combinations sound good and all sound bad.

It's ironic to note however that Nika has not been in the audio world for years now and last I heard he was practicing law in New York City. I'm guessin' the fuel ran out on his sample rate furnace as well

Rehashing the same stuff over and over does become less interesting. My posts tend to be of shorter and shorter length on many topics.

Nika's book is well written and I have enjoyed reading it. Aliasing of frequencies above the sampling frequency to harmonic content below the sampling frequency is particularly relevant to the direction this thread has taken.

[Trevor B.]

I found this video on youtube and would like to know if it provides a reasonable overview of sampling, aliasing, and the Nyquist Theorem for those of us new to the concepts?

[Psalad]

Originally Posted by rick-slo

Nika's book is well written and I have enjoyed reading it. Aliasing of frequencies above the sampling frequency to harmonic content below the sampling frequency is particularly relevant to the direction this thread has taken.

I haven't read his book, but I watched that legendary thread on the George Massenburg forum called "George Watch This" unfold live... where Nika brought up many of the issues we are discussing in this thread. I was a part of that forum (which I love and miss).

Very interesting that he's practicing law right now... not surprising, as he is a smart, articulate, and passionate guy... probably a great lawyer.

Kev, why not do the test you have outlined? I am in norcal and would enjoy participating, or even just watching. I am always willing to learn something new.

BTW, I did the test you reference, with the following modification since I didn't have two of the same converters.

mic (think it was either TLM103 or my stereo Avantone) -> mic pre (Great River 2mh) -> audio d/a -> simultaneous recording on both MOTU traveler (44.1) and Steinberg 816x (88.2). It was very easy to do. I played acoustic guitar through it and other instruments. Played triangle. Played my drum kit.

In no case could I pick out one from the other. Obviously it's not a scientific test as there are two different interfaces in play, but... it should have been easier, considering really the difference between the two interfaces should be obvious. Even then.. no.

I had a couple different monitors connected, including my Dynaudios and Tannoys. I could reliably hear no difference in blind testing.

Anyway, point is, as I said, it's pretty reasonable if one cannot hear a difference in some basic tests, the difference that might exist is irrelevant. Combine that with the results of the tests many people have done. So anyway, that's my take.

[rick-slo]

Originally Posted by Trevor B.

I found this video on youtube and would like to know if it provides a reasonable overview of sampling, aliasing, and the Nyquist Theorem for those of us new to the concepts?

Good video.
Human hearing up to 20,000 hertz.
44,100 sampling rate above double that leaves some headroom for the aliasing filters to operate to fully attenuate frequencies at 22,050 hertz and above while letting 20,000 thousand hertz and below fully through.

When those aliasing filters are substandard to doing that task is where you might hear some differences in the between recording at 44,100 versus a higher sampling rate. Most even lower end digital gear is so much better now than it was a few years ago that it is unlikely to be a problem in any half decent new gear you purchase.

Regarding added effects, say an in the box software reverb, they work at a higher bit depth, they up-sample, do their thing, and then drop back down. Possibly with some software something may occur audibly there between a base sampling rate of 44,100 hertz versus higher. Don't know, but I hedge my bets and thus usually record at 88,200 and keep it there until I am all done.

[Trevor B.]

Originally Posted by rick-slo

Good video.
Human hearing up to 20,000 hertz.
44,100 sampling rate above double that leaves some headroom for the aliasing filters to operate to fully attenuate frequencies at 22,050 hertz and above while letting 20,000 thousand hertz and below fully through.

When those aliasing filters are substandard to doing that task is where you might hear some differences in the between recording at 44,100 versus a higher sampling rate. Most even lower end digital gear is so much better now than it was a few years ago that it is unlikely to be a problem in any half decent new gear you purchase.

Regarding added effects, say an in the box software reverb, they work at a higher bit depth, they up-sample, do their thing, and then drop back down. Possibly with some software something may occur audibly there between a base sampling rate of 44,100 hertz versus higher. Don't know, but I hedge my bets and thus usually record at 88,200 and keep it there until I am all done.

Thanks for the input and the advise. I found the video easy to understand so it's reassuring to know the information is accurate.

[KevWind]

Originally Posted by Psalad

Kev, why not do the test you have outlined? I am in norcal and would enjoy participating, or even just watching. I am always willing to learn something new.

BTW, I did the test you reference, with the following modification since I didn't have two of the same converters.

mic (think it was either TLM103 or my stereo Avantone) -> mic pre (Great River 2mh) -> audio d/a -> simultaneous recording on both MOTU traveler (44.1) and Steinberg 816x (88.2). It was very easy to do. I played acoustic guitar through it and other instruments. Played triangle. Played my drum kit.

In no case could I pick out one from the other. Obviously it's not a scientific test as there are two different interfaces in play, but... it should have been easier, considering really the difference between the two interfaces should be obvious. Even then.. no.

I had a couple different monitors connected, including my Dynaudios and Tannoys. I could reliably hear no difference in blind testing.

Anyway, point is, as I said, it's pretty reasonable if one cannot hear a difference in some basic tests, the difference that might exist is irrelevant. Combine that with the results of the tests many people have done. So anyway, that's my take.

While the test I purposed definitely would be a very interesting test, and perhaps even serve to inform the discussion, which would be great.
However, since I do not own 2 Avid Omni's , nor 2 Avid HDN licenses and PCIe cards, nor 2 Mac Pro towers and do not know anyone I could borrow 1 more each of those from, it ... as they say... "aint gonna happen".

Honestly I see the test I proposed as the very minimum criteria elements to be met to even be considered a valid "basic test".

But I must admit I am confused you offer your test as being the same test as my proposed one, but "modified"only by having different brands of converters
BUT is that the only difference ? I ask because I am not clear as to your actual methodology. And particularly if the signal flow of the two different sample rates remains discrete and totally separated and unconverted , from the first A/D conversion from mic pre ... to the final D/A conversion to output to the monitors. In other words each of the two different sample rates, gets only one Analog to Digital conversion, and One Digital to Analog conversion.......

In your signal flow what exactly does "> audio d/a >" mean ? Can I assume you meant an audio a/d unit ?
Since a (d/a) in that position of signal flow doesn't seem to make sense. And Is this unit outputting two different sample rates going into the two different interfaces ? And why not just split the outputs from the GR and go directly into the inputs on the Steinburg and MOTU directly ?


Also
One computer or two ?... 2 computers seems straight forward, but 1 computer seems a bit tricky .. I get it can have multiple USB or Firewire ports for two different interfaces but from that point ???? what exactly is the work flow to keep the two sample rates totally discrete an unconverted until final output ?

[Psalad]

Originally Posted by KevWind

However, since I do not own 2 Avid Omni's (snip)

Why would you need two of those? Two of any of the same converters would work, no? Wouldn't you think it would be somewhat easy to put out a note to friends to borrow two that match?

Even if they didn't match... might be worth a try to see if people can even hear the difference between two good quality modern converters. I doubt if I could tell.

And particularly if the signal flow of the two different sample rates remains discrete and totally separated and unconverted , from the first A/D conversion from mic pre ... to the final D/A conversion to output to the monitors. In other words each of the two different sample rates, gets only one Analog to Digital conversion, and One Digital to Analog conversion.......

Again, if you think it needs a test to that degree, then you should do it. That would be great!

In my case, the one computer that was doing the playback and running a/b/x software was running at 88.2 (so the software was up converting for the test). I know that's not good enough for you, and that's fair enough, you are welcome to create your own test.. but it's good enough to me for a basic test.

In your signal flow what exactly does "> audio d/a >" mean ? Can I assume you meant an audio a/d unit ?

"Audio d/a" is "distribution amplifier." Active splitter. line level.

One computer or two ?... 2 computers seems straight forward, but 1 computer seems a bit tricky .. I get it can have multiple USB or Firewire ports for two different interfaces but from that point ???? what exactly is the work flow to keep the two sample rates totally discrete an unconverted until final output ?

Two computers for recording, and one for playback as it means you can simply run software a/b/x testers. But the more complex two computers/two interfaces test would be even better if you have the ability to level match, have someone switching and tabulating in a way that ensures there are to tells, etc.

[Doug Young]

Originally Posted by Psalad

Even if they didn't match... might be worth a try to see if people can even hear the difference between two good quality modern converters. I doubt if I could tell.

This stuff gets difficult really fast.... There was a serious attempt just to compare different converters a few years back on gearslutz. Had some very qualified people involved - AES speakers, recording engineers, etc. Turned into a giant flame war that ended with the entire thread not only being locked, but deleted. I imagine some others here saw it at the time before it disappeared. Somehow, these discussions/tests are not only never conclusive, but they never end well. Seems easy enough, but clearly it's not.