Does the sampling rate have any effect on the sound quality produced by the audio interface other than effecting the overall latency?

If so what do most of you use-96k-48k-44.1k?

Thanks

Jack

i run my DAW at 44.1KHz

can't hear the difference between 96KHz and 44.1KHz anyways... and 44.1KHz is easier on my processor for latency purposes

I use 96 because it doesn't seem to affect my performance or latency and I'm always superstitious about whether there is an audible difference. I think software pianos (at least some of them) are sampled at faster than 44.1, so in principle there could be a difference. Can I personally hear it? No.

In my case as I use higher frequencies I seem to get lower latencies seems counter intuitive-- IE 96 instead of 48 etc!!!

Jack

Yes, that's correct. For any given audio buffer size, as you increase the sample rate, the latency will decrease (improve). If you're on the minimum buffer size possible for your audio interface, you might be able to improve the latency even more by increasing the sample rate. My own USB interface can go as low as 128 samples for the (ASIO) buffer size, and I find that the latency is fine at 44.1kHz, so I just leave it on that.

Greg.

128 Samples and 44100 sample rate.

No issues with latency.

I'm actually interested in an answer on the very reason 96kHz sample rates may be better. From a math pov, 48kHz is enough to perfectly sample any sound ranging from 0 Hz to 24kHz, the latter being beyond the ability of most people's ears. Why is 96kHz useful then?

Some folks say that even frequencies which we cannot directly hear contribute to the sound, because we can hear the heterodyning (beats). (whether it is actually "musical" to hear this beating is another question!) Also, part of the process of converting digital audio to analog is that it must be filtered to remove unwanted high frequencies (anti-aliasing filtering), and this filtering is easier to do when the sample rate is higher, and thus, a small improvement in fidelity may theoretically result. (just as an aside, I stumbled on the NFB-12 DAC/headphone amp, which actually allows this filtering to be adjusted by the user)

Greg.

From a math POV 48kHz is enough for sample up to 24kHz, but not perfectly. Sound waves are something like this:



Amplitude is the vertical axis and time is the horizontal one. To sample is to take the the amplitude of the wave at a given time. If you do that at a constant frequency, then you get a digital and easily reproducible representation of the analogue wave.

The digital representation is an approximation of the real thing as it has 'edges' (it's quantized) while the original is rounded. The resolution (16 bit, 24 bit, 32bit) will improve the accuracy of the vertical axis, while the sampling frequency will improve the accuracy of the horizontal.

So, at 48Khz, you can represent 24Khz waves but only with a sample at the top of the wave and a sample at the bottom for the best case, but with nothing in between. That's far from perfect. And depending on the phase, the samples will be taken at other points at the middle of the wave, thus given a distorted representation.

In practice, 24Khz is inaudible for most of the people (certainly I cannot hear it) and the difference between 48Khz and 96Khz is also very hard to hear, but in theory 96 Khz gives a better representation of the wave even at low frequencies. You will also need a very high quality amp and speakers to try to hear such differences.

There is good reason for using 96Khz in this case: if you are going to digitally process the sound (that is applying complex math functions to the wave), the better the resolution, the better the result will be because the inaccuracies of the sampling get amplified by each processing, and in a studio sound can be precessed lots of times while being produced (delay, reverb, eq, compressor, limiter, aural enhancer, dithering, etc).

I use 96Khz in software pianos, but mainly because that halves the buffer latency for the same sample size (compared to 48Khz), but not for audio quality reasons.

Regards,
Kurt.-

The imperfections you speak of don't exist. You've omitted any mention of low-pass filtering, which eliminates those "edges".

Believe it or not, the interpolator (reconstruction low-pass filter) will take care of that and give you a perfect reconstruction of the original continuous input signal.

The Nyquist-Shannon sampling Theorem states: If a function x(t) contains no frequencies higher than B hertz, it is completely determined by giving its ordinates at a series of points spaced 1/(2B) seconds apart.


In fact sample rates above 44.1kHz have no audible effect (LINK).

A better question to ask is how many bits you need for your sampled application. If you are recording you probably want more than 16 bits to give yourself some headroom before the digital clipping point. If you are filtering the extra bits can make the filter design easier by lowering accumulator truncation noise, though even very wide accumulators normally employ some form of truncation noise energy suppression.

[edit]If you are mixing down then a final 16 product should be sufficient to give you almost 100dB of dynamic range.

If you are just listening to audio content then there's no reason to go beyond 16/44. So the cynical answer to your question is: the music industry needed something other than more songs (=hard work) to fill future physical standards that will almost necessarily accommodate increased data. More bits and at higher rates mean more data. The average consumer thinks more is better (and they are generally right). Certainly upgraded physical formats used to bring all kinds of benefits to the consumer, but lately they seem more like a patent / standards / DRM lock-in by large corporations, and a pointless reason to "upgrade" entire expensive collections. All they have to do is spread FUD that the old standard is somehow inferior to the new.

Thanks for the link about the sampling rate study Dewster - very very interesting indeed!

Greg.

Hi - it is my impression - and also experience - that there are many aspects to whether or not sample rates are actually audible. There may be a given person's training in listening, whether or not the person is relaxed, the person's state of feelings on a given day, not least the person's general level of sensitivity to subtleties as such, the audio system's capabilities of reproducing nuances etc.

I used to work with high-end audio and over the years my realization was that there are indeed differences between e.g. sample rates but much depends on the above factors as well (and there's even more factors like e.g. the room's qualities). I usually recommended customers that they listened themselves and made their conclusions based on this. In this way many of the above factors would automatically be accounted for because either the person noticed the differences - or not ...

And then maybe some of you have noticed something similar to this: In my experience music really can be like massage or like other of the finer/subtler joys of life: it takes time, patience, respect, relaxation & a person's ability to connect with his or her "inner", be open. And when this happens even the subtlest nuances can he heard or sensed ....

And then a couple of practical comments on the sample rate issue:

- there's been made a study examining the psychological effects of including sounds above 20 kHz. This study found that there's indeed a positive effect of doing so. See e.g.

Journal of Physiology

- The CD's 44.1 kHz allows for sampling sine waves of up to 44.1 Khz/2. When the signal is not a sine wave - which is the case with music - the samples most likely will be incorrect possibly both in terms of actual frequencies and amplitudes sampled. Music also contains signals above 20 kHz. See e.g.:

Music above 20 kHz

Reconstruction filters may be implemented but as far as I know for them to precisely reconstruct the original signal their algorithms need to be infinitely complex which is not (yet?) the case as it requires a very high data processing power.

A low-pass filter may also be implemented but for it to be completely transparent it needs to add no amplitude and phase distortions - or other possibly unmeasurable distortions.

See e.g. here for a subjective comparison of capacitor sounds (capacitors are often used in low-pass filters e.g. in a cd player).

capacitor comparison

A couple of years ago I spoke with Tim de Paravicini of Esoteric Audio Research (a high end audio company) and he said that until the cd - or a similar media - had a resolution of 24 bits/600 kHz sampling rate it would not be as good as the analog media. To me an interesting observation - and very much in line with my own observations, although I believe that there's really no limit to what the human ear - or other senses assisting the ear - can register.

And then - to revert to the beginning ... my experience is that the sensitivity to subtleties/nuances really is key when "experiencing". In a period of my life I was indeed very sensitive and during this period I had superb experiences with playing back music on my current mini-stereo (with reasonably good speakers). Experiences that - when it comes to feelings, joy, intensity, presence, nuance etc. - in many respects were more life-like than what I had experienced with my previous ~ USD 30.000 stereo system.

Mind-boggling to me

Best wishes for your Sunday,

Jesper

Pfff...I'm in this for the long-haul!

Well from a math pov every signal is actually a sum of sine waves, I'm not an audiophile but that I know

The key phrase here is "up to" - you can't include one-half the sampling frequency.

Also, any waveform presented to the DAC with a fundamental above one-quarter the sampling frequency will necessarily be a sine wave, otherwise it's harmonics would alias back down into the passband.


True, but using a low pass filter for reconstruction is generally fine.


If you look at the specs for modern oversampled DACs the passband ripple is something like 0.1dB. The majority of the low-pass filtering is done digitally in the oversampled domain, which makes the design of the filter more tractable.


No personal offense intended, but that link is an excellent example of what is wrong with the dizzying subjectivity so egregiously rampant in the audiophile community.

Also, modern DACs mainly use digital filtering.

"Everyone is entitled to his own opinion, but not his own facts" - Daniel Patrick Moynihan.

I really appreciate all this technical talk about sample rates. In the end. All that matters is how it sounds. Yes...my dizzying subjectivity that is rampant. Really don't mean to argue. On capacitors...David Hafler started much of this with polypropylene capacitors. Before one can comment. You have to listen to female vocals through such an amplifier. Absolutely beautiful, obscenely intimate ....so clear you can hear her mouth move...you can hear every little nuance of her breathing...you can almost hear her sweat... is that subjective enough?

Is there anything to tie the type of capacitors to your being able to hear "every little nuance"?

I have two objections to your argumentation:

1. Believe it or not the interpolator is not perfect so you don't get exactly the same wave, just an approximation. Certainly a very good one, but not exactly the same.

2. The key words on the Nyquist-Shannon sampling Theorem are "contains no frequencies higher than B hertz" that is simply not true when dealing with real sound.

My point is: 96 Khz sampling gives a better representation of the original wave than 48 Khz (and if I am not missing something, this is true, with or without interpolators). Working at 96Khz (and even higher) definitely worth it if the sound is processed. Is it worth it for the sake of reproduction and nothing else? Not sure... I think that for some people it's worth it, and for the majority (me included) it is not... everyone can choose what best fits to him/her and having choices is good.

Regards,
Kurt.-

The point is ... when digitizing audio you must filter out all signal components above 1/2 the sampling rate. This is handled by a low-pass filter.

When sampling at 44 kHz, you must filter out all sound at or above 22 kHz. Easy.
For 48 kHz sampling, filter at 24 kHz.
For 96 kHz samping, filter at 48 kHz.

It is well known. David Hafler built a famous reputation on it.

Of course you need everything else. Even though David's stuff wasn't super expensive. There are people who believe his amplifiers were the best mid range...period. I owned a pair of DH220's. Bridged over to mono.

Again....I thank you all for the technical talk. It is interesting.

"It is well known" doesn't address the question. I'm wondering HOW is it known?

Please excuse me. But...This sounds a little to me like: "How is it known that the sky is blue?"
Well....David said so. Then other manufacturers of high end audio tried the same thing and got similar results. Then we all lived happily ever after.
As far as everyone asking David how he got such great mid range. They listened to his amplifiers. Were very impressed. Then they asked David why his amplifiers were this way. David told everyone. Then we all lived happily ever after.
Except David...as I understand he has passed away.

Shall we now try for: HOW is it known that Brittany Spears is dangerously cute?

Doh!

D'oh back to you.

Anyone can see the sky. But who has seen (heard) "David's" equipment?

If all they said was that A sounded better than B, then surely someone else will say B sounded better than A.

It would be more meaningful to see the measurements. The test results. Can you reference them (and drop the sarcasm)?

Many audiophiles bought stuff from David Hafler. I bought two of his amplifiers.
What was special on his amplifiers was his mid range. Was hard for anyone to argue beating his mid range. Ones that did argue, cost multiples more. His highs and lows weren't as super special as his mids. Weren't at all bad though.
I really have no idea where to find any kind of test measurements to rationalize what goes on with polypropylene capacitors in amplifiers versus other capacitors.

Well, now we're back on point.

So the question is ... Does his equipment sound better purely because of the caps? Were A/B comparisons made in which the only difference between A and B was the type of caps ... and everything else the same?