I have what I reckon is possibly a quick question that I would appreciate help with …
As it is I would like to be able to record & playback at two different sample rates, e.g. record at 768 kHz/32 bits and at the same time play back this recorded 768 kHz signal at e.g. 192 kHz/32 bits. Any of you know if this is possible/practically feasible?
Also, does the recording clock need to be the same as the playback clock, i.e. come from the same oscillator although it may e.g. be divided to allow for 768 kHz recording & 192 kHz playback?
It’s not impossible, but without anymore context I’d have to remark that its a totally crazy idea that I can’t see any purpose for.
Ardour can resample all of its output, so if you had two devices (each with their own clock, or a single clock with an appropriate divider/multipler), it could theoretically resample the output to match the device.
Also, there are no 32 bit signal paths. There are no converters that use 32 bits and if they did they’d be picking up atomic brownian motion.
Just briefly I can say that my reason for asking is that my DAC works best at 384 kHz whereas the ADC works best at 768 kHz (or 1.536 MHz. It is in the design phase so not yet completed). So I was at least considering using different sample frequencies within Ardour for the two converters.
And, yes, I know 32 bits are at least currently illusory … but who knows if in say 15-20 years it may - by means of e.g. an “artificial intelligence quantum computer” - be possible to extract the sound that today is considered hidden in either distortion of noise … ? Anyway, just thinking aloud here as both the ADC (to be) and the DAC works with 24 bit resolution (which is also currently ambitious in practice but I reckon not within reach in the years to come).
Yes, there’s limits with modern electronics, and these limits can be lowered with supercooling.
These limits are caused by the physical properties of materials at normal temperatures and how we use them for electronic circuits at a fundamental level. We would need a completely new replacement technology for electronics to remove these limits, and I don’t see any currently even in sight.
I don’t see how quantum computing will help in any way; it’s not even currently being proposed (by those who actually understand it) as a complete replacement for current digital computing, and certainly isn’t looking like a replacement for modern analog electronics.
But the real limits of this are human hearing and, for all practical purposes (music, home theatre) the noise floor of even CD standard 16-bits is significantly below that required and well below the background noise floor of any listening environment you are likely to encounter.
Which is why most commercial music, even carefully recorded and produced jazz or classical music, rarely has a dynamic range of more than around 20-30dB.
24-bits exceeds the resolution of the human ear, assuming you push human hearing to it’s limits, risking permanent damage. Reduce the SPL to safe levels, and well-dithered 16-bit material can capture just about everything.
And even if there was a way in the future to enhance our brains to experience ultra-sonic, hyper-sensitive hearing, anything you record today is going to be limited by the noise-floor of the equipment you are using today.
Anything you record in 32-bit integer formats with current generation electronics (pending some revolutionary and fundamental replacement for electronics) will contain at least 8-10 bits of noise.
… First I would like to thank you for reading and considering my post - & also replying.
And then I feel like treading carefully here because it is my impression that we come from very different points of view. Which is entirely fine but I have just been close to such a talk before and it is my experience that it may either not be very fruitful personally or in terms of the end “outcome”.
And since being on good terms and respecting other people matters to me I think I will choose to leave it here … But, again, thanks for considering & replying - I appreciate you took the time and attention to do so
I know the original poster has bowed out since this isn’t really germane to the original question, but out of curiosity I spent some time investigating from a system perspective what the “true” limits are (e.g. if you posit that some magical…err…“quantum?” technology is developed in the future which eliminates the electronics as a limitation).
The accepted brownian noise floor of room temperature air molecules is calculated to be around -24dB SPL (I will stick with room temperature values for the sake of the musicians).
The point at which the rarefaction cycle of a sound waves hits 0 mb pressure is 194dB (above that you get a lot of distortion of the air itself, so the air starts making those crackling noises you get when rockets launch).
So to cover from brownian motion to limited by air pressure is around 218dB, which you could cover in 36-ish bits.
With 32 bits you would have to limit yourself to covering brownian air molecule motion to “only” 168dB SPL. I guess you could shift your window and cover from 194dB SPL down to “only” 2dB SPL.
That is above the best case minimum audible level for a young person who has lived a sheltered life, who may be able to barely detect a 3kHz tone at -6 dB SPL, but of course once such a person has been in the vicinity of 168 dB sounds that person can no longer detect -6dB tones, so mostly academic in the context of audio recording and playback.
All good and interesting stuff. Thanks for that fascinating analysis.
By the way, the following article is illuminating:
I will point out (and I was very careful to use the word “fixed” in my post), that, if one wants massive dynamic range (and there are cases where it’s helpful), there are 32-bit floating point recorders already on the market which give a dynamic range of up to 1,528 dB, which is more than enough to capture, literally, any sound on earth, and which already exceed the capability of the OPs hypothesised system.
The advantage of this, of course, is not that we need such dynamic range for human hearing, but that it negates the need for setting gain levels when capturing as you will never clip.
It’s kind of like automatic gain control, except it works by providing a massive dynamic range.
When I have questioned the need for this I have been educated by others who have the need to capture audio in challenging environments where, previously, multiple microphones might have been needed (which can present it’s own problems).
One such environment, apparently, is recording church organs in some cathedrals where the SPL can range from pretty loud to very quiet, and where the acoustics are challenging for recording with multiple mics.
Of course, at some point the audio has to be converted to something that can be rendered and heard by humans, and those are all fixed-point formats, so some work is required in the DAW to reduce the dynamic range to something rational, and to prevent clipping.
Also note that the “no clip” isn’t a feature of 32 bit float ADC (though it helps), despite Zoom’s marketing: the use of dual ADC would allow 140+ dB dynamic range even in 24 bit fixed.
You could potentially do this. However, you would need a driver that would allow for this. I have used mixed sample rates with zita-a2j and zita-j2a. This is a hacky solution though, and probably won’t achieve the best sound quality due to resampling occurring. Recording at 768khz seems like you are trying to record very high frequency signals, slowing them down, then playing them back at a slower speed. If you are using this for audio frequencies, then this is massive overkill. 192khz at both recording and playback is more than enough and would be far more practical for audio. using those things you could do it. I don’t think JACK could even go that high without massive overhead. So there you go.
…