Well you see, my train of thought went like this - typical thermal noise in a dac circuit is about a microvolt, so on a 5v DAC you surely can't even reach 23 bit resolution. maths: 1 microvolt is of course one five-millionth of 5v, and 23 bit audio has 8 million odd voltage levels. Above five million, your noise exceeds the distinction between levels, so even an idealised perfect ADC couldn't recover the data accurately. With 24-bit audio, noise is more than three times the signal resolution.
Obviously this then gets amplified. What I figured was that if you cut out that 1 microvolt of thermal noise, then you have greater resolution. But of course the various means of cutting out thermal noise are not easy to achieve.
However, I then realised that once the DAC signal is amplified, say, to 50v, that 1 microvolt is equivalent to a third of the signal resolution. maths: 50v divided by 16 million (which is 2^24, ie 24 bit res) is about 3 microvolts. This is of course 3x the thermal noise.
Soooo... if you get cryogenic superconducting circuit (and we all know that for some people, the money they spend on hifis could easily accomodate this!), thus eliminating thermal noise from the DAC, amplify it, and then no more need for cryocooling! Since thermal noise doesn't depend on voltage, the thermal noise on the 50v stage is still about 1 microvolt, ie 1/3 of the signal resolution!
Cryogenising a circuit could be easily and silently done with a Stirling cryocooler. What do you think, am I mad or is there a spark of genius in it somewhere?
Obviously this then gets amplified. What I figured was that if you cut out that 1 microvolt of thermal noise, then you have greater resolution. But of course the various means of cutting out thermal noise are not easy to achieve.
However, I then realised that once the DAC signal is amplified, say, to 50v, that 1 microvolt is equivalent to a third of the signal resolution. maths: 50v divided by 16 million (which is 2^24, ie 24 bit res) is about 3 microvolts. This is of course 3x the thermal noise.
Soooo... if you get cryogenic superconducting circuit (and we all know that for some people, the money they spend on hifis could easily accomodate this!), thus eliminating thermal noise from the DAC, amplify it, and then no more need for cryocooling! Since thermal noise doesn't depend on voltage, the thermal noise on the 50v stage is still about 1 microvolt, ie 1/3 of the signal resolution!
Cryogenising a circuit could be easily and silently done with a Stirling cryocooler. What do you think, am I mad or is there a spark of genius in it somewhere?
24 bit is just a marketing ploy. People always want bigger numbers. Don't fall for it.
Can you hear the thermal noise from your dac? If no, then you also can't hear the bits "below" it, so no point in trying to reconstruct them.
Can you hear the thermal noise from your dac? If no, then you also can't hear the bits "below" it, so no point in trying to reconstruct them.
Hi,
the incoming signal to the ADC is a preconditioned one. There will be analog stages sitting before the ADC´s fronted. Those will set the limits not the resolution of the converter. Most of the analog frontends will probabely not even reach 20BITs of resolution. You should rather fight noise and distortion here. The ADC is not the prime patient.
jauu
Calvin
the incoming signal to the ADC is a preconditioned one. There will be analog stages sitting before the ADC´s fronted. Those will set the limits not the resolution of the converter. Most of the analog frontends will probabely not even reach 20BITs of resolution. You should rather fight noise and distortion here. The ADC is not the prime patient.
jauu
Calvin
edit: first off, very interesting tangent 🙂
yes, and what about resistors in dacs?
maybe they must be external, or maybe voltages supplied to the dac.
imagine superconducting temps, no internal voltage drops ideally, resistive at least.
yes, and what about resistors in dacs?
maybe they must be external, or maybe voltages supplied to the dac.
imagine superconducting temps, no internal voltage drops ideally, resistive at least.
Last edited:
interesting responses folks 🙂
so....
@kavermei - surely an audio system is inherently lossy. You get your source material at a certain quality, and at every stage, for example, CD reading, decoding, analoguing, amplifying, and playing through the speakers and ultimately through the air to your ears something is lost. I see this as cumulative - ie, your CD read head is not bit perfect, thus limiting the whole system, or your dac is not perfect, thus limiting, or your wires are not shielded so susceptible to interference... it goes on. At every stage, quality is lost, that's why every bit of a hifi is worth upgrading...? My proposal is simply to eliminate one of the causes of poor quality. Just like using higher and higher quality cables, people will argue forever about whether you can hear a difference, but the science is there. My hope is just that if implemented, my music will sound better.
@Calvin - sorry, I'm a bit confused. You're talking about ADCs, where my system is about DACs. Are you referring to the source material having effectively less resolution? If so, I take your point, but to recover it inaccurately will only add to existing errors, surely? I suppose the question of source error being an order of magnitude higher than DAC error means DAC error is sort of irrelevant, but it does ultimately add to the existing error.
@woody - yes, pretty much, a superconducting circuit would have to be created, you can't just supercool an existing one.
@jechentau - thanks 🙂 My knowledge of cryotronic design is very limited though I am studying hard. Certainly right now it seems resistors are the kind of obvious problem, my knowledge is more about cryocooling than electronics. I'll see what I can find out
so....
@kavermei - surely an audio system is inherently lossy. You get your source material at a certain quality, and at every stage, for example, CD reading, decoding, analoguing, amplifying, and playing through the speakers and ultimately through the air to your ears something is lost. I see this as cumulative - ie, your CD read head is not bit perfect, thus limiting the whole system, or your dac is not perfect, thus limiting, or your wires are not shielded so susceptible to interference... it goes on. At every stage, quality is lost, that's why every bit of a hifi is worth upgrading...? My proposal is simply to eliminate one of the causes of poor quality. Just like using higher and higher quality cables, people will argue forever about whether you can hear a difference, but the science is there. My hope is just that if implemented, my music will sound better.
@Calvin - sorry, I'm a bit confused. You're talking about ADCs, where my system is about DACs. Are you referring to the source material having effectively less resolution? If so, I take your point, but to recover it inaccurately will only add to existing errors, surely? I suppose the question of source error being an order of magnitude higher than DAC error means DAC error is sort of irrelevant, but it does ultimately add to the existing error.
@woody - yes, pretty much, a superconducting circuit would have to be created, you can't just supercool an existing one.
@jechentau - thanks 🙂 My knowledge of cryotronic design is very limited though I am studying hard. Certainly right now it seems resistors are the kind of obvious problem, my knowledge is more about cryocooling than electronics. I'll see what I can find out
quick update - it's all more than possible, cryotronic dacs exist, they work using what's called rapid single flux quantum electronics. it's fascinating, transistors are replaced with things called josephson junctions, and it's compatible with conventional logic - signals are converted to extremely short pulses.
My thinking really is that designing lossless digital circuits is comparatively simple. Harmonics and the inherent problems of square functions are still issues, but can be de facto removed in well-designed circuits. But the analogue stage, which can never be removed, is by definition susceptible to everything out there and surely that's where all innovations are needed!
My thinking really is that designing lossless digital circuits is comparatively simple. Harmonics and the inherent problems of square functions are still issues, but can be de facto removed in well-designed circuits. But the analogue stage, which can never be removed, is by definition susceptible to everything out there and surely that's where all innovations are needed!
Soooo... if you get cryogenic superconducting circuit (and we all know that for some people, the money they spend on hifis could easily accomodate this!), thus eliminating thermal noise from the DAC, amplify it, and then no more need for cryocooling! Since thermal noise doesn't depend on voltage, the thermal noise on the 50v stage is still about 1 microvolt, ie 1/3 of the signal resolution!
Cryogenising a circuit could be easily and silently done with a Stirling cryocooler. What do you think, am I mad or is there a spark of genius in it somewhere?
Oh dear, so complicated. Here's an alternative: preemphasize the digital signal before it hits the DAC, since high frequency noise is more audible than LF noise, and apply de-emphasis in the amplifier feedback loop. Voila! Your perceived DAC noise is reduced a bunch of dB, and internal amplifier noise is also knocked down.
This, by the way, is not a serious suggestion, although I'd like to find out the result if anyone tries the technique. Hmmm... It could be interesting for systems with digital volume control, actually, although the extra high frequency energy would work the reconstruction filters pretty hard and you'd lose some headroom depending on how much HF energy was in the source. Oh well, something else to work on in my copious free time.
- Status
- Not open for further replies.