At what point is "low noise" low enough for dac supplies?
Where does the audible threshold end?
Where does the audible threshold end?
It isn't simply a matter of noise. There is a problem we appear to have since all the ultra-low noise LDOs started appearing. Sometimes people seem to focus too much on noise, to the extent of failing to adequately consider other important factors. In addition, dacs are particularly complex in the sense that there is audio, RF, digital, analog, etc. Different subsystems might reasonably be expected to have somewhat different power supply considerations.
Last edited:
Totally agree with you Markw4. This may sound crazy but certain parts of circuit will sound good with linear regs while another part of the circuit will sound good with shunts. It’s not all about low noise & speed. It’s more of using the right component to get the best performance.
Finding out which is best is the thing - no one can give a definitive answer.
Given the amount of rf present in the environment from phones, wifi etc (and getting worse thanks to the IoT) using some copper screening around the pcb may have more effect than a change of regulator.
Speaking of which Batteryman, I tape an copper foil on top of my TDA S1 shouldered a wire on the foil & connected it to digital gnd of chip, guess what you hear an immediate improvement in clarity. Yes it is long & tiring to be testing & testing making changes but if you want to attain great sound, this is about the only way to go.
What other important factors?
I'm trying to decide whether it's worth my time and money to make a discreet power supply or to just use the LDOs. The LDOs have super-reg level output impedance and less than 1uv of noise. I can do better with discreet but I don't understand how this is going to be audible, I feel like we are chasing ghosts.
If it is audible, what is the technical reason that allows it to be audible?
Can I assume that no one has an actual answer as to why 1uv noise power supplies are needed for dacs? I'm tempted to call BS.
You do not 'understand', you 'experience', how something is audible.
If you want to know the 'technical' reasons, I suggest you do your own study of it. Don't know anyone to answer the question, I don't know with any degree of certainty, and I don't care all that much, since I have heard the 'proof' of experience in my own audio system.
Dave
You are preaching to the quire.
But no progress can be made without understanding.
Whether or not a 1uv noise regulator can be heard is not the core issue.
I've never seen any evidence anywhere as to why such excessively low noise is required for dacs.
Without a solid understanding, the reasoning for an audible improvement can be any numbers of things unrelated to noise.
Presumptuousness is not wise.
But no progress can be made without understanding.
Whether or not a 1uv noise regulator can be heard is not the core issue.
I've never seen any evidence anywhere as to why such excessively low noise is required for dacs.
Without a solid understanding, the reasoning for an audible improvement can be any numbers of things unrelated to noise.
Presumptuousness is not wise.
It would seem to depend on which particular dac power supply being discussed. ESS Sabre dacs use a reference voltage called AVCC that has zero PSRR. Noise on Sabre dacs is specified to be down at -122dB or lower (depending on model and configuration). -120dB is one part in a million. AVCC is recommended to be run at +3.3v. For AVCC, it could be that more than microvolt noise levels deteriorates performance. What may be worse in the case of AVCC is that noise there is not just additive with the audio output, it is multiplicative. That mean noise on AVCC can intermodulate with the audio signal causing increased intermodulation distortion above the minimum, too. Can anybody hear it? Perhaps related to claimed audibility of noise floor modulation behavior of delta-sigma dacs, it appears that some people possibly could hear AVCC-supply-sourced noise/side-effects and find it objectionable. Would the same consideration apply equally for the DVCC supply? Seems very unlikely, IMHO. Still don't know what the spec should be for that one, though. Up to the designer to figure it out.
Last edited:
So AVCC powers the internal output stage of the dac + a voltage reference or is AVCC itself the voltage reference?
There is a rapidly switched resistor network that determines the analog output. AVCC supplies current to the resistor network and needs to be maintained at a stable fixed voltage. You might want to read some of the material on the ESS website downloads page to see what you can pick up from that. For one example: http://www.esstech.com/files/4314/4095/4318/sabrewp.pdf
Last edited:
Are there audio circuits where noise is desirable? Or, at least where 'higher' amounts(of noise) doesn't necessarily matter?
Noise has been used by some designers of low cost equipment to mask distortion. Not unusual for noise to sound better than high-order harmonic distortion.
I wouldn’t say noise is desirable, its more of achieving a senergy be it regs, caps etc. What I feel is that we should not be caught up just nick picking on one particular area.
Quite. I've been harping on about load-induced noise for a very long time. Unless your load is drawing a truly signal-invariant current then the likelihood is that your regulator's output impedance is the first of your worries, not its noise credentials.