Decoupling digital supplies
There are a few puzzles over the practice of 'audiophile' decoupling - for example why NP0 are preferred over hi-k types (piezo-electricity seems to be a no-no but I haven't heard an explanation as to why). Yes, hi-k are non-linear in many ways, but how does that affect the sound when they're in the power supply?
Perhaps you're thinking of 15mm lead spacing ones. Wima does polyprops (FKP02) which have a self-resonant frequency above 10MHz, they're 2.5mm lead spacing. Can't think why I'd want to use them for decoupling though as their ESRs are none too spectacular ( 0.1ohm-ish ). I've played with a Kemet software package (free on their website) in the past week - found that high value Z5Us have a very low ESR which make them (on paper at least) excellent high freq (say 1-10MHz) decouplers. Above 10MHz, getting low ESR turns out to be quite hard...
Yes - 0.22uF ceramics will crap out around 10MHz, we need to go to a lot lower value (100X lower as the SRF depends on the square root of capacitance) to reach up to 100MHz.
There are a few puzzles over the practice of 'audiophile' decoupling - for example why NP0 are preferred over hi-k types (piezo-electricity seems to be a no-no but I haven't heard an explanation as to why). Yes, hi-k are non-linear in many ways, but how does that affect the sound when they're in the power supply?
Perhaps you're thinking of 15mm lead spacing ones. Wima does polyprops (FKP02) which have a self-resonant frequency above 10MHz, they're 2.5mm lead spacing. Can't think why I'd want to use them for decoupling though as their ESRs are none too spectacular ( 0.1ohm-ish ). I've played with a Kemet software package (free on their website) in the past week - found that high value Z5Us have a very low ESR which make them (on paper at least) excellent high freq (say 1-10MHz) decouplers. Above 10MHz, getting low ESR turns out to be quite hard...
Yes - 0.22uF ceramics will crap out around 10MHz, we need to go to a lot lower value (100X lower as the SRF depends on the square root of capacitance) to reach up to 100MHz.
Hi,
With a big value capacitor and an emitter follower.
I'd post the complete circuit, but there are too many chinese copiers watching and too many diy'ers that will copy circuits without understanding and promptly misapply them
Ciao T
With a big value capacitor and an emitter follower.
I'd post the complete circuit, but there are too many chinese copiers watching and too many diy'ers that will copy circuits without understanding and promptly misapply them
Ciao T
Hi,
I usually use 0603 0.1uF ceramics or 0.1uF X2Y capacitors for such decoupling, local as close to the chip pin as I can. Then bigger ceramics and Os-Cons. I normally poke the differential probes of a > 100MHz 'scope around the PCB to check my decoupling works.
I fond for generic digital audio stuff sound quality is often directly inversely proportional to the noise at the PSU Pin's (for non-generic stuff I expect it to have been minimised as the first step).
Ciao T
I usually use 0603 0.1uF ceramics or 0.1uF X2Y capacitors for such decoupling, local as close to the chip pin as I can. Then bigger ceramics and Os-Cons. I normally poke the differential probes of a > 100MHz 'scope around the PCB to check my decoupling works.
I fond for generic digital audio stuff sound quality is often directly inversely proportional to the noise at the PSU Pin's (for non-generic stuff I expect it to have been minimised as the first step).
Ciao T
For the WM8805, it has separate PLL 3.3V VDD at 1.7mA so I was going to use a 47uH SMT inductor and 160uF and 0.1uF caps for decoupling, running off the digital VDD.
Separate voltage regulators seems to be popular, but I have trouble with that on a 2mA load as most vregs get unstable and DVDD and PVDD must be within 0.3V of each other. That means clamp schottky diodes also.
The Wolfson W8805 evaluation board either ties it all to USB power (ugh!) after a 3.3V reg OR it's off to banana posts for PVDD. Until I get a proto running and scope everything, I'll have to guess at how to get the best jitter performance.
Separate voltage regulators seems to be popular, but I have trouble with that on a 2mA load as most vregs get unstable and DVDD and PVDD must be within 0.3V of each other. That means clamp schottky diodes also.
The Wolfson W8805 evaluation board either ties it all to USB power (ugh!) after a 3.3V reg OR it's off to banana posts for PVDD. Until I get a proto running and scope everything, I'll have to guess at how to get the best jitter performance.
Hi,
Try 100uH Ferrites and add a big value (several uF) SMD Y5U Cap in addition ot a 0.1uF SMD in the smallest format you can solder (I'd use 0402 if I could still see them reliably).
Ciao T
Try 100uH Ferrites and add a big value (several uF) SMD Y5U Cap in addition ot a 0.1uF SMD in the smallest format you can solder (I'd use 0402 if I could still see them reliably).
Ciao T
For those of us with poor close-in eyesight there is some relief available in the newer 'reverse geometry' caps. Instead of being 0805 (about my limit for seeing the beasts clearly) they now also come in 0508 where the terminations are along the longer edges. These have a worthwhile decreased inductance than their corresponding 'widescreen' variants.
Yes, you are right. I care about the sonic performance and I will wait untill you finish the project.
Thanks!
Just to keep you updated on the project.
The ESS board has been playing for some time with our own discrete analog stage, but the result was not really pleasing, which might be due to other things than the chip itself.
Ie. we tried to use I/V conversion done by the AD797 originally used by ESS, and from there the signal was sent into our analog stage.
But the result was not in anyway "high endish" or just close to that.
So more investigations are to go into the project.
As a spin off from it, we are considering to do a small jitterkiller for use together with amongst others a digital streamer as i.e. the Squeezebox.
The jitterkiller will be done with a Wolfson transciever and a precissionclock with smooth supplies, and maybe a low noise supply for the Squeezebox itself.
By now I´m waiting for my Squeezebox touch to arrive, then we will go on from there.
If it turns out right, we might start a new DIY project based on that.
But I must admit, my hopes are not really that high.
The ESS board has been playing for some time with our own discrete analog stage, but the result was not really pleasing, which might be due to other things than the chip itself.
Ie. we tried to use I/V conversion done by the AD797 originally used by ESS, and from there the signal was sent into our analog stage.
But the result was not in anyway "high endish" or just close to that.
So more investigations are to go into the project.
As a spin off from it, we are considering to do a small jitterkiller for use together with amongst others a digital streamer as i.e. the Squeezebox.
The jitterkiller will be done with a Wolfson transciever and a precissionclock with smooth supplies, and maybe a low noise supply for the Squeezebox itself.
By now I´m waiting for my Squeezebox touch to arrive, then we will go on from there.
If it turns out right, we might start a new DIY project based on that.
But I must admit, my hopes are not really that high.
Hi Kurt,
Please look closely at the analog supplies of the ESS chip. A Op-Amp regulator there makes "Op-Amp" sound. Also, try using passive I/V conversion (resistor) and simply turn your discrete analog stage's gain up to compensate.
Ciao T
Please look closely at the analog supplies of the ESS chip. A Op-Amp regulator there makes "Op-Amp" sound. Also, try using passive I/V conversion (resistor) and simply turn your discrete analog stage's gain up to compensate.
Ciao T
Hello again,
A short note about the ESS demo board filter: it is quite far from a linear phase type (made with noise rejection in mind), the step response has about 5% overshoot and in this particular implementation the sensitivity for two of the components (C4, C5) is quite high. Eg. if these values are 3% 'offcenter' the overshoot simply doubles.
Of course something better is needed for high-end, including supplies and the rest...
Hey you fine Guys
I do not really know where the bug might be burried, but I think we might get further on trying to finish the design as it should be done all over.
And I think a passive I/V conversion stage might do some of the trick.
As I recall our first "well" functioning prototype of the earlier design DAC featuring the CS4398, this sounded almost just as bad as the ESS does by now
, except that the earlier one had a much more realistic low end, which actually made us persue the potential of it, but the lack of that in this design, might be a result of the I/V stage AD797.
So it looks as if it might have some potential after all
I do not really know where the bug might be burried, but I think we might get further on trying to finish the design as it should be done all over.
And I think a passive I/V conversion stage might do some of the trick.
As I recall our first "well" functioning prototype of the earlier design DAC featuring the CS4398, this sounded almost just as bad as the ESS does by now
, except that the earlier one had a much more realistic low end, which actually made us persue the potential of it, but the lack of that in this design, might be a result of the I/V stage AD797.So it looks as if it might have some potential after all
Last edited:
It is not magic.
But pure chemistry to solder SMD components.
YouTube - Lead-Free Multi Lead Soldering | Soldering Tips | B.E.S.T.
I think this project died R.I.P.
Yes, I consider this project dead
You can't get enough of a group consensus on what to build, then it seems to have become one designer's baby. Not really an open project.
I would have done a modular DAC design- split up the PSU, the digital receiver, the DAC chip, output stage etc. so people can choose what they want, instead of some single-board.
Yes, I consider this project dead
You can't get enough of a group consensus on what to build, then it seems to have become one designer's baby. Not really an open project.I would have done a modular DAC design- split up the PSU, the digital receiver, the DAC chip, output stage etc. so people can choose what they want, instead of some single-board.
What you would have done is exactly what the TPA guys have done with the Buffalo II DAC. State-of-the-art ES9018 DAC chip, no need for a digital receiver and you can choose whatever PSUs and output stage you fancy. Difficult to make something better or more flexible at a significantly lower price.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.