Hi !
sorry for a double post here but this issue is very important to me.
Power supplies and their influence on digital devices.
I have been advised not to loose the sleep ... that just decent PS would do very fine because digital is more robust to noise that analog for instance. 🙄
Then i see amazing projects of ps for digital devices trying to push down the noise up to tenth of uV 😱 (especially for clock i would say but also usb receivers and dacs)
Who is right ?
Just this. I have to get a definitive answer. My sanity is at stake here.
Are the designers pushing down the noise to uV levels playing a game ? is that a competition an end in itself ?
Thanks a lot.
Have a nice weekend, gino
P.S. i tend to think that noise makes a difference
but i could be very well wrong ... with digital i mean.
sorry for a double post here but this issue is very important to me.
Power supplies and their influence on digital devices.
I have been advised not to loose the sleep ... that just decent PS would do very fine because digital is more robust to noise that analog for instance. 🙄
Then i see amazing projects of ps for digital devices trying to push down the noise up to tenth of uV 😱 (especially for clock i would say but also usb receivers and dacs)
Who is right ?
Just this. I have to get a definitive answer. My sanity is at stake here.
Are the designers pushing down the noise to uV levels playing a game ? is that a competition an end in itself ?
Thanks a lot.
Have a nice weekend, gino
P.S. i tend to think that noise makes a difference
but i could be very well wrong ... with digital i mean.
Last edited:
Suspend your question for just a moment. Your sanity is not at stake.
I have not designed any power supplies, so this is only coming from an application point of view.
Class D devices tend to have power supply ripple rejection engineered into the device for flexibility of application.
Chip devices, can have some of this robust nature too.
Solid State devices can have these attributes, but it depends on the design and implementation, but they can also have large power requirements, as can tubes depending on their design as well.
Bottom line, if you are using anything other than Chip or class D for your project, it can be a noticeable problem if poorly done.
Power supply is one of the parts of the chain from the wall socket to the speaker. Everything influences something to different degrees.
The question is; do your parts work well together for your needs.
So, before you decide who is right, decide what your needs are.
Then you can decide what is important and who knows enough about your needs that you can consult.
I have not designed any power supplies, so this is only coming from an application point of view.
Class D devices tend to have power supply ripple rejection engineered into the device for flexibility of application.
Chip devices, can have some of this robust nature too.
Solid State devices can have these attributes, but it depends on the design and implementation, but they can also have large power requirements, as can tubes depending on their design as well.
Bottom line, if you are using anything other than Chip or class D for your project, it can be a noticeable problem if poorly done.
Power supply is one of the parts of the chain from the wall socket to the speaker. Everything influences something to different degrees.
The question is; do your parts work well together for your needs.
So, before you decide who is right, decide what your needs are.
Then you can decide what is important and who knows enough about your needs that you can consult.
digital has noise margins on a decision whether it's a 0 or 1 AND makes its own noise on the PS rails and even the ground, so it's silly to provide a clean supply lower than what will be corrupted. now if there are analog functions built in hopefully the designer took his own precautions. modern digital has moved towards serial data buses to further improve noise margins between subsystems and unbelievably good error correction schemes. so rest easy they've got you covered.
edit> if it has a clock recovery subsystem, there are bad ones that are more susceptible to PS modulation so IMO the crappy stuff needs better engineering to make up for the bad original ones.
edit> if it has a clock recovery subsystem, there are bad ones that are more susceptible to PS modulation so IMO the crappy stuff needs better engineering to make up for the bad original ones.
Last edited:
gino, again it's straightforward ... the only thing that really matters in our audio world is that the analogue signal being fed to the speakers is as clean as we possibly can engineer. And in an ideal world the part of the system that treats the signal as an analogue waveform, after being processed by various bits of digital circuitry at the start of the chain, is completely independent of what happens in that digital area. However, we live in an imperfect world, the analogue circuitry is not perfect in its ability to reject interference effects originating in the digital area - so, making the digital PSs better may help things, because their operation no longer creates as much spurious noise, which means less nastiness has a chance of channeling through to the analogue side by some means.
The short answer is, that everything done, even though it seems very distant in its possible impact, may help; simply because electronic circuitry is not perfectly robust.
The short answer is, that everything done, even though it seems very distant in its possible impact, may help; simply because electronic circuitry is not perfectly robust.
What people call "digital" thesedays covers a lot of things.
If you're storing or retrieving from digital memory or doing "digital processing" with a microprocessor or DSP, power supply noise doesn't mean much - it doesn't affect anything as long as it stays within the specified operating voltage ratings of the devices, and it doesn't change signals from the intended "1" and "0" thresholds of the input circuits.
But I've never heard such a digital circuit. My ears won't directly interface to them...
To be made audible, these digital "things" go through D to A converters of some such, and those things ARE sensitive to power supply noise. Even Class D (which isn't digital, but that's a separate rant) is sensitive, but it's usually inside a high-gain feedback loop much like in an analog power amplifier, and the loop substantially reduces the sensitivity to PS noise (or said another way, it increases power supply rejection ratio).
Last edited:
snup, do you know why you never see elephants hiding in trees?
It's because they're REALLY good at it!
It's because they're REALLY good at it!
Hi and thanks a lot for the very valuable reply.
But why your bottom line is in the middle of your statement ? 🙄
😀 that is indeed the point.
I am focusing these days only on just one task ...
to get a decent sound (i mean music not a bunch of noises) out of my pc with both
streaming from the net (Netflix and Youtube) and from a nas where i have stored most of my cd collection.
Let's say that i would like to get a sound similar to the one from my cd player.
For now i still prefer the cd player 🙁
So i am using an usb to spdif converter and send the signal out from this converter to a good external dac.
The problem is that i have the strong feeling that they could have done better with the converter power supply in terms of isolation from the mains.
It is just a trivial toroidal, poor integrated diodes bridge and some caps
After there is the regulator.
I am sure that a lot of noise from the mains reaches the regulator without any barrier, like i am sure the regulator is not able to block all of this noise.
I am sure of this even if i do not have scope.
So the question is really if this noise leaking through the regulator can have some kind of nasty effects on
1) the usb receiver chip (by Xmos)
2) the clock generator circuit
My feeling is that this noise have a nasty effect and this translates in some kind of noise embedded in the sound.
something like when something is frying in the pan.
I have also the solution ... the batteries.
I am sure that the batteries would be exceptional for this purpose.
But they are so unconvenient ...
The batteries are absolutely the best way to isolate the device from the mains.
I have to try them one day ... just to hear.
So my bottom line 😉 is that i am only interested, now, in a very very good power supply solution for my digital converter.
The rest of the chain can wait. Actually with the cd player it sounds already not excellent by decent.
Not with the pc ... is not decent. Yet.
Thanks a lot again, gino 😀
Last edited:
Hi and thanks a lot for your very helpful advice.
But my question is even more basic.
Let's take a dac. How is important that the VA supplied to its circuit is clean ?
and how much clean it must be ?
This is the question.
Someone say exceptionally clean is a must.
Others say reasonably clean.
Others say just clean ... do not make an issue of things that an issue are not.
For instance these people here deem that it must be very very clean.
JS-2 Linear Power Supply – UpTone Audio
an interesting reading ... are they right ? that is my question.
Because someone say that this is ********.
my converter has clock circuit that sets the master clock for the dac after it.
This is the point. It is the master clok generator.
And actually a master clock generator power supply is what i would like to study if only i had the chance.
I would look at the best performing clockers on the market and copy their power supply confident that they are excellent for any digital converter. 😉
I think i have reached something here. 🙄
Thanks a lot again, gino 🙂
Hi and thanks and this is clear. I have no doubt this.
Problem is when the sound starts as a group of 1 and 0. This is not sound.
The conversion AD and in my case DA is critical. Is not straightforward at all. That's the point. The conversion.
Thanks again for the kind and precious advice, but i need example of something very very well done. Something that can be used as reference of conversion DA done very very well.
And then i will look at their PS in order to get some hints.
But thinking a little more i have come to a very important conclusion, at least for me.
There are units around called master clock generators ... one example here
Isochrone 10M | Antelope Audio
this is in my mind a very demanding task. The same ps must be perfect for any other digital application. And it does not look extremely complex. It is even SMPS ! 😱🙄
Here is a picture of the internals ...
I wonder even if it is an OTS part. I would buy it immediately. Immediately !
I have to discover this. I feel that if i can get that PS i am done completely. No more problem at all.
Anything out of my pc will sound at least decent ... even streaming from Youtube, just think of that.
Completely.
Thanks a lot indeed.
Kind regards, gino
Attachments
Last edited:
Wenzel associates make some of the lowest noise clock sources on the market (For applications where it REALLY MATTERS like radar and sat comms), and actually have some papers in their library that may be of interest.
Finesse Voltage Regulator Noise! |
What you will find is that the usual practice is to have a very good regulator dedicated to the clock (possibly more then one if there is a PLL in play as well) located right next to the clock module, this is not something where 6 inches of wire and common impedances in the ground net makes any sense...
Usually there is also some ferrite as regulator CMRR falls with frequency, and usually this regulator is fed from a regulated higher voltage rail (Maybe a 3.3V LDO fed from a 5V rail for example).
Now there is a fashion in some parts of the hifi scene for massively overdone power supplies replete with R core iron, belly bands, huge amounts of capacitor and shunt regs running at amps, none of which IMHO is sufficiently close to the point of loads that matter to do much good, and interestingly you don't tend to see this stuff in the gear in the studios where the records were made....
Critical rails need regulation right next to the part, with due consideration for things like hierarchical grounding.
Consider the (utterly excellent) ADM7150 regulator has a noise level in a 100KHz bandwidth of <2uV, and PSRR > 60dB to 1MHz under most operating conditions.
Now what do we have to do to get that level of goodness:
Well, if we want to preserve that noise performance then even 1mOhm of common impedance between the load ground and any other current will trash it if the current is even a milliamp, so layout is really going to matter and the reg clearly must be right next to the load.
Further we should study the datasheet to decide how much CMRR we are getting under our operating conditions, and make sure the input is quiet enough, if it is 70dB for example then we need to pre regulate the input to have less then 10mV of noise if PSRR is not to dominate, nothing hard, but it is not exactly a retrofit to an existing board.
However for the digital bits as opposed to the analogue and mixed signal parts (Clocks are actually analogue) it is mostly unimportant as long as the rails stay within tolerance, I mean fast digital introduces so much crap onto its rails anyway and it is so fast that there is nothing a regulator a few inches away can do about it (Hence the bucketfull of 10nF or 100nF caps you always see scattered around such things).
The power supply architecture must be DESIGNED, it is generally very hard to retrofit in a way that really makes sense after the fact.
Regards, Dan.
Finesse Voltage Regulator Noise! |
What you will find is that the usual practice is to have a very good regulator dedicated to the clock (possibly more then one if there is a PLL in play as well) located right next to the clock module, this is not something where 6 inches of wire and common impedances in the ground net makes any sense...
Usually there is also some ferrite as regulator CMRR falls with frequency, and usually this regulator is fed from a regulated higher voltage rail (Maybe a 3.3V LDO fed from a 5V rail for example).
Now there is a fashion in some parts of the hifi scene for massively overdone power supplies replete with R core iron, belly bands, huge amounts of capacitor and shunt regs running at amps, none of which IMHO is sufficiently close to the point of loads that matter to do much good, and interestingly you don't tend to see this stuff in the gear in the studios where the records were made....
Critical rails need regulation right next to the part, with due consideration for things like hierarchical grounding.
Consider the (utterly excellent) ADM7150 regulator has a noise level in a 100KHz bandwidth of <2uV, and PSRR > 60dB to 1MHz under most operating conditions.
Now what do we have to do to get that level of goodness:
Well, if we want to preserve that noise performance then even 1mOhm of common impedance between the load ground and any other current will trash it if the current is even a milliamp, so layout is really going to matter and the reg clearly must be right next to the load.
Further we should study the datasheet to decide how much CMRR we are getting under our operating conditions, and make sure the input is quiet enough, if it is 70dB for example then we need to pre regulate the input to have less then 10mV of noise if PSRR is not to dominate, nothing hard, but it is not exactly a retrofit to an existing board.
However for the digital bits as opposed to the analogue and mixed signal parts (Clocks are actually analogue) it is mostly unimportant as long as the rails stay within tolerance, I mean fast digital introduces so much crap onto its rails anyway and it is so fast that there is nothing a regulator a few inches away can do about it (Hence the bucketfull of 10nF or 100nF caps you always see scattered around such things).
The power supply architecture must be DESIGNED, it is generally very hard to retrofit in a way that really makes sense after the fact.
Regards, Dan.
Ah, the old Rubidium clock for audio chestnut....
Word clock generators are really not that critical, and actually don't even need to be all that low jitter (The loop filter bandwidth in the receiving gear suppresses most external jitter if it is halfway sane), not that this ever stopped the marketing department....
That power supply is an off the shelf 2 * 4 inch module, TDK, XP Power, someone like that.... Plenty of them out there.
Have you measured and actually established that the power supply is a problem? As opposed to say ground plane layout, decoupling or even just ringing on an incorrectly terminated clock line?
Even if the power is a problem, have you sought to establish HOW it is a problem? You need to know what to fix before you can fix it....
Regards, Dan.
Word clock generators are really not that critical, and actually don't even need to be all that low jitter (The loop filter bandwidth in the receiving gear suppresses most external jitter if it is halfway sane), not that this ever stopped the marketing department....
That power supply is an off the shelf 2 * 4 inch module, TDK, XP Power, someone like that.... Plenty of them out there.
Have you measured and actually established that the power supply is a problem? As opposed to say ground plane layout, decoupling or even just ringing on an incorrectly terminated clock line?
Even if the power is a problem, have you sought to establish HOW it is a problem? You need to know what to fix before you can fix it....
Regards, Dan.
Hi ! and thanks a lot for the very precious advice.
It is very high level but i think i have got something. Like that if the pcb has some design flaws no external supply will fix that.
I of course will leave absolutely the pcb as it is ... apart some caps replacemente because i have already taken the original out ...
Of course i have no evidence that the stock PS is an issue ... 😱
it is just that i see in the units i dream of other type of transformers, for instance (EI instead of toroidals). So i conclude without even thinking that they are better parts for this application.
Let me read the article first. It is very important for me.
Thanks a lot again.
Kindest regards, gino
It is very high level but i think i have got something. Like that if the pcb has some design flaws no external supply will fix that.
I of course will leave absolutely the pcb as it is ... apart some caps replacemente because i have already taken the original out ...
Of course i have no evidence that the stock PS is an issue ... 😱
it is just that i see in the units i dream of other type of transformers, for instance (EI instead of toroidals). So i conclude without even thinking that they are better parts for this application.
Let me read the article first. It is very important for me.
Thanks a lot again.
Kindest regards, gino
Attachments
![IMG_0444[1].jpg](/community/data/attachments/444/444360-f7032871bb524a9c5df93c8f3ada21c0.jpg?hash=9wMocbtSSp)
Lots of fashion in high end audio, and huge amounts of merde talked about power supplies.
Well there is nothing obviously wrong with that board design, how the USB interface is set up is impossible to tell, but end of the day the interesting bit is the DAC hung on the back end surely?
I know this makes me one of 'THOSE' people, but it pays to establish that you have a problem before you start tearing into functional electronics.
BTW: Nothing you can do will make UTube audio sound great in a technical sense, great performances and great writing might make for a captivating performance where you ignore the artifacts, but that has to be where you look for greatness in such things.
Regards, Dan.
Well there is nothing obviously wrong with that board design, how the USB interface is set up is impossible to tell, but end of the day the interesting bit is the DAC hung on the back end surely?
I know this makes me one of 'THOSE' people, but it pays to establish that you have a problem before you start tearing into functional electronics.
BTW: Nothing you can do will make UTube audio sound great in a technical sense, great performances and great writing might make for a captivating performance where you ignore the artifacts, but that has to be where you look for greatness in such things.
Regards, Dan.
Yes ... you are right and my ability to screen out garbage is low very low 😱🙁
also this i did not know but when it was working i liked what i was listening quite. But i think that only specific measurements can tell that.
problem was that i have a dac (maybe two) that potentially could provide good sound. But they have not usb input so i needed an interface.
And i have decided to stick with AES/EBU digital connection after reading many comments advising this from famous designers.
This interface have them all. It is very handy on this aspect.
you are right. Sometimes i tend to create problems. And i have the feeling that i could have addressed the problem in a very different way.
Maybe just a better sound card for my pc is just what i needed.
let's take MP3 ... is clearly a lossy standard but i have listened to MP3 tracks on different systems with very different results.
The aim is to squeeze the most possible out of anything ...
Thanks a lot again for the very kind and helpful advice.
Kind regards, gino
Last edited:
gino, just on that last point, about squeezing the most out of compressed audio - which is AAC on YouTube, not MP3, a better encoding algorithm - this is certainly possible. I've done experiments which demonstrate this to me, but it entailed doing processing of the track beforehand - so, extra effort, storage space, it all gets messy very quickly. The ideal would be a system that could transparently get the most out of a track of any quality, but this adds yet another layer of difficulty to the whole exercise - there are no easy answers that I've discovered yet, 🙁.
Hi and thanks for the very helpful advice.
I rephrase then ... squeeze a "musical" outcome from youtube
something that would make the experience satisfactory
In the waiting of a paid service with better sound quality
Thanks again, gino
Last edited:
- Status
- Not open for further replies.