Medical 12V 3A SMPS

[jabbr]

 

I just got this from Mouser ... anyone tried CUI ... advertised as Medical Class VI ??‍♂️

 

I needed a 12V 3A supply for my Espressobin and I’ll compare this against an LPS

[Ralf11]

Good to hear - how many nV can you resolve?

 

Price?

 

Model #?

[jabbr]

49 minutes ago, Ralf11 said:

Good to hear - how many nV can you resolve?

 

Price?

 

Model #?

MFG Part No: SWM30-12-NV-P5

 

Isolation resistance: 50 MOhm at 500V DC (this is presumably the inductance that corresponds to leakage current)

Leakage current: 100 microAmps at 240V/50 Hz

Max ripple: 250 mV

$20

 

Selected for reasonably low leakage not nV of ripple  

[Ralf11]

$20 ??

 

too low to be a neuro-typical audiophool device

 

and... OMG!! you mean we have to do 2 factor optimization???

[jabbr]

See the graphs I posted on new thread ... more than simply 2 factor optimization  

[Superdad]

On 9/4/2018 at 12:53 PM, jabbr said:

I just got this from Mouser ... anyone tried CUI ... advertised as Medical Class VI ??‍♂️

 

I needed a 12V 3A supply for my Espressobin and I’ll compare this against an LPS

 

Are you going to shunt its DC output -ve to AC ground?  If not then it will still transfer a good deal of "high impedance source component leakage."  John has measured a few "Medical" SMPS units and while they are low on typical "touch current" leakage they still put out plenty of the other sort.

 

Here are graphs he made of a Mean Well medical-classed adaptor at different bandwidths.  Still plenty of leakage, but much of this form goes away when shunted.

[sandyk]

39 minutes ago, Superdad said:

Are you going to shunt its DC output -ve to AC ground?  If not then it will still transfer a good deal of "high impedance source component leakage."  John has measured a few "Medical" SMPS units and while they are low on typical "touch current" leakage they still put out plenty of the other sort.

 

Here are graphs he made of a Mean Well medical-classed adaptor at different bandwidths.  Still plenty of leakage, but much of this form goes away when shunted.

 

Alex C

 

Very informative ! And this is from a Premium type SMPS PSU.

[jabbr]

43 minutes ago, Superdad said:

Are you going to shunt its DC output -ve to AC ground?  If not then it will still transfer a good deal of "high impedance source component leakage."  

Units please. By a “good deal” of leakage current, do you mean 2 nA? 10 nA? 

 

In amperes, what is a “good deal” of current? For example what is the current noise in nA of the supply itself?

[jabbr]

@Superdad, I think if I want to go in and hack the SMPS, I'd just build my own PSU, and go linear because:

 

1) lets not go overboard with "leakage" to the extent that the leakage currents are already less than the background noise of the power supply. I mean noise is noise. Of course different types of noise matter to different circuits, but we are speaking generally, so unless you draw a circuit where the noise goes, then voltage noise is voltage noise, current noise is current noise and phase noise is phase noise.

2) as above, if the background noise is in the millivolt range, why would I be concerned about leakage current noise in the nanoamp range?

3) we are dealing with a CPU board and processor running at lets say 1Ghz. There is going to be noise in the millivolt range I expect.

4) if we are talking about "optimizing" the board layout, I know for a fact that the manufacturer here is using Mentor HyperLynx to optimize board SI (espressobin)-- I'd love to be able to afford that, certainly beats me with my hand calculator or slide rule  

 

1) CMOS "popcorn" current noise in the 40 nA range (just one type of CMOS noise): http://web.mit.edu/klund/www/papers/UNP_noise.pdf

2) https://ieeexplore.ieee.org/document/918240/

3) http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.479.4689&rep=rep1&type=pdf

4) ooohhh ... look here: this shows what is possible with CMOS when low current noise is actually important: http://ims.unipv.it/FIRB2006/pub/Manghisoni07.pdf picoamp range  ... now that was with 90nM CMOS process, and as the die shrinks and the transistors get smaller, the leakage noise goes up ?‍♂️ -- that's a real problem these days, and what I'm harping on actually

 

So I'm not trying to be theoretical here, rather very practical. Why would I want to spend $1000 on a power supply for a specific application where $10 or $20 would do? ... and again I'm going to use a $100 12V 3A linear power supply from China/Ebay because it has a reasonable r-core transformer and reasonable regulation and a reasonable case -- i.e. cheaper for me to buy this that build it myself, maybe not parts but my time is >$20/hr.

 

So the way to be practical is to set a design target for voltage, current and phase noise. As a manufacturer you of course make these decisions every day, so shouldn't we consumers?

[barrows]

11 hours ago, sandyk said:

 

Alex C

 

Very informative ! And this is from a Premium type SMPS PSU.

I would not consider these examples "premium" in terms of SMPS!  Just because something meets the "medical" grade does not mean that it is the best it can be.  There are much better performing SMPS out there!  But, they do cost much more than $20.

[Superdad]

15 hours ago, jabbr said:

Units please. By a “good deal” of leakage current, do you mean 2 nA? 10 nA? 

In amperes, what is a “good deal” of current? For example what is the current noise in nA of the supply itself?

 

I'll have to check with John--when he returns from vacation--with regards to the actual current range of the graphs.  Of course like most measures--and evident from the graphs--it is not a single number as the leakage current varies with frequency.

 

4 hours ago, jabbr said:

@Superdad, I think if I want to go in and hack the SMPS, I'd just build my own PSU, and go linear because:

 

Well hacking is not required--just a wire from the output barrel to AC ground.  Of course an LPS would be preferred, but it was you who chose this SMPS and started this thread. 

 

Quote

1) lets not go overboard with "leakage" to the extent that the leakage currents are already less than the background noise of the power supply. I mean noise is noise. Of course different types of noise matter to different circuits, but we are speaking generally, so unless you draw a circuit where the noise goes, then voltage noise is voltage noise, current noise is current noise and phase noise is phase noise.

 

"Noise is noise." Really surprised to hear you say that!  Broadly perhaps, but that ignores how different forms of it travel and how hard or easy it is to block.  Leakage current being the present example.  It is common-mode AC and readily makes its way past a lot of regulators.  What do you think Amir is so often showing in his [single-ended] DAC measurements?

 

Quote

1) CMOS "popcorn" current noise in the 40 nA range (just one type of CMOS noise): http://web.mit.edu/klund/www/papers/UNP_noise.pdf

2) https://ieeexplore.ieee.org/document/918240/

3) http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.479.4689&rep=rep1&type=pdf

4) ooohhh ... look here: this shows what is possible with CMOS when low current noise is actually important: http://ims.unipv.it/FIRB2006/pub/Manghisoni07.pdf picoamp range  ... now that was with 90nM CMOS process, and as the die shrinks and the transistors get smaller, the leakage noise goes up ?‍♂️ -- that's a real problem these days, and what I'm harping on actually

 

So I'm not trying to be theoretical here, rather very practical. Why would I want to spend $1000 on a power supply for a specific application where $10 or $20 would do? ... and again I'm going to use a $100 12V 3A linear power supply from China/Ebay because it has a reasonable r-core transformer and reasonable regulation and a reasonable case -- i.e. cheaper for me to buy this that build it myself, maybe not parts but my time is >$20/hr.

 

I always enjoy reading the 1/f noise articles you link, even if they often go over my head, thanks.  You are the engineer here, not me, so I'll not attempt to argue too technically (will leave that to @JohnSwenson if he feels incline to take the time).

 

But more broadly I am just surprised that you--of all people--would be questioning the need for low-noise/low-impedance power supplies as well as the banishment of AC leakage.  Even leaving aside the widespread reports of differences people hear with quality power sources and attention paid to connections, isolation, and leakage.  It was you who only a day or two ago posted the Walton graphs comparing the PSRR, noise, and impedance of a range of integrated and discrete regulators.

So color me confused about what you are getting at now.  

 

Quote

So the way to be practical is to set a design target for voltage, current and phase noise. As a manufacturer you of course make these decisions every day, so shouldn't we consumers?

 

Sure!  But picking a $6 (wholesale at 1,000 pcs.) SMPS with 250mV of output noise seems to be setting the bar pretty low my friend. 

[jabbr]

30 minutes ago, Superdad said:

"Noise is noise." Really surprised to hear you say that!  Broadly perhaps, but that ignores how different forms of it travel and how hard or easy it is to block.  Leakage current being the present example.  It is common-mode AC and readily makes its way past a lot of regulators.  What do you think Amir is so often showing in his [single-ended] DAC measurements?

 

Yeah I know that John started chasing this concept of "high impedance leakage" in response to Amir. Um frankly it depends on the level and if e.g. < 100 nA then maybe I'm not too concerned, or maybe I am, as I said, it would depend on the circuit to determine where the current flows, or doesn't flow. If the relevent circuit has current noise in the picoamp range, then yes nanoamp leakage is relevent.

 

That's my point here. Can we save $$$ by using appropriate parts and circuits and components in appropriate places? (of course)

35 minutes ago, Superdad said:

I always enjoy reading the 1/f noise articles you link, even if they often go over my head, thanks.  You are the engineer here, not me, so I'll not attempt to argue too technically (will leave that to @JohnSwenson if he feels incline to take the time).

 

 

I have a goal of trying not to argue by assertion, which is why I try to post references to back my statements.

38 minutes ago, Superdad said:

But more broadly I am just surprised that you--of all people--would be questioning the need for low-noise/low-impedance power supplies as well as the banishment of AC leakage.  Even leaving aside the widespread reports of differences people hear with quality power sources and attention paid to connections, isolation, and leakage.  It was you who only a day or two ago posted the Walton graphs comparing the PSRR, noise, and impedance of a range of integrated and discrete regulators.

So color me confused about what you are getting at now.  

 

 

I'm always glad to see measurements in datasheets and find that when a part is claimed to be designed for a specific purpose, the datasheet will more likely contain such relevant measurements. AC leakage currents are currents that travel across parasitics and there will always be low level parasitics -- so there is really no way to ever eliminate leakage currents, rather minimize they below significance.

 

I'm asking: what level of leakage currents are significant for different circuits? In one case we have a DAC design, and at the other end  a NAS. For the DAC I care very very much, for the NAS: not at all as long as it serves the bits as fast as I've designed it to.

 

47 minutes ago, Superdad said:

Sure!  But picking a $6 (wholesale at 1,000 pcs.) SMPS with 250mV of output noise seems to be setting the bar pretty low my friend. 

 

My friend, I'm trying to define situations where your ISO Regen provides a significant benefit, so I'm picking the "worst" or shall I say I'm picking very specific parameters: both output noise (higher) vs leakage current (lower). Which is more important? Maybe when the iFi is operating on battery mode, and I've got complete WiFi isolation, perhaps leakage current is less relevant?

 

Again, I'm not attempting to create the best possible system, rather trying to test specific variables ... that's what I'm getting at...

 

[Ralf11]

maybe this is a good time to discuss what levels of noise might matter at different points in the chain of components from music file storage device... to analog end of the DAC

 

what noise levels have been measured in the power supplies of expensive high-end DACs?

[Superdad]

12 minutes ago, jabbr said:

I'm asking: what level of leakage currents are significant for different circuits? In one case we have a DAC design, and at the other end  a NAS. For the DAC I care very very much, for the NAS: not at all as long as it serves the bits as fast as I've designed it to.

....

Again, I'm not attempting to create the best possible system, rather trying to test specific variables ... that's what I'm getting at...

 

I'm with you there!

Saul Good-man...

[marce]

One thought, theses are basically EMC measurements and need to be done in a shielded EMC lab...

The other thought is the effect this noise will have depends on:

How much actually gets onto the DAC PCB, the designers have probably thought about EMC protection...

The effect of any current noise that does get through will depend on the impedance of the impedance of any nodes affected. High impedance nodes will suffer more than low impedance nodes, so you protect sensitive high impedance nodes, the first rule being reduce thier capacitance (cutting the ground from directly under pads, thinnest trace width taking into account other factors, etc. etc.)

Power delivery system design plays a BIG part, notice I say POWER DELIVERY SYSTEM, not just a LPS at the front end, isolation, power islands,  etc. etc. use local LDO's with high PSRR for sensitive devices, buffer data across the different supplies.........

 

[Superdad]

16 minutes ago, marce said:

Power delivery system design plays a BIG part, notice I say POWER DELIVERY SYSTEM, not just a LPS at the front end, isolation, power islands,  etc. etc. use local LDO's with high PSRR for sensitive devices, buffer data across the different supplies.........

 

+100!

[sandyk]

7 hours ago, barrows said:

I would not consider these examples "premium" in terms of SMPS!  Just because something meets the "medical" grade does not mean that it is the best it can be.  There are much better performing SMPS out there!  But, they do cost much more than $20.

 

 

I am not disagreeing with you, but these are a cut above the rest of the generic SMPS plugpacks in the amount of crap going back into the A.C. mains.

[Ralf11]

Let's look inside a good sounding, expensive DAC.  If it has 3 power supplies, do they differ in noise?

[sandyk]

8 minutes ago, Ralf11 said:

Let's look inside a good sounding, expensive DAC.  If it has 3 power supplies, do they differ in noise?

Look inside your 205.

 

 Does it have a toroidal transformer to power all of it, or like some more upmarket ones does it use a toroidal transformer for the Audio area ONLY, and SMPS for the rest ?

[Superdad]

12 minutes ago, sandyk said:

I am not disagreeing with you, but these are a cut above the rest of the generic SMPS plugpacks in the amount of crap going back into the A.C. mains.

 

Not by much.  And certainly not by the amount of the form of leakage that John (and Amir) has showed can sail on through a DAC via digital and analog connections!

AC-mains-connected linear supplies also have leakage (due to parasitic capacitance of their transformers), but not near as much as much as SMPS warts and bricks with their required 'Y' capacitors.  

Of course I am not telling you anything you did not already know Alex K. You've been railing on about SMPS evils for years. 

[sandyk]

16 minutes ago, Superdad said:

Of course I am not telling you anything you did not already know Alex K. You've been railing on about SMPS evils for years. 

 Hi Alex

 Those very interesting graphs from John complement very nicely the original articles from Silicon Chip magazine in this area,  although their suggested small value capacitor (22nF 275VAC X2) from case to A.C. mains earth didn't work as well for me as John's direct grounding when done with my STB, which resulted in improved Audio from Coax SPDIF Out., not just reducing the nasty bites when plugging/unplugging cables.

 

Regards

Alex

[Ralf11]

1 hour ago, Ralf11 said:

Let's look inside a good sounding, expensive DAC.  If it has 3 power supplies, do they differ in noise?

 

by expensive, I mean $10k - unless one finds a $2k one sounds better

 

by DAC I mean a DAC only, not a disc player 

 

- an empirical way to see what near state of the art DAC manfs. do...

[Superdad]

1 hour ago, Ralf11 said:

Let's look inside a good sounding, expensive DAC.  If it has 3 power supplies, do they differ in noise?

 

My old friend Jeff Kalt at Resolution Audio produces a well thought out DAC (apparently he has a new version out based on an instrumentation DAC chip since he likely can't get PCM1704 R2Rs anymore).  Here is a photo of model that I and others here have heard. Multiple independent power supplies...

 

[Ralf11]

do they differ in noise?

[PeterSt]

7 hours ago, Ralf11 said:

what noise levels have been measured in the power supplies of expensive high-end DACs?

 

Not really qualifying what expensive and what high-end is, but this is from the Phasure NOS1 (not "a" yet and not "G3" yet) ...

 

This is measured on the 3.3V oscillator supply (this is part of the digital supply (separate from the analogue) with separate regulation for the audio clocks ) :

 

 

Or in FFT format :

 

 

I drew these out of a topic which was about spurea on the power supply in general. Red is measured in the normal DAC output and where you can see that the anomalies could be guided away (as how I call it) and white is the 3.3V clock supply again. The right one is my dreaded heat system water PWM-controlled pump and this goes through air (so is picked up somewhere (or everywhere) in the system (chain)). Left one is the fridge (unknown whether through air or not but on a different mains ring anyway).

 

 

Same but further down the line at a base frequency of close to 65KHz; I dedicated that to a switching power supply somewhere, with the notice that the audio PC I used there had a normal SMPS. Btw the topic as a whole was about how to create a Linear Power Supply for the "XXHighEnd PC" (that, a few years later, resulting in the Mach II and today's Mach III PCs, both with LPS indeed).

 

Subjectively and maybe off topic, you can see (and I could for the whole spectrum, back at the time) that the output of the DAC itself was free of anything (completely ruler flat meanwhile), but that the supply to the clock was not.

If you look at the first picture again, a trained eye can see that there's a virtual DC offset (general line is a bit below zero) but which is caused by 50Hz mains influence; look a bit later and you'll see the offset above zero (the line wobbles a bit).

This was all eliminated in the "a" version of the NOS1 (I don't have pictures of the supplies, I think), knowing that the "a" version was about (a first) galvanically isolated means of the clock section from the (USB) interface. And guess what ? that expressed in enormously lower jitter figures (measured, thus not theory, but the theory you see surface in the above text ).