PSU noise pollution: iFi, Meanwell, Teddy Pardo, UpTone JS-2 and others

[Elberoth]

Every PSU kicks some noise back into the AC line, which may have an influence on the rest of your gear connected to the same AC line. 

 

I still remember a review written by Martin Colloms of some gear using onboard SMPS, where he claimed that simply connecting said component (otherwise not connected to the system) to AC line, made his system sound considerably worse.

 

So how bad is the problem ? Having bought a wideband AlphaLab EMI Power Line Noise Meter I decided check that out on various PSUs I had on hand.

 

The setup

 

To get a reference/base noise level I have used my old and trusty Audio Power Industries Power Wedge Ultra 215 power conditioner, which I normally use to 'sink' all of my computer related PSUs - main server PSU, SSD battery SMPS charger, UpTone Audio JS-2 powering two LPS-1 PSUs and a computer monitor.

 

The noise level measured at the output was just 23mV. I have used one of outputs to connect a simple power strip (with no filtering):

 

[/img]

 

The PSU under test was connected to the same power strip, next to the EMI meter:

 

 

I had 8 different PSUs on hand: 3 linear ones and 5 SMPS (switch mode) types:

 

[/img]

 

1. Teddy Pardo linear PSU

2. UpTone JS-2 linear PSU

3. SOtM supplied SMPS (PowerTek)

4. Meanwell SMPS (supplied by UpTone with Regen and LPS-1)

5. Some generic linear PSU I found lying around

6. Ktec SMPS that I got with an external HDD

7. iFi iPower SMPS 

8 Co Ming Data SMPS 

 

Since I have observed the level of noise beeing kicked back to the AC line varies with PSU load, I have measeured those PSUs in the following conditions:

 

1. Idle - PSU connected to the power strip, but nothing connected to the PSU (no load)

2. Noise under light load - PSU powering Regen, which in turn was powering AQ DragonFly DAC

3. Noise under moderate load - PSU powering SOtM sMS-200

4. Noise under heavy load - PSU powering the LPS-1 Ultracapacitor PSU, which in turn was powering the SOtM sMS-200.

 

The results I got are as follows (green = best case; red = worst case):

 

 

Please note that the smallest, generic linear PSU (last column) was not tested with LPS-1, since its power rating was too small.

 

Some things I have noticed:

 

The SMPSu in general injected much more noise than linear ones. Even the cheapest generic wall wart linear PSU was better than the best SMPS.

 

In many cases, the noise figures dropped at idle a few mV below the reference level. It is my understanding that this is due to filtering built at the input of those PSUs, which interacts / lowers the noise on the AC line. 

 

The PSU noise readings were jumping up and down when the LPS-1 PSU was connected (chence the results are in ranges). You can see this on a short video I uploaded on youtube.

 

 

It is my understanding that noise spikes were caused by the LPS-1 bulit in (capacitor) charger - when the charger was running, the noise was higher; when the charger was idling, the noise was lower. I'm not sure if higher noise readings are caused by higher current draw or the noise is comes from the LPS-1 charger itself. Whatever the reason - some PSUs were much more sensitive to that than the other.

 

For kicks, I have also measured several other PSUs and even light bulbs:

 

 

- iPhone, iPad and MacBook chargers are pretty harmless when at idle. I wouldn't want to have one connected to my powerstrip when charging though (all of you using MacBooks and other laptops connected to your system take note).

- CFL light bulbs (and LEDs) are huge noise generators (unlike regular light bulbs, which generate close to zero noise).

- chineese iPhone charger knock off from ebay, creates some absurd amounts of noise -  many times more than the original Apple accessory (one more reason not to buy knock offs).

- TeraDak linear ATX PSU that powers my CAPS v4 Pipeline server generate generates only 48mV of noise - suprisingly litte.

 

Final notes:
 

If you have lots of those generic SMPS around your audio system (powering DACs, HUBs, routers, external HDDs etc), it probably makes sense to replace them with something designed with the AC noise pollution in mind, like the $49 iFi iPower SMPS for example. Linear still seems to work best though.

 

Having said that, please remember, that the best PSU is not necessarily the one which kicks the least amount of noise back into the AC line. The best PSU is the one that has low AC noise footprint and offers stellar DC output performance. 

[Superdad]

Hi Adam:

 

Since you brought the topic over there, I just replied to you in the UltraCap Listening Impressions thread.  But since your report here seemed lonely, I'll cross-post my thoughts here as this really is the more relevant thread for the conversation.

------------

I read your report a week ago, but have been too busy to comment.  While your effort is interesting, I caution against drawing much conclusion from it.  Here is why I say that:

 

1) Your $130 meter is distilling all the noise it measures (with unknown weighting across the band) down to one number, and as such is not an entirely useful result for qualitative comparison. For example: a supply with a bunch of of large nasty spikes at the lower harmonics of its switching frequency (typically 80KHz-100KHz) might read out on your meter a lower number than an SMPS which spreads most of its noise across a very broad band ("spread spectrum") at much higher frequencies--in the MegaHertz range and at lower levels. Thus, the unit with a higher reading on your summing meter might well be the more "line benign" back to the wall.

 

2) Your meter "measures" noise beginning at 10KHz and is therefore inappropriate for comparing the harmonics kinked back into the mains by linear power supplies.  The harmonics from LPS units will always be at multiple of the line frequency (50 or 60Hz), and your meter will miss most of those.  The diode switching (especially from cheap LPS units) and transformer ringing (on units that don't use an RC snubber on the secondary) will result in quite a bit of harmonics back to the wall. [As an aside, the filtering effect of the inductance and capacitance of AC cables--of what a component's PS kicks back in harmonics--is the primary reason people hear differences between AC cords; assuming decent wire gauge for current draw, it is not at all about the "quality" of the power delivered TO the component.]

Traditional LPS units (transformer>diodes>caps>regulator) draw current over only about 50% of the AC wave cycle--hence their harmonic distortion of the mains, even if relatively benign.  That effect is part of what has driven the decades long push for "power factor corrected" power supplies worldwide (most easily accomplished with SMPS designs.  While one rarely sees a power factor corrected linear power supply, our own JS-2 is an example of one.  Its large DC choke results both in a smooth half-cycle wave delivered to the regulators (making their job easy), and in a power-factor of about 0.97--as it draws current over the entire AC wave cycle, thus not generating harmonics back into the line.

 

If you really wish to measure what various supplies kick back into the wall, then a more elaborate scope or spectrum analyzer should be used.  But what I expect you will find is that what is coming out of your wall in the first place (the spectrum of that residual 23mV you started with) is far nastier (i.e. at undesirable frequencies) than what is being added by the SMPS units you are testing.

 

None of the above in any way is meant to minimize the problems or differences in:

a) the quality of the DC output from various SMPS and LPS units;

b) the inherent "evil" of ALL SMPS units--very high leakage currents (due to their required use of Y-capacitors) (AC>DC linear supplies also have leakage, and while it is MUCH lower, it still will form "loop" with leakage from other connected components.)

 

Of course, the whole point of the UltraCap LPS-1 is to render both a) and b) above irrelevant! 

 

Adam, please do not take this post as any indictment or dismissal of your fine efforts to shed light on the subject of what SMPS units put back into the mains.  This is just my attempt to show how the matter is more complicated than it might seem (and wait until John corrects me and tells of even more complex factors!).

 

By the way, short of having a 100X scope probe and needing to do other things to safely look at noise on wall voltages with a O-scope, I did find this neat (if somewhat pricey at $389) adapter that plugs into the wall (there is a version with leads too) and provides a BNC to go directly to a scope or other analyzer. http://www.onfilter.com/products.html?s=MSN01

 

Best,

--Alex C.

[Elberoth]

Alex,

 

I value your feedback very much and highly encoureage everyone to join the discussion.

 

I know that a single digit meter is not a perfect solution. However, with all its limitations you can see a certain patterns here - namely, cheap SMPS in general generate much more noise than more expensive ones, designed with EMI radiation in mind (iFi iPower for example or genuine Apple iPhone charger vs chineese knock off). 

 

BTW - Cealin Gabriel (the CEO of Shunyata) have posted this info on another forum:

 

I have several of the Alpha and the Stetzerizer units. Personally, I prefer the old Entech unit. Entech was actually a very technically advanced group of engineers. Both the Alpha and Stetzerizer are similar in that they have a bandwidth that extends quite low for measuring RFI/EMI. They drop down to 15-30 kHz in there measurements. While this may seem to be an advantage it actually has a fatal flaw. Both aggregate noise from about 15 kHz to around 10 MHz and lump them all together in a single reading. Some frequency bands are more important to the performance of audio equipment specifically our research indicates the the most harmful frequencies are in the 100 kHz to around 1 MHz bands and then secondarily from 1 MHz to 10 MHz. Interestingly, I believe that the Entech engineers knew this and designed their device to be sensitive in the frequencies from 300 kHz to 700 kHz range. This makes the Entech particularly useful for finding noise sources in the specific ranges that are most noticeable in the context of music systems. 

I have done direct side-by-side comparisons using all of the units listed. I would then plug a device into the power line that we 'know' produces particularly harmful noise effects as measured by our power and spectrum analyzers. In many cases, the Alpha and Stetzerizer would barely change their relative readings while the Entech would go off the charts. This is because the Entech is more narrowly focused and the other two are less so.

And then conversely I would connect filters that reduce power line noise in a known manner as verified by the power and spectrum analyzers and in many of these cases the Alpha and Stetzerizer would show only as small or marginal difference while the Entech would show a much larger and significant difference. 

Look these are all very inexpensive devices and all of them can be useful. But they have severe limitations from a technical point of view. I found the most useful for the hobbyist to be the Entech. As I said earlier, we are developing a device that will be much more useful and will show noise in specific bands much like the old audio spectrum displays. This is being developed for our medical division where this device will be quite useful in identifying specific types of noise pollution in hospitals. I hope this is helpful.

 

You may have noticed that he places much greater importance on the freq in the 100+kHz bands as more important to the performance of audio equipment. In other words - he doesn't seem to think 10kHz meter limitation is a problem.

 

 

 

 

 

 

[Superdad]

3 hours ago, Elberoth said:

You may have noticed that he places much greater importance on the freq in the 100+kHz bands as more important to the performance of audio equipment. In other words - he doesn't seem to think 10kHz meter limitation is a problem.

 

Great post Adam.

 

Even speaking only of their "pollution" back to the wall (and without even getting into discussion if/why it matters and how that can be dealt with), I think that SMPS units for small devices should be measure and considered separately from linear power supplies--especially large LPS units.

I say that for two reasons:

1) The spectra both kicked back noise and distortion of the waveform due to design is far different for an LPS, but even for those an undamped transformer will pass back some 100KHz+ ringing.  And the unregulated typical LPS of a big power amp can have a huge impact on the line.

2) The criteria (and performance measures) for a high current supply are quite different from those for a small wall wart.

 

Really, in the end I think we can agree that what comes out the DC side is much more important than the few tens of millivolts of noise that gets put back into the mains.  The latter is not so hard to filter out (though of course large leakage current from SMPS still persist).  Yet heavy truncation/distortion of the AC wave (from a shocking number of components as well as some household items) is something serious to guard against.

 

Cheers,

--Alex C.

[Jud]

3 hours ago, Elberoth said:

Alex,

 

I value your feedback very much and highly encoureage everyone to join the discussion.

 

I know that a single digit meter is not a perfect solution. However, with all its limitations you can see a certain patterns here - namely, cheap SMPS in general generate much more noise than more expensive ones, designed with EMI radiation in mind (iFi iPower for example or genuine Apple iPhone charger vs chineese knock off). 

 

BTW - Cealin Gabriel (the CEO of Shunyata) have posted this info on another forum:

 

I have several of the Alpha and the Stetzerizer units. Personally, I prefer the old Entech unit. Entech was actually a very technically advanced group of engineers. Both the Alpha and Stetzerizer are similar in that they have a bandwidth that extends quite low for measuring RFI/EMI. They drop down to 15-30 kHz in there measurements. While this may seem to be an advantage it actually has a fatal flaw. Both aggregate noise from about 15 kHz to around 10 MHz and lump them all together in a single reading. Some frequency bands are more important to the performance of audio equipment specifically our research indicates the the most harmful frequencies are in the 100 kHz to around 1 MHz bands and then secondarily from 1 MHz to 10 MHz. Interestingly, I believe that the Entech engineers knew this and designed their device to be sensitive in the frequencies from 300 kHz to 700 kHz range. This makes the Entech particularly useful for finding noise sources in the specific ranges that are most noticeable in the context of music systems. 

I have done direct side-by-side comparisons using all of the units listed. I would then plug a device into the power line that we 'know' produces particularly harmful noise effects as measured by our power and spectrum analyzers. In many cases, the Alpha and Stetzerizer would barely change their relative readings while the Entech would go off the charts. This is because the Entech is more narrowly focused and the other two are less so.

And then conversely I would connect filters that reduce power line noise in a known manner as verified by the power and spectrum analyzers and in many of these cases the Alpha and Stetzerizer would show only as small or marginal difference while the Entech would show a much larger and significant difference. 

Look these are all very inexpensive devices and all of them can be useful. But they have severe limitations from a technical point of view. I found the most useful for the hobbyist to be the Entech. As I said earlier, we are developing a device that will be much more useful and will show noise in specific bands much like the old audio spectrum displays. This is being developed for our medical division where this device will be quite useful in identifying specific types of noise pollution in hospitals. I hope this is helpful.

 

You may have noticed that he places much greater importance on the freq in the 100+kHz bands as more important to the performance of audio equipment. In other words - he doesn't seem to think 10kHz meter limitation is a problem.

 

 

 

 

 

 

 

Hi Adam.  I would be very interested to know if any reason is given as to why these MHz frequencies should be important.  A link is fine because I don't want to get off topic.

[Superdad]

1 hour ago, Jud said:

I would be very interested to know if any reason is given as to why these MHz frequencies should be important.  

 

Maybe your DC-to-light Spectral amps are receiving antennas for such Jud! 

[Elberoth]

19 hours ago, Jud said:

 

Hi Adam.  I would be very interested to know if any reason is given as to why these MHz frequencies should be important.  A link is fine because I don't want to get off topic.

 

The original thread:

 

http://www.audioshark.org/power-conditioners-regulators-24/measuring-ac-power-line-noise-11713.html

[plissken]

What was the frequency component of the noise? What was the measured transfer function of the noise? 

 

How do we know the noise is affecting audio output?

[Jud]

47 minutes ago, Elberoth said:

 

The original thread:

 

http://www.audioshark.org/power-conditioners-regulators-24/measuring-ac-power-line-noise-11713.html

 

Thanks, very much appreciated.