Sunday, 28 July 2019

MEASUREMENTS: "Noisy" Switching Power Supplies vs. Lithium Battery Packs (12V battery with RME ADI-2 Pro FS & Topping D10 DAC results)

Perhaps we don't think about it much or give it enough credit. I believe one of the advances in the last few decades that has provided the foundation for technological progress we enjoy today is the lithium ion battery. Li batteries were commercially released in 1991 and since then, thanks to the energy density available, combined with increasingly efficient electronics, all kinds of things these days can run off battery power - cars being the poster-child of this advancement...

A few weeks ago, I showed the little Class-D amplifier system running off a battery. How about we try out my RME ADI-2 Pro FS on batteries:


What you see here is the device connected to the Talentcell 6000mAh 12V Li-Ion battery pack. Since the RME ADC has become my standard for measurements in the last year, I have wanted to make it "portable". Talentcell makes an even smaller 3000mAh 12V battery which would work fine. While a little larger in size and weight, the 6000mAh unit is still lightweight and will be able to power the ADC/DAC for >5 hours - enough for any recording and measurement I would want to do in a single sitting.

Just like with other equipment upgrades, especially when I'm modifying my measurement gear, I like to make sure that objectively, the changes are proven not to harm the results obtained at the very least and ideally will improve resolution if possible. Imagine if switching to a battery, one actually finds more noise or that it adds distortions!

As much as audiophiles tend to have the belief that switched-mode power supplies (SMPS) are "bad" or that one should change to linear supplies or batteries (an extreme example would be these Stromtank battery power supplies), other than my discussions here and here, there seems to be a paucity of actual data to support these assertions.

Today, there are 3 things I want to do:

1. Confirm that using the Li battery vs. stock SMPS with the RME ADI-2 Pro as ADC does not mess up measurements. Perhaps doing this might result in improvements; let's not be presumptive that it "will".

2. Compare performance of a streamer, DAC, and ADC using battery vs. stock/inexpensive SMPS. For this purpose, the other day I took over my dining room table to lay out the gear:


As you can see, everything is running off Li-ion batteries. Laptop obviously has its internal battery, RME ADC running off the 12V battery, and the Raspberry Pi 3 / touchscreen / Topping DAC all running off the 5V battery.

3. Since I didn't do it a month back, get more precise measurements of the Topping D10 DAC using the RME ADI-2 Pro (I had used my old Focusrite Forte ADC at that time).

As you can see in the picture above, that's what it looks like with the whole system running off batteries. Since I'll compare results with the SMPS, here's a closer look at the power supplies and batteries themselves:


As you can see, we'll be comparing what's essentially a <$10 5V/2.5A SMPS with the RAVPower 22000mAh 5V battery for the Raspberry Pi "Touch" / Topping DAC . And then we have the stock 12V/2A RME SMPS compared to the Talentcell 12V/3A 6000mAh lithium battery.

That should give us at least a good idea to see if indeed SMPS's are really all that "noisy" and bad for these low-power digital audio devices. (Not making any claims here for devices needing higher power like amplifiers.)

Part I: Did switching the RME ADI-2 Pro FS ADC to the battery change measurement characteristics?

This is for me the most important part of today's post because practically, if it works fine, this will allow me to do some "field" measurements and not worry about whether there's a convenient AC outlet around. Plus it takes away another variable in measurements - the "quality" of the power outlet wherever I might be.

First, just starting without even connecting the ADC inputs to a device, let's see if we can register a noise floor difference between the SMPS and battery. To do this, I connected my computer to the RME ADI-2 Pro, put the RME settings to the most sensitive values (input level +4dBu), turned off any digital gain. Here's the noise level from the ADC sampled at 384kHz with frequencies all the way to 192kHz captured. Using WaveSpectra, I turned on the "max" display and put "average" setting to "100" which interestingly did not really show average amplitude, but conveniently an average of the minimum calculated FFT:


Notice the red line represents "max" level, blue line as average "min" level, and the green as the current signal plot when the screen was captured. I let the max/min levels average out over 1 minute.

As you can see, indeed there's a difference between the SMPS and battery. Most evidently, there's a little cluster of noise just beyond 40kHz contributed by the SMPS. It's low level, peaking at -120dBFS. That's really all to see in the extended spectrum.

Remember that a linear scale like this misses out on low-frequency differences which are more important for audibility... Check this out, plotted with log frequencies (I removed the green line as it was a bit distracting):


Nice! The "obvious" superiority of the lithium battery compared to switching power supply.

Before we get too excited. A couple of observations:

1. Most of the difference is in the form of the 60Hz (N. American electrical system here) hum and harmonics. It's not the high frequency switching noise that is of highest level with the RME stock SMPS. Understandably, the battery powered system is free from the hum.

2. The RME power supply is doing a good job keeping the hum low however. It's only peaking at -110dBFS with the highly-sensitive low-level input setting of the ADC.

3. As for the high frequency anomaly, it's stuff beyond 40kHz. Again of low level - even lower than the 60Hz hum - with a peak of -120dBFS. The RME stock SMPS is good but is it that special compared to most other SMPS's out there? Unless you think that the stock RME power supply is somehow much better than most out there, so far, the evidence does not suggest that switched-mode supplies pose a problem for high fidelity.

Part II: SMPS vs. Li battery with the RME ADI-2 Pro FS - measurement differences? Raspberry Pi 3 "Touch" with Topping D10 DAC system tested...

The test above is artificial in that we're not connecting anything to those inputs! What's supposed to happen is that we connect a signal from a low-impedance output to those inputs for recording / measurements. So, now let's hook up the Raspberry Pi 3 "Touch" system with the Topping D10 DAC and connect the RCA output from the D10 to an XLR-to-RCA to feed the RME ADC as shown in the dining room picture above.

When we do this, a difference picture emerges...


Click on the image to have a better look. As you can see, the noise floor in fact looks fine with either the SMPS or lithium battery now.

In fact, we can confirm that noise level is low by measuring the battery-powered Raspberry Pi / Topping D10 DAC at 24/96 with either the SMPS or battery-powered RME ADI-2 Pro FS. As you can see, there is very little difference in the results:


These are tiny differences and IMO insignificant; the kind of minor variability I might see between measurements turning devices on/off and cable switching. From 20Hz to 20kHz, there was no evidence that the switched-mode power supply added noise. In fact, the SMPS test showed marginally better noise level but the battery-powered RME ADC measurement showed slightly lower THD and IMD distortion.

Summary graphs:


What this means is that whether I'm using the SMPS or battery with the RME ADI-2 Pro FS, my results are equivalent at least down to the resolution of the D10. I suppose it's possible that with a higher resolution DAC like the Oppo UDP-205 using balanced XLR output, there could be more of a difference (not tested thus far). From a practical and "good enough" perspective, already these results indicate that differences at even lower levels would be absolutely inaudible!

Part III: Does it now make a difference if we measure the Raspberry Pi / Topping D10 DAC with SMPS vs. Li battery?

Okay, so, having determined that there wasn't much difference for the RME ADI-2 Pro FS whether the SMPS or battery was used, how about we now check whether the very inexpensive switch mode power supply vs. battery makes a difference to the analogue output from the Raspberry Pi 3 "Touch" / Topping D10 DAC.

Here is the RightMark result summary for all 4 combinations of ADC and DAC using either a switched-mode power supply or lithium battery measuring from the Topping D10:


On the far left we see the combination of lithium batteries for both the RME ADC as well as for the Raspberry Pi/Topping DAC. And on the far right is when we run the ADC/DAC system plugged into the wall with the power supplies.

As you can see, there's numerically not much difference at all! Here are the summary graphs:


Again, these are trivial differences between the various combinations of SMPS and Li battery used. In fact, the only consistent anomaly I see is the slight increase in noise at around 180Hz (3rd harmonic of 60Hz hum) which appears when we use the switched-mode power supply with the Raspberry Pi 3 "Touch" / Topping D10 DAC (red arrow).  Considering that this is way down at -135dBFS or so, no wonder this doesn't even really register in the numerical results.

Since the RightMark measurements above are at 96kHz samplerate (48kHz bandwidth), what about even higher frequencies? Let's turn the ADC input to the highest sensitivity +4dBu setting again, and samplerate at 384kHz to look at frequencies out to 192kHz:


There you go, beyond 48kHz, we now see that there is indeed a little bit more ultrasonic noise injected into the system from using the switching power supplies at ~55-60kHz. I can confirm that this is from the $10 SMPS being used for powering the Raspberry Pi / Topping D10 DAC and is not from the RME's stock SMPS. Notice again how low level this noise is!

Part IV: Summary...

Underwhelmed?

1. Yes, I can demonstrate quite readily that in certain circumstances like when the ADC input is disconnected, the switched-mode power supply can be seen to introduce noise that's picked up by the RME ADI-2 Pro FS. But in actual use connected to a DAC, the anomalies appear to be very low. Anomalies like 60Hz mains hum and high frequency noise did not intrude on high-resolution measurements (ie. at least of such low amounts in my test that they were irrelevant down to ~19-bits of resolution).

2. With the small Talentcell 6000mAh 12V battery pack, the results obtained with the RME ADI-2 Pro FS appear to be interchangeable with those using the power supply. I don't see any concern with direct comparisons using my standard measurement technique whether with the stock SMPS or Li-ion battery.

3. In further experiments using an inexpensive <US$10 switching power supply for the Raspberry Pi 3 with touch screen "streamer" connected by USB cable to a Topping D10 DAC, there was no evidence to suggest that the SMPS significantly impaired fidelity. Despite "audiophile wisdom" suggesting how "noisy" everything should be whether it's the generic cables used, supposedly "noisy" USB interface, "non-audiophile" board like the Pi 3, or even the fact that there's a 7" touch screen attached, I see no need for concern at all.

As perfectionists, we can see that there are some small anomalies, such as that little 55-60kHz cluster of low-level noise using the inexpensive switching power supply with the Pi 3 / Topping DAC. However, this is far out of the audio band and inconsequential for sound quality plus of low magnitude below -100dBFS.

If we allow ourselves to look at the writings of audiophile commentators on the Internet, we will no doubt run into all kinds of claims about the evils of switching power supplies (versus linear supplies and batteries). The results here support Benchmark's views a few years back that the anxiety around SMPS noise is actually an "Audio Myth".

I must admit that it's a bit annoying when some in the audiophile world claim things, and magnify effects, without any evidence or apparent curiosity to confirm their beliefs (a form of dogmatic "received wisdom"). This only worsens the image of the audiophile endeavor and hobby. For example, Mr. B, supposedly a longtime audio "journalist" expresses his beliefs in no uncertain terms and even claims he hears significant differences. Notice in 3:30 he claims switched-mode supplies are "extremely noisy" and "a lot of effort to reduce the noise to level for critical audio applications" is needed. Don't tell me, show me dude with that nice oscilloscope you have sitting behind you apparently perennially unused!

For what amounts to product infomercial video reviews like that, I have no problem believing that the Sbooster is a fine power supply. However, at least give us all some results showing that the Sbooster was able to lower this noise he spoke of, ostensibly a significant part of why there was an "immense increase" in sound quality over the years in his "Setup 1". Amazing.

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

With that, in the days ahead, you'll see more of my measurements being done with the battery-powered RME ADC.

All the best and happy listening!


Addendum:
Here's a picture of the back of the Pi 3 streamer with touchscreen using the SmartiPi case as per Mark's Blog. As you can see, I'm just using a Y-splitter for the USB power for the screen and Pi board. The splitter is part of the SmartiPi package.

Remember that the Pi, touchscreen, USB stick, and DAC will draw power from the battery. With a 2.4-2.5A power supply, you'll see some low-power warnings on the screen (lightning icon) especially during startup as it initializes the screen (indicator that the voltage dropped momentarily below 4.7V or so). I have not had issues thus far with file system corruption and can set the screen brightness lower to reduce load.

BTW, you can see the 16GB USB stick where I put my test signal files...

13 comments:

  1. How did you connect the "Touch" and the DAC to the 5V battery pack. My battery has two 2.4 amp USB ports. When I connect two low-draw audio devices, one to each port, I get tons of noise of the in-your-face humming/crackling variety. I had to resort to using a second battery pack. I thought of trying some sort of USB splitter cable connected to just one port, but did not have one.

    ReplyDelete
    Replies
    1. Yeah Mark,
      It's just a simple Y-cable for microUSB. Will include a picture above...

      Delete
  2. Dear Archimago: I like/admire most of your work, however here I am surprised you did not are aware of something relevant. There are NOT such things as "Lithium 12 or 5 V batteries": Lithium cells are 3'3-3'6 V nominal each, thus they add up to a bit more of that nominal PSU volatges. What you tested here and what many people uses, incorrectly calling them "batteries", are in reality power supplies BASED on batteries. You do not have 60 Hz hum from those, but you DO have a SMPS inside: the regulator circuit who lowers voltage from the Li pack to the output. And not all are exactly the same nor exactly clean... here you have a nice project to investigate ;)

    ReplyDelete
    Replies
    1. Thanks Unknown for the clarification.

      Indeed, as you say the lithium cells are ~3.3-3.6V nominal as per the chemistry and these "batteries" do include an internal voltage "switching regulator" as Mans also said below.

      I guess the point is that generally/colloquially we speak of these "lithium power packs" as "batteries" for our phones and tablets and such...

      An interesting table with some basic numbers and various types including LiFePO4, etc.

      https://www.houseofbatteries.com/battery-chemistry-chart

      Delete
    2. BTW Unknown,
      As you say, there are many types of Li packs and obviously I'm making some generalizations with the packs I'm testing here. Have you ever come across Li packs where the switching regular would actually cause noise / other problems for something like the Raspberry Pi / USB DAC I used here?

      Delete
  3. That battery pack almost certainly includes a switching regulator.

    ReplyDelete
    Replies
    1. Hi Mans,
      Do you know whether there are typical switching frequencies these work at?

      Delete
    2. Low-voltage DC-DC converters usually have a switching frequency somewhere in the 100 kHz to 1 MHz range.

      Delete
  4. I concur with your results, I recently tackled this issue with the power supply designs I'm doing where I've used switchers for everything. I made measurements against some simple op-amp buffer circuits as my test platform. I found that for the lowest noise designs I did need to add a small LC filter which got me another 3 dB, or -120 dBV RMS (unweighted) at the output.

    I did test against batteries as well (two 9V) and got the same performance. One thing to watch out for with switchers is if they are lightly loaded they start cycling at a few kHz and I have seen that be a problem.

    My testing notes are posted to https://clk.works/technotes/ but they're intended more as a reminder to myself of what I did vs. making blanket statements about supplies. TN004 includes battery operation as a comparison. I couldn't test all combinations but IMHO enough to show it should be OK.

    Though there's a clear caveat: The circuits being powered need reasonable PSRR, which if using op-amps shouldn't be a problem. If you custom design something discrete then I would certainly add a low noise linear regulator. Or if you want to get noise at the output below 1uV RMS, I think I would add the LDOs just be on the safe side as when you start putting together a bigger system interconnected to the real world noise can show up in unexpected places.

    Still be done is testing with the ADC/DACs and see if there's issues.

    ReplyDelete
    Replies
    1. Thanks Brewster,
      Great to hear from a hardware designer! Had a look at some of the technotes - lots of work put into those. Thanks for dropping by and adding to the discussion :-).

      Delete
    2. Thank you for the kind words.

      After I posted I realized that rather than someone having to go through all of the stuff I could post a plot that summarizes the battery vs. switcher noise measurement. I can't seem to post images, but here's a link:

      https://clk.works/wp-content/uploads/2019/07/smooth_fft_noise_compare.png

      The 5532 20-20k Hz unweighted noise is -107.8 dBV for both battery and power supply. The 797 is -120.3 dBV for battery and -119.4 dBV for the power supply. (moving cables around on the bench can easily change measurements at that level by over half a dB or more; there's a lot of stray RFI present)

      Details of the test setup:
      This used an AD797 and NE5532 op-amps in a non inverting (unity gain) buffer configuration as the power supply noise impact test circuit.

      The input was shorted at the (RCA) input to the test fixture (which connected through ribbon cable to the input and output boards) to be more "real world" like in terms of places to pick up noise from.

      The switching supply was a Meanwell NDS10-15D +/-15V (which uses very old tech) which was supplied by a CUI SW125-12-N wall wart (in other words one switching supply feeding another, which should generate all sorts of ugly IM products between the two - and it does, but at a low level). The Meanwell outputs were filtered by a 10uH choke and 100uF. An additional 100 uF was on the test fixture along with small caps for HF noise.

      16 K point FFT, average 32 times, and (approximately) 1/3rd octave smoothing.

      Caveat: The battery test is only +/-9V vs +/- 15 for the supply, so internal IC noise effects would be expected to shift with the different voltages. No effort was made to see what that effect was since the results really didn't change much.

      One final thought: when previously researching another (battery operated) project the question of "how noisy are batteries" came up. They do have measurable noise, but from the research papers I found it's going to be < 150 dB in level, as well a cap can filter it out.

      (and for those that want to see all the details see the link in the original post)

      Delete
  5. I did this a few years ago but no measuring equipment just my ears. Good to read your findings its certainly a cost effective way of improving sound
    A few thoughts if a SMPS is designed right it can be a lot less noisy
    If you want to take this even further try A Supercap PSU it takes this approach to another level John

    ReplyDelete
  6. Hi Archimago-san,

    Great topic. I use Radio controlled car batteries, 2x Tamiya 3700HV Ni-MH battery, to power Lynx Hilo for measurement purpose.
    Recently I purchased portable power pack Suaoki PS5B and testing its DC output socket and jump start outlet.

    ReplyDelete