Sunday, 13 May 2018

MEASUREMENTS: Computer USB +5V Power Noise by Raoul Trifan & MQA's Bob Stuart Speaks Again...


Editor's Intro:
Hey there guys. One of the joys of running this blog has been the opportunity to meet some of you and connect through E-mail with folks from around the world with shared interests.

Every once awhile, I get the chance to see some of the measurements and work you guys have done. Today, I want to show you some work from Raoul Trifan from Romania.

In the last year, we've been in contact about his work with USB ports using the various computers he has at his disposal. What he's sharing today is literally the "tip of the iceberg" in terms of the data and measurements collected over the last while. He has done all kinds of other work with power supplies, modifying audio gear and such which I'll leave for future discussions.

Today, let's just consider the noise on the +5V USB line from computers:
Using a PicoScope 2204A, he has produced some interesting screen shots to consider with the various machines.

A Look At The Computer USB +5V Noise...


By Raoul Trifan

In high fidelity, we are aware that we should minimize the noise coming from our power supplies as this could improve the ultimate sound produced by the audio equipment. For example, Analog Devices has shown us how an AD9445 ADC can be affected by the ripple and noise created by removing the decoupling capacitors from its power supplies. So, indeed, very clean power does matter at a certain level, but just how much attention and money do we need to get into this may be up for some debate.

Because many audiophiles are spending money on USB power injectors and power cleaners, I decided to see how noisy the USB ports from some laptops I happen to have around here can be. I’m using for this test my PicoScope 2204A digital scope with its original probes, my Windows 7 and macOS Sierra desktops and a few laptops with the original power adapters and original lithium batteries. Some of the laptops having also their original docking stations available to measure. The machines are the Lenovo X201, Lenovo T430, Lenovo T440, HP2560p, HP8460p, MacBook Pro (mid-2012), MacBook Pro (mid-2014), iMac 27” (late-2013), and my DIY desktop computer based on the Gigabyte GA-B85M-D3H motherboard.

First, let's have a look at the Lenovo Thinkpad T430 with its multitude of ports:


As you can see, the vertical scale is in mV with each division line set at 10mV. Horizontally is time with each division line as 20μs. We can see the amount of noise in the DC power line fluctuation. For this computer, I like the left side 1st USB port and also the rear USB port (by the power connector), but I totally hate how dirty the power is on the docking station’s ports! Notice that the voltage fluctuations go above 30mV at times with the docking station. Definitely not recommended to connect your DAC into this docking station.

The next machine is the Thinkpad T440:


Similar to the T430, I like the left USB ports, but I do not like how the power on the docking station’s USB ports perform. Again, not recommended to connect your DAC into this docking station!

Moving along to the next Lenovo machine, this is the old Thinkpad X201 from around 2010:


Surprisingly, the old X201 has the best USB power from all tested laptops in this survey of devices! Again, the docking station seems to show a bit more fluctuation with spikes ~10mV in magnitude but even then you could probably connect your DAC without concern. That right side USB port looks excellent. Quite impressive results from an old device like this one.

Next is a Gigabyte GA-B85M-D3H motherboard used in my desktop machine and the desktop Apple 27" iMac from late 2013:


I’m impressed by the clean USB power from my DIY 4-years-old desktop with the Gigabyte GA-B85M-D3H motherboard when the rear USB ports (soldered onto the motherboard) are used. As for the wired front USB port, please don’t even try using these. Clearly the cheap USB connector and long cable connecting the motherboard to the front of the computer case is picking up noise from inside the case.

Given the audiophile perception (maybe general perception) that Apples are good machines for audio, I was hoping for a cleaner result from the late-2013 iMac, especially since the rear USB ports are soldered directly onto the motherboard (no USB cables involved). It looks like the ripple and noise spikes are not so good!

Despite the spikes from the iMac above, I was unable to hear any issues with the sound itself, but I had no time to do a detailed A/B comparison and have no decent ADC to run RMAA. Anyway, most likely any engineer designing audiophile products seeing the iMac’s USB power waveform above will probably recommend using a good external +5V power supply for the USB or a Y-cable with clean +5V injector.

Next we have a couple of HP notebooks - the Elitebook 2560p and newer 8460p:


The HP 2560p might not having the purest USB power ever, but it’s definitely much better than the bigger brother 8460p! Unlike the Lenovo computers, the HP docking station shows much lower noise level. This just goes to show that I cannot generalize and say that "docking stations are bad" because it obviously depends on the specific device.

As mentioned above, Apple has created quite a following and there are audiophiles who feel that the Macs are better for audio purposes than Windows-based options. To some extent this is understandable given that Macs are often rather quiet sounding machines and aesthetically more pleasing (especially going a few years back). But what about the USB port power noise level? Here are two MacBook Pro laptops:



As you can see, on the 2014 MacBook Pro the USB port near the power plug is noisier than the other port. Apple machines are clearly not special in any way in terms of power noise from what I can see. The Lenovo laptops have cleaner +5V on many of the USB ports.

Finally, let's have a look at a decade-old Fujitsu-Siemens Amilo Pro V3205:



The right side port from my old Amilo Pro is quite decent, too bad the left port is significantly poorer.  The rear port sits in between with the amount of noise seen on the scope.

Archimago's Thoughts and Conclusions:

Thanks Raoul for letting me show these measurements you've collected! As I mentioned in the introduction, Raoul has actually done a bunch of other measurements he has graciously shared over the last couple of years. It's great that someone out there has taken the time to measure the USB power line noise level from a selection of laptops and desktop machines to demonstrate the variability out there... Furthermore, it's interesting to see just the amount of variability between the different USB ports on the same computer. I suggest very obsessive "high end" audiophiles should break out the oscilloscope and check to make sure they're using the quietest USB port on whatever device they're using! :-)

I think this is interesting because after years of being told as audiophiles by the likes of AudioQuest (and their Jitterbug device) or the cottage industry of USB "cleaner" devices like UpTone (and their USB Regen) or iFi Audio and their host of "purifiers", there really doesn't seem to be much out there showing what the noise looks like or at what levels they are to be found from a typical computer USB port. I would have thought that measurements like these from USB ports would be part of the promotional material to demonstrate the "problem" and then the company showing how their devices clean up these voltage imperfections. Would that not be a "no brainer" from the perspective of marketing and public education?! Also, as usual, where is the audiophile press in all of this over the years to educate the hobbyists and buying public? Would it also not be nice to have simple USB power cleanliness measurements on very expensive audio computers like this ~£3000 Innuos Zenith or >£2000 Melco N1A/2 to double check that at least they're performing in ways superior to just a standard off-the-shelf computer?

Unfortunately Raoul noted that he has not conducted A/B listening tests and doesn't have an adequate ADC at this time to measure the effects these USB voltage anomalies have on the analogue output from USB DACs. From measurements over the years and as far as I can tell, DAC output does not appear to vary significantly with the machines I've used. At least I can say that I have found no measurable difference using even inexpensive DACs powered from the USB port like AudioQuest's own Dragonfly V1.2 awhile back, or the AudioEngine D3 - never mind the more expensive wall-powered USB DACs! There's other evidence of this littered throughout my measurements over the years beyond those 2 examples.

Hopefully at some point Raoul will do some measurements with his own DACs. For example, it would be very interesting using something like a Dragonfly plugged into the poor quality T430 laptop dock USB port and comparing the output with the relatively clean left USB port.  Since we know the noise level is much poorer with the dock, are we then able to correlate this with DAC noise levels for example? (Maybe someone wants to take on this little project! :-)

For now, as far as I can tell with my computers, even with inexpensive DACs these days, the power management and filtering is doing a good job without any extra help from audiophile accessories.

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

Well, I see that the latest issue of Stereophile (June 2018) has been released in the digital subscription form. As you might expect, MQA is yet again discussed (I guess they just have to keep it in the minds of audiophile consumers for some reason). At the time of this blog post, I have not seen the article released online yet (I'm sure it will since it is very much a "message from the manufacturer") [it's posted now]. This time, Jim Austin writes yet another MQA Q&A with Bob Stuart's comments.

Not unusually, some more terminology gets mentioned - this time B-spline interpolation is tossed out for the reader to consider and they're talking about the "center of gravity" being in the right place in regard to temporal accuracy. There is acknowledgement also of both aliasing and imaging present with MQA but of course said to be inaudible. I don't know about what you guys are perceiving but there does appear over time to be a softening of how actually "revolutionary" this technology is; rather than bold and hyperbolic statements of greatness, it's now widely acknowledged as lossy and more of "yeah, we know there are things like imaging and aliasing... But it ain't all that bad... And the 'center of gravity' is in the perfect place".

I see that the only measurement offered is another FFT of one of 2L's tracks (2L-078). A new "improved" version of the encoding algorithm was done on the music with less noise compared to the DXD playback. So what? The FFT shows that the portion of the track they're demonstrating has no content above 23kHz! So maybe the new encoding may have utilized steeper filters to reduce ultrasonic leakage or maybe they applied lower dithering / noise shaping levels compared to before? Good to see that potential "bug fixing" can be done and there's flexibility in the processing I suppose. However, a simple downsample from DXD to standard 24/48 would have resulted in just as beautiful of an FFT (likely with an even lower FLAC-compressed bitrate)!

The final sentence of the article goes like this (about the "de-blur" and time smear claims):
This smear, we believe, can be material for the human listener who is extracting multiple cross correlations, as well as envelope and nonlinear measures of the audio.
I don't understand exactly what Mr. Stuart is claiming here; what "cross correlations" specifically? What "envelope"? Which "nonlinear measures"? Yes, the mechanisms of listening and hearing are complex, but would it not be good to be a little more specific? But the point remains that after all these years (and despite earlier in the article Stuart claims they did "hundreds of experiments"), at best he can say is that he "believes" what they did with MQA was helpful! Surely, out of these hundreds of experiments, he can spare just a little more information about what beneficial evidence they may have found. Given countless opportunities, it's rather clear that they've got nothing. And writers like Jim Austin likewise has nothing to add other than uncritically repeating company claims. At this point, I really hope the McGill University folks have something "material" to talk about (maybe some actual data by the Milan AES in late May?).

Well folks, MQA's marketing began to get heated a few years ago (eg. here and here) with Bob Stuart's Q&As and I hope it ends with his Q&As. Fascinating that there are not other technical "evangelists" talking in support of MQA. It always just ends up about Bob Stuart! It would be unfortunate if the legacy of Mr. Stuart is so strongly tied to this questionable (and IMO unnecessary) codec.

If in the days ahead, we see music companies and streaming services actually adopt MQA as a distribution format, I think it's pretty clear that this is in spite of public acceptance of the format rather than what companies would prefer to see - acceptance because of public demand.

This week, we were told that ESS has decided to incorporate MQA "rendering" into their DAC feature set. It looks like they're targeting this for "small and lower power solution" DACs meant for mobile devices and headphone users. Something like the Dragonfly Black rendering on-a-chip. IMO, with actual high performance DAC chips like the ESS 9018 and recent 9038Pro, it's hard to imagine why anyone would want to use those MQA upsampling digital filters on playback as they can only serve to reduce sonic fidelity from these high-end parts.

Anyone getting really tired of MQA yet? :-) I noticed that there was little talk of MQA at the AXPONA 2018 coverage and so far not much from Munich High End (just a couple new partners and their insistence that this represents a "growing momentum"). On the software side, I guess it's finally about time that Audirvana Windows version is released with software MQA decoding; about a year after the Mac. Also, Roon 1.5 now handles MQA including the ability to perform DSP on the unfolded data. Nice to see that MQA is relaxing not only the insistence on achieving "revolutionary" sound quality, but also on how rigidly they're "end-to-end" "authenticating" the playback chain... Yet again, why bother with all the hassle in the first place? I think it is good that Roon has MQA support now given the asking price and the fact that they're closely tied to Tidal. It would be rather unsatisfactory to pay quite a bit for the Roon service and not be able to access the much-hyped Tidal "Masters" feature!

Have a great week ahead folks! As you can imagine, I do quite a bit of keyboarding for work and blog writing of course. This past week I got a new mechanical keyboard; the Kingston HyperX Alloy Elite in my home office and its little brother the HyperX Alloy FPS at work. It has been awhile since I last used the clicky keys of the Cherry MX Blues (thankfully my wife doesn't mind)! I think the last time I owned a keyboard with Cherry switches was with my early generation Amiga 2000 computer in the late 80's :-). Nice to be back with a great tactile keyboard rather than the usual low-profile, shallow-travel "Chiclet" keyboards... In retrospect, I should have done this years ago!

BTW, for those who want something quieter, consider the Cherry MX Brown switches; still feels great with less noise. Also, one could dampen the MX Blue switch noise a little with <$10 O-rings under the key caps which I might do if the wife complains!


Until next time, enjoy the music everyone...

31 comments:

  1. MQA

    Now it begins....

    http://www.cdjapan.co.jp/feature/Worlds_First_HiRes_CD_by_Universal_Music_Japan

    Ian

    ReplyDelete
    Replies
    1. Hey Ian,
      Yeah, it'll be interesting to see how many folks "bite" at the opportunity to own so-called "hi-res CD" this time round. In Asia, every few years we see yet another iteration of higher resolution CD - whether it was XRCD, XRCD24, HQCD, UHQCD, Blu-Spec, DSD-sourced CDs... I'm certainly all for better quality materials, better mastering and more precise stamping of the bits. But MQA is of course something else; yet another logo to add to the list. Does anyone still get affected by this anymore considering we're still ultimately talking about a 16/44 PCM audio stream!?

      Good luck to Universal. I'd be surprised if they can sell many "yet-another-version" of The Who's Tommy or The Stones' Sticky Fingers :-). At some point, the gig's up.

      Delete
    2. Oh yeah... Especially when the asking price is >$US25 per CD on old music!

      That's nuts.

      Delete
    3. MQA CD reminded me of CCCD, HDCD and with collaboration of Universal, the horrible DRM watermarking.

      https://en.wikipedia.org/wiki/Copy_Control
      https://www.mattmontag.com/music/universals-audible-watermark

      The most distinct "feature" of HDCD is squashing the peaks on non-compatible CD players rather than "expanding" them on compatible players.

      https://en.wikipedia.org/wiki/High_Definition_Compatible_Digital

      Basically all of these either worsen compatibility or fidelity.

      Delete
    4. Yes, for Y3000 you get MQA (with all it's limitations) while Y3500 gets you the SACD version. Not a difficult decision.
      Especially for me - I already have 33 high res. versions of the 65 in the Pop and Jazz sections.

      Delete
  2. I think the next step is plugging a reputable DAC or interface and measure how these USB noise patterns affect the actual audio signal but yes, I would generally avoid front panel USB apart from stuff like keyboard or mouse since plugging at the rear usually weaken their reception.

    Sometimes my thumb drives are not being recognized if I plug them into the front panel jack, maybe I am just a poor DIYer :P

    MQA CD... historically everything ended with "CD" with proprietary stuff eventually failed apart from the original CDDA, or the so-called Red Book. Time will tell.

    ReplyDelete
    Replies
    1. Agree, the next step is to determine if the power irregularities can be shown to affect sound quality significantly (beyond the usual testimonies and casual listening of course).

      I can certainly dream of interesting experiments to try out. My hope is that this post will spur on more curiosity and measurements/listening tests to elucidate claims even further.

      Delete
    2. So it means we should plug USB DACs into rear ports on the desktop and into direct (not docking station) USB ports in notebooks. Or have USB DACs with their own power supply (USB used only for data connection). And when we feel bad about sound from particular connection, especially on notebooks, it is worth to try another port. Thanks for measurements.

      Delete
  3. I am sorry to be critical about the USB measurements, but measuring a USB port's + pin with an unbalanced and most probably earthed DSO, from machines that have switching power supplies and also rely on earth, and backfire their switching energy into the mains (AC leakage), I know from my own experience that you have to isolate your measuring gear in every possible way: use battery operation, galvanic isolation for the Picoscope's USB etc. The noise/signals shown above might just be a problem of the measuring setup and might look much different when measured in a different setup.

    Is the author aware of these typcial problems with his Picoscope, and what did he do to prevent them?

    ReplyDelete
    Replies
    1. All measurements done with PicoScope connected to battery powered laptop. Thank you for pointing this out!

      Delete
    2. Thanks for the clarification!

      Delete
    3. Raoul, in the (IMHO terrible) ASR forum one moron ignored my question and your answer. But he came up with another unclear point: were your measurements done unloaded or loaded? My experience is that the difference is zero. Simply because there is no dedicated power supply per output port, and the internal ps is already loaded with other stuff.

      Delete
    4. Hi again, unloaded USB ports. Please give at least 1 weekend, so I can do some loaded vs. unloaded measurements, OK? I'll choose one of the above laptops, but not an identical one (same model, but different P/N), so unloaded measurements could differ a bit from the above posted already.

      Delete
  4. Interesting idea, and good work, to measure the noise on the USB supplies. A scope is really the best way to do that. What I'd like to see is measurements done with some kind of load. Often, power supplies will have more ripple or noise as their load goes up, though some may get quieter. Testing with a real world load would be best. (and might change thi picture entirely!)

    ReplyDelete
    Replies
    1. > A scope is really the best way to do that

      That's a point for discussion. Meanwhile I prefer to do FFTs with standard audio interfaces instead. You see a lot more in these, and you see more of the really audio relevant stuff that potentially comes over USB into your DAC etc. It won't show how stable the 5V are, and which voltage exactly is there, but it shows (much better) what this is all about - how clean it is...

      Delete
  5. My first USB DAC was the ODAC and it was sensitive to voltage fluctuations on the USB port. The standard fix was to add a powered USB hub but I used one of those inline USB power monitors to test all of my devices and find the ones that didn't have a problem. The differences between machines and even between ports on any given machine were amazing.

    Power was certainly a issue in the performance of early USB powered DACs. Howwever, the latter USB DACs don't seem to have a problem with dirty power and the advantage you gain by powering off the USB is you aren't introducing anything new like potential ground loops into your system.

    ReplyDelete
    Replies
    1. Thanks for the note George.

      I've never had a chance to play with the ODAC. I can certainly imagine some of the older designs having more issues. It has been awhile though since the ODAC came out from NwAvGuy back in 2012. In the intervening period, we've also updated significantly from USB Class 1 and into asynchronous interfaces...

      It would actually still be interesting to measure the analogue output of the ODAC between clean and dirty USB power to see if one can correlate the measurements!

      Delete
    2. Yeh in NwAvGuy's desing write up he stated, "Worst Case – If someone does encounter a USB port that is either extremely noisy or can’t provide 125 mA of current, the solution is simply to use a powered USB hub."

      I easily verified that current fluctuations will kill your sound quality (no measurements required). When we first hooked the ODAC up it would distort badly on music peaks. Once I got the tester I quickly identified that the USB on the front of Dell desktop was trash. The ports in the rear were a little better but none of them were as good as my old tank Dell Latitude laptop.

      I just recently retired the Dell for a cheap Asus but not until I spent hours chasing a system interrupt on that machine. It was some sort of battery management routine that added a nice "pop" about every three seconds. Stuff like that makes the case for something like your OPPO that's dedicated to music.

      Delete
  6. Some. DACs do not require the 5v supply from usb. Does it make sense/is it possible to measure noise on the data and/or ground? I have no technical knowledge so please forgive me if this question seems idiotic!

    ReplyDelete
  7. Hi Archimago, is there a way to contact you by email? I have a question about another topic you wrote about last year. Thanks.

    ReplyDelete
    Replies
    1. Hi Josef,
      The best way to get hold of me is actually through Computer Audiophile's private messaging.

      Too much of a pain with E-mail these days since I try to just use one address mainly for work.

      Delete
  8. "I don't understand exactly what Mr. Stuart is claiming here; what "cross correlations" specifically? "
    Do not worry, he does not understand it either :)

    ReplyDelete
  9. One should realize that most of the better USB DACs, that use the +5V from the USB side (for instance to power the USB receiver), have filtering of the incoming USB power supply voltage.
    The better (usually more expensive) the DAC the better the filtering in general is.
    This means that higher frequencies in the +5V line are filtered out.
    With the cheapest USB dongle DACs less filtering or even no filtering is used.

    Certain issues are of far greater importance than the +5V power supply.
    For instance: ground loops, USB receiver implementations, USB drivers, PCB design, used receivers or soft-/firm-ware to name but a few.

    It would be more interesting to see how much the analog output quality (noise/distortion level) actually changes with these different USB ports.

    ReplyDelete
  10. Hi, I'm just discussing in an Italian forum about the effectiveness of https://www.amazon.com/Wyred-Sound-Recovery.../dp/B01CUTMKYM.
    Me, having read your previous comments on the subject, I qualify that "thing" as "snake oil".
    And you?

    ReplyDelete
    Replies
    1. For the price it can be considered snake oil.

      It is basically an USB hub (not a switch) with just one output and perhaps a better clock in it with perhaps a bit more powerline filtering.
      There is no isolation against groundloops which, most likely, is a bigger problem when running into weird sounds coming from a music system.

      It will be a lot cheaper to buy a normal hub with a small (linear) 5V power supply feeding it and connect only the DAC to one of its outputs.
      That has the same functionality, and maybe even the same chip.
      Don't worry about it not having a femto clock. Not important for USB transport despite 'audiophile experts' claiming otherwise.

      Whether it may bring benefits depends on the chain it is used in.
      Very early USB DACs or DACs with a very crappy USB implementation may possibly benefit somewhat.
      The better DACs won't.
      The improvements heard by some will mostly fall into placebo domain.
      In some rare occasions it may be possible real improvements could be had.

      Delete
    2. Thanks. I checked the Service Manual of my Audiolab 8200CD.
      Port 1 (+5) is not used.
      Why should I worry about the noise on that line?

      Delete
    3. Indeed the +5V is not connected in the 8200 CD, only the datalines and ground are.
      So.. in your case it does not matter.

      Delete
    4. Hey Teodoro,
      Yeah as per Solderdude, it looks like a "one-port-hub" like the USB Regen products that came out a few years back.

      Whether any of them make a difference to the sound is IMO questionable. I have yet to see any measurements to suggest significant change to noise or jitter amount. If "better-placed sound stage" and "blacker backgrounds" are audible, then these changes will be measurable through the analogue output of whatever DAC they feel can benefit.

      The claim on that link that this device will result in "much improved sonic quality" is clearly over-promising :-). In that regard, until proven otherwise, this is "snake oil" in my opinion.

      Delete
  11. the docs coming with my UD-503 say:
    Isolation circuit that completely separates the grounding of the digital and analog sections
    Between the digital and analog sections, the circuit design uses a digital isolator to completely isolate both the power supply and the ground. So, all noise coming from digital input sources, starting with noise from the computer via USB, is prevented from entering the analog section by way of the power line or the ground. An isolation circuit offers significant benefit particularly when playing high-resolution audio sources with high sampling rates. The UD-503 is the first product in our Reference Series to include one.

    ReplyDelete
  12. I'd love to see measurements of what something like a Jitterbug does to this noise... eliminating that horrendous dock USB result and turning it into a nice flat line would definitely be a good thing to put on the marketing materials :)

    ReplyDelete
  13. I wonder if the Gigabyte DAC-UP USB ports are even better than the ones in the B85M-D3H board.

    ReplyDelete