Saturday 1 December 2018

MEASUREMENTS: JustBoom Digi S/PDIF HAT for Raspberry Pi. (Plus "Why isn’t digital audio 'just ones and zeroes'?" anxiety...)


I mentioned a few months back in September that I updated my Raspberry Pi 3 B+ Touch device with the JustBoom Digi HAT for S/PDIF coaxial and TosLink digital audio output. Over the last few months, I've been using it more regularly for my nightly music listening.

As I noted previously, this device is essentially a 1-chip solution featuring the Cirrus/Wolfson WM8804 transceiver with jitter attenuating PLL (rated intrinsic jitter of ~50ps RMS) released in 2009. It also uses the Murata DA101JC isolation transformer for the coaxial output.

The device is rather plain otherwise as you can see. It does sound good and is capable of up to 24/192 with both coaxial and TosLink outputs. And at <US$40, it's also very affordable. Let's see how well it measures!

Here it is attached to the back of my "Touch" Raspberry Pi 3 B+. I have the 7" official Raspberry Pi touchscreen out front using the SmartiPi Touch enclosure.



This will be my first time evaluating a S/PDIF digital audio board like this. You might recall that awhile back, I measured the HiFiBerry DAC+ Pro HAT as analogue audio out.

First things first, here is the general measurement set-up I'll be using:
S/PDIF streamer (Raspberry Pi 3 B+ with JustBoom Digi or Squeezebox Touch) → 6' generic TosLink / coaxial → DAC (TEAC UD-501 or RME ADI-2 Pro FS or Oppo UDP-205) → 6' XLR balanced cable → RME ADI-2 Pro FS ADC → generic USB → Windows 10 measurement laptop
I've highlighted in bold the major components to the set-up. Notice I'll at times use the old SB Touch for some comparisons. As usual, piCorePlayer (current version 4.1.0) is used for playback with the Pi 3 B+. I'll focus mainly on the TEAC and RME DACs for the conversion to analogue demonstrating the differences that DACs make. Finally, we'll be using the RME ADI-2 Pro FS as the ADC for the "3rd Gen" measurements. For simplicity, I'll just focus on 24/96 high-resolution measurements unless otherwise specified. Other than saving some time, I can tell you that I found no significant differences at "standard resolution" 16/44.1. As with most measurements over the years using decent gear, 16/44.1 did not pose a challenge for modern gear.

For the Raspberry Pi streamer, I'm just using a generic 3A 5V switching power supply plugged into my wall socket here in a residential neighborhood in the city of Vancouver, my house wiring is about 10 years old and there is no special dedicated circuit to my sound room where the testing took place. In sum, nothing fancy or expensive used.

Let's do this like a Q&A as we explore whether the inexpensive JustBoom Digi does a good job...

Q: Compared to asynchronous USB, how well did the S/PDIF output from this HAT board perform?

Let's jump right into some RightMark 6.4.5 Pro results:


Notice that I spread out the results into 2 clusters - the TEAC DAC cluster to the left, and the RME ADI-2 Pro FS as DAC on the right. Results with "JB Digi" obviously means I'm using the JustBoom Digi HAT either with coax output or TosLink. The USB results were obtained with the Raspberry Pi 3 B+ USB port plugged directly into the DAC.

Numerically, we can see that the performance is excellent across the board. It would be remarkable for anyone to claim that any of the results would stand out audibly. I personally cannot hear a difference. Nonetheless, the measurement system can indeed show us some small differences. We can see that the asynchronous USB measurements achieve slightly better low-noise performance. Distortion values for THD and IMD are also very similar except for the IMD+N sweep where there's more of a spread in results with the TEAC DAC. Here are some graphs to examine. I zoomed into the frequency response and noise level graphs to better see the minute details:


As you can see, the differences are rather benign looking. Frequency response differences +/-0.1dB between DACs and interfaces. No real difference in the noise floor except overall lower noise level with the RME DAC (no surprise as it is a newer device). Low stereo crosstalk overall. Notice that for the IMD+N sweep, the TEAC UD-501 shows elevated results with the JustBoom HAT and the TosLink was worse than coax. Essentially no difference for the RME DAC.

If we are to generalize, it looks like the JustBoom Digi's coaxial output achieves slightly lower noise level than the TosLink. Likewise, specifically for the TEAC UD-501, we see that THD and IMD distortions seem just a tad higher with TosLink. No evidence here that the inherently galvanically isolated TosLink connection works any better than the coax cable.

Q: You're using an LCD touchscreen with your Pi 3 B+ playback device. Doesn't that make a difference such as by contributing extra noise to the system?


Bottom line, no evidence of difference whether the Pi LCD touchscreen is on or off using coaxial connection to the RME DAC (again the TosLink interface would inherently be galvanically isolated). The frequency response was exactly the same. Noise floor also same whether 96 or 192kHz sampling rate with no evidence of higher noise when touchscreen plugged in and turned on.



Q: Well then... Since "bits are not bits", what about jitter?

Ah yes, jitter. As we depart from primarily tests of bit-perfection and noise level into temporal performance, let's see what we find. Starting with the 16 and 24-bit J-Tests from the RME ADI-2 Pro being fed the TosLink input...


Not bad at all! A single tiny sideband pair in the 24-bit J-Test down below -140dBFS. Impressive performance from the RME device and not surprising as we saw this behaviour before. What this shows us is that paired with a good DAC with accurate temporal performance, the JustBoom Digi certainly is an excellent performer.

Now, check out the TosLink output fed into the Oppo UDP-205 as DAC:


We see the Oppo UDP-205 low-level sidebands presumably due to voltage modulation as discussed previously, but otherwise no significant jitter anomaly to speak of.

However, not all DACs are as "perfect" as what we see above! Here's the TEAC UD-205 (the machine circa 2013):


As you can see, the TEAC UD-501 shows more jitter than the above Oppo and RME DACs. In fact, with the JustBoom Digi, the 24-bit J-Test pattern looks worse than the 16-bit test! If we now change over to the coax S/PDIF output from the JustBoom Digi to the TEAC, a different pattern emerges:


It looks like with the coaxial output, the 16-bit J-Test got worse, but the 24-bit J-Test improved compared to the TosLink. In any event, the strongest amplitude sidebands were found with the 24-bit TosLink test which is consistent with the general impression that all things being equal, TosLink tends to have higher jitter than coaxial S/PDIF.

Overall, the jitter-rejection capability of the TEAC UD-501 with S/PDIF input is inferior to the newer Oppo and RME DACs; asynchronous USB is not an issue however. I didn't bother capturing the J-Test result with the coaxial connection to the Oppo and RME as they didn't look any different from TosLink. As previously measured (Oppo, RME), jitter results are excellent for both devices regardless of TosLink or coaxial S/PDIF input.

Q: Is there anything else we can keep an eye on?

Glad you asked :-).

Remember, S/PDIF data transmission is unidirectional and the DAC has to use a PLL to "follow" the data stream and synchronize with the sending device; in this case, the DAC has to follow the JustBoom Digi clock. This synchronization process and clock imprecision can result in jitter as demonstrated above.

We can also ask, however, beyond the temporal anomaly from jitter, just how accurate is the clock in the JustBoom Digi? That is, if the device is sending out an "exact" 1kHz or 10kHz signal, how precisely does it "hit" that frequency target? And can we measure the relative difference between different S/PDIF devices?

Yes... We can! Using the RME ADI-2 Pro FS as the reference measurement device, if we feed a 10kHz signal to the TEAC UD-501 through either a Squeezebox Touch coaxial output or the Raspberry Pi 3 B+ with the JustBoom Digi, how precise is the output frequency to the ideal 10kHz  tone (as judged by the RME with its "femtosecond" clock)?


The above was captured with SpectraPLUS-SC running a 1M-point FFT, with the ADC running at 192kHz. The target frequency is 10kHz and I'm zooming into the 9.9-10.1kHz narrow range.

What we see is that compared to the RME reference, the JustBoom Digi is -0.061Hz (-0.00061%) off the reference, whereas the Squeezebox Touch's S/PDIF signal is +0.122Hz (+0.00122%) "faster". Of course these are not audible differences but I think it's nice that even such minutiae can be measured. I doubt even the most capable human with "golden eared absolute pitch" would claim an audible difference! :-)

By the way, we can see that the Digi has slightly less "skirting" at the base of the 10kHz signal compared to the Squeezebox Touch when fed to the TEAC DAC suggesting less low-level jitter present. Obviously, nothing of concern when listening.

In summary...


Some generalizations then:

1. We can see very slight pitch variations with S/PDIF. We see that there is an insignificant difference between the Squeezebox Touch and JustBoom Digi HAT board for the 10kHz output for example. This is to be expected since the on-board clocks on different S/PDIF out devices will be close but not exact. No surprise. In fact, we can expect that as devices warm up, the clock will fluctuate slightly. For example, years ago (2013), I showed the samplerate drift when I did some Audio DiffMaker tests over time; good to be aware of this when doing testing even though in day-to-day listening, this would not be a problem.

By the way, speaking of frequency stability, for those who have never tried, with a good turntable and cartridge, have a listen to a stable test-tone (like a 1kHz tone) played back. You might be shocked at how audibly obvious even low amounts of wow and flutter sounds as that vinyl disk spins! I assure you that pitch-perfect hearing is not needed to detect the fidelity issue. [For more about this, remember the PlatterSpeed results from a few years back here, here, and here. One would be very fortunate to have a turntable and LPs with +/-0.25% frequency deviation!]

2. Clock fluctuations, which results in jitter anomalies, appear to be primarily a function of the DAC itself. Looking at the results from the DACs above, we can see that the Oppo and RME performed significantly better based on the J-Test results than the TEAC UD-501. My recommendation over the years has been that if one is to build a good digital streaming system, focus on the DAC.

I do not recommend spending a ton of money on the computer/streamer itself (much of this discussed a couple years ago) despite the dramatic claims of interviews like with Innuos (and this one) when it comes to sound quality. Just make sure the usability and features are there for what you need! Stuff like the "importance" of linear power supply, SSD vs. RAM playback, OS, software, etc. make no sense and are without evidence. Even if we can measure minute differences, what needs to be judged ultimately is audibility. As far as I can tell with measurements and my own controlled listening, there are no differences between decent streaming devices and computer systems unless one purposely subjects the device to terrible conditions (eg. unreasonably running CPU and GPU full tilt whole measuring a DAC!). Of course, make sure to maintain bit-perfect data transmission, no buffer under-runs, resolve any issues with stuttering, gaps, or other digital data corruption. As usual, when it comes to the Innuos interviews, this is all talk with no evidence from a company that sells what amounts to low-power computers with a 1TB hard drive starting at £699 (~US$900).

3. With the JustBoom Digi HAT board, there is no evidence of excess noise with the coaxial output versus TosLink in the measured DAC output. In fact, the objective results suggest that noise level was actually a tiny bit lower with the coaxial output. No evidence either that noise level changed whether the 7" LCD screen was turned on or off (the HAT board is maybe 2 cm behind the screen with the Pi 3 B+ motherboard between). What this means is that I have no evidence to support the idea that fancy galvanic isolation is needed beyond the inexpensive Murata DA101JC isolation transformer (<US$2 a piece) built into this inexpensive HAT board.

4. Ultimately, when it comes to measurements of digital S/PDIF output, we can say that "bits are not just bits" - but let's not go nuts about what this means. Yes, we can see that a "more jittery" device like the TEAC UD-501 will show more frequency anomaly on the J-Test and in other tests like the IMD+N sweep which probably is related. But as we see here, even from a <US$40 HAT board for the Raspberry Pi, performance is excellent and there is no evidence to suggest that much if any improvement is needed, especially if we select a capable modern DAC as per Item 2 above.

So, as you can see, the JustBoom Digi HAT board does an admirable job for its intended purpose of bit-perfect S/PDIF playback with TosLink and coaxial outputs. Remember that as a "system", the sound is very much ultimately dependent on the DAC it's paired with. Anomalies like jitter are low and IMO inaudible as shown with the RME, Oppo, and TEAC devices despite it being objectively worse with the TEAC. Sure, one could pay more money for other HATs -  maybe the HiFiBerry Digi+ Pro (~US$50), the multifunction Wingoneer WX4000 (DAC + digital out, <US$70), or Allo Digione (US$99). I suppose it could be interesting to compare results. But given this excellent level of performance for a <$40 product, I think time is best spent listening to music than being concerned about likely minute differences!

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

Before the closing bit, permit me to digress a little into what Mr. Vitorino says in his videos linked above. Specifically, notice in the Darko interview at ~9:10 there's talk about noise and interference - perhaps showing an example of the noise could be useful. He also starts trotting out the idea of the superiority of linear vs. switching power supply with no evidence. Maybe even easier, perhaps he can show us the "noise" of the USB system that shows up in DACs these days (~15:00)? How about concerns about noise from wireless data transfer versus ethernet (~27:00)? I also don't understand why latency is such a big deal with this guy (~31:00, mentioned a couple times there and earlier as well) - no lip sync issues with hi-fi playback, right?

Rather disturbing that audiophile websites keep "advertising" this kind of "information". This doubtful chatter is why recently I posted this on the Steve Hoffman Forum about "keeping it simple" for a music lover starting off in this hobby. There are many unsubstantiated claims out there and this can create all kinds of fears, uncertainties and doubts instead of recognition that one does not need to go into the world of esoteric "high end" audio to achieve excellent fidelity! Sadly, I believe one could go down the proverbial "rabbit hole" and end up with a less-than-high-fidelity system or waste time and money on snake oil. It's tragic that so-called "journalists" in audio have no apparent ability to show some critical thought when fed all kinds of unlikely claims without even simply asking if the (supposedly knowledgeable) designer/engineer has any proof at all that they can share! If one were to ask about the concept of "truth in advertising", where would one look when it comes to verifying advertisements such as these?

Look folks, let's just call it what it is and what it's about. Some individuals and companies in the "high-end" audiophile industry need to create the fear in order to induce the interest of those who are susceptible to audiophilia nervosa. The last thing "they" want is to see many audio lovers sense contentment that the hardware is just fine. And even worse, to have potential customers even think for a moment that the gear is already "perfect" - as in capable of exceeding the auditory perceptual limits either of oneself or more broadly, human perception - is anathema.

And of course, there are "journalists" willing to work on the above industry intent of presenting every new product cycle as something better and desirable. This is the "job" of some writers and websites; they've gotta make a living, right? Within this circle of industry sponsored websites and perhaps magazines, contentious opinions are problematic. Web sites like Darko's have shut down public comments for more than a year and some places heavily censor responses citing incivility while trying to maintain some superficial claim of being "open minded". Sometimes we don't want to hear the truth but it must still be expressed nonetheless.

IMO, these are obviously not signs of health when beliefs cannot be challenged nor obvious truths allowed to be expressed. At best they can be "protective" of the companies who depend on such services to advertise their goods for a time. In a world that is increasingly interconnected and information more readily available, this will not advance a hobby nor create "good faith". Rather, they will exacerbate the fractures that exist and widen the extremes. At worst, create cliques hanging on to potentially delusional, or cult-like "statements of faith". Sadly, I would not be surprised if the survival of some companies depend on this.

Good luck audiophiles... Remember, the advertising industry depends on emotional extremes. Euphoric hype works just as well as intentionally stoking anxiety and fostering uncertainty as with the example from Innuos (and in years past, iFi's Gremlins). There is already much contentment to be found in the multitude of excellent choices of hardware we have these days and certainly for many years already. As digital technology continues to advance, the price of entry for high quality sound continues to drop but there is a "ceiling" beyond which no audible differences will be experienced - this too is another truth that some do not want to acknowledge. When this happens (I think we're already there when it comes to DACs), many will sing the praises of devices that "sound different" not because of better fidelity (as per discussions of NOS DACs last week). Stay cool, rational, and remember to enjoy the music first and foremost.

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

I'm enjoying Michael Bublé's new album Love. Easy listening, non-challenging vocal pop as I get stuff done around the house. The rendition of "La vie en rose" is quite nice... It's Cécile McLorin Salvant who "steals the show" for me in that duet though.

Honest question. Did this album need to push amplitude levels such that "When I Fall In Love" ends up looking (and sounding) like this?


Lots of audio limiting going on here...

The DR meter is giving me a reading of 5. At least it doesn't sound too distorted, but the loudness gets a bit wearing after a few songs. This would be a good album where Reprise's hi-res release could be less stingy with a few dB's. Come on Mr. Record Producer, could you give a brother a few more dB's of dynamic range with an "Advanced Resolution" version?

For "fun", we could run this album through iZotope's "De-Clip" algorithm as discussed previously to artificially reconstruct those peaks: -6dB Gain, "Extreme Analog Clipping" De-Clip, and then dithering back to 16-bits. Voilà:


Not gonna vouch that the post-DSP version sounds "much better" since this is an artificial reconstruction and what is lost is gone forever, but it does have an effect for highly compressed audio like this. The average value for the album has been bumped to DR8 in any event.

That's all for this week... As suggested above, I hope you're all enjoying the music! Happy December.

13 comments:

  1. Nice review .... That Wolfson WM8804 spdif chip is a winner for sure with its ultra low jitter spec. New to your site BTW and have been spending the weekend reading your blogs. I mostly appreciate your objective take on the detailed measurements and what they mean. As an EE, my congratulations on your due diligence and skill in taking consistent measurements.

    I,being an Oppo UHD-205 owner (still new and unopened in box for 2-3 months now!), I read through all your tests and they were very thorough. But I was disappointing to see that you never did any DSD tests on it, like you so thoroughly did for the Oppo Sonica DAC (or was it the HA-1 amplifier?). Any reasons for this? No other reviewer (from other magazines that perform technical tests) has done so.

    Would you be so kind as to test the 205 for DSD 64/128/256/512 and even PCM 768KHz? I have no clue how much work this would entail and I'm sure its quite a lot. I just wanted to have these tests done, if possible, as one of the ways to pay final tribute to a king of all players that we'll miss for sure...but also see how well they designed it for DSD since I have quite a SACDs in my collection.

    Thanks for the great blog...
    David

    ReplyDelete
    Replies
    1. Greetings David!

      True, I did not measure the Oppo UDP-205's DSD output. At most I went up to PCM 192kHz.

      These days, I can run RightMark PCM tests up to 24/384 with the RME ADC. And sure, I'll run some DSD64/128 numbers in a follow-up post. DSD256 and above could be done as well.

      Welcome to the blog!

      Delete
  2. That would be most welcome,if you did testing all the way to DSD512 and PCM 384KHz. I'm hoping its a simple(???) automated macro that you run to do all these tests so that your precious time is not wasted on testing :-).

    Please let us know when these tests are up. Thanks for the great blog!

    ReplyDelete
    Replies
    1. No worries David.

      The truth is, testing doesn't actually take too much time even though we don't see many audiophile writers actually bother to do it. It's the write-up and trying to convey the findings that take time :-). If the posts and ideas can be of benefit to others, I'm certainly happy to contribute. Instead of spending time mindlessly surfing the web or watching YouTube videos which I'm apt to do... :-)

      Delete
  3. Soooo...do you have a rough eta on these DSD tests? :-) Thanks.

    Sadly, Oppo refused to give me their DSD test data when I asked :-(

    ReplyDelete
  4. I'm very confident many DSD/SACD listeners find your tests very beneficial. No doubt!

    ReplyDelete
    Replies
    1. Yup, will post on the DSD results in early January I think!

      Delete
    2. Amen to that brother! Can't wait for the test results :-)

      Delete
  5. You note the poor jitter performance of the TEAC UD-501 coax and toslink inputs relative to those of the RME and Oppo units. Might that have been the cause of the Chromecast Audio’s poor jitter performance? Would it make sense to retest the Chromecast Audio with either the RME or the Oppo?

    ReplyDelete
    Replies
    1. Hi Steve,
      Yes, that is certainly a strong likelihood.

      If one were to connect the Chromecast Audio to something like the RME and Oppo, no doubt the J-Test would improve.

      If I have some time with the Chromecast Audio, will demonstrate this...

      Delete
  6. Hi,
    When you say ' compared to asynchronous USB' is that means computer to USB dac?

    ReplyDelete
  7. Thank you for this incredible article and all the precise work you do.

    I have powered speakers with a built-in DAC and a 100% digital amp. I wanted to keep the signal digital throughout the entire pipeline. The speakers have two digital inputs: USB & Toslink. I'm trying to decide what would give me the best audio quality (or if there would be a noticeable difference) while playing USB Thumbdrive music with Raspberry Pi 4:
    A) The USB output on the Raspberry Pi 4.
    B) JustBoom HAT w/ Toslink
    C) HiFiBerry HAT w/ Toslink

    Thanks again! I appreciate any wisdom.

    ReplyDelete
  8. A nice and thorough review as usual (I discovered your site yesterday, and have been enjoying reading lots of Pi streamer related posts).

    With regard to power supplies, I've had an interesting few days. I've got a Model B+ Pi running piCorePlayer with a HiFiBerry Digi+ S/PDIF HAT.

    I initially used it with the standard Pi 2.5A power supply. It sounded good at first glance, but I soon noticed that it sounded harsh. Some shrill high frequency instruments had a bit of an edge to them that I found hard to listen to. The final little flourish at the end of the second track on Carey Bell's Harpslinger album is a good example, should anybody fancy trying to reproduce this on their system.

    I started looking into why this might be, thinking "but surely they're just 1s and 0s" and "why doesn't it sound just like my CD?" My external DAC is also my CD player (an Audiolab 8200CDQ), so it seems reasonable to assume that they should sound pretty similar.

    I should mention at this point that I'm quite sensitive to fatiguing systems. A couple of years ago I got (permanent) tinnitus when testing a pair of speakers that were particularly harsh in this regard, so I suspect I'm probably more sensitive than some.

    Anyway, I recently discovered the chat about linear power supplies, noise filtered switched mode power supplies, etc. I also noticed somebody recommending running a streamer off a 5V USB battery pack. I have a 10,000mAh battery pack, so I tried it. The Pi + Digi+ went from hard to listen to (and tinnitus inciting - seriously, a listening session makes my tinnitus worse) to smooth at the high end and fuller/weightier in the midrange. And it didn't exacerbate my tinnitus.

    I was happy with the sound, and thought I should get a linear PSU to resolve the problem more permanently.

    But before I bought one, I swapped my speakers. I'd been using my Mission stand mount speakers (which are great), but swapped them out for my Neat floorstanders. The Neat's do airy and spacious treble better than the Mission's, and I immediately noticed that the streamer had lost (relative to CDs) a lot of the high end frequencies that give you that sense of space.

    For a while I thought maybe the Digi+ just couldn't produce those frequencies. Then I switched back to the 2.5A PSU and sure enough, the high frequencies were there. They were just a bit unpleasant and made my tinnitus worse pretty quickly.

    So I started trying other wallwart power supplies that I had in the house, comparing them to my CD player by listening to the same track on both. I made detailed notes as I went.

    The Pi 4's standard supply delivers 3A, and so is fine for running the older Pi's if you get a USB-C to micro USB adapter. It sounded better; the harsh edge wasn't immediately obvious, but I could tell after about half an hour that my tinnitus was getting worse and my ears were getting fatigued. Midrange and bass was slightly fuller than on the 2.5A supply too.

    Then I tried the 3A supply that came with my Google Pixel 3a. Immediately I could tell that it had the high frequencies, and was much easier on the ear.

    Whether the differences are down to "noise" or not, I couldn't tell you. It's clear to me that changing how the power is supplied has an impact on the sound. And it's clear from your measurements Archimago that noise levels don't look like they should be able to explain it.

    I think the next thing for me to do is to order a linear PSU and see if I prefer the sound to the one I'm getting from my Pixel phone charger (if I don't I won't keep it). I am very happy to say though that when playing FLAC files the Digi+ is now in the same ballpark as my Audiolab CD player, which it clearly wasn't when running on a battery.

    ReplyDelete