MEASUREMENTS: Raspberry Pi 3 "Touch" music streamer

The music being played is the soundtrack from Japanese anime Your Name (Kimi No Na Wa). Saw it with the family last week - great flick!

A few weeks ago when I described the building of the Raspberry Pi 3-based music streamer with touchscreen (I'll just call it "Pi Touch" in this post), I promised that I was going to publish some measurement results.

Well... Let's have a look at those results! :-)

Remember that the measurements in this post will be built upon what we know already about the quality of the Raspberry Pi 3 as USB streamer to my TEAC UD-501 DAC and also the analogue output from the Pi 3 with the HiFiBerry DAC+ Pro. Please have a look at those posts for descriptions of sound quality and subjective impressions... This post is about finding objective change.

For completeness if anyone is wondering, here's the "Pi Touch" with the inexpensive switching power supply I'm using (it's the US$8 Enokay 5V 2.5A power supply - convenient with a push-button switch you see on the left):

The measurement set-up is very simple...

For analogue output (HiFiBerry DAC+ Pro, Squeezebox Touch):

Device being tested --> 6' generic RCA --> Focusrite Forte ADC --> 6' generic USB --> Windows 10 laptop

For digital output (USB):

Device being tested --> 6' generic USB --> TEAC UD-501 DAC --> Focusrite Forte ADC --> 6' generic USB --> Windows 10 laptop

The "Pi Touch" device had the screen brightness up at level "50" (unless otherwise indicated) using the latest version of piCorePlayer available currently (3.20 Normal Version; the "Audio optimized" version seems to not recognize the touchscreen properly for changing brightness in Jivelite). I made sure the screen stayed on with the "Currently Playing" album cover displayed during testing. These measurements are made with default piCorePlayer configuration settings, not the CRAAP tweak :-).

The Squeezebox Touch was updated with the EDO (Enhanced Digital Output) plugin that allows output to 192kHz with the custom kernel. The TEAC UD-501 I believes uses the Tenor (TE8802L?) USB chipset so you need to use the Triode Kernel #12 to get it working - check out this "how-to" page to get it working.

Below, you see the simplicity of this testing setup with both the "Pi Touch" and an actual Squeezebox Touch connected over WiFi to my Logitech Media Server machine in the basement, playing various test tracks. Obviously to get the touchscreen working on the "Pi Touch", I activated the Jivelite interface in piCorePlayer. On the computer screen is the FFT of the 24-bit jitter test probably from RCA output of the SB Touch based on how the cables are arranged. Very straight forward test of the analogue outputs from these devices. Remember that the Focusrite Forte can run completely USB-powered off the laptop which makes it convenient that I don't need to have the computer or ADC plugged into the wall. This also reduces or removes 60Hz AC hum in the measurements.

And here's what the devices looked like hooked up to the TEAC UD-501 DAC for testing USB digital output:

A. RightMark test results

Remember that these tests are mainly to see if the "Pi Touch" device with the official Pi touchscreen assembled in the case affected the already excellent objective results from a basic Pi 3.

16/44 (Analogue output from HiFiBerry DAC+ Pro):
Starting with standard CD resolution as usual...

What you see is the output of the "Pi Touch" using the HiFiBerry DAC+ Pro analogue output in the first column. This is followed by the actual Squeezebox Touch analogue out, then the Pi3-HiFiBerry DAC+ Pro without touchscreen, and finally on the right for reference my TEAC UD-501 DAC fed by a Windows 10 Surface Pro 3 computer using TEAC's ASIO driver that I have been using for years to represent a more expensive desktop DAC system.

Here are the graphs:

What's pretty obvious is that 16/44 poses no challenge for any of the devices really. We see that comparatively the old Squeezebox Touch has higher noise in the low frequencies; especially more 60Hz hum than the others. However, the HiFiBerry DAC+ Pro as I had shown before has a bit more IMD+N.

Notice that the touchscreen interface made no substantial difference to the measurements compared to the Pi 3 + HiFiBerry without touchscreen and the Jivelite GUI interface running.

16/44 (USB digital out to TEAC UD-501 DAC):
Likewise here are the measurements for USB digital output from the "Pi Touch" to the TEAC UD-501 DAC:

As you can see when you communicate with the digital USB asynchronous interface, what you end up with is "bitperfect" transmission and timing parameters inherent to the DAC itself... Clearly evident in the graphs:

Not only is 16/44 no big deal these days, but when using the same DAC, the graphs are essentially exact overlays of each other.

24/96 (Analogue output from HiFiBerry DAC+ Pro):
Here we go into the hi-res territory of 24/96...

Notice that in this series of tests, I wanted to see if the screen brightness made any difference to the measurements. With a 24-bit signal, we want to make sure no extra noise is affecting the output.

The 25% and 100% designations indicate just how bright I made the screen in the settings using Jivelite. Audiophile "common sense" might suggest that the higher the brightness, the more "noise" there should be, especially when measuring the analogue output from the HiFiBerry DAC+ Pro HAT board situated behind both the Pi 3 motherboard and the LCD touch screen of set brightness.

Well, numerically we see no evidence that there's much if any noise difference between 25% and 100% screen brightness; only 0.2dB difference in the noise level and if you compared with the old "Pi3 - HiFiBerry DAC+ Pro" results without the touchscreen (which was measured months ago), there's barely any difference to report.

Here are the graphs again:

Notice the frequency response graph is zoomed in to show the variation between devices in the high frequency roll-off. Very little difference down <1kHz.

24/96 (USB digital out to TEAC UD-501 DAC):
As for the digital USB output between the devices:

Again, we see just how robust the asynchronous USB interface is! There's just no measurable difference between the devices feeding the TEAC UD-501 whether it's the "Pi Touch" with screen brightness at 25% or 100%, the old Squeezebox/Logitech Touch with EDO plugin/kernel, or the Microsoft Surface 3 feeding the TEAC UD-501 with ASIO.

24/192 (Analogue output from HiFiBerry DAC+ Pro):
Finally, let's just have a look at the 24/192 test. As usual, RightMark messes up some of the graphs but I can at least show the frequency response and noise floor measurements.

First up is the analogue output with the HiFiBerry DAC+ Pro compared to others. Note that there is no Squeezebox Touch in this series because the internal DAC maxes out at 24/96:

Again, I have comparisons of the touchscreen at 25% and 100% brightness, then there's just the Pi3-HiFiBerry combo without the touchscreen/Jivelite, finally on the right is the Surface computer feeding the TEAC DAC via USB.

No significant difference between 25% and 100% brightness. In particular I don't see any unusual noise issue. In fact the TEAC shows a couple of noise spikes around 120 and 180Hz in the noise floor not found with the HiFiBerry DAC+ Pro output.

24/192 (USB digital out to TEAC UD-501 DAC):
And as for feeding digital signal to the TEAC DAC using the different devices through USB (Squeezebox Touch with EDO kernel supports 24/192 USB digital output):

Yet again, asynchronous USB simply "works"; no significant difference in the numerical results nor in the graphs. Again, no evidence of extra noise from the touchscreen between 25% to 100% brightness across the 96kHz audio bandwidth.

B. Jitter

Here's the J-Test with the HiFiBerry DAC+ Pro HAT attached to the "Pi Touch" device:

Now with the "Pi Touch" sending asynchronous USB out to the TEAC UD-501:

Obviously there is no worry at all about jitter being a problem! You can see the low level 16-bit jitter modulation signal with the TEAC UD-501 better because it has a lower noise floor (higher resolution).

In fact, I'll even show you the J-Test result with the Squeezebox/Logitech Touch feeding the TEAC UD-501:

Again, that's simply beautiful and a nice demonstration of the ability of asynchronous USB in terms of its ability to reject jitter!

Want to see some jitter, though?

Notice the difference when we connect that same Squeezebox Touch to the TEAC DAC but using a 6' coaxial S/PDIF cable instead! Clearly, we can see the effect of jitter and the effect of the temporal inaccuracies in the S/PDIF system. Remember that although this is easily demonstrable, I would not put any bets on blind-testing audibility!

C. A Closer Look At The Noise Floor

As suggested above with the 25% and 100% RightMark tests of screen brightness, there was no evidence that the screen added noise. For completeness, I wanted to have another look using a higher resolution FFT playing a "digital silence" track through the HiFiBerry DAC+ Pro HAT board while the screen is displaying an album cover at various brightness levels. Remember, the HAT board is just attached right behind the Raspberry Pi 3 motherboard and in quite close proximity to the touchscreen.

As you can see, there's no evidence of noise whether the screen is off at 0% brightness, dim 25% or bright 100%. Note that the 37kHz noise is from the Focusrite Forte ADC and not from the Pi 3 device.

D. Conclusions...

Well, there you have it. The "Pi Touch" device sounds great already and the objective results simply confirm what I've been hearing from this device with a touchscreen LCD grafted on and using UI software (Jivelite). Remember too that I'm just using the inexpensive US$8 switching power supply, no fancy USB reclocker, no expensive linear power supply, and simply generic cables. In summary:

1. The analogue output from the HiFiBerry DAC+ Pro is as measured previously. No evidence that the addition of the touchscreen affected the sound quality. No evidence that increasing screen brightness adds any extra noise or causes sonic deterioration. Compared to the original Squeezebox Touch, the HiFiBerry DAC+ Pro has lower noise floor, less stereo crosstalk, but a bit higher intermodulation distortion.

2. Digital output through USB to an asynchronous DAC (the TEAC UD-501) is free from jitter and measures just as well as other devices. Subjectively and objectively, I have no reason to believe that a digital source changes the sound through a good asynchronous USB DAC. I realize that this statement is in contradiction to some articles claiming otherwise - such as this one. I am basing my opinion on the measurements above plus my own experiences over the years... I would certainly like to see some objective evidence from those making the opposing claims.

3. Remember that whereas asynchronous USB can be free from jitter, S/PDIF interfaces (TosLink, coaxial) are generally not. This is I think nicely demonstrated above comparing USB output vs. coaxial cable with the Squeezebox Touch to the TEAC DAC. This is contrary to some opinions I have seen that because USB is a "general" computer interface, it is somehow "bad" for audio. Nonsense. A good asynchronous DAC will be able to handle noise and provide better time domain output with USB in my experience.

As a point of comparison, the old Squeezebox Touch coaxial S/PDIF seems to be a little more jittery than the Chromecast Audio's TosLink output at 16-bits but better at 24-bits.

One last thing... How much CPU is being used with the "Pi Touch" when streaming high-resolution music while the Jivelite user interface is running?

As you can see, it's 3.7% here while streaming a 24/88 track over ethernet from the Logitech Media Server computer in the next room. In fact, 3.7% is on the high side. Most of the time the CPU usage is around 1.5-2% with 24/88 and 24/96 music. With 24/192, I see CPU utilization hovering mostly around 3%. Needless to say, the Raspberry Pi 3 isn't particularly strained with piCorePlayer and Jivelite running! This means there's plenty of CPU cycles available for other things which we'll talk about another time :-).

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

Okay folks... As I said last week, life is busy :-). Having said that, I have been sitting on these results and just had to get them out since it has been a few weeks already since the previous "HOW-TO" build article.

Other than the "High End Munich 2017" show this week, in audio news recently, I see that MP3 has finally been deemed "patent free" as of April 16, 2017 (although formally there is this one that runs through to Dec. 30, 2017). Well, finally open-source software like Linux distributions will be able to have full releases with MP3 capability now that IIS Fraunhofer/Technicolor have terminated their licensing program ("Full MP3 support coming soon to Fedora"). Certainly nice to have the convenience now of full functionality included in the Linux distros.

This is good news right? It is fascinating the perspective however from the side of the rights holders. For example, check out this title from Billboard magazine - "The MP3 is Officially Dead: License Terminated by Developer"!

Well yes, the licensing program might be "dead", but IMO MP3 itself is as alive as ever :-). Of course license holders want to steer the industry to use file formats with extant licensing agreements. Sure, the newer lossy formats may be "better, [with] lower-bit files", but it's not that much better by the time we're looking at 256kbps and above. Remember the MP3 blind test done here years ago suggesting that MP3 ~320kbps was transparent (and might even be subjectively preferred by some!) for music compared to lossless 16/44.

I think we can look at JPEG as an example of how an old, lossy format continues to thrive today as the most used compression standard for 2D images (JPEG patents expired by the end of 2006). Despite technically superior contenders like JPEG2000 or BPG, good 'ol JPEG is going strong and remains essentially unchallenged. So too I suspect will be the fate of lossy audio encoding... There's just no compelling need to change. MP3 is not dead, and in fact it has just become stronger as the de facto, now free, and universally compatible lossy audio format for the foreseeable future.

We should all remember the contributions of open software developers, especially those who worked on LAME, in maturing the code and advancing the psychoacoustic tuning.

Hope you're all having a lot of fun with the audio hobby!