 |
| Battery-powered Raspberry Pi 3 B+ with JustBoom Digi. Bohemian Rhapsody soundtrack. |
Remember that within audiophile circles, there are often all kinds of
claims about noise affecting streaming devices; especially
computer-based devices. I already touched on this last week when referencing the videos by Innuos. Over the years, I had addressed software playback, different OS's, lossless formats, "need" for ethernet galvanic isolation, and explored the lack of difference between machines used for playback.
In a similar way this post will also explore some "received wisdom" which seem so common among mainstream industry-supported audiophile websites.
Since I had my set-up running for last week's evaluation of the JustBoom Digi, I figure I might as well do a few more tests to demonstrate for myself (and you, fellow readers) whether these beliefs hold any truth.
Belief 1: Power supplies to digital streamers make a difference to the sound. "Noise" is particularly an issue with switch mode power supplies because they add all kinds of nasty high frequency noise.
Just to be clear. This claim is not about the DAC itself, but the streaming machine/computer being powered by a switching power supply and then connected to the DAC.
Recently, I bought one of those power banks to use on my overseas trip back in late October. I needed something that was acceptable in airplanes but big enough for a few phone and tablet recharges. I settled on the RavPower Ace 22000mAh pack which goes for around US$40:
It worked well for its intended purpose overseas and I had no problem with crossing international carry-on security checks. It comes with a couple short USB cords and soft pouch. I guess the only thing I could have wished for was a USB-C cable... No big deal.
I wondered. How about if I powered the Raspberry Pi streamer with this? This thing is said to be equipped with a good Panasonic power management system, can provide +5V and 2.4A so should be enough to power the Pi 3 B+ with touchscreen and the S/PDIF HAT. Indeed, when I tried hooking it up, it does work as shown in the image at the very top streaming music off WiFi through the HAT S/PDIF output.
As a bonus, I rummaged through my drawer of trinkets and found this:
It's a "free" battery I got awhile back when I attended a conference. 5V out with up to 1A current, 2200mAh capacity (1/10th that of the RavPower). Enough to power the Pi 3 B+ (CRAAP-undervolting disengaged, just standard clock speeds and voltage) but with the screen disconnected. Let's see if I can discern a measurable difference when powered with this "cheap" power source in comparison.
So, would using lithium-ion batteries like these make any difference compared to an inexpensive wall-wart switching power supply like this "generic" Aukru 5V 3A with on/off switch. Let's find out. Remember, even if we argue that the lithium batteries could have noisy voltage converters internally, it would be nice to see a difference in noise especially since we're also going to be using two quite different batteries!
Belief 2: WiFi (wireless in general) is "bad" according to certain subjective audiophiles. Presumably because of the need for complex data transmission and radiation of signals and this ostensibly will increase "noise". Some even go so far as to predict that wireless data transmission will also increase jitter. Okay then, let's do some tests between ethernet from the Raspberry Pi and using the internal WiFi.
Belief 3: Devices like screens can cause problems because they add to the complexity of the device and are supposedly "noisy".
Belief 4: While we didn't show any special superiority of TosLink to coaxial S/PDIF last time, maybe TosLink, being optical in nature can still be beneficial for galvanic isolation. Let's see if this results in improvements during these tests.
In the testing today then, let's try to address these since I can run tests with one or more of the above conditions relatively easily and see if there are distortions and noise to be found.
Here's the testing set-up:
You might be wondering - "why are you using the old TEAC UD-501 DAC for this"? Well, precisely because it's a 5-year old DAC! :-)
You see, my newer DACs like the Oppo UDP-205 and RME ADI-2 Pro FS are just too good to show differences. When I can't even demonstrate jitter anomalies and at best very slight distortions with S/PDIF input as discussed last time, what potential do I have to show anything for a test like today's unless changes are quite obvious? What I know is that the TEAC UD-501's XLR output is still excellent with very low noise floor superior to RCA outputs and I know it is susceptible to jitter and has distortion anomalies when I compare USB input compared to S/PDIF. This sensitivity to anomalies makes it the most likely platform to show differences if there are any to be found.
Test Set 1: Comparing USB to Coaxial S/PDIF - switching power supply vs. battery, varying ethernet/WiFi, Screen on/off
As you can see, I have the results of 5 RightMark runs above with the conditions defined in the caption. All tests are done at 24/96 since if there are detectable changes, it should show most clearly in high-resolution. The first 2 tests on the left are USB measurements from the Pi 3 B+ to the TEAC UD-501 asynchronous USB input. The only difference is whether WiFi is on - as you can see, no difference.
The last 3 columns are using the JustBoom Digi's coaxial S/PDIF. The middle column test is still using the switching power supply while the last 2 tests on the right are with the RavPower battery whether with WiFi or ethernet and LCD screen on or off (disconnected).
Hmmmm... Anyone see any improvement using the lithium batter? Not to my eyes. These are essentially small differences due to inter-test variability one sees when looking at noise levels down at almost -120dB and distortions below 0.001%! I guess one could say the the tests with the battery may have improved the noise level by about 0.4dB over the switching power supply. I was able to reliably show about 0.3-0.5dB lower noise level over 4 runs with the switching power supply vs. battery. This is of course simply academic and ridiculous to suggest audible differences (certainly I did not hear a difference playing some music).
The only clear objective difference really is with the IMD+N sweep values and we can see this in the composite graphs; again, I would not suggest that this is audible:
We saw this IMD+N sweep deviance previously and I assume this is related to higher jitter with the TEAC UD-501 receiving S/PDIF because we didn't have this issue when testing the Oppo UDP-205 and RME ADI-2 Pro FS where jitter was much better controlled across the different digital inputs.
Okay then, how about with the TosLink test set?
Test Set 2: Comparing USB to TosLink - switching power supply vs. battery, varying ethernet/WiFi, Screen on/off
Again we have 5 results on the board. The left 2 columns are the same USB results as above for comparison. And to the right we have the TosLink measurements.
Yet again, there's really not much to see. My suspicion is that the TEAC DAC's TosLink interface is simply more noisy, hence higher noise level overall compared to USB and coaxial inputs. Whether due to the noise floor being higher or the galvanic isolation with optical, notice there is no difference between the noise level and dynamic range measurements of the switching power supply compared to the battery.
As demonstrated last week, TosLink is prone to higher jitter. The IMD+N sweep has consistently higher distortion than either USB or coaxial S/PDIF.
Again, no evidence that having the LCD screen on or totally disconnected made a difference. As for WiFi vs. ethernet, if this changed anything, the difference seems insignificant.
Test Set 3: Comparing the high capacity RavPower vs. the "cheap" freebie low-capacity battery
Alright, then... Would an inexpensive, low power 2200mAh lithium battery make a difference? Are there noise differences between power packs presumably introduced by the battery or power management system?
On the left are the numbers measured from the DAC fed with coaxial in. On the right the TosLink numbers. Remember that at 1A output, the "cheap" battery had to have the touchscreen disconnected otherwise the Raspberry would not boot.
Nothing remarkable with the numbers and just as unremarkable composite graphs:
No evidence of any extra noise nor distortion whether I used the RavPower or "free" lithium battery. Notice that the IMD+N sweep results segregated based on which interface was used (coaxial vs. TosLink).
Test Set 4: Wideband noise floor
Fine. So far we don't see any extra noise with 24/96 tests... How about checking to see whether powering the Raspberry Pi with a switching wallwart vs. lithium ion batteries, turning the touchscreen on/off, using WiFi or ethernet will affect the DAC's output all the way to 192kHz. Easily done by using the RME ADI-2 Pro FS ADC sampling at 384kHz (using WaveSpectra, 128k FFT points for the graphs):
Nothing of significance. At most, we can make out the slightly noisier TosLink (eg. the orange tracing) hovering higher than some of the others. Alright, maybe we can look at a log scale and see if there are anomalies down in the lower frequencies?
Yet again. Nothing of significance. Whether using the inexpensive Aukru switching power supply or using the lithium batteries with the screen turned on or if WiFi was used, as you can see, none of the conditions using the Raspberry Pi 3 B+ streaming device affected the analogue output noise level from the downstream DAC; even in this case with a 5-year-old TEAC UD-501.
Test Set 5: Jitter differences!?
There's always jitter, right? ;-).
Here's what I got previously for TosLink input using the switching power supply:
And here is TosLink to TEAC UD-501 using the RavPower lithium battery:
No real difference... The only way to improve this substantially is to use the less jittery asynchronous USB interface to the TEAC DAC:
There you go. "More ideal" looking J-Test results; especially the 24-bit J-Test.
Summary...
I think the results speak for themselves. I do not believe nor have heard a situation where the streaming device makes much if any appreciable difference assuming one is using a reasonably good DAC and the data transmission is bit-perfect. What we've seen in these measurements using the Raspberry Pi 3 with just a simple switching power supply is that:
1. There is no evidence that using the switch mode power supply adds to "noise" when connected to a 5-year-old DAC even down to very low noise levels using XLR analogue output.
2. Lithium battery packs do not seem to improve anything. No evidence between a new high-capacity battery pack (RavPower) compared to the "cheap" freebie battery stick.
3. No evidence that a 7" LCD screen adds any noise to the output even from the coaxial S/PDIF connection. This was reported in my JustBoom Digi measurements last time.
4. No evidence using the JustBoom Digi S/PDIF HAT board that the coax output was significantly susceptible to electrical noise. At most, we're only seeing 0.5dB difference down at the noise floor between switching power supply and battery power.
5. No evidence using the Raspberry Pi 3 B+ as a streamer that ethernet or WiFi made any significant difference to the DAC output (even with the streamer sitting literally right on top of the DAC during testing transmitting/receiving wirelessly).
6. Even with a wide bandwidth analysis of the noise floor up to 192kHz, there was no evidence that using a switching power supply, lithium batteries, WiFi connection, ethernet connection, or LCD screen turned on in the streamer device contributed to any extra noise down to -140dB in the audio band and below -110dB all the way to 192kHz.
While I believe XLR is the best analogue connection for those desiring highest fidelity, we can of course wonder if issues may only show up with RCA... Perhaps we'll look at this in another post.
Well, hopefully that relieved anxieties for some. The reader might now say... "Okay, so the Raspberry Pi 3 as a streamer is rather immune from power supply issues and such, but maybe if someone were using a computer, I bet the switching power supply and noisy computer would still result in a ton of noise to your DAC!"
Maybe we can consider that a bit more next time even though in the past we've had some discussions in this regard.
Finally, one could as - "Why don't you try a linear power supply?" Well, indeed, we could try that (maybe something like the inexpensive Nobsound 25W adjustable linear supply), or maybe a specially advertised "low noise" power supply like the iFi iPower, SBooster, or the myriad of choices on eBay. But here's the question - why bother? We know they're more expensive, run warmer, and are less power efficient. Unless there is evidence to show that indeed noise levels are kept lower or they result in lower distortions when connected to the DAC in certain conditions, why should consumers even consider doing this other than for experimentation purposes? I think the onus is on the manufacturers to give consumers a reason if they insist that upgrades or components are anything more than a waste of time and money for audiophiles. As usual, perhaps the audiophile "press" should do a little investigative journalism before declaring that any difference can be heard (especially this guy here, LOL)?
I trust everyone survived the Thanksgiving and "Black Friday" shopping mayhem a few weeks back. I find it amazing how consumerism has continued to expand during the end-of-year holiday season over time! Anyhow, I don't see many "must have" tech gadgets this year but I did grab a deal on some Corsair Vengeance LPX 16GB DDR4 RAM for ~US$100 to boost my workstation memory to 32GB RAM. Not that I absolutely need this much memory, but sometimes I see that I hit >80% memory usage when doing stuff like video editing and encoding with 4K content.
Also, anyone else notice that web browsing these days can take up a ton of RAM?!
Wow. 7 tabs on Firefox ate up >3GB RAM and one single tab open to edit this blog post used up 1GB in Chrome. Seems rather inefficient to me!
I don't remember when I first updated my main workstation computer to 16GB of RAM. Probably back in early 2012 when I put together my Intel Ivy Bridge i7-3770K machine. Since I suspect the speed of upgrades appear to be slowing due to generally a lack of need for most consumers, I would not be surprised if 32GB will last me awhile... At least this post will serve as a reminder for me when I made this upgrade.
For those unaware, most of the time memory is sold as kits; usually something like 8GBx2 or 16GBx2. This allows manufacturers to component match and make sure the RAM can run at the "overclocked" speeds. For DDR4 RAM, the base clock rate is typically 2133MT/s (1066.5MHz). The Corsair 8GBx2 pair is rated at DDR3000 which is 1500MHz using the overclocked XMP profile.
Alas, even though I have 4 sticks from the same brand, on my AMD Ryzen 1700 workstation (overclocked of course), I could not get the MSI motherboard to POST beyond 2400MT/s speed. After a few trials, I didn't feel it was worth pursuing this much further even with RAM voltage pushed to 1.4V. Maybe I can try again later if I'm bored.
However, I do like efficiency, low voltage, and squeezing a little more performance out of the RAM. So what I did was "tighten" the timing on the RAM while keeping it at 1.2V. Best JEDEC DDR4 2133 timing for these RAM modules are identified as: 15-15-15-36-50. I pushed it down to 14-14-14-28-45 while running at DDR4 2400. I've kept Command Rate at 2T but 1T was stable with voltage up to 1.25V but this provided little benefit. Runs stable with Intel Burn Test ("Maximum" stress level for all memory) and Prime95 (24 hours). Even if I don't have the benefit of higher clockspeed on the memory, at least latency is better by a few nanoseconds for short random access bursts of data...
Cruising into December and year end!
Speaking of "cruise", I enjoyed the latest Mission Impossible: Fallout the other night with the family; probably my favourite of the MI movies. Very good action flick with a fantastic demo-worthy Atmos soundtrack. However, I was a bit surprised at the quality of the video. Despite a supposed 4K digital intermediate and some fantastic high-resolution 16:9 aspect ratio digitally filmed portions (Red Weapon Dragon camera used I believe), most of it was clearly from 35mm film presented at 2.39:1. It looks muddy, blurry at times and grainy. Remember the 4K comparisons between 35mm film (eg. The Prestige, Blade Runner) and true 4K-worthy digital or 70mm IMAX film (Blade Runner 2049, Dunkirk) earlier this year. I guess some might like this kind of look or there's some artistic merit to this. It's a little like thinking about "hi-res" audio in 24-bit and 96+kHz "containers" but not recorded or produced with high-resolution equipment or technique. Other than the fine HDR contrasts and highlights, most of this movie would have looked just as well on 1080P Blu-Ray or streaming online. Anyhow, a bit disappointed that a big budget movie like this didn't look substantially better for much of it.
Lastly, on the music release front, The Beatles' White Album 50th Anniversary 2018 Remix is now out along with the Super Deluxe package that includes a Blu-Ray with the 5.1 surround mix. The surround mix is excellent! Whether it's hearing an old favourite like "Ob-La-Di, Ob-La-Da" as if it's the first time, or tripping out on the technically manipulated "sound collage" of "Revolution 9", the 5.1 version is pure fun... Don't miss this if you have a surround system!
One last thing for those who have a music archive and want to save space, IMO feel free to convert the 24/96 5.1 mix to 16/48 5.1. Most tracks look like they have just various amounts of noise beyond 24kHz. Here's "Ob-La-Di, Ob-La-Da" during playback:
As an analogue-sourced album from 1968, one cannot expect "true" high resolution audio quality.
Enjoy the season, movies and music everyone...
Since I had my set-up running for last week's evaluation of the JustBoom Digi, I figure I might as well do a few more tests to demonstrate for myself (and you, fellow readers) whether these beliefs hold any truth.
Belief 1: Power supplies to digital streamers make a difference to the sound. "Noise" is particularly an issue with switch mode power supplies because they add all kinds of nasty high frequency noise.
Just to be clear. This claim is not about the DAC itself, but the streaming machine/computer being powered by a switching power supply and then connected to the DAC.
Recently, I bought one of those power banks to use on my overseas trip back in late October. I needed something that was acceptable in airplanes but big enough for a few phone and tablet recharges. I settled on the RavPower Ace 22000mAh pack which goes for around US$40:
It worked well for its intended purpose overseas and I had no problem with crossing international carry-on security checks. It comes with a couple short USB cords and soft pouch. I guess the only thing I could have wished for was a USB-C cable... No big deal.
I wondered. How about if I powered the Raspberry Pi streamer with this? This thing is said to be equipped with a good Panasonic power management system, can provide +5V and 2.4A so should be enough to power the Pi 3 B+ with touchscreen and the S/PDIF HAT. Indeed, when I tried hooking it up, it does work as shown in the image at the very top streaming music off WiFi through the HAT S/PDIF output.
As a bonus, I rummaged through my drawer of trinkets and found this:
It's a "free" battery I got awhile back when I attended a conference. 5V out with up to 1A current, 2200mAh capacity (1/10th that of the RavPower). Enough to power the Pi 3 B+ (CRAAP-undervolting disengaged, just standard clock speeds and voltage) but with the screen disconnected. Let's see if I can discern a measurable difference when powered with this "cheap" power source in comparison.
So, would using lithium-ion batteries like these make any difference compared to an inexpensive wall-wart switching power supply like this "generic" Aukru 5V 3A with on/off switch. Let's find out. Remember, even if we argue that the lithium batteries could have noisy voltage converters internally, it would be nice to see a difference in noise especially since we're also going to be using two quite different batteries!
Belief 2: WiFi (wireless in general) is "bad" according to certain subjective audiophiles. Presumably because of the need for complex data transmission and radiation of signals and this ostensibly will increase "noise". Some even go so far as to predict that wireless data transmission will also increase jitter. Okay then, let's do some tests between ethernet from the Raspberry Pi and using the internal WiFi.
Belief 3: Devices like screens can cause problems because they add to the complexity of the device and are supposedly "noisy".
Belief 4: While we didn't show any special superiority of TosLink to coaxial S/PDIF last time, maybe TosLink, being optical in nature can still be beneficial for galvanic isolation. Let's see if this results in improvements during these tests.
In the testing today then, let's try to address these since I can run tests with one or more of the above conditions relatively easily and see if there are distortions and noise to be found.
Here's the testing set-up:
S/PDIF streamer (Raspberry Pi 3 B+ with JustBoom Digi HAT) → 6' generic TosLink / coaxial → DAC (TEAC UD-501) → 6' XLR balanced cable → RME ADI-2 Pro FS ADC → generic USB → Windows 10 measurement laptopAs typically done here, I'm streaming with piCorePlayer. The RME ADI-2 Pro FS has been updated to the newest firmware (185/89).
You might be wondering - "why are you using the old TEAC UD-501 DAC for this"? Well, precisely because it's a 5-year old DAC! :-)
You see, my newer DACs like the Oppo UDP-205 and RME ADI-2 Pro FS are just too good to show differences. When I can't even demonstrate jitter anomalies and at best very slight distortions with S/PDIF input as discussed last time, what potential do I have to show anything for a test like today's unless changes are quite obvious? What I know is that the TEAC UD-501's XLR output is still excellent with very low noise floor superior to RCA outputs and I know it is susceptible to jitter and has distortion anomalies when I compare USB input compared to S/PDIF. This sensitivity to anomalies makes it the most likely platform to show differences if there are any to be found.
Test Set 1: Comparing USB to Coaxial S/PDIF - switching power supply vs. battery, varying ethernet/WiFi, Screen on/off
 |
| USB = USB connection, Coax = S/PDIF coaxial digital, SPS = Switching Power Supply, Bat = using the RavPower lithium battery, WiFi = Pi communicating with server over WiFi, Eth = wired ethernet connection to server, Scr = 7" Touchscreen ON, NoScr = 7" Touchscreen disconnected (OFF). |
The last 3 columns are using the JustBoom Digi's coaxial S/PDIF. The middle column test is still using the switching power supply while the last 2 tests on the right are with the RavPower battery whether with WiFi or ethernet and LCD screen on or off (disconnected).
Hmmmm... Anyone see any improvement using the lithium batter? Not to my eyes. These are essentially small differences due to inter-test variability one sees when looking at noise levels down at almost -120dB and distortions below 0.001%! I guess one could say the the tests with the battery may have improved the noise level by about 0.4dB over the switching power supply. I was able to reliably show about 0.3-0.5dB lower noise level over 4 runs with the switching power supply vs. battery. This is of course simply academic and ridiculous to suggest audible differences (certainly I did not hear a difference playing some music).
The only clear objective difference really is with the IMD+N sweep values and we can see this in the composite graphs; again, I would not suggest that this is audible:
We saw this IMD+N sweep deviance previously and I assume this is related to higher jitter with the TEAC UD-501 receiving S/PDIF because we didn't have this issue when testing the Oppo UDP-205 and RME ADI-2 Pro FS where jitter was much better controlled across the different digital inputs.
Okay then, how about with the TosLink test set?
Test Set 2: Comparing USB to TosLink - switching power supply vs. battery, varying ethernet/WiFi, Screen on/off
 |
| USB = USB connection, TosLink = S/PDIF TosLink digital, SPS = Switching Power Supply, Bat = using the RavPower lithium battery, WiFi = Pi communicating with server over WiFi, Eth = wired ethernet connection to server, Scr = 7" Touchscreen ON, NoScr = 7" Touchscreen disconnected (OFF). |
Yet again, there's really not much to see. My suspicion is that the TEAC DAC's TosLink interface is simply more noisy, hence higher noise level overall compared to USB and coaxial inputs. Whether due to the noise floor being higher or the galvanic isolation with optical, notice there is no difference between the noise level and dynamic range measurements of the switching power supply compared to the battery.
As demonstrated last week, TosLink is prone to higher jitter. The IMD+N sweep has consistently higher distortion than either USB or coaxial S/PDIF.
Again, no evidence that having the LCD screen on or totally disconnected made a difference. As for WiFi vs. ethernet, if this changed anything, the difference seems insignificant.
Test Set 3: Comparing the high capacity RavPower vs. the "cheap" freebie low-capacity battery
Alright, then... Would an inexpensive, low power 2200mAh lithium battery make a difference? Are there noise differences between power packs presumably introduced by the battery or power management system?
 | |
| CheapBat = inexpensive "freebie" 2200mAh battery. |
Nothing remarkable with the numbers and just as unremarkable composite graphs:
No evidence of any extra noise nor distortion whether I used the RavPower or "free" lithium battery. Notice that the IMD+N sweep results segregated based on which interface was used (coaxial vs. TosLink).
Test Set 4: Wideband noise floor
Fine. So far we don't see any extra noise with 24/96 tests... How about checking to see whether powering the Raspberry Pi with a switching wallwart vs. lithium ion batteries, turning the touchscreen on/off, using WiFi or ethernet will affect the DAC's output all the way to 192kHz. Easily done by using the RME ADI-2 Pro FS ADC sampling at 384kHz (using WaveSpectra, 128k FFT points for the graphs):
Nothing of significance. At most, we can make out the slightly noisier TosLink (eg. the orange tracing) hovering higher than some of the others. Alright, maybe we can look at a log scale and see if there are anomalies down in the lower frequencies?
Yet again. Nothing of significance. Whether using the inexpensive Aukru switching power supply or using the lithium batteries with the screen turned on or if WiFi was used, as you can see, none of the conditions using the Raspberry Pi 3 B+ streaming device affected the analogue output noise level from the downstream DAC; even in this case with a 5-year-old TEAC UD-501.
Test Set 5: Jitter differences!?
There's always jitter, right? ;-).
Here's what I got previously for TosLink input using the switching power supply:
And here is TosLink to TEAC UD-501 using the RavPower lithium battery:
There you go. "More ideal" looking J-Test results; especially the 24-bit J-Test.
Summary...
I think the results speak for themselves. I do not believe nor have heard a situation where the streaming device makes much if any appreciable difference assuming one is using a reasonably good DAC and the data transmission is bit-perfect. What we've seen in these measurements using the Raspberry Pi 3 with just a simple switching power supply is that:
1. There is no evidence that using the switch mode power supply adds to "noise" when connected to a 5-year-old DAC even down to very low noise levels using XLR analogue output.
2. Lithium battery packs do not seem to improve anything. No evidence between a new high-capacity battery pack (RavPower) compared to the "cheap" freebie battery stick.
3. No evidence that a 7" LCD screen adds any noise to the output even from the coaxial S/PDIF connection. This was reported in my JustBoom Digi measurements last time.
4. No evidence using the JustBoom Digi S/PDIF HAT board that the coax output was significantly susceptible to electrical noise. At most, we're only seeing 0.5dB difference down at the noise floor between switching power supply and battery power.
5. No evidence using the Raspberry Pi 3 B+ as a streamer that ethernet or WiFi made any significant difference to the DAC output (even with the streamer sitting literally right on top of the DAC during testing transmitting/receiving wirelessly).
6. Even with a wide bandwidth analysis of the noise floor up to 192kHz, there was no evidence that using a switching power supply, lithium batteries, WiFi connection, ethernet connection, or LCD screen turned on in the streamer device contributed to any extra noise down to -140dB in the audio band and below -110dB all the way to 192kHz.
While I believe XLR is the best analogue connection for those desiring highest fidelity, we can of course wonder if issues may only show up with RCA... Perhaps we'll look at this in another post.
Well, hopefully that relieved anxieties for some. The reader might now say... "Okay, so the Raspberry Pi 3 as a streamer is rather immune from power supply issues and such, but maybe if someone were using a computer, I bet the switching power supply and noisy computer would still result in a ton of noise to your DAC!"
Maybe we can consider that a bit more next time even though in the past we've had some discussions in this regard.
Finally, one could as - "Why don't you try a linear power supply?" Well, indeed, we could try that (maybe something like the inexpensive Nobsound 25W adjustable linear supply), or maybe a specially advertised "low noise" power supply like the iFi iPower, SBooster, or the myriad of choices on eBay. But here's the question - why bother? We know they're more expensive, run warmer, and are less power efficient. Unless there is evidence to show that indeed noise levels are kept lower or they result in lower distortions when connected to the DAC in certain conditions, why should consumers even consider doing this other than for experimentation purposes? I think the onus is on the manufacturers to give consumers a reason if they insist that upgrades or components are anything more than a waste of time and money for audiophiles. As usual, perhaps the audiophile "press" should do a little investigative journalism before declaring that any difference can be heard (especially this guy here, LOL)?
------------------------------
I trust everyone survived the Thanksgiving and "Black Friday" shopping mayhem a few weeks back. I find it amazing how consumerism has continued to expand during the end-of-year holiday season over time! Anyhow, I don't see many "must have" tech gadgets this year but I did grab a deal on some Corsair Vengeance LPX 16GB DDR4 RAM for ~US$100 to boost my workstation memory to 32GB RAM. Not that I absolutely need this much memory, but sometimes I see that I hit >80% memory usage when doing stuff like video editing and encoding with 4K content.
Also, anyone else notice that web browsing these days can take up a ton of RAM?!
Wow. 7 tabs on Firefox ate up >3GB RAM and one single tab open to edit this blog post used up 1GB in Chrome. Seems rather inefficient to me!
I don't remember when I first updated my main workstation computer to 16GB of RAM. Probably back in early 2012 when I put together my Intel Ivy Bridge i7-3770K machine. Since I suspect the speed of upgrades appear to be slowing due to generally a lack of need for most consumers, I would not be surprised if 32GB will last me awhile... At least this post will serve as a reminder for me when I made this upgrade.
For those unaware, most of the time memory is sold as kits; usually something like 8GBx2 or 16GBx2. This allows manufacturers to component match and make sure the RAM can run at the "overclocked" speeds. For DDR4 RAM, the base clock rate is typically 2133MT/s (1066.5MHz). The Corsair 8GBx2 pair is rated at DDR3000 which is 1500MHz using the overclocked XMP profile.
Alas, even though I have 4 sticks from the same brand, on my AMD Ryzen 1700 workstation (overclocked of course), I could not get the MSI motherboard to POST beyond 2400MT/s speed. After a few trials, I didn't feel it was worth pursuing this much further even with RAM voltage pushed to 1.4V. Maybe I can try again later if I'm bored.
However, I do like efficiency, low voltage, and squeezing a little more performance out of the RAM. So what I did was "tighten" the timing on the RAM while keeping it at 1.2V. Best JEDEC DDR4 2133 timing for these RAM modules are identified as: 15-15-15-36-50. I pushed it down to 14-14-14-28-45 while running at DDR4 2400. I've kept Command Rate at 2T but 1T was stable with voltage up to 1.25V but this provided little benefit. Runs stable with Intel Burn Test ("Maximum" stress level for all memory) and Prime95 (24 hours). Even if I don't have the benefit of higher clockspeed on the memory, at least latency is better by a few nanoseconds for short random access bursts of data...
Cruising into December and year end!
Speaking of "cruise", I enjoyed the latest Mission Impossible: Fallout the other night with the family; probably my favourite of the MI movies. Very good action flick with a fantastic demo-worthy Atmos soundtrack. However, I was a bit surprised at the quality of the video. Despite a supposed 4K digital intermediate and some fantastic high-resolution 16:9 aspect ratio digitally filmed portions (Red Weapon Dragon camera used I believe), most of it was clearly from 35mm film presented at 2.39:1. It looks muddy, blurry at times and grainy. Remember the 4K comparisons between 35mm film (eg. The Prestige, Blade Runner) and true 4K-worthy digital or 70mm IMAX film (Blade Runner 2049, Dunkirk) earlier this year. I guess some might like this kind of look or there's some artistic merit to this. It's a little like thinking about "hi-res" audio in 24-bit and 96+kHz "containers" but not recorded or produced with high-resolution equipment or technique. Other than the fine HDR contrasts and highlights, most of this movie would have looked just as well on 1080P Blu-Ray or streaming online. Anyhow, a bit disappointed that a big budget movie like this didn't look substantially better for much of it.
Lastly, on the music release front, The Beatles' White Album 50th Anniversary 2018 Remix is now out along with the Super Deluxe package that includes a Blu-Ray with the 5.1 surround mix. The surround mix is excellent! Whether it's hearing an old favourite like "Ob-La-Di, Ob-La-Da" as if it's the first time, or tripping out on the technically manipulated "sound collage" of "Revolution 9", the 5.1 version is pure fun... Don't miss this if you have a surround system!
One last thing for those who have a music archive and want to save space, IMO feel free to convert the 24/96 5.1 mix to 16/48 5.1. Most tracks look like they have just various amounts of noise beyond 24kHz. Here's "Ob-La-Di, Ob-La-Da" during playback:
As an analogue-sourced album from 1968, one cannot expect "true" high resolution audio quality.
Enjoy the season, movies and music everyone...
Hi Archimago
ReplyDeleteexcellent article as usual.
There are two trains of thought that I would be interested to hear your views / experiences on:
- Many things audiophiles care greatly about (PSUs, cables etc.) don't seem to matter. I agree with this and your measurements seem to support this. My question: What DOES matter and make a difference?
- Along this line of thought: RP3 has a shared controller for Ethernet and USB whereas Odroid C2 / Asus Tinker Board / Intel NUC have separate controllers. Does this result in differences regarding measurable audio quality for sample rates up to 24/384kHz?
Hi jacobacci,
DeleteYup, I think the results are showing us that many of the things audiophiles seem to value and get criticized for are indeed mainly "snake oil" when advertised and hyped up unreasonably when put to the test.
For me, the "usual" time-honoured components do matter... A decent sound room, low ambient noise level, good quiet pre-amps, clean amp, good speakers. On the digital playback side, a good DAC will cover over a multitude of "sins" when it comes to suboptimal source devices (which we'll cover this coming week). And of course with analogue playback, clean records, good turntable and the usual TLC are a must!
Beyond that, just make sure the cables do the job (ie. reliable, error-free digital transmission), adequate gauge speaker cables, and shielded cables for long runs...
As for ethernet on devices like the ODROID/Tinker/NUC, I honestly have not noticed much difference. Remember this quick test with the Raspberry Pi awhile back showing the difference between ethernet vs. WiFi:
http://archimago.blogspot.com/2016/11/measurements-on-value-for-ethernet.html
The difference was barely notable. And my old ODROID C2 testing back in the day using the separate ethernet controller:
http://archimago.blogspot.com/2016/05/measurements-odroid-c2-with-volumio-2.html
Did not argue for any major issues...
Having said this, I do hope that future Raspberry Pi devices will have a separate ethernet controller so as not to saturate the USB 2.0 infrastructure!
Another winner.
ReplyDeleteThis is now the only audio site/blog that I read seriously now.
You are a dangerous man, Arch, puncturing the unsubstantiated claims that unprincipled manufacturers and their golden-eared lackeys put forward to maintain their market share in a difficult world. You're not about creating irrational desire for the latest under-performing and overpriced piece of garbage, so bravo to you. This article covers equipment that I don't (and probably won't) own but still makes very interesting reading and comes to really valuable conclusions.
This hobby should be fun and interesting, an opportunity to enjoy beauty and learn about art and technology, but it has devolved into mysticism in the service of very questionable business practices.
Your rationality and scientific approach (even into the subjective aspects of these things) are really needed: please keep it up.
Thanks for the note Phil.
DeleteUltimately it's just good to know the "truth", or at least try to look for it in whatever way and with whatever capacity one can.
Hopefully in some way, I can encourage other hobbyists to try testing claims out themselves. Once we hit a certain threshold of fidelity, it's hard for the mind to differentiate the various influences whether it be word-of-mouth, the appearance of a device, or countless other variables.
I find it funny when people say it's about "fun" as if testing things out and seeking the truth isn't fun :-). As a hobbyist, I certainly do not consider it "fun" to read all kinds of articles and reviews IMO essentially written to appease the Industry over the years. At some point, I think many audiophiles have had enough and it's time to look at the facts and figure it out ourselves...
Hi Archimago,
ReplyDeleteGlad to see you always make good use of your equipment to try something out. Not long ago I saw some people criticized your test methodology by using prosumer interfaces and "amateurish" software like RMAA. I did not reply them but I was really unhappy to see those comments.
While something like AP is necessary for verifying product specs, prosumer interfaces with an array of software tools are much more flexible and practical for identifying potential issues in a specific signal chain for individual users.
Some people "trust their ears" with sighted tests, "audio without numbers" (except MSRP). On the other hand some people make buying decision based on 0.00x% vs 0.000x% distortion without noticing some glaring flaws like bad firmware, compatibility and such. Some people believe NOS = bit perfect, on the other hand some people believe 64-bit float PCM volume control or upsampling to DSD512 are essential for software players to minimize "quantization error" and avoid "out of band noise" bleeding into audible range. Yeah, some of these things are measurable, but how meaningful they are?
Good point Bennett.
DeleteAs I noted to Tony above, I hope the posts and measurement results get hobbyists thinking and trying to verify claims themselves. At this time in history, we are very fortunate to have access to the hardware and software tools to understand what's going on.
To have hobbyists questioning claims is good I believe because it pushes the manufacturers and the audiophile press to seriously evaluate how they advertise and what kind of message they're trying to "sell". I believe that educating consumers is the only way for the hobby to move forward and have legitimate advances be rewarded and snake oil discarded.
I agree, we should NOT be making decisions simply based on 0.000x% distortion. That's not needed. But it's just as nonsensical for a brand to advertise a rather poorly measuring NOS DAC as if that should be revered as some kind of engineering achievement or a magazine to trot out the $10,000 "cable of the month" as if there's anything noteworthy.
Ultimately it will be about one's philosophy and values. After all, we are talking about consumer goods... Toys that can bring joy :-).
But did you put a doorstop on the Raspberry Pi at any point? You know, for jitter. Just joking, great article!
ReplyDeleteLOL.
DeleteGood one Daniel. I saw the article you're referring the other day :-).
I'm sure jitter and distortion would have been much lower if I purchased the door stop. Plus reduced the horrors of microphonics during playback. Alas, the touchscreen makes it a little difficult to keep the doorstop balanced :-).
Would love to see measurements with and without doorstop though!
Hi Archimago,
ReplyDeleteA very good read! Maybe some audiophiles feel lost with the memory timings part but I also think that it is essential that we bring out the max from the equipment we have (within stability and temperature limits of course).
I had many systems where neither the XMP profiles, nor manual configs didn't get me the rated clocks on the memory modules, despite having good "overclockable" motherboards.
While timings should be pushed down on the modules one already owns, rarely do the more expensive modules with tighter timings justify the extra price (if that difference can be spent on faster cpu or gpu).
Some real world examples here.
Hey there Turrican,
DeleteBTW was your handle inspired by the game back in the 90's? Good memories of my old Amiga computer :-).
Yup, agree that some might not get the memory timing bit in this post. As a blog, although I mostly talk about audio stuff, I do like touching on other bits like 10Gb ethernet, overclocking CPUs, computer builds. To me this is important stuff to understand and makes sense of why digital audio works and helps dispel "snake oil".
The audiophile hobby (especially computer audio and streaming), isolated within its own niche, dissociated from the other techy geek hobbies I think is a bad thing!
Thanks for the link. Indeed, I don't think the tighter timings make much difference. Better on the whole to run higher clockspeeds than improve latency marginally these days.
Indeed, it was inspired; however, in my case it was a Commodore 64. :)
DeleteFor a while I was running with ethernet over power, and to be honest I couldn't hear a difference. Now, I wasn't looking for a difference, and therein lies the rub; if someone wants to hear a difference, and if they strain their sinews doing so, then they will convince themselves that a difference exists. What could be noisier than ethernet over power? Maybe it does sound 'noisy', I don't know, but I never had a problem with it.
ReplyDeleteGood point!
DeleteI've never run PoE but certainly imagine the metres upon metres of cabling with power running through them in such a setup throughout a house.
It's a good think you're "Unknown"... Otherwise you'd have your Golden Ears Audiophile Membership Card immediately revoked :-).
Again it is important to stress that ethernet was designed in a way that it contains error handling (within used protocol). In other words, the exact packets sent are the exact ones to arrive. If one gets corrupted during transmission, it will be resent until it comes correctly on the other side.
DeleteDoes the noise influence it? Of course! Especially when used in wifi or ethernet-over-power scenario.
Is this important? Yes if you crave maximum throughput. In example, your gigabit link may be capped to max 200 mbit.
Does it really matter? only if throughput speed is critical. for audio listening applications the concern is non-existant as the data and packets can be transmitted 1000 and more times faster than they can be played. so if a packet comes damaged, it will be resent way way before comes the time it needs to be used.
In pro audio production you want to use wired ethernet not only because of throughput, but because of low latency (i.e. DAW controls synchronised equipment in the rack).
for listening it doesn't matter, just as long the connection is stable.
Audiophiles will tell you lineair power supplies performan better because they generate less HF noise on the net, which can be picked up by your preamp for example. So still some myths to debunk ☺ Keep up the good work!
ReplyDeleteThat is not myth, that is a fact. The Myth part is that this HF noise is not directly listenable (as hum or noise), but causes all kinds of degradation of audiophool sound signatures, like stage width, glitter, space, detail, bla bla bla...
DeleteYes, there are some noise that can indeed be picked up by other devices like the pre-amp.
DeleteOne example I discussed awhile back was the 8kHz USB PHY noise I picked up with my preamp in analogue "home theater" passthrough mode:
http://archimago.blogspot.com/2015/05/measurements-usb-hubs-and-8khz-phy.html
whenever I connected the computer to my USB DAC. This could be modulated with different USB hubs (this was around the time the UpTone USB Regen first came out) and the Corning optical USB extender back then.
While the question of how much noise a switching power supply could add noise to one's system compared to linear PS may be quite different between systems (I personally have not run into any issues), I absolutely agree that the myth of "degradation of audiophool sound signatures" is real :-).
Of course it's no myth about, and it's measurable. It has nothing to do with audiophiles or golden ears or anything. It's noise generated, in this case by the internal Raspberry DC-DC regulators, which are almost impossible to bypass. Even using a linear regulator will change nothing on that noise.
DeleteKK Batteries is one of the best Inverter Service centre in Chennai.We provide batteries with guarantee thus making us the best Car Battery Dealers in Chennai. We are Inverter Battery dealers in Chennai with various accessible models. Inverter Battery dealers in Chennai
ReplyDeleteI am just starting to read your review, but I do have to comment on something you claim at the beginning. You said "So, would using lithium-ion batteries like these make any difference compared to an inexpensive wall-wart switching power supply". But you are just comparing two switching supplies, one inside the lithium ion batteries and the one inside the wall-wart supply". That's a fact, not an opinion. Now if you tell me that you are using a 5v battery, not boxed or anything, connected to an usb cable or socket, then the comparison should be between different thing. Besides that, and you will probably mention this further one, the problem is the DC-DC regulators INSIDE the Raspberry, which have no easy way to eliminate and feed linear power all over. THAT is what adds noise to the supply lines.
ReplyDeleteEven though it’s an afterthought for most builders, the power supply is actually one of the more important parts of a build. Picking a quality power supply can mean the difference between a well running system and one that suffers from crashes and boot failures. Worse yet, cheap generic models can literally explode into flames, taking the rest of your computer with it. ny526
ReplyDeleteIt doesn't seem surprising that the external supply you used changed almost nothing, as the Raspberry weak point, IMHO, is its on-board DC-DC regulation. That is which should be eliminated and I don't think is possible. Is it?
ReplyDeleteIs there any other of this "small PCs" that allows replacing the internal regulation for a linear one?
Thank you for clearing that up. Unfortunately it leaves me even more confused. Albeit that my case concerns recording with a pi rather than streaming/listening. I found that i would get significant noise depending on upload rate, even when using shielded mics and cables (without shielding i would also pick up wifi, screens ticking and mains humming. In my search for a quick and cheap fix i rambled on to your blog.
ReplyDeleteAny ideas pointers to improve recording quality are much appreciated, but please dont tell me to buy a usb soundcard and mic for 200 $ a piece.
This comment has been removed by a blog administrator.
ReplyDelete