huge job...!!!! and it's not over!!At a moment I can tell that my sample unit has an increased space around the coax socket. And a glossy finish of the chassis.
Stay tuned
Part 0 - a test gear possibly involved
RME ADI-2 DAC FS, ESS edition
miniDSP Flex, RCA edition
E1DA Cosmos ADC, A grade
PicoScope 4444
Intona 7055-D isolator or ISOUSB211 based one - when needed
and some other less important stuff - powerbanks, Allo Nirvana SMPS, Icron Ranger 2201 USB extender, SPDIF Audio Switch by Beni, etc.
Part 1 - analog output - Pro vs Pro Plus combat
The fight begins. Both devices powered by Xiaomi battery packs, both use the strongest 2 Vrms outputs.
Pro's FW is 4.8.523533.
For the Pro Plus is was 4.8.521533 initially, and later 4.8.523842 with a fix for the reported bug.
Round 1 - power consumption
The Pro Plus is more power hungry. It needs almost 2.6 W (0.5 A / 5.1 V) when idle with a wired network. It rises to above 4.5 W (0.9 A / 5.1 V) when sine 1 kHz at 96 kHz is played. On the other hand the Pro needs 1.3 W (0.25 A / 5.1 V) and almost 1.7 W (0.32 A / 5.1 V) respectively.
Round 2 - output signal voltage
The Pro Plus real output voltage level is 1.973 Vrms for both channels, for sine 1 kHz at 96 kHz. For the Pro it's slightly unbalanced 2.012 for the left channel and 2.006 for the right channel. Values measured with a multimeter.
Round 3 - REW measurements
My Pro unit measures slightly better on its left channel while for the Pro Plus it's the right one, so better channels are compared unless stated otherwise.
All the measured values are presented as Pro vs Pro Plus.
Sine 1k 48 kHz:
View attachment 1454
View attachment 1455View attachment 1456
Significant 22.5 dB difference in SINAD.
SMPTE 48 kHz:
View attachment 1457
View attachment 1458View attachment 1459
Clear advantage of the Pro Plus but low freq harmonics higher than expected.
CCIF 48 kHz:
View attachment 1460
View attachment 1461View attachment 1462
Another test for intermodulation distortions, this time with a wider bandwidth included in the calculations. Looks like Pro's Achilles heel, it's performance is horrible here. Pro Plus can shine.
Multitone 48 kHz:
View attachment 1466
View attachment 1467View attachment 1468
The test which focuses more on the real life performance. IMD tests issues mentioned before are clearly visible.
DAC filters:
The Pro seems to use a kind of fast linear phase filter with the response as below:
View attachment 1470
The Pro Plus uses rather a fast minimum phase filter:
View attachment 1471
Attenuation of the Pro Plus filter is much more effective (48 kHz):
View attachment 1472
Internal headroom management:
A test to see how the unit can handle intersample peaks. Lets imagine we have a test signal with sample peaks at the max allowed 0 dBFS level, but these samples are not sine wave peaks. On the picture below we can see possible clipping areas if the device cannot provide a sufficient amount of headroom to handle it.
View attachment 1482
To test it I've prepared 4 test signals with sample peak levels from 0 dBFS to -5 dBFS, and intersample peaks from +3.02 dBFS to -1.98 dBFS.
The Pro results:
View attachment 1485
1 and 2 are clipping, 3 and 4 remain the same. Looks like there is no internal headroom or it's just minimal.
The Pro Plus results:
View attachment 1486
Only 1 is clipping. The headroom wasn't big enough to handle +3.02 dBFS peaks, but it's above 1 dB.
Edit:
One more test added for intersample +2.02 dBFS peak:
View attachment 1488
It clips for such level, so the headroom is somewhere in the range 1-2 dB.
Dynamic Range:
It's calculated using a sine 1 kHz signal at -60 dBFS level.
View attachment 1491View attachment 1492
Pro: 44.1 + 60 = 104.1 dB
Pro Plus: 56.3 + 60 = 116.3 dB
Being idle / playing silence:
The noise level when device is doing nothing or when it plays a silence stream - all the samples are zeroed. Also the cases where the PSU used usually impacts measurement plots, especially when the device is idle.
For the Pro idle and silence look different:
View attachment 1495
No idle mode artifacts coming from the PSU, but it's a battery pack used here.
There is no such noise level difference for the Pro Plus, although an alone component appears:
View attachment 1497
The Pro Plus seems to be perfectly muted when playing a silence.
THD vs frequency:
A sweep signal can deliver some information regarding the relation between harmonic distortions level and a frequency of the fundamental.
View attachment 1499
A kind of trouble with low frequency distortions is visible here for the Pro Plus. Bad Pro's performance with high frequencies also shown.
Part 2 - a closer look onto Pro Plus analog output
Output voltage levels:
There are 4 output voltage levels available, 2 Vrms, 1 Vrms, 0.8 Vrms and 0.5 Vrms. How does it affect the performance?
View attachment 1503
While the voltage level decreases a SINAD decreases as well (sine 1k 48 kHz).
2 Vrms - SINAD 111.7
1 Vrms - SINAD 109.4
0.8 Vrms - SINAD 107.6
0.5 Vrms - SINAD 103.8
Voltage levels measured by PicoScope using sine 1k 48kHz signal:
1.977
0.990
0.784
0.495
There is a lot of headroom on 1 Vrms output. No clipping at all for intersample peak tests regardless of the signal level:
View attachment 1504
What if I change the power source?
An ordinary Huawei phone charger is the first candidate. Labeled as 2A maximum, is strong enough to power the Pro Plus.
Did I notice performance decrease in my 38 synthetic tests? Not at all.
The Allo Nirvana ultra low noise SMPS is next. No serious improvement again, but at least I've caught a difference:
View attachment 1513
It's a removal of insignificant ultrasonic components in the idle test which is the most sensitive for any PSU changes. That's all.
I can guess that the Pro Plus PSR ratio is high enough to make it quite immune to common SMPS ripples and noise.
I will give the Nirvana SMPS a second chance during listening tests.
Low frequency THD:
A closer look reveals a behavior similar to the infamous ESS hump, more common for implementations of ESS chipsets, where distortions level rises in a specific way:
View attachment 1517
"Better optimized" 1k signal doesn't grow so rapidly:
View attachment 1518
Unexpected behavior of low frequency distortions is also visible on IMD graph below, which uses SMPTE test signal:
View attachment 1519
It should be noticed that although these higher than expected harmonic distortions levels are significant in terms of measurements, their audibility is rather questionable and probably should not bother.
Part 3 - the digital domain
Streaming to digital outputs with bit-perfect capable interface:
The Pro Plus has been connected to RME ADI-2 DAC FS, which will act as the spdif-usb interface to let me analyze the Pro Plus output. All the DSP features of the Pro Plus are disabled.
Nothing unusual here, the Pro Plus is fully digital transparent on both coax and toslink outputs, just like any other WiiM. It also passes all RME bit-perfect tests up to 24 bit depth.
Toslink input to digital outputs:
The Pro Plus still connected to RME ADI-2 DAC FS, but an additional device appears here - SMSL PO100 Pro, which acts as the usb-spdif interface to let me feed the Pro Plus.
The Pro Plus is again fully digital transparent on both coax and toslink outputs.
There is also a surprise here. The Pro was suspected to perform a signal reclocking of the input stream to its internal clock, which could result in some unwanted behaviors like clicks and pops. But what do I see in REW when using its generator:
View attachment 1534
An audio chain is:
PC (REW generator) -> usb-spdif -> Pro Plus input -> Pro Plus output -> spdif-usb -> PC (REW capture)
For such scenario the source signal is clocked by usb-spdif interface (PO100 and XMOS clock). There is no discrepancy with the clock captured on the PC, I believe it can be a sign that the reclocking on the Pro Plus wasn't made. The same behavior can be seen with both toslink and coax outputs.
Another possible argument for this assumption is a drift calculation between the input and the output in Multitone Analyzer:
View attachment 1531
I will evaluate this theory later with a different capture interface.
What about the latency between digital input and output?
For 192 kHz signal, the latency measured in Audacity, toslink in -> toslink out, is 44 ms.
View attachment 1533
For toslink in -> coax out, it is 49 ms.
I'll be back
;-)
