Pi2Design Mercury Streamer / Server Measurements

Content and analysis made possible thanks to support from https://headphones.com, and https://patreon.com/goldensound supporters

This unit was loaned to me by Pi2Design for testing

ChitChat:

The Pi2Design Pi2AES was a product that earned a well deserved reputation amongst the audiophile community for incredible performance per dollar.
It was a HAT (with accompanying case) that could be installed on any Raspberry Pi to turn it into a high performance I2S/AES/SPDIF capable audio network streamer.

And whilst this product was pretty fantastic, it had two main drawbacks:

– It required a little bit of DIY and setup, which some users were averse to.
– It was mostly intended as a streamer, and if wanting to use it as a server via the Volumio software, could only utilise relatively slow USB or SD card storage.

To address both of these concerns, Michael from Pi2Design has created the Mercury, a standalone, ready to go device that requires no setup, looks great on a desk or in a rack thanks to having all I/O on the rear and a sleeker case, and most importantly; an M.2 SSD slot! Allowing it to be used as a fast, responsive, and compact music server, all for $599.

The device is still pi based, utilising a CM4 module as the brain of the operation, but the main product board provides most of the functionality for the device.
This includes reportedly better jitter performance than the original pi2aes, the aforementioned M.2 storage support, and also now utilising a 12v power supply instead of the 24v supply used on the original pi2aes. This is a welcome change as it allows people to use their own 12v linear or upgraded power supplies if they wish. 24v linear supplies are much harder to come by.
In fact, the Mercury is using a linear pre-regulator (LT2941 LDO) which allows for the entire power chain to be linear when an external linear PSU is used.

The rear of the device has all I/O neatly arranged, and includes any digital audio connection you might need including both BNC and RCA coaxial SPDIF.
There are two HDMI ports, one which is for actual HDMI video to connect a display, and one for I2S audio output.
Make sure not to get these mixed up!

Unfortunately the I2S port does not have configurable pinout, and is fixed to the PS Audio pinout (detailed in the manual).
However a lot of DACs themselves are now adding configurable pinout so it is less necessary to have it included on the source itself.

There are also two USB 2.0 ports should you wish to connect external storage or a USB DAC.
Multiple DACs can be connected to the Mercury simultaneously.

Lastly, when ordering the Mercury, you have the option of choosing whether you would like to have Volumio or Ropieee preinstalled. Though as it is pi based, you are also able to install your own OS should you wish to do so.

I opted for Ropieee as it’s what I personally use on other pi based streamers due to ease of use.

Measurements:

Test Setup:

– Audio Precision APx555 B-Series Analyzer
– AudioQuest Carbon Coaxial Cable and AES cable(1.5m)
– RopieeeXL via Roon used for OS/Streaming
– Device powered by an ifi iPower 12v power supply
– 44.1khz and 48khz real music played through device during measurement
– Jitter analysis is configured to use an AES3 standard 700hz high-pass filter, and a 100khz butterworth low-pass filter
– Coax output used unless otherwise specified

Jitter:

To measure jitter, the device is connected to the digital input of the APx555, and the analyzer is set to analyze jitter, not audio content.
These measurements do not show audio/analog info, but instead show the spectrum of jitter, ie: time-domain inconsistencies.
This is the primary factor that a good streamer or DDC will seek to improve.

Jitter for 48khz base rate playback does indeed seem to be slightly improved from the original pi2aes. Going from 209ps RMS / 617ps Peak to the results shown above.
And what’s more, it has almost no random spikes whatsoever below 40khz. Something that hardly any DDCs I’ve tested manage to achieve.
Even other very well performing DDCs like the Audio-GD DI20HE have some spurious content below 20khz.

Jitter performance is very similar for 44.1khz base rate playback, but has one spike at about 3600hz.
This does mean that the original pi2aes actually beats the Mercury for 44.1khz jitter performance, but this was due to the original pi2aes having better 44.1khz jitter performance than it did at 48khz. Whereas the Mercury is quite consistent for both.

How about noise?

Now at first, noise seems to be quite good, though with some 50/150/250hz etc content visible.
But there are two important things to note. Firstly, even when tested as-is, the <20khz content is still much lower than the original pi2aes:

But also, the outputs on the Mercury are galvanically isolated, which means they are ungrounded.

As a result of this, some of the content shown here is not from the device itself, but rather being picked up by the interconnecting cable.
If we manually ground the Mercury to the analyzer using one of the gnd pins as shown below, then we get a clearer picture of the true noise from the source.

Ultra clean!
This is an incredibly low noise source.
Whether it’s worth manually grounding the Mercury in your setup will be situational. The galvanic isolation has the benefit of blocking any DC offset but the interconnect may pick up some noise on its own.

From Michael Kelly:
“Professional studios use it by default because of the possibility of significant ground differential between various pieces of equipment across studio rooms. The benefit in consumer setting is not as high, but it does prevent ground noise from coupling over to other systems.

I have to admit that your testing of the actual noise on the various ports did not seem very useful to me. The reason is as you’ve identified it doesn’t actually correlate with the signal quality. And as long as the actual signal is not distorted or having jitter imposed on it, it won’t affect the downstream audio.“

And on that latter point, he’s absolutely right. Noise levels this low are incredibly unlikely to have any effect on the performance of the receiving device. And if it did, it would actually indicate that the DAC you are using has an awfully designed digital input section.
But it’s worth checking to ensure that a device does not have an unusually high noise output, and also because regardless of whether it provides a genuine sound quality benefit, sometimes it’s just nice to know that something is clean as can be.

In future, I will be adding some measurements of the I2S output of this and other DDCs directly. Unfortunately the APx555 cannot accept LVDS I2S signals natively, so cannot test them yet. But the jitter on other outputs is still a very good indicator as they’re still using the same clocks as the I2S output.
Originally I had planned to use the Holo May NOS DAC as a medium for this test, as playing a 12khz square wave through a NOS R2R DAC effectively acts as a 256x clock divider. However the May also has some compensation for ladder inaccuracies as well as an analog low-pass filter and so it may not be the most fair device to use for this test.
I am working on a flip-flop based clock divider and once finished will add direct I2S measurements to this and other DDC posts.

Conclusion:

The Pi2Design Mercury is a fantastic device.
For only $599 you get a high performance digital source with any kind of output you could ask for, in a compact form factor, and that can also be used as a standalone server both wired or wireless.
There really is little to complain about here at all other than that it does cost twice as much as the original pi2aes, but given the increase in prices of almost all electronics globally in recent times, this is to be expected.

If you’re wanting the absolute lowest possible jitter, there are some dedicated (but more expensive) DDCs like the DI20HE that will outperform the Mercury, but it’s hard to think of any other streamer that competes with the Mercury on a price to performance level, and I highly recommend it.

Congrats on another fantastic device Michael!