Due to consecutive time constraints I had to wait a bit before be able to do some deep listening of the NanoPat Streamer (just to name it 
Before telling how the small animal is sounding, I want to thanks PSAudio for its openness for sharing very rich information in its forum (especially Ted and Paul) and all contributors on this thread who presented their projects or questions. My turn to share here all the decisions which drove to chose the components and assemble this setup. I’m not a god in electronic and all choices have been made in a context of time/price and self judgement (surely not a super reference, thought). So I guess there’re in my choices surely things not aligned with some experts considerations or targeted expectation from anyone.
Hope, this could help anyone to build its own solution but sorry in advance for the long post
.
The initial target: Build a streamer as an endpoint feed by either a Roon or UPNP Server or LMS elsewhere on the network, paying as most as I can most attention to the power supply, with the aim to use SuperCaps on the last stage. The streamer must also be very low in power consumption to minimize EMI and its output must be isolated electrically to avoid ground loop with the DAC (a point I believe only partially reached). Take the opportunity to have a PSU for my network switch closed to the Hifi system. The whole built by assembly of a set of cards, no PCB to design and produce.
The project took 1 year and lot of equipments not necessarly in the toolbox of everyone (maybe a FabLabs at proximity would be a great option).
To remind, the result at this time (NanoPat):
An external linear power bloc (with multiple outputs with filtering and classic regulation lines) powering a separate SBC with DietPI, an USB/I2S interface and I2S LVDS/HDMI module, the last 3 being themselves powered by ultra low noise regulators (2 Linear LT3045 modules). The SBC, LT3045, interfaces are in a second metal box (see pictures).
The SBC:
The RPI2 and 3 are of course good candidates or even better the Allo Sparky board better designed for Audio but
the NanoPI is a bargain, it has sufficient CPU/RAM power to assume its role but that’s not all:
- 2 USB ports on its PIN header, so you can solder your own and very short (less than 10cm) USB cable to the I2S interface
- Low power consumption, only few hundred mA if you reduce the CPU clock (did it at 624MHz on quad cores, works perfect)
- No feature totally unnecessary for the project (no Wifi nor video output)
- Small size
- Supported by the great DietPI OS (unfortunately not by Ropieee).
Note that the download area of the DietPI website is a good reference for a list of SBC on the market.
The I2S interface:
Here, a long dilemma took me between many options, but the Armature Hecate seemed to come with all I wanted:
- Support of all formats (native DSD included),
- A galvanic isolation of its output with its own dedicated power supply (unfortunately not able to correctly feed the AudioGD HDMI module)
- Two good oscillators,
- Like the oscillators, the USB processor XMOS U208 is reported by several feedbacks on audiophiles forums as a reference
- The SPDIF output gives out of the box an idea of the performance of the card
The internal LT3045 regulators of the NanoPat:
The LT3045 is recommended for medical applications, thought. As an obsessed audiophile, this sounds good for me .
There’s a lot of regulators to provide 5V and 3.3V. The LT3045 seems to come as the one having an ultra low noise and may be one having the best regulation (PSSR, the output variation).
Besides that, it manage overheating, over current and undervoltage detection. It is able to drive 500mA but if you double it you gain in noise rejection and power can go up to 1A. Hum…just sufficient for both the NanoPI and the Hecate board.
The external PSU:
Many of us (audiophile) reject all switching solutions because of the pollution they can produce on on our Hifi setup. Its an established fact, don’t know if it is always justified as I used with pleasure a switching IFI Audio PSU in 5V and the result was positive. But the doubt … So the purpose was to build a multiple PSU to power the streamer and my network siwtch. External, because it is in a separate metal box which can be placed at a reasonable distance from the other “sensible?” elements.
About SuperCaps:
As the HDMI module needs a very low current (30mA), it sounds to me that it would be a good idea and relatively easy and cheap way to setup a bank of supercaps with their low ESR characteristic in order to meet may be the benefits of a LPS1 Ultracap and to really isolate the last stage in contact with the DAC. I can’t achieve this goal for the moment, while the amount of work to complete what is already done was huge for me. Moreover the benefits of Supercaps, as they are presented here, seems finally not guaranteed to me.
Does it worth it without powering all the other electronics components (SBC + Hecate board) too ? Is seems easy to power a 3.3V module while many of the SuperCaps are 5V (so we are in the range). But powering a 5V SBC, require to find out more than 5V Supercaps (rare and costly) or introduce a kind load balancing control to avoid over voltage of the Caps. I’ve brought a symmetrical +4V/-4V regulated power to the SBC/Hecate part (up to the LT3045 5V), so putting few 5V SuperCaps on positive and negative lines before the LT3045 regulators coud do the trick. But a real ground isolation would need 2 banks ? One being charged, the second discharging and powering cleanly the streamer. I’ve also no measurement solution to check if it worth it, except may be the SQ it could produce. See that later…
Some open questions:
- Why not use the 3,3V of the LT3045 regulator in replacement of the embedded power circuit in the isolated part of the Hecate board ?
- All these internal regulators on the NanoPI and the Hecate board, what to think about them…
- Is it easy to unsolder the native I2S cables on the HDMI module to replace them with shorter and shielded silver cables ?
- Is there any interest to replace the HDMI module by one of the new modules I discovered few days ago on the electronic dealer I use ? Some of them require a 5V instead of 3.3V but they have their own embedded regulator (not identified)?
- Would it be an improvement to place some some grounded metal separation inside the box of the streamer? For example to isolate the NanoPI SBC from the output module ?
Now how the NanoPat is sounding ?
The first impressions was right, the result is awesome. So it may be not the best, I would have to compare it, but the improvements over the several steps is clearly at a new level. As already said, I lived few weeks with the NanoPat streamer prototype, i.e, without the full 2 new power lines (external PSU and LT3045). It was powered by a classic PSU and even a better AQVOX PSU + a battery with a small regulator for the HDMI module. The other version before was just the NanoPI alone or the RPI3 alone. Each step brought a better SQ.
This new setup is stunning, the amount of new micro informations is incredible and the sound stage seems now far more accurate. I gain a bit of clarity too. The different vibes between each track is well perceived as ever. Very very Happy !!!
To sum up, such project takes energy, time, money and tools. Not sure if I would not reconsider the idea with more importance given to the option to buy a good streamer « on the shelf ». It is certainly possible to do simpler or go even further. If there were one or two components which seems to be the keystone, who they are ? Not sure, but one hypothesis: in the chain from the ROON (or UPNP) server, up to the last stage before the DirectStream, the more you go closer to the DAC, the more it is important to take care to the power quality and interconnections (cables). Just an hypothesis from a neophyte.
), who argumented on reducing the number of transformations so targeting a more straightful I2S chain. So far, at the moment the Kali Reclocker was limited and had some problems if I remember well. We was waiting for a solution from Audiophonics (which come far far later as expected) and from Allo. Seems now we have both. But the time being, I decided to test the Armature Hecate board and was stunned by the result so I decided finally to build something around.
. I’ve put a low ESR audiophile capacitor 3900uF 16V from Panasonic (the one you noticed on the picture) in order to filter the last stage. I believe the LT3045 has its own filtering cap, so I don’t know if the Panasonic brings something or not. In my open questions in a previous post, I still have the interrogation about the importance to power the Hecate’s input (as well as the SBC that feed the Hecate with data) with as much care as we take for the HDMI module…