Adding an Amp into Thin-Client PC

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Perhaps this will come useful to someone...

A small fanless thin-client PC (FS Futro, HP Txxx, Dell Wise etc.) powered by a typically 20V adapter, for nothing from ebay, ideal for home audio (Daphile). Some models offer room inside, integrated Intel HDA soundcard codecs have improved a lot, digital amp boards are small, efficient and use single-voltage power - let's make a small dirt-cheap all-in-one machine.

Sounds good, but my try with tripath amp Home * pavhofman/plabs-player Wiki * GitHub was a bit disappointing. The ground loop within the PC board and the amp was just unfixable. Every CF card access, every letter typed on screen, every mouse move, all was clearly audible no matter what I tried. I ended up with two adapters External Relay Activated Power Supply for the Amp * pavhofman/plabs-player Wiki * GitHub - one for the PC and another one for the amp. Clumsy.

Then I learned about the TPA3116 chip which (unlike tripath) offers balanced inputs. Most TPA3116 boards do not pull out the negative inputs from underneath the heatsink but there are some which do, e.g. DC 12V 24V TPA3116 Dual Channel Stereo 50W*2 BTL Mono 100W Audio Amplifier Board | eBay

Just disconnecting the -IN capacitor from ground makes a perfect balanced input.

Unbalanced connection + power from the PC adapter proved just as disappointing as the tripath.

Then I connected the amp in the usual way for unbalanced-balanced interface, using shielded microphone cable, powering the amp directly from the jack power socket of the board:

balanced-amp.png


futro_TPA3116.jpg


Voila, the noise was completely gone, dead silent at full volume. Amp for 8 dollars, thin client incl. the adapter for 10 euro, 18W/8ohm THD-N 0.05%, reasonable performance even for 2.7ohm.

I did some measurements ( Headless Amplifier Measurement Workstation ), they fit the amp datasheet http://www.ti.com/lit/ds/symlink/tpa3116d2.pdf very close, which means the noise from the ground loop is eliminated. All the figures are available here Player v.2: TPA3116 with Single Adapter * pavhofman/plabs-player Wiki * GitHub

Perhaps some daphile or other small audio PC builders will find this useful. It works :)
 
Last edited:
Thanks. It is a regular stereo board, with bridge switches at each channel output.

Its performance in absolute terms is doubtlessly quite inferior compared to good analog amps. But these do not come with single supply voltage, tiny footprint, high efficiency allowing use of the PC adapter + low heat losses. Not even talking about the price :)

The project is done by a few school boys here who are building a piece each for themselves to use it at home. Another piece plays in kitchen. There is really no ambition to replace a living-room setup.

Yet the question is how such device would fare against a full-blown hifi gear in a blind listening test at normal listening levels...
 
It works because the power adapter has only two pins, so no ground connection.

If you look at the diagram, the same power adapter is powering both the PC and the amp. This ground loop is not related to mains ground (PE wire), but to the common ground line of PC and the amp. The common-mode rejection of the amp's balanced inputs eliminates the common-mode noise voltage of both soundcard outputs against the common ground. Logically no elimination occurs when using an amp with unbalanced inputs.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.