Recording audio from PC sound card



I am trying to use the sound card in PC 1 as a synthesizer from PC 2, which sends MIDI commands to PC 1 through Ethernet, which is not a problem. The problem appears when I connect line out to the analogue input, for audio capture.


Here, PC1 is a Dell Dimension XPS d333 machine, and the device I am using is TurtleBeach montego (vintage gaming sound card).

Doing so, results in a bee swarm sound (not 50 Hz mains hum, but a tone with a fundamental around 160 Hz), with or without the shield connected.

What I have tried or thought of:

  • Record "What you hear" on PC 1. This would add to much latency. I need Windows 98 to host the card, and thus, I do not have access to a low-latency API. Also, even if the synthesizer worked on ALSA, streaming data back through TCP would also add too much latency (MIDI commands uses a custom UDP based protocol with fixed sized packets).

  • Use the soundfont from PC 1 in a soundfont player. It should work, but the file is in a proprietary format that needs to be reverse engineered.

  • Separating the two grounds solves the problem, but introduces other strange artefacts like unexpected voltage on chassis.

  • From the schematics, you can see that there is no true ground, since the setup is in a room without grounded sockets. I temporarily (and probably illegally) tried to power the system from from a socket in another room with ground, to see if that would help. Doing so did not help, so paying an electrician for adding grounded sockets to the room would not give any payback.

  • Another solution could be an audio transformer. I tried that, but the problem remains + it adds 50 Hz mains hum to the signal.

  • Use some RF modulation and a pair of antenna. This requires a good choice of modulation frequency, or some kind of isolation box (for 2.4 GHz, a cavity surrounded by water would do the trick.

  • Add a linebox. The USB audio device can drive 48 V phantom, if required by the linebox. Maybe, however, the manual says that I need properly grounded mains power, which I do not have.

  • Replace the power supply in PC 1


Posted 2016-12-15T10:29:33.850

Reputation: 203

If Record "What you hear" on PC 1 works, you can remove latency by using an external audio interface. – Dalv Olan – 2017-02-16T16:07:06.360

@DalvOlan No, the latency comes from the internal buffer of the API. I successfully launched JACK with 4 ms latency (after compiling it without SSE). Also the "what you hear" stuff should be fed back through TCP, which may require 50 ms to run stable. – user877329 – 2017-02-16T16:24:22.413



If one or both of these is a portable computer, then your are dealing with the NOTORIOUS noise from the power bricks that come with virtually all portable devices these days. Those things are TERRIBLE and make audio interfacing a nightmare (as you are discovering).

Yes, by all means try plugging both computers into a common "power strip" or equivalent which will at least connect the grounds together, assuming they BOTH have grounded mains plugs. If they don't BOTH have grounded mains plugs, then this is of doubtful benefit.

A proper isolation transformer will NOT add hum (presumably from magnetic pickup). Try moving the transformer away from the computers and/or orient it in different directions to minimize magnetic coupling. And/or use a better transformer with proper magnetic shielding (which likely means expensive).

I would try just recording on Computer #1 and eliminating the whole intrigue of sending the audio output to another computer (which seems the cause of the situation.)

If you will get too much latency by attempting to record on PC1, then you are back to improving the traditional analog input/output/isolation/coupling fundamentals of the problem. A proper isolation transformer will have magnetic shielding that will prevent picking up ambient magnetic fields. Exotic "direct-boxes" with phantom power, etc. are typically needed for low-level signals such as from electric guitars. It should NOT be necessary to use an active "direct box" for a line-level source (out of PC1).

Improving the power supplies for both PCs would also be likely to reducing the noise problem. Switching power supplies are notorious and generally avoided in audio gear for that reason.

Richard Crowley

Posted 2016-12-15T10:29:33.850

Reputation: 1 849


Recording from a digital source by going through D/A-A/D conversion may be the most universal format converter you can come up with.

It is, however, also by far the most problematic since it involves filtering and resampling at highly correlated frequencies. So you really need high-quality converters (which don't rely on other equipment and hearing for filtering out inaudible frequency components) to pull that feat off without introducing significant resampling artifacts.

If you have any means to just transfer the data digitally (S/PDIF is still pretty versatile to hook up when some Ethernet connection does not cut it) you should try going that route.

There might be some ground loop problem: if that is the case, a DI box will likely help. But a 160Hz buzz does not sound like ground loop: it could well be some resampling artifact.

However, you state separating the grounds helps, but using an audio transformer doesn't. That does not make sense unless you are using the audio transformer wrong (namely, not as a ground lift, or by having other connections still making a ground loop).


Posted 2016-12-15T10:29:33.850

Reputation: 1