or Connect
Head-Fi.org › Forums › Equipment Forums › Sound Science › Simulated crosstalk listening example (audio files inside)
New Posts  All Forums:Forum Nav:

# Simulated crosstalk listening example (audio files inside)

In discussions of amp soundstage and other topics in different parts of the forums, some people have quoted amp crosstalk figures as a hypothesis for differences in perceived sound quality.  Also, some people (I've got to include myself) may not really have an intuitive handle on what crosstalk sounds like in practice.

Thus, I've tinkered around with a snippet of music, processing it to artificially add L -> R and R -> L channel crosstalk.  Suppose that L_o[n] is the original left channel signal and R_o[n] is the original right channel signal; L[n] is the new left channel signal and R[n] is the new right channel signal (n indexes samples, time), with A as a constant.  To simulate crosstalk, I just did a simple multiply and add.

L[n] = L_o[n] + A * R_o[n]

R[n] = R_o[n] + A * L_o[n]

with a constant A being 0, 0.0001, 0.001, 0.01, or 0.1 (original, -80 dB, -60 dB, -40 dB, -20 dB).  There's a file corresponding to each value of A.

edit: with edited versions, that'd be 0, -0.0001, -0.001, -0.01, -0.1

In practice, is the crosstalk really phase-aligned, as what I've done here?  (It's definitely at least a function of frequency usually, though the transfer function would have a different shape for different systems.)  Note that this is different than intentional crossfeed processing, which may use a time delay and frequency selectivity.

Anyhow, I've compiled the few files with different levels of simulated crosstalk according to the very simple procedure above, all based on the same original 30 second clip.  I did not match the levels.  If you want to listen and compare, there could be an argument made for level matching here, so use ReplayGain.  Files are labeled; this is for self-educational purposes, so no benefit to "cheating".  Rename files or use some kind of ABX test if you want to test yourself fairly.

There should be a decent amount of percussion in the snippet which might be easy to "locate" spatially.  Are there issues with placement or anything else with soundstage with the different levels of crosstalk?  At what levels?

Files here, encoded as FLAC 44.1 kHz 24-bit (original was CD audio, 44.1 kHz 16-bit):

edit: decided to invert phase, see discussion below, still flat frequency response of mixed signal

Files here, encoded as FLAC 44.1 kHz 24-bit (original was CD audio, 44.1 kHz 16-bit):

edit2: mirror to above, individual links on a different host

https://www.box.com/s/q5ej41g4ss3lgyshgy9a

https://www.box.com/s/vv4waa8367m30un8oblt

https://www.box.com/s/ju0gftr7eeb4eabagw50

https://www.box.com/s/4fowxtmymoa582hyeu5n

https://www.box.com/s/vcoygz2g9j9y4w8emk2w

Edited by mikeaj - 3/26/13 at 2:16pm

Quote:

Originally Posted by mikeaj

In practice, is the crosstalk really phase-aligned, as what I've done here?

It can be inverted if it is because of high (common) ground resistance, like in the case of TRS headphone outputs. Also, it often increases with frequency because of capacitive coupling, which can easily be simulated as well.

Edited by stv014 - 3/26/13 at 12:42pm

Oh yeah, should've just drawn the circuit myself in half a minute.  If one channel has a positive signal and the other is zero, then the midpoint voltage (past the left and right side loads, before the common ground impedance) is clearly positive so voltage across the load on the other channel is negative.

Well it's not like any of this processing requires any actual effort to do... let me update the files.  I'd go with inverted by default, I suppose.

You should host these files on a service like Google Drive, Box, or Dropbox. I hate file lockers

Quote:
Originally Posted by Tus-Chan

You should host these files on a service like Google Drive, Box, or Dropbox. I hate file lockers

Well if you say so, doesn't really take much effort so might as well...

OP updated

New Posts  All Forums:Forum Nav:
Return Home
Back to Forum: Sound Science
• Simulated crosstalk listening example (audio files inside)