The whole mixing as done by the MSX hardware causes such an offset. Just connect an oscilloscope to the MSX sound output.
Would the same argument apply for the two MAME
drivers, Bombjack and Nemesis, that I quoted in that message? I can't help but feeling that it seems "suspect" that only the drivers containing the ay8910 (be it in MESS or MAME) seem to have these offsets, but not others.
On the Audacity site
the cause for a DC offset is said to be (almost always) a fixed voltage offset somewhere in the audio chain
. That page also lists the problems that a DC-offset introduces for sound quality.
"The cause is almost always a fixed voltage offset somewhere in the audio chain before the analog signal is converted to digital values. For example, the voltage may be directly caused by a faulty sound card, or may come from some other device that is attached to the card. Any offset is normally so small as to not be noticeable, but with defective or poor quality hardware it may become large enough to be a problem. "
Interestingly I see that a fixed voltage offset is purposedly introduced in the ay8910 driver:
"For the AY measurements cited in e.g. openmsx as "Hacker Kay" for a single channel were taken. These were normalized to 0 ... 65535 and consequently adapted to an offset of 0.2V and a VPP of 1.3V.
Could this be the culprit?
If so, what changes would I need to make to remove that voltage offset in the source code? Just so that I can re-run the audiogram test.
MAME Nemesis ay8910:
MAME Bombjack ay8910: