[Nut-upsdev] Powercom black knight 800AP (serial)

David Gardner david-sari_gardner at sil.org
Thu Oct 4 08:46:28 UTC 2007

Nut versions: 2.2.0 and 2.0.2 (powercom -x type=KIN1500AP -x linevoltage=220)

 I've got the above mentioned serial UPS and have found some problems using

1) under linux 2.6.12 select() timed out on reading the serial port 
in NONBLOCK mode.
Not sure about this one, seems most likely a kernel problem, since it's
working OK in 2.6.20. But on the other hand it was a nasty one.

For what it's worth, the solution under 2.6.12 was to rewrite the read
code to not use select(), instead using time() calls for the timeouts.
Or upgrade the kernel..

2) The maths for the output voltage is wrong when on battery
power, but right(ish) when on line power. I've not tested it enormously,
but (comparing with the windows program, volt meters, etc) 
it looks like the multiplyer is OK, but the offset is wrong.
My (220 volt) version needed a slight tweak to the multiplyer when on
line voltage (1.954,-0.0977)
But on battery, the offset needed to be pushed up to 162.1. (otherwise
it displays about 57 volts. The windows software showed there was 
still about 2v resolution, so its almost certainly an offset error, not 

One option is that the logic should switch the maths based on 
the power status rather than on line voltage. 
This obvously would give problems since the maths are different with
line voltage, but that one could be solved with a different model type.

Is this a offset error present on any other powercom devices, or just
the Blacknight 800AP? The mfr obviously codes for it in their windows
(The mfr-supplied linux driver just reports zero volts all the time....)


More information about the Nut-upsdev mailing list