Hi Paul,
Am 06.07.2017 20:13, schrieb Paul Nicholson:
> channel 1 seems to show some minor glitches, see attachment.
Looks bad. Run the signal through vtstat -i to see
if channel one is clipping at all.
vtread ... | vtstat -i
Maybe both channels are close to hitting 1.0 amplitude in
the stream and ch1 is slightly higher gain. You need to
look out for clipping if using integer sample formats.
-First checked the signal with vtread | vtstat -i and got:
format: float8
channels: 2
samples: 3006464
clips: 0
breaks: 0
sample rate: 24000, correction 1.00000000
duration: 125.2693333 seconds
start: 2017-07-05_03:19:11.000000
end: 2017-07-05_03:21:16.269333, interval 125.2693333 seconds
mean,rms,peak: 1 4.538e-03,5.045e-03,7.630e-03
mean,rms,peak: 2 4.507e-03,5.010e-03,7.691e-03
----
...so this looks fine, as expected.
-Next i took the original audio interface delivered together with the
card (not my homemade 3 channel BNC input interface). Repeated the test,
got the same (bad) effect on CH1.
-Next i run the vt processes from the console (not from the script) just
to see if there are differences: No difference, the same effect.
-Next i recorded a file without vttime and without vtresample. Converted
to wav, displayed in SpecLab: Still the same effect!
-Next i made a wav file from CH2 and CH3, just to see how CH3 looks in a
spectrogram, when injecting the same 500 Hz signal: CH3 looks better
than CH1 but some minor glitches are still there.
Well, we're discussing about signals of > 60 dB SNR. Will this cause a
significant performance reduction at all? Did you do similar large
signal tests as well and observed the results?
Should i ignore these faint effects? Maybe it is more useful to finally
connect two antennas and record a first VLF signal with some natural
noise in the background...
73, Stefan
|
