Klaas Robers wrote:Another strange thing are the "half way" samples, or have you programmed it in that way?
No, the "half way" samples are not part of the file, the values are not contained within it at all. Being 10uS long it implies 96kHz being used, perhaps as a digital filter.
If you play back the file and normalise it to 1Vp/p the values in the file should reveal the following:-
00h = 0.000V (Sync tip)
4Ch = 0.298V (Black, nominally 0.300V)
66H = 0.400V (1st grey step)
7Fh = 0.498V (2nd grey step, nominally 0.500V)
99h = 0.600V (3rd grey step)
B3h = 0.702V (4th grey step, nominally 0.700V)
CCh = 0.800V (5th grey step)
E6h = 0.902V (6th grey step, nominally 0.900V)
FFh = 1.000V (Peak white)
If you look at the 'data chunk' (yes, that is the correct term), from byte 45d onwards, it only contains those values. The preceeding 44d bytes are the header information.
I will try the same exercise at a much lower sample rate and see if the same thing happens. If it were, at a sample rate of 8kHz (far too low for NBTV) there might be these "half way" steps with a duration of around 60uS.
I'll keep you updated.
Yes, a simple low-pass filter would remove them, but even without one I doubt they would be visible, I am just curious as to the reason they are there.
Steve A.
Postscript.
In fact it seems that some processing is being done in the digital domain to simulate an analogue filter. Of what sort I'm not sure, compare the two photos and you can see the similarity, although there are differences.
The analogue version (Clean.jpg is a Sallen and Key filter at 15kHz at -12db/octave), whereas the other seems to be a digital version of a Chebyshev filter, possibly around the Nyquist cut-off of 24kHz, hence the ringing.
Forget the fact that one is ascending, the other decending, it shouldn't make any difference.