[Lowfer] SDRadio and phasing exciters : status update

[Alberto di Bene]

I have done some tests, hardware and software, and have drawn some 
conclusions.
We must differentiate the input side from the output one, as they have 
different
solutions for the problem of skewed sound cards, where one channel is 
delayed
wrt the other by a given amount of time.

For the input side, I thought in the beginning that the skew amounted to 
a fixed
value of degrees of phase difference. This can be compensated in at 
least a couple
of ways in software, the most simple and effective being the skewing of 
the I and Q
components of the NCO used for the final downconversion to zero IF. I 
put up a very
simple RC network, that allowed me to feed the L and R channels of the 
sound card
with the output of a 10 kHz oscillator, with an adjustable phase 
difference, and in effect
I was able to null the unwanted response by varying the phase angle 
between the LO
I and Q components.

Unfortunately, from the tests of Lyle what resulted is that what is 
fixed is not the phase
skew, but the time lag of one channel wrt the other. At this point I 
thought at the possibility
of inserting a software time delay in the other channel, for 
compensation. But, unless
there is a solution that I haven't thought of, this seems to not be 
feasible, for a very
simple reason. Even sampling at 48 kHz, the fastest rate supported by 
the majority of
sound cards, the smallest delay granularity that can be achieved is the 
inverse of that
value, which means about 21 microsecond. Lyle measured about 14 us in 
his case,
so 21 us is by far a too coarse step... so I think that for the input 
side, if you have such
a bad behaved card, the only solution is a hardware all pass, like that 
designed by Lyle,
to be inserted in the channel opposite to that which is suffering the delay.

For the output side, of interest to those who are playing with the idea 
of building a phasing
exciter, may be the situation is not that gloom. Here we don't have to 
deal with a range
of frequencies, but with just a monochromatic signal, which, at any 
given point in time, is
just a single frequency.  This make possible to dynamically adjust the 
phase angle used
as correction in the NCO, in function of the measured time delay of the 
sound card, and the
frequency being generated.

To sum up, for SDRadio I am afraid that those who happen to have a 
skewed sound card
must ignite their soldering iron...
For TCube, which can be used as VFO for the phasing exciters, I will 
insert into it the
possibility to specify the amount of microseconds of delay to be corrected.

If you have comments or suggestions, please express them, thanks.

73  Alberto  I2PHD