G'day Jim,
Thanks for this. It is a very thorough treatment of the method - including
the limitations. It will be one of my "print out and file" reference
documents and I will alert the VK LF-ers to the link.
While in the general area of soundcard frequency accuracy - there was a time
in the distant past when worrying about sampling frequency errors was less
of a concern to me as all the cards I tested were well within 100ppm (<
0.1Hz @ 1000Hz). In that distant past soundcard sampling rates were
44.1ksps, 22.05ksps, 11.025ksps and 5.512ksps (usually implemented as
5.5125ksps). That is, all sampling rates were "harmonically" related.
Now we have the 48ksps, 96ksps (and higher?), which, while self related
harmonically, are not harmonically related to the legacy sample rates.
The practical consequences of this caused much head scratching on my part,
until the fog cleared, when I first dabbled in W*S*P*R ("*"'s added to avoid
spambots...;-). Everything was going swimmingly until I cross-checked the
frequency of my transmission as reported by WSPR against ARGO. There was a
discrepancy of some ten's of Hz. For someone who is a bit AR about
frequency accuracy this caused mild apolexy.
To make an already long story not grow too much longer, I always use
external USB soundcards (EDIROL UA-1eX) which have a physical hardware
switch to select sampling rates (44.1ksps, 32ksps, 48ksps and 96ksps) built
into them. For some reason I had switched the sampling rate to 48ksps.
This is the "native" rate for WSPR - so it was reasonably accurate. When I
booted up ARGO, the OS drivers had to somehow downsample the 48ksps hardware
sampling rate in software (?) to the nominal 5.512ksps rate used by ARGO, a
process it did with sufficient roughness to produce the very significant
frequency error. I proved this by hardware switching to 44.1ksps on the
external soundcard, upon which the discrepancy from the 48ksps setting WSPR
reading disappeared. Note that now (when switched to 44.1ksps), the WSPR
reading had moved a few ten's of Hz as now the OS drivers had to upsample
from 44.1ksps to 48ksps. Worryingly, at the same time, the decoding
sensitivity on WSPR took a dive.
IMHO - bottom line to this is - if you have a newer PC you are likely to
have a card based around 48ksps. This means you are probably good to go
with WSPR, but may have frequency errors in ARGO. If you have an older PC
and/or soundcard it may be based around 44.1ksps, in which case you are good
to go with ARGO, but there may be frequency errors and poor decoding
performance on WSPR.
How much of this effect you see will depend on the quality of the
downsampling/upsampling/interpolating process in the drivers and/or OS.
You may not see any discrepancy at all - but it is worth checking if you are
having poor decoding on WSPR.
In my case I can avoid all problems by hardware switching to 44.1ksps when
using ARGO and 48ksps when using WSPR.
Sorry for the long post.
73 Steve VK2XV
----- Original Message -----
From: "James Moritz" <[email protected]>
To: <[email protected]>
Sent: Friday, October 02, 2009 3:09 AM
Subject: LF: RX and Sound Card
Dear LF Group,
I have just uploaded "RX_Soundcard_cal_v1.pdf" to the files page at the
UK500kHz group web site (go to
http://groups.google.com/group/uk500khz/files - it is down at the bottom
of the list). This describes a way of calibrating receiver frequency and
sound card sampling rate using an off-air standard frequency such as MSF
or 198kHz R4, without requiring an accurate audio frequency to calibrate
the sound card. This is addressed primarily to WOLF mode transmission and
reception, where this calibration has to be done fairly accurately in
order for the mode to work properly, but could also be generally useful
where sound card sample rate calibration has to be done, and there is no
precision audio frequency source available in the shack.
Any comments would be welcome,
Cheers, Jim Moritz
73 de M0BMU
|
