Message-ID: <27356.200103211226@gemini>
From: "James Moritz" <j.r.moritz@herts.ac.uk>
Organization: University of Hertfordshire
To: rsgb_lf_group@blacksheep.org
Date: Wed, 21 Mar 2001 12:31:44 +0000
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII; format=flowed
Content-Transfer-Encoding: 8bit
Subject: LF: <Tech> Soundcard calibration
Precedence: bulk
Reply-To: rsgb_lf_group@blacksheep.org
Sender: <majordom@post.thorcom.com>

Dear LF group,

Rik's method of producing an accurate audio tone for calibrating 
the soundcard is very similar to what I am doing, and I can confirm 
this gives an accurate and unambiguous result. I did have some 
problems, however - the logic level square wave (in my case 1kHz) 
contains many harmonics, and I found that my soundcard did not 
have a very good anti-aliasing filter. The result of this was that 
many spurious frequency components were present in the output 
data, which could confuse the calibration program. I overcame this 
by adding a simple filter - 10k in series with the divider logic output, 
connected to a parallel - tuned circuit consisting of a 100mH 
inductance in parallel with 220n + 22n capacitors. This made the 
output a fairly clean sine wave. You may have to fiddle with the 
capacitors to peak up the output, or for different audio frequencies.

If you don't want to build a calibrator, there are a couple of ways of 
using a receiver as a calibrated audio source. If the reference 
oscillator output in the RX is reasonably accessible, it can be 
coupled with a capacitive probe wire to the RX input. If the RX is 
tuned to the reference frequency, the audio output will then be the 
nominal value within the tolerance of the reference; errors in RX 
tuning due to the reference will cancel out. This should work well 
with an RX like the RA1792, where it is easy to get at the 
reference output.

If the reference is not available, the frequency offset can be 
eliminated by a differential method; Tune the RX to a stable carrier, 
so as to produce a fairly high frequency audio output - I used 
1.3kHz. Record several minutes at 8k samples/sec, and measure 
the exact frequency using the Wolf -m option. Re-tune the RX on 
the same carrier to produce a low frequency audio output - I used 
300Hz. Repeat the frequency calibration using Wolf -m. Calculate 
the difference between the two RX frequency settings, and the 
difference between the frequencies as measured by Wolf. The 
soundcard sample rate is then:

8000 x (difference in RX tuning setting) / (difference in measured 
audio frequency)

This gave me almost the same results as calibrating with an 
accurate audio frequency. Both methods assume the receiver 
reference oscillator is accurate, but it's accuracy will normally be 
pretty good compared to the soundcard. Note, however, that both 
methods may be defeated by receivers using DDS synthesisers, 
which tune in odd-frequency steps, but display the output to the 
nearest Hz.

Note also that the soundcard sampling rate error depends on the 
actual rate selected - for example, measuring 800Hz using 
Spectrum Lab with 8k sample rate gave me a frequency error of 
about +4.5Hz, but with the 11k rate, the error was about -1.6Hz. 
The errors are different with different types of soundcard. So using 
other software to measure frequency offsets will not be helpful, 
unless it also uses an 8k sample rate. I think Argo uses the 11k 
rate. I don't know about the PSK31 software. Oh joy.....

Hope this is helpful,

Cheers, Jim Moritz
73 de M0BMU