Content-Transfer-Encoding: 8bit
Message-ID: <65AECDF1F89AD411900400508BFC869F0D7543@pdw-mail-1.dera.gov.uk>
From: "Talbot Andrew" <ACTALBOT@dera.gov.uk>
To: rsgb_lf_group@blacksheep.org
Subject: LF: Frequency Calibration Argo etc.
Date: Mon, 22 Jan 2001 14:46:41 -0000
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1; format=flowed
Precedence: bulk
Reply-To: rsgb_lf_group@blacksheep.org
Sender: <majordom@post.thorcom.com>

The recent talk about soundcard accuracies set me thinking.   Tuning
error and soundcard accuracy do not have exactly the same effect on the
apparent observed frequency line in Argo or any DSP system where the
sampling rate is not locked to the same frequency as used for tuning.

Tuning error is reflected as a frequency shift which applies equally to
all tones within an audio bandwidth.
Sampling rate error appears to change the tones by an amount
proportional to frequency.

eg at 8000 Hz sampling rate + 10ppm  = 8000.08 Hz
a 1kHz tone will appear at to be 999.99Hz (0.01 Hz low),    1500Hz will
appear to be 1499.985Hz (0.015Hz low)   etc
If the receiver tuning is 0.1 Hz in error these would be 999.9 ands
1499.9 respectively.

So to calibrate out both uncertainties a minimum of two measurements are
needed.  Here is a procedure to do this :

Tune into a transmission whose frequency is known EXACTLY - such as MSF
at 60kHz or the centre of Loran at 100kHz - using your favourite piece
of narrow band software and receiver as used normally.   Adjust tuning
for an audio tone of 500Hz and measure the exact tone frequency
resulting.     Then alter the receiver tuning to get a tone of 2500Hz
and measure this figure exactly.

The difference between the two measured frequencies  expressed as a
fraction of the wanted separation (here 2kHz) is the soundcard sampling
error rate.   Any error that applies equally to both tones is due to
receiver tuning alone.  Obviously the widest tone separation as possible
is desired to minimise measurement error. 
This test does pre-suppose that the error in the receiver is determined
only by its internal oscillators and not by errors in transferring the
desired frequency to the display.  Most modern PLL and DDS controlled
receivers do in fact generate exactly what the dial says, subject only
to their oscillator calibration.

In most cases the receiver frequency setting error should swamp that of
the soundcard but this may always not necessarily be the case.   For
example an IC746 with TCXO option and trimmed to 0.2ppm accuracy (it can
be done) used to receive directly on 137kHz will exhibit an error of
0.027 Hz as all frequencies within the reciver are locked to the master
source.   A laptop integral soundcard could easily be 30ppm out even if
the oscillator is 'meant' to be exact ie. not allowing for poor
implementation which can give errors up to 0.2 percent !!!   30ppm would
give 0.03Hz at 1kHz tone and 0.06Hz at 2kHz tone frequency, error
exceeding that from the receiver.

Andy  G4JNT


> 
> Assuming an overly optimistic stability of 10 ppm over the 
> whole temperature excursion range
> for that crystal, this means a deviation of 0.01  Hertz for 
> an audio signal of 1 kHz.
> At this frequencies, even cheap crystals can prove adequate.
> 
> Thanks for your tests Alan,
> 
> 73   Alberto    I2PHD
> 
> 
> 


-- 
The Information contained in this E-Mail and any subsequent correspondence
is private and is intended solely for the intended recipient(s).
For those other than the recipient any disclosure, copying, distribution, 
or any action taken or omitted to be taken in reliance on such information is
prohibited and may be unlawful.