Frequency Accuracy/Stability

Frequency Accuracy:

Frequency accuracy refers to the absolute frequency accuracy of the equipment or process.  The frequency accuracy of two stations at each end of a communication link impacts on the difficulty of the initial tuning in of the signals from the other station.  Advanced microwave enthusiasts using CW for weak-signal work are looking at better than 0.001ppm as they are looking for weak signals at 10GHz!!!    Some even use Caesium standards!!!

For LF narrowband at least a 2ppm TCXO is needed, although even the standard 5ppm crystal in a rig like the FT-847 will do.

For Laser Comms DX the accuracy of a standard soundcard should be sufficient to generate and capture the sub-carrier tones for some modes of narrowband work.  Tests I have done on several soundcards show that the initial accuracy of the standard sample rates (11025Hz, 22050Hz and 44100Hz) should be within 100ppm of the nominal frequency.

Here is the results of calibrating two different soundcards against a local time/frequency standard station.

I.  Creative Labs Soundblaster in 200MHz Pentium.

Above is the calibration results for a Creative Labs Soundblaster soundcard for 11025Hz sample rate, resolution BW of 0.001314Hz.   The actual sample frequency is 11024.963Hz which is an accuracy of -3.320ppm.  The two pips either side are the sidebands were produced by the 60 second cycle of the pips which apart from marking the seconds (those sidebands are way off the screen at either side) carry information in codes formed by the duration of the pips.  The pips are at 0.01667Hz offset either side. The computer had been on for 4 hours before the calibration was done.

II.   ESS Soundcard in a Compaq Presario Pentium 266MHz Laptop

Above here is the calibration results for a ESS soundcard in a Compact Presario Pentium 266MHz laptop.  This was taken straight after switch-on for its 11025Hz sample rate.  Here the actual sample rate is 11025.558Hz which is an initial sample rate error of +50.568ppm.

I make the assumption that the frequency of a soundcard would be within +/- 100 ppm for the standard samples rates.

Note: The non-standard rates (eg 5500Hz) are an entirely different matter as the actual sample rates are more likely to be the closest sample rate which can be produced from the same reference as the standard rates.  For example, my Creative Labs Soundblaster soundcard has an actual sample rate of 5512.5Hz (11025 / 2) for a set sample rate of 5500Hz.

Frequency Stability:

Frequency stability refers to the drift in frequency over the period of a record of the equipment or process.  For narrowband work this impacts on the maximum gain in S/N ratio that can be achieved by going to narrower BWs.  If, for example, you are working with an FFT bin BW of 0.001Hz and over the period of the record time the signal drifts by 0.1Hz, the S/N gain by using 0.001Hz BW is lost.

Here is the results of re-calibrating the Presario soundcard after 5 hours of warm-up.

III.   Re-calibration of Presario Soundcard After Warm-up

Above here is the re-calibration results for a ESS soundcard in a Compact Presario Pentium 266MHz laptop after a 5 hours warm-up.  Now the actual sample rate is 11025.601Hz which is a sample rate error of +54.511ppm.

This shows a drift of +3.943ppm from switch-on to warm-up stabilisation.

Conclusion:

Subsequent measurements 2 hours later on both soundcards showed that the sample rate had drifted by less than one FFT bin (0.001314Hz).  At 1000Hz this translates to less than 1.3ppm drift.

Soundcards without software calibration should therefore have sufficient stability to allow use of both AFK and FDK modes using record lengths of one minute (BW about 0.02Hz) with character code spacings of 0.1Hz.

However, they do not have sufficient accuracy without software calibration for the AFK mode.  To this mode it will be necessary to calibrate the soundcard against a standard of at least 10ppm.

 E-mail me with comments, suggestions and corrections.