[Elecraft] Sound Cards

[Jack Smith]

While the subject is sound cards internal versus external, It's worth 
mentioning an oddity I experience with a Dell laptop's sound card. At 
least one other ham has an identical problem with his (different model) 
Dell laptop.

Both Dell machines use a Sigmatel sound chip, integrated on the mother 
board. This chip has a hardware sample  rate set in multiples of 24 K, 
up to 96 ksamples/sec. (Might well be multiples of 12 Ks/s but my tests 
have all been at 24, 48 and 96 ks/s.)

There are many programs in use written for sample rates tied to the 
CD-ROM 44.1 ks/s rate, and sub-multiples such as 11.025 ks/s.

 From my experiments, it seems that the Sigmatel chip cannot sample at 
44.1 ks/s related rates, and instead samples at the a 24K multiple and 
synthesizes the requested rate by interpolation. This process is done, I 
believe, in the sound driver supplied by Dell with the laptop.

The problem is that the synthesized sample  rate is not close to the 
right speed. This shows up as a frequency error in the software.

I recently ran a test with the newest version of Spectragram, a program 
I have great confidence in. It allows a selection of sample rates and I 
cranked an 1850 Hz tone into it. (Believe me, the 1850 Hz  tone is 
accurate to well below +/- 0.01 Hz.) Spectragram reports the following 
frequencies, based on the sample rate, with the internal A/D chip.

(Ks/s)            (Hz)
Sample Rate   Reported Frequency
22.050            1889
44.100            1838
48.000            1850
96.000            1850

These figures are +/- a couple Hz because the FFT bin width has to be 
considered.

I then tried the same experiment with my external E-MU 0202 sound card. 
No point printing a table, all the readings were 1850 Hz regardless of 
sample rate selected.

So, why is the important? First, it means that you should calibrate your 
software program, in the event that you have software that allows 
calibration.

Programs known to have odd results when the sample rates are so far off 
include ARGO and Frisnit's NAVTEX. The NAVTEX decoder not only displays 
the frequency wrong (1850 Hz tone shows as 1898 Hz) it must use  sound 
card-based timing internally (integrate and dump post-detection 
filters?) as it has excess errors even for strong clean signals. 
Switching to the E-MU 0202 external sound card shows essentially perfect 
decoding under even worse signal conditions.

So, it's important to know how your sound card actually behaves.

I'll likely write up this, along with some sample files, as a web page 
over the next couple days.

Jack K8ZOA