Hi Andy,
IMO, you are being too much of a perfectionist to worry about
sin(x)/x. I believe that x must be pi * fsig / fsample,
because the spectrum goes to zero at multiples of fsample.
So at 20 kHz the spectral density is 20*log(sin(pi*20/5000)/(pi*20/5000)).
This is about 0.00023 dB down from the level near DC.
The remaining question is what's the scale factor.
I believe that you can ignore the high frequency behavior,
and pretend that the power is spread over half the sample
frequency. If your shift register puts out +/- 1V square
wave into 1 ohm (total RMS is one watt), then the low
frequency power density is 10*log(1/2.5E6) = -64 dBW/Hz.
My math is really weak, so I verified this by simulating a
random square wave, and looking at the spectrum with Cool Edit.
73,
Stewart KK7KA
----- Original Message -----
From: "Andy talbot" <[email protected]>
To: <[email protected]>
Sent: Wednesday, December 10, 2003 6:59 AM
Subject: RE: LF: Related Technical Query - Soundcard calibrator
Yes, the power spectrum of a SIN(X) / X pattern.
To Brian 'GVB, the limits will be (theoretically at least) - infinity to
infinity, but in practice a few lobes will be sufficient, say about +/- 5
Andy
|
