[Elecraft] KDSP2 and Spectrogram

[Lyle Johnson]

Hello Dave!

> My KDSP2 arrived yesterday, took a couple of hours to build and install.
> Works fine on air...

This is what it was designed for, success on the air :-)

> However, tonight I hooked up Spectrogram and a noise
> generator. The filter doesn't look as narrow as I would have expected and
> appears no narrower than about 200 Hz - even when set to "50 Hz".
> Settings, CW, 700 Hz centre frequency and de-noiser off.

When you say no narrower than 200 Hz, do you mean at the top (3 to 6 dB
down), or at the "bottom" (30 to 60 dB down) of the Spectrogram plot?

At the 30 to 50 dB down area, using the default (not "soft") filters, I
would expect 250 to 300 Hz to be displayed by Spectrogram.

I wouldn't believe Spectrogram for anything below about -30 dB to -40dB from
its peak response (driven to the limit before distortion) without actually
calibrating it.  It is a good relative tool, but not a lab instrument. And
with a noise signal it may be hard to see the distortion. Try backing down
the RF gain so the signal is about 20 dB weaker than you are currently using
it, and see what you get.

With my unit and a noise generator, I see a peak of - 37 dB at 611 Hz,  280
Hz wide at -60 (or 23 dB down from the peak
***

A better test might be the following.

1) tune in the 4.000 MHz internal K2 calibration signal.
2) disable AGC
3) adjust RF gain so nothing is overdriven
4) adjust AF gain for a reasonable indication on an AC voltmeter - or
Spectrogram if you wish, in which case go for a fairly high level
5) tune the radio from the marker signal in 50 Hz steps (use the .1 kHz
tuning rate and note that there are two steps to change digits - the radio
actually tunes in 50 Hz steps, or at least mine does :-)
6) note the amplitude
7) tune to the next step, etc.

This may give you a better idea of what is happening.

For example, in my unit I just measured the following:

Freq Amp (dB)
477  -60
506  -40
571  -32
614  -32
668  -38
711  -54

Note that since an FFT is being done in Spectrogram and the radio may not be
tuned to exactly the center of a bin, the steps aren't exactly 50 Hz.

What we see from this is the following

Peak is about (614-571 =) 43 Hz wide.  OK for a 50 Hz filter with limited
test gear applied.
8 to 10 dB width is about (668 - 506 =) 162 Hz.  This is a reasonable roll
off.
Approx 30 dB down is (711 - 477=) 234 Hz.

The design is to have the filters be about 300 Hz wider at -40 to -60 dB
than the selected width.  At +/- 100 Hz, we're about 30dB down, so it is
working pretty close to spec.

As always, the important thing is not what Spectrogram says, it is what your
ears tell you :-)

73,

Lyle KK7P