[Elecraft] Filter Challenged...I'm so confused

[Ron D'Eau Claire]

You have it right Andy.

Since the K2 "returns" to FL1 when you go to xmit, the other station
won't sense any change in tx frequency no matter what you have done to
your various filter settings. 

It's easiest for me to "picture" in CW mode, in which I might zero beat
a signal with the SPOT tone. Then, if I have set the BFO frequency
differently in a different filter setting, the K2 has adjusted the LO
frequency so I will stay in zero beat as I switch through the filters. 

I FOUND the original article on the Elecraft web site. I'll copy it
below for you. 

Ron AC7AC

Date: Tue, 30 Nov 1999 00:21:58 -0800
From: Wayne Burdick <n6kr@elecraft.com>

...I misunderstood your situation: I thought you were using the
spectrogram program to plot the center frequency of the *noise* output
of the RX--I didn't realize you were using a carrier and tracking its
pitch, which the K2 will preserve regardless of how you move the BFO! I
can assure you that what you're experiencing is not related to the
firmware rev., by the way.

An example may help:

- VFO dial 7000.0
- XFIL 4913.6 (center of xtal filter at present bandwidth)
- Sig gen 7000.0
- BFO 4913.1
- Sidetone 0.6
- VCO 11913.7 (VFO + BFO + sidetone)

The signal generator's carrier is not quite centered in the filter
passband with the present setup -- it's 100 Hz too high: 11913.7 -
7000.0 = 4913.7 = IF carrier freq.

But you'll hear the signal at a 600 Hz pitch because that's where you
have the sidetone set (BFO - IF): abs( 4913.1 - 4913.7 ) = 0.6 kHz

What happens when you run CAL PLL and move the CW-N BFO down 100 Hz to
4913.0? The new VCO value is calculated as soon as you switch filters or
modes (in your case, you moved from CW-N to CW-R, then back again): VCO
= 7000 + 4913.0 + 0.6 = 11913.6

The IF carrier signal now *is* centered in the filter passband, which is
the desired result: 11913.6 - 7000.0 = 4913.6 = new IF carrier freq.

BUT, you still hear it at 600 Hz: abs( 4913.0 - 4913.6 ) = 0.6 khz

So what we've done is to move the signal in relation to the filter,
while *preserving* pitch. (This is similar to IF shift on other radios.)
No matter how far you move the BFO, this will always be the case. Hence
your spectrogram program will think the signal has not moved--although
it *will* be better centered in the crystal filter passband, at least if
you've correctly determined what the center frequency of the filter is.
You can now move your signal generator or the K2's VFO around a bit to
check if the filter center corresponds to the desired pitch. Actually,
if you can find the filter center frequency at each bandwidth used
(using the "zero-pitch" method), you can then just set the CW-N and CW-R
BFOs directly to Fcenter - sidetone and Fcenter + sidetone. That's why
we added the BFO DISPLAY mode to CAL FIL.

HOWEVER, the bigger problems are still (1) the counter's quantization
error, and (2) a small amount of drift with temperature. No matter how
much you tweak the BFO, you'll potentially always be off by as much as
the sum of these two.

I hope this helps!