[Elecraft] Thoughts about K2 freq shift & FINE RIT...

[David Woolley]

Steve Kallal wrote:

Further to yesterday:

> The common fix for this problem is to use the tx frequency lock feature in
> most of the digital comm software. Can I assume that the tx and rx
> frequencies will match if I use the same filter for tx and rx? In my case,
> this would be the OP1 filter. I could tune to the other station's frequency
> in FL1 (OP1), before locking the tx frequency in the software. Once I lock

You must use FL1 (and the same mode and reverse setting).  Using OP1 on 
a different filter number may produce an offset because:

- you deliberately chose a different position of BFO with
   respect to the passband;
- it's possible, that two different BFO control voltages,
   at one end of the range, show the same frequency to
   within 10Hz;
- the measured frequency normally has an uncertainty in the last
   digit, and, depending on when the frequency measurement is captured,
   and maybe whether there is any filtering to improve the results,
   either of two values 10Hz apart might actually get stored.

Actually, the last point means that, as well as measuring the BFO 
frequency to better than 1Hz, you also need to perform some measurements 
to determine which of the two candidate measured frequencies was 
actually stored.  I'm assuming that the firmware doesn't use a longer 
gate time when capturing the value to store, and actually try and set 
the local oscillator more finely than 10Hz, in which case RIT step size 
might be the primary source of error!

Some of the theory I missed yesterday:

As I understand it, when you use the Display key mode in CAL FIL, you 
are stepping the BFO varactor control voltage.  The steps vary from less 
than 10 Hz, to a few 10s, depending on which end of the range you are 
at.  No calibration data is stored for this, but it means that the BFO 
will have discrete possible offsets from the nominal 10Hz tuning points.

What also happens is that the internal frequency counter measures the 
actual BFO frequency, but only to 10Hz precision and with a tolerance of 
about 10Hz, although, as noted elsewhere, it could do a more precise 
measurement when storing, compared with what is displayed.

When you tune to a frequency in SSB or RTTY mode, the firmware looks up 
the stored BFO frequency for the current filter, adds or subtracts it 
from the requested frequency and gets a 10Hz resolution desired local 
oscillator frequency.  This will have an error due to:

- differing thermal environments (which will tend to affect all
   filters the same, if calibrated at the same time);
- the uncertainty in the last digit from the frequency counter;
- the offset between the actual BFO frequency step and the corresponding
   10Hz step.

Assuming that it doesn't try to be very clever, it will then set the PLL 
  to the nearest 5kHz(?) and set the reference oscillator varactor 
voltage to the RIT step that is closest to the target 10Hz step.

Some of the uncertainties here are:

- the calibration data is based on measurements which have a 10Hz
   uncertainty in the  measurement, although it might smooth those out;
- calilbration data isn't stored for every possible control voltage, so
   there will be some interpolation errors, although these are probably
   limited so as not to dominate other error sources;
- the RIT steps will not fall on exact 10Hz points and the offset will
   be frequency dependent.

The firmware may try to be clever and be more accurate than this, but 
with closed source firmware, one can't really tell.  One way it could be 
clever is by relying on the frequency reading errors being random over 
large ranges and try to get a fit for the reference oscillator frequency 
versus control voltage step curve that is much more accurate than any 
individual reading.  It might then choose a 5kHz (?) setting that is not 
the closest.  However, I'm not sure that the pulling range is large 
enough to allow that on all bands (although it is the bands which would 
  have the largest RIT steps that would most likely allow it).  By doing 
that, it might be able to find a step that was closest to the desired 
10Hz.  (I suspect it would exacerbate the popping problem though, ss the
PLL would need stepping up and down at frequent intervals.)


> the tx frequency, I should be free to choose a narrower filter for rx.
> 
> Again, can I assume the tx and rx frequencies will be matched if I use the
> tx filter?

As long as you mean FL1, same mode and same reverse setting.


-- 
David Woolley
Emails are not formal business letters, whatever businesses may want.
RFC1855 says there should be an address here, but, in a world of spam,
that is no longer good advice, as archive address hiding may not work.