At 04:19 AM 6/8/2006, Mike, G3XDV wrote:

[..]

I am now convinced my problem is caused by frequency error in my TS-
850. Looking at several standards, the errors are as follows
(relative to an arbitrary zero):
60.0kHz:        -1.0Hz
75.0kHz:        -0.8Hz
77.5kHz:        -0.7Hz
162kHz: -0.1Hz
198kHz: +0.1Hz
490kHz (using my LF Tx DDS) +3.0Hz
Plotting this on a graph shows approximately a straight line. This
indicates a progressive error as frequency increases.

The error isn't strictly linear over all frequencies but it is repeatable
so once you know the error for any given frequency it can be allowed for.

The TS850 uses a combination of PLL's and DDS chips to generate all the
internal signals it needs - everything is derived from a single 20 Mhz
reference.  One interesting thing is that the error can be completely
characterised by measuring it at a particular frequency within any 500 Khz
block (e.g. 0 - 500 Khz, 500Khz - 1 Mhz, 1.5 Mhz - 2.0 Mhz etc).  Once
you've measured that error in any block you know it for the same frequency
modulo 500 Khz (ie the offset for say 1.290123 Mhz is the same as the offset
for 290.123 Khz, 790.123 Khz etc).

Another interesting thing is that in many cases the hardware is capable
of tuning closer to the optimum value than the rig's firmware actually
allows!  In other words, the algorithm Kenwood uses to figure out the best
possible values to load into their four (yes four!) 28-bit DDS chips is
not optimized to always tune to the nearest frequency to the one shown
on the display.

Based on Kenwood's block diagram of their synthesizer, I wrote a program
to calculate the actual frequency the rig tunes for any given displayed
frequency - but I had to offer several choices near the *best* frequency
because I never did figure out Kenwood's algorithm.  To find the *actual*
freq the rig's firmware tunes to, I still have to measure the audio output
with a known frequency input.  Fortunately the steps between possible freqs
are big enough that it is is obvious which one they're using given that the
error in my measurements (due to sound card sampling uncertainty) is around
1 mHz.  The worst mis-tune I've ever seen on the TS850 was around 58 mHz,
which would not normally be noticed by HF operators.

Assuming CW mode and 800-Hz tone out, here are the actual offsets (rounded
to 3 decimal places) for your freqs:

Frequency     Best possible tune          Actual tune
-----------------------------------------------------------
 60.000 Khz        799.983            799.960  (40 mHz low)
 75.000 Khz        799.997            799.967  (33 mHz low)
 77.500 Khz        799.999            799.962  (38 mHz low)
162.000 Khz        799.984            799.984  (16 mHz low)
198.000 Khz        800.008            799.972  (28 mHz low)
490.000 Khz        799.983            799.983  (17 mHz low)

I'd love to hear from anyone who may have access to the Kenwood firmware
code so I can make my little program show the "right" frequency every time
instead of saying "it must be one of these possibilities..."  I wrote Kenwood
asking for details of their algorithm but they did not reply to my email.

Bill VE2IQ