[Elecraft] re: K2 crystal grounding and blowby.

[Ferguson, Kevin]

Well, the subject has come up again.

As the guilty party that got this whole thing started,
allow me to give the history of the mod:

My previous rig was a TS-520 that leaked IF through the SSB
filter when the CW filter was selected...after fixing that,
I felt the K2 was suffering from a similar problem, and set
about to see if I could fix it.  When I came up with the mod, 
my idea was to improve the inter-crystal sheilding as much as I could
easilly do.

I did this by adding a few wires to the board, and using the
second set of grounding pads, and adding a second ground wire
to the top of the crystal cans, wrapping the ground leads so that
the wires would not mechanically interfere with the SSB
option board.


I also added some additional bypass capaciters to the variactor
bias resistor network, and some grounded wires laying on the 
underside of the board.  These changes made a notable improvement to 
the filters ultimate rejection, as tested by  my ears.  I was delighted
with the result and wrote up the mod for others to enjoy.

Then all heck broke loose.  Some hams tried the mod and noted a similar
improvement....notably Fred, KT5X, an accomplished contester. He made
measurments using spectrogram that confirmed his and my own "by ear"
estimates of the improvement. I compared the improvement to the difference
between good NR analog tape, and CD audio. The tape can be really good, but
the CD will be notably better. 

Fred is an FT-1000 MP owner, and his rig is equipped with inrad filters. He
has compared the performance of the modified K2 filter, and was impressed.
Spectrogram revealed that the inrad filters have a slightly better shape
factor as I recall...However the modified K2 gives a lower noise floor...and
of course the K2 does not offer the option of adding
DSP filtering to that provided by the IF filters.

So why all the fury?...well, some hams noticed NO improvement. And a huge
debate insued.  In at least one case, (possibly John KI6WX ??) it turned out
that the builder had grounded the base of the crystals when the rig was
built. Also note that this mod improves the filter performance at a point
where it is already providing about 35-40 dB of attenuation.  You aren't
going to hear a difference if there is a passband signal that is pulling the
AGC down by that much, and you may well not hear a difference if you don't
have good hearing. If you have less than 40 dB of dynamic range left outside
the CW passbands, you CAN'T possibly detect the improvement.  Stock, _I_ can
hear ~S7 or stronger signals on the opposite side of zero. (other sideband)
Modified _I_ can't. Unmodified, I can tune through strong signals and still
hear them up to the frequency limitations of my hearing. Modified the
amplitude fades to undetectability long before a frequency limitation is
detected.

Furthermore, when measuring with spectrogram, it is easy to overdrive your
sound card, causing distortion products which will mask any improvement.

Enter John, KI6WX who has access to some excellent test equipment. His
measurements indicated that, up to the limits of his equipment, (~100 dB
IIRC) ALL the improvement was due to improved crystal grounding, and
grounding the crystals at the base yielded superiour blowby performance.

My rig currently has all the CW filter crystals grounded on both sides at
thier bases, using short leads. In addition, I shortened the ground leads on
the 2nd IF filter, as well as the SSB filter on the option board.  The wider
filters do NOT allow the blowby to be as easilly heard as the narrow CW
filter, however, I figure every little bit helps. 

I have not removed the capacitors nor the shield wires.  It has been
mis-stated that these were shown to be ineffective. A correct statement
would be that no improvement could be measured. An equally correct statement
would be that no degradation could be measured.  

For those who doubt the efficacy of improved crystal grounding: It is a FACT
that ultimate rejection is NOT established by the "intended" circuit design.
You can't analyze the circuit schematic and predect what it will be.  It is
all in the physical implimentation.  Parasitic terms, and stray coupling are
the culprits. Of course some really industrious soul might be able to model
this. Not me.

So how does the mod improve ultimate rejection? Can't prove any of this,
but here is what my experience tells me:

My thinking was that doubling up the ground wires should halve the
reactance between the cans and ground...thus yealding a ~6 dB reduction
in blowby at each crystal. Don't know if I was correct on the reason
behind it, but the expected improvement was (is) real. Another VERY
effective way to reduce inductance is by shortening the wires.  Grounding
the crystals at the base reduces the lead length by much more than half.
Adding a second such wire gives a further halving.  It should be possible to
reduce the inductance between the crystal case and ground to about 1/10 that
provided by the instructions. That should result in about 100 dB reduction
in the coupling between the cans. Does this improve ultimate rejection by
100 dB? NO! there are other ways signal can sneak past the filter. As you
plug the big leaks, the smaller ones contribute a greater share.  Also, 100
dB is about the dynamic range limit of good test equipment.  Improvement
beyond that point won't be measureable. Human hearing has similar
limitations...so you probably can't hear any difference between 100 and
120dB ultimate attenuation.

Also, I have finally figued out one thing that at first puzzled Fred and
myself.  On both CW and rev CW the mod improved the low frequency skirt more
than the high frequency skirt, as measured using spectrogram.  The reason is
that the blow-by allows the noise from the opposite sideband to 
alias into the lower skirt more than the upper skirt, so there is more
improvement to be had on that skirt than on the high frequency side.  This
is becuase the outer skirts still fall off at ~6dB/decade without the mod.
(modified they are at around 30 dB/decade, continuing the slope of the
intended response)

-73-
ko0b