This photograph shows how I mounted the Collins Mechanical CW filter.  It works, and the band pass as measured with a commercial signal generator is within the filter specs.   I had to play with the matching capacitors on the input and output to
get a decent match, though.  This modification was a waste of time and energy.  Given the R2000 uses 50Hz tuning steps, a filter with a bandwidth of about 160Hz is just too narrow to be useful.  A technical success at best.

R-2000 Filter Commentary

Some additional thoughts on adding "better" filters to the R2000 in order of improvement:

1.  Standard LC filters form the core, but have limited skirts.  That is they suffer from wide bandwidths unless specialized design and construction techniques are used to produce filters narrow enough for serious radio reception.  LC filters suffer from being physically large, for acceptable bandwidths.

2.  Next comes ceramic filters.  A true breakthrough.  Small, roughly 1/"*1/2"*1/2" with very good bandwidths.  Good skirt factor (how fast the bandwidth decreases as you move from the center frequency).  And best of all very cheap.

3.  Then there's the Collins Mechanical Filters.  A few other companies made mechanical filters, but Collins made 99.99% of them.  These filters can be fabricated with very narrow (500Hz) bandwidths, and have very steep skirts.  They are expensive.  The really narrow one can ring with "impulse" (static) noise. They wear out and fail for no obvious reason.

4. The "ultimate" is a crystal filter.  This is most often made by placing a crystal (the same device used to stabilize oscillators) in an electronic bridge so that a very narrow bandwidth can be achieved.  The design is deceptive.  Precision construction techniques, with precision parts must be used.  This is a true art and my few attempts have failed.  Any stray capacitance will allow signals to leak around the filter, degrading its performance.  They can also suffer from ringing in the more narrow bandwidths.

I had a Collins F455-FA-05 6535-526-9494-00 mechanical filter given to me right before I purchased my R2000.  When I got the receiver and a copy of the service manual, I was disappointed to find out that the design engineers at Kenwood "cheated".  They used the CPU to offset the first local oscillator so that one filter could be used for AM-narrow, LSB/USB and CW-wide reception.  To be fair, the R2000, like all devices is a compromise between what we would like to have and what we can afford.  By using a few lines of machine code to do away with a filter and the associated switching, the engineers at Kenwood met their need to control costs while not effecting most users.  After all, the R2000 is not a dream receiver, it was intended to be a good, solid general purpose receiver.  Most people who want really superior SSB/CW reception are not going to bother with a receiver that has 50HZ as the minimum tuning step.

Kenwood does offer a 500Hz CW filter, by the way.  It is called the YG-455C and has a pass band width of 500 Hz (-6 dB), with a center frequency of 455.7 kHz.

I decided there was nothing to loose by adding my Collins filter.  As mentioned, the LO is offset from the normal frequency by +1.7KHz for USB, -1.7KHz for LSB and +0.7 KHz for CW.

As I have described in my modification article for the BFO, I first re-designed the BFO so that USB and LSB could be tuned properly.  To me that means, if you tune to WWV at 10 MHz (during the quiet period), and tune to exactly 10.000.00 MHz, there should be no (or nearly no) tone when the mode is changed from USB to LSB.  Further, if the radio is tuned to 10.000.00 +100Hz then the recovered audio should be exactly 100Hz when measured by a good digital frequency counter.  Similarly, when tuned to 10.000.00 -100KHz in the LSB mode, the audio should also be exactly 100Hz.  When receiving FSK (RTTY or FAX), it is very nice to have a radio that you can just change from USB to LSB (or the reverse) and the frequency just inverts, no retuning is needed.  My BFO modification does that with ease.

Back to the CW filter issue. When tuned to a displayed frequency of 10.000.00 MHz, (always tune from below upwards  to the desired frequency as the last 50 Hz is not shown.  Otherwise, you could end up 50 Hz high if you tune down to the frequency!) and the mode is changed from AM to CW, the IF center frequency is no longer 455 KHz, it is now 455 +.7KHz, or 455.7.  The BFO is tuned to 456.7KHz (in the LSB and CW modes), so the recovered audio will be 1000Hz.  My filter has a center frequency of 455 kHz and a bandwidth of 500 Hz or .5KHz.  The filter only allows RF from (455-0.5 KHz)~~(455+0.5 KHz) or 454.5KHZ~455.5KHz.

But the R2000 center frequency is now 455.7 kHz.  So the center frequency is now shifted toward the edge of my CW filter.  The filter works just fine, but the band pass is just too narrow to be very user friendly.  The 50 Hz tuning steps make tuning very difficult.  I have included a wav file that shows the effect of engaging the CW N filter on WWV and on a CW QSO.  The effect is pretty dramatic.  But it would be a lot more useful if the bandwidths matched up.  I think my bandwidth is about 200 Hz.  It is possible to rewrite the machine code in the CPU, but I have enough projects so I am very unlikely to ever even attempt that type of solution.  A better solution would be to purchase the correct filter from Kenwood.  However at this late date I wonder if they are still available.

I have often thought about adding a Collins Mechanical filter for improved SSB reception.  Before I got the
R2000 I used a military surplus R392/URR.  A 28V "portable" version of the famous R390/R390A.  The R392 was a great receiver.  But it did not have very decent SSB reception.  It had been designed and built back in the days of AM.  The R392 did have a buffered 455 kHz IF output.  So I built an outboard product detector with its own BFO.  Worked great.  I had intended to add the Collins CW filter to that unit, but good fortune allowed me to buy the R2000 instead.  I considered adding a buffered 455 kHz IF output to the R2000 and just use my outboard detector/BFO.  No big deal, a JFET source follower would work great. But somehow I never found time to build something that would defeat the whole concept of a portable, self contained radio that even my wife likes to use.

Bruno Haineault's recent inquiry concerning filters on the R2000 discussion list has got me thinking again. A very dangerous prospect.  The stock filters for the R2000 are not all that bad.  AM-wide is 6 kHz @ -6dB and 18 kHz @ -50dB.  Now, a Collins filter with 4 kHz @ -6dB with 8.5 kHz @ -60dB would be GREAT.  But expensive.  For AM-narrow, CW-wide and SSB the filter is 2.7 kHz @ -6dB with the -50 dB @ 5kHz.  This is not too shabby. The Collins filters for 3.1 kHz (they don't have one for 2.7 kHz) are down 60 dB @ 6.5 slightly more attenuation at a slightly wider frequency.  With the 1st LO +/-1.7 kHz offset, it would be difficult to use a filter with much less bandwidth then 2.4 kHz.  Collins does make one with a bandwidth of 2.1 kHz @ 6 dB points.  This might be a killer filter.  But at a cost of several hundred dollars (US) just to try it and see, it is out of my price range.  I also suspect that with a filter that narrow, human speech would be difficult to understand.  For FSK (RTT and FAX) it would be great, though.  But all and all, the R2000 filters are not too shabby.

Collins filters are pricey.  Several hundred dollars.  And rather delicate.  They wear out and fail for no obvious reason.  OK what else is there?  The original R390 used a crystal filter.  But the R390, R390A and R392 all had several IF stages for gain to make up for the loss introduced by some very well designed, and rather large, LC interstage coupling.  The R2000 just does not have the selectivity ahead of the filter section to really benefit from a crystal filter.  Nearby impulse noise will make it ring like a church bell.  Hams need narrow RF filtering when they design and build their own SSB equipment.  One favorite method uses a crystal lattice array, constructed of commonly available (and cheap!) crystals.  That might
work if 455 kHz crystals were all that common.  I have a very well equipped junk box.  I have been messing around with electronics for over 40 years.  In all that time I have found only one 455 kHz crystal.  Another method might be to stack several inexpensive ceramic filters like are used in the R2000 already.  I bought several CBs and salvaged the ceramic filters.  They all have roughly the same response as the 2.7 kHz filter.  This should give the same bandwidth, with much steeper skirts.  That would reduce the level of nearby interference.  But again, this represents a whole lot of work and will make the R2000 a lot more complex to use.   And again, the R2000 has a few other limitations.  Being fully synthesized, it has a lot more “phase noise” then a R392, for example.  The mixers do not have the dynamic range that a good tube/valve mixer exhibits.  Please don’t get me wrong.  The R2000 is a great receiver.  I love mine and rate it at 5 stars.  But, the sad truth is that the R2000 is a compromise, a well integrated set of subsystems.  Trying to make major improvements in any one area will bring you up against the limitations imposed by other design characteristics.  Mechanical and Crystal filters work best in receivers that already have superior RF characteristics.  The R390, R390A, and R392 are like a Rolls Royce.  They were built for the military where cost was not an important consideration.  The R2000 was designed to be purchased by normal people.

The only modification that I would consider would be one of the filters offered by Kiwa.  They have a good reputation.  I know of several people who have had various receivers “tweaked” at Kiwa and the results have been very good.  For less then $100 (US), it looks possible to make a reasonable improvement in the filtering for the R2000 (or many other radios with a 455 kHz  [or 450 kHz] IF chain).

But after all the schemes, I think I will live with the stock R2000 filters.