KX3 Heat Sinks for WSJT and other Digital Modes
The popular Elecraft KX3 is a very flexible, small footprint HF+6m SDR (Software Defined Radio) that can be used without a computer.  However, if one wishes to use the radio for digital modes such as WSJT, some interface between the computer and the radio must be used to isolate audio between them and provide for other computer communication functions.  This is outlined elsewhere.  The purpose of this site is to discuss my experiences with use of the KX3 on the full duty digital mode of JT65A. 

When I first received my KX3 in the summer of 2012, I ran JT65A  for an hour (3w output, 10.0 VDC during transmit), and the OSC temperature peaked at 42C and finally cooled to 39C just before the next transmit period started.  Since my primary interest is to use the KX3 on JT65A mode for 6m EME, good stability essential.  Initially, the drift during transmit and receive periods was too  extreme for reliable JT65A operation on 6m under weak signal conditions.  Elecraft came up with a great temperature compensation routine which became part of the KX3 firmware, and it made the frequency very stable during full duty cycle digital mode operation on HF. 

However, I still noticed a very pronounced frequency jump as things began to warm up at the start of each transmit sequence up on 6m, plus a slow steady drift during the rest of the cycle.   I thought that if the OSC could be tied to a larger mass, it might be less prone to change frequency so readily.  I also wanted to make sure that the KX3 could function well even in higher ambient temperatures than my very cool basement ham shack, since I have tropical DXpediton locations in mind for it.

I found what appeared to be a manageable aluminum heat sink on EBAY that would be large enough for me to put across the top of the KX3 (to cool the PA) and still provide enough material for a heat sink on the bottom case of the KX3 for the OSC.  The particular piece that I chose was offered by seller "goldpart" and described as "100x220x18mm Aluminum Heat Sink for Electronics Computer Electric equipment H168", weighing 472 grams.  It cost $19.63 and was quickly shipped from China.

I cut off a 7.5" long x 1.7" wide piece to mount on top of the KX3 in place of the standard flat heat sink.  I still was able to cut a piece 6" long x 1.45" wide to fit on the rear/bottom of the KX3 to act as a heat sink for the OSC.  The heat sinks were mounted with additional 4-40 x 3/8" long screws with split ring lockwashers and nuts on the inside and two washers and a lockwasher on the surface of the heat sink.   This made the tip of the screw flush with the nut surface on the inside of the KX3.  The only exception was the 4-40 x 3/4" long screw used in the top to pass through the hole in the KXFL3 board.  Silicone "Type Z9" heat sink compound was used between the case and both heat sinks.


Locations for the screws were chosen so they would avoid interfering with any components inside the KX3.  The two existing holes in the rear and four in the top were used, with additional screws added (4 more in the top and 3 in the rear).  AJ7LL was kind enough to donate his time and milling machine to remove sections of heat sink fins so I could fit the screws and washers into the mounting holes. The installed heat sinks are shown.  Note that the heat sink on the rear still permits the KX3 to be supported by the feet.  The heat sink on top of the KX3 is for the final transistors, and the heat sink on the rear is to stabilize the temperature of the oscillator.

 The location of the OSC on the RF board was kindly provided by Elecraft engineering.  In order to couple the OSC to the rear case and heat sink, product called "Gap Pad®" from Bergquist Company was used.  Gap Pad® is a thermally conductive while electrically insulating material that is pliable and somewhat compressible, depending on which particular product is used.  For this project, I chose their Gap Pad®  5000S35 material, which had the highest thermal conductivity.  The pads come in small sheets with a paper backing on one side and a clear backing on the other.  Both sides are somewhat tacky, and are meant to adhere to a case or circuit board.  However, the side with the clear film backing is somewhat stickier.  

The KX3 oscillator is the silver colored device inside the red circle in the photo above.  Also inside the red circle, but just to the lower left of the oscillator is U4, the temperature sensor for the oscillator. 

Immediately to the upper left of the circled oscillator in the above photo, you can see the 200 ohm 1 watt resistor R40 (with the large "201" marked on it).  This is connected in parallel with identical value resistor R96 on the opposite side of the PC board.  The instability problem is caused by current passing through those two resistors during transmit, and rapidly heating them up.  Their proximity to the oscillator causes its temperature to change, creating rapid frequency drift at the  beginning of each transmit cycle.  It was hoped that by better coupling the oscillator to the heat sink, the rapid temperature excursions caused by those two resistors could be minimized.  

The photo below shows the rear cover with the heat sinks installed.  You can see where located the additional screws.  In order to couple the heat sink to the PA transistors, heat sink compound was applied to the case before re-installing the RF board.  

As you can see from the photo above,  piece of .100" thick Gap Pad® was installed on the inside of the case after thoroughly cleaning it with steel wool and degreaser. The stickiest side was placed against the case.  A 1.5"x.875" piece of .100" thick Gap Pad® was placed as shown over the OSC, but positioned so it did not cover any components that were higher off the RF board than the OSC.  When the RF board was re-installed, the gap pads were the proper thickness to connect and bridge the thermal gap between the OSC and the case, with easy compression of the Gap Pad® materials. 

The temperature calibration routine was done again on the KX3 with the heat sinks installed.   Then it was set up as shown in the photo below, transmitting JT65A into a dummy load.  During transmit sequences, the KX3 display showed 10.0 VDC at 1.7A.  After two hours of JT65A operation (in my 10C basement ham shack) at the KX3's three watt output setting, the OSC temperature was peaking at 35C after a few seconds, and rapidly cooling down to 33C (within 5 seconds after ending the transmit sequence).  The PA temperature peaked at 41C and also cooled down to 33C.  Note that the maximum temperatures were much lower than those experienced without the heat sinks, and the coolest temperature was very quickly reached as soon as the transmitting sequence ended.   In order to successfully use the KX3 Temperature Compensation routine, I needed to add heat to the KX3 to get the OSC up to an "ambient" temperature of 35 C for the initial calibration.

The trace of the transmitted signal (JT65A sending "73") still seemed to have a  slight "hook" of a few hertz a during the first few seconds of each JT65A transmitting sequence as the OSC temperature quickly ramps up to 35C.  I suspect there may be a way to correct for this hook when the temperature suddenly increases, by re-running the temperature compensation routine again more slowly and carefully.  However, I did not go any further initially, since the KX3 appeared to be functioning well enough on 6m, at least with strong signals. 

Note that the settings shown on the WSJT SpecJT waterfall below are not typical normal operating conditions, but were set to help attenuate the KX3's tranmit signal, in order to be able to see the most detail.

When I initially added these heat sinks in 2012, I  had not verified the currently configured KX3 performance through actual on-the-air weak signal 6m EME tests, but expected that it will also be able to receive the JT65A signals without drifting since the OSC so quickly returns to its resting cool temperature, before the other station's transmit sequence begins.  There was, however,a quick jump in frequency at the beginning of each transmit sequence, as you can see by looking at the bottom end of each left hand trace.
As you can see from the above waterfall photo, there still was an instability during the initial transmit period of each JT65A sequence.  In fact, during my 2015 6m EME DXpedition to V6M, a number of stations reported excessive drifting and difficulty copying my signals, despite the fact that I seemed to receive them quite well with the KX3.  I suspect there were additional other factors involved that combined to make decoding difficult, and perhaps the KX3 heated up above the highest temperature in the temperature compensation .   However, there is no doubt that there still was frequency instability during transmit, when current flows through R96 and R40, heating up the area next to the oscillator.  Itis my understanding that Elecraft has since changed their main RF boards to avoid this problem, but since I still had one of the first KX3's, I had to fix the problem myself.

With the even narrower bandwidth and shorter operating periods of FT8, I think it will be even more important to improve stability.  The major problem with the instability on 6m was the location of R96 and R40.  With the encouragement and guidance from AJ7LL and PA4K, I removed the offending R40 and R96 resistors (each 200 ohm, 1 watt mounted in parallel on opposite sides of the RF board) and replaced them with an AVX brand flange mount 100 ohm 20w planar resistor (AVX Part #RP60300R0100JNBK) that was directly coupled to the case/final amplifier heat sink.  I found that a 3mm stainless steel metric size screw and nylock nut fit the flange quite neatly, and in my case a 10mm screw was the perfect size to mount the resistor through the case and heat sink on top of the KX3.   As you can see from the photos, a pair of short wires were soldered to the through-plated holes where each of the resistors had been.  In my case, I happened to have a mounting screw from my top  heatsink nearby, and I just stuck the 3mm screw in the old 4-40 hole.   Because I had to rotate the resistor up to avoid touching components on the filter board when it is installed, I also had to nibble out a corner of the front cover so it would properly close without binding on the resistor.
Rather than using a second large .1" thick piece of heat pad mounted over the PC board as I had done before, I decided it would be better to more tightly couple the oscillator to its temperature sensor, U4, and limit its coupling to other components.  As shown in the photos below, I tightly packed heat pad material between U4 and the oscillator, as well as all around U4, taking care not to touch any of the other circuit components.  This packing also brought the height of the heat pad over U4 up to the same level as the oscillator, so another small .1" pad could be placed over top of both devices, coupling them closely.  When the RF circuit board was mounted back into the bottom case, this new small heat pad joined with the larger .1" thick heat pad to couple both the oscillator and its temperature sensor to the rear panel heat sink. 

I did not plan to use internal batteries with my KX3, so I was not going to re-install the battery holders.   In order to make sure that the board was held tightly in place and provided good pressure between the oscillator and the heat sink through the tightly compressed heat pad, separate 2-56 screws, washers and lock washers were used to secure the board where the battery holder screws had been used before.   These are shown in the final photo below in which the filter board is also installed in front of the new AVX 20w resistor. 



After resistor replacement, heat pad modification and re-assembly, the Elecraft Temperature Compensation routine was done again on my KX3.  The initial frequency reference calibration was done with the KX3 at a temperature of 25C.   The frequency correction compensation values were collected and stored so the KX3 frequency will be stable between 12C and 36C.  The oscillator temperature now closely follows the gradual changes in PA temperature, since both large heat sinks are joined via the rear panel cover.   Because of the large heat sinks, it is not anticipated that the oscillator temperature will ever exceed 36C.

For testing, the KX3 was set up transmitting the two tone "shorthand" 73 message in JT65A mode at 5w into a dummy load.  JT65A was again chosen for the testing since its 48 second transmit periods would create the most sustained heating and show any frequency instability due to temperature fluctuations.  Reception of the 50.200 MHz test transmissions by my K3 is shown at right.  The traces are certainly stable enough for FT8 mode on 6m, and a number of contacts have been made using FT8 with my KX3. 

However, more detailed comparisons were done here, comparing the stability of JT65A on 6m between the K3, FT857 and KX3:


Based on the above charts, it appears that there is still a wobbling from one frequency "bin" to another on JT65A mode with the KX3 on 6m.  So, although it is stable enough for FT8 mode on 6m, the reliability on JT65A mode is in questionable.  Of course, the KX3 is now stable enough for all modes on the HF bands.

The use of the Elecraft Temperature Compensation routine that was developed and included in the KX3 firmware, plus the addition of heat sinks and the replacement of R40 and R96 by the flange mounted resistor, appear to assure a stable frequency for the FT8 digital mode on 6m.  Of course, this means that it is also stable on lower frequencies.  However, it does not appear to be stable enough for reliable weak signal communication using JT65A mode on 6m.   Many thanks to the patient and creative folks at Elecraft for their ingenuity with the firmware, and also for their suggestions regarding the magic Bergquist heat transfer products. 

It is not known whether the stock KX3 with only the Temperature Compensation firmware and the resistor modifications alone would provide similar performance.  However, the addition of heat sinks in combination with these modifications certainly solved the stability problem, and also seems to permit higher power to be run without overheating.   There now are a number of commercial heat sinks available in the event the user still needs any additional stability that is not provided through the resistor change alone.  I also understand that Elecraft has introduced a revised RF printed circuit board design in the newer KX3's to address the heating issue with R40 and R96, so if you don't have the original board pictured above, these resistor modifications may not apply to your KX3.
Since this page was originally published, Elecraft has revised its main RF board in the KX3 to provide much better frequency stability. The more recently manufactured units are stable enough to operate JT65A on 6m. However, the new standard mode used for weak signal operation on 6m is Q65-60A, which uses frequency "bins" practically half as wide as in JT65A mode. Therefore, even with the increased stability afforded by the newer KX3, I cannot recommend the use of the rig for Q65 modes on 6m. The stability is certainly adequate for use of any other modes on 6m, and any digital modes on lower frequencies.And since most of the activity now on 6m in on FT8 mode, the KX3 is very well suited for general use on 6m.

Lance Collister, W7GJ
Revised 11 August, 2022