During a lot of digital mode QSOs in the beginning of 2013, I regularly got feedback from the other party that my signal seemed to be drifting slightly. PSK31 having a very narrow bandwidth, even a small transceiver drift can be an issue during a QSO. This is usually partially hidden thanks to AFC on the receiver side, but good hams never fail to notice the problem and kindly point it out – at any rate, I was grateful several people did, since it is not always easy to spot this by yourself.
The Elecraft KX3 is a transceiver which uses a Silabs Si570 oscillator. While this is a good quality frequency source, it also drifts significantly depending on its internal temperature. It is part of the design of the chip, which contains a temperature sensor to compensate for this in software. The KX3 therefore is able to compensate for this drift by using an internal table.
Unfortunately, the drift compensation table is specific to individual units, and populating it requires connecting the radio to a known and very stable frequency source, and go over the whole oscillator temperature range. Doing this in factory is not practical as this requires way too long.
But Elecraft being Elecraft, they released a very detailed procedure on how to do this at home, using their XG50 precision frequency source. I ordered it a few weeks ago, and went through the calibration procedure. This involves taking your time, putting your KX3 in the fridge, and using a hair dryer. Did I mention you also need to take things slowly and make sure you have a couple of free hours ahead of you?
If you use digital modes regularly, my advice is that you definitely should go ahead and go through this extended calibration procedure, the results are well worth it. I did not get any negative ‘drift’ comment since I did it, and my waterfall is way more stable on fldigi.
While doing the extended calibration, you should definitely let the KX3 utility run in terminal mode: during the procedure, the KX3 will output the calibration value for each temperature value (ADC connected to the temperature sensor, really), and you can store this table and graph it for future reference in case you ever want to either improve calibration or re-run the procedure at a later stage.
Below is a screenshot of the results on my own KX3, I would be very interested to know what this looks like on your own unit if you calibrate your own unit!
As a bonus, here is a link to the Google Docs document which contains the above chart, if you are interested to go deeper into number.