A week or so ago I downloaded the latest version of K1JT's WSPR software. It has a few interesting features including a facility to set your transceiver's frequency using computer control and a way to compensate for any frequency readout inaccuracy - important when you are working within a sub-band only 200Hz wide. The process involves measuring the error at various frequencies by tuning 1.5KHz below stations of known accuracy and measuring the frequency of the heterodyne. The program will then calculate and apply the correct offset for each amateur band. However a better solution if you don't mind opening up the case is to adjust the master oscillator in the transceiver so that the readout is spot-on in the first place.
I did this first with the K3, which was easy because the master oscillator is adjusted digitally using a menu. I then decided to calibrate my FT-817ND. The reference oscillator is accessed by removing the top cover. With the radio facing towards you it is on the left hand side. The trimmer you need to adjust is shown by the arrow.
I don't have the optional TCXO module which is rather expensive, but at shack temperature the standard oscillator seems pretty stable. However, the calibration adjustment is very touchy. The tiniest hint of movement can change the transmitter frequency by as much as 10Hz at 10MHz.
I tuned 1.5KHz below WWV on 10MHz and then adjusted the trimmer so the audio frequency of the heterodyne was measured as 1500.00Hz. Adjustment was an iterative process. The audio frequency might be 1497.35Hz, then I'd nudge it up to 1497.85Hz, then the next time it would go up to 1505.43Hz and I'd have to start nudging it back again. This went on for some time, getting within a fraction of a Hz and then overshooting by several Hz in my attempt to get it spot on and having to start again. I probably spent over an hour on this and was getting a bit frustrated by the end of it, but eventually I got to within 0.15Hz which is probably as good as you can expect with an FT-817ND.
The audio frequency is measured using the computer sound card, which is not a calibrated device, and I wondered if errors in the sound card clock frequency could make this calibration method invalid. I asked about this in the WSPRnet forum and Joe, K1JT replied: "If the crystal oscillator controlling your soundcard's sample rate is off by, say, 100 ppm then your measurement of a 1500 Hz audio tone's frequency will be off by 1500 * 100 / 1000000 = 0.15 Hz. It's a pretty lousy crystal oscillator that is off by more than that. There are many ways to test your soundcard's sample rate. All of them boil down to measuring a supposedly known audio frequency. One good method: put your radio in AM mode and use WSPR's "Measure an audio frequency" button while listening to WWV, during one of the periods when they broadcast a standard tone."
I can't receive WWV that well here and I have never heard it broadcast a continuous audio tone, so for the time being I'll accept Joe's assurance that my soundcard is unlikely to be out by more than the degree of precision that I'm able to adjust the reference oscillator to anyway.