It's in my opinion the dark side of the project because it's a bit difficult to understand for newbies.
There is several ways to calibrate the radio and not easy to understand.
Finally, as you know Here, there is a WSPR beacon function inside the uBitx. So I red the blog post of KD8CEC and decided to try it with WSPR reports.
Here is the link of the post of KD8CEC : KD8CEC post
So, for making some calibration, there is a formula :
875000000 / Test Frequency * Error Hz ⇒ Calibration value
(Before Calibration Value + Adjust Calibration value = new Calibration value)
(Please read the whole article of KD8CEC, it's well explained).
So, I've tried this this a spreadsheet to make the calculations with the mean frequencies and you will find this in the DropBox folder of the whole project with the filename Master calibration.ods in the Tests folder.
It seems to work but the new calculated value is like a range which is getting narrower the more you test.
After 5 - 6 tests, I was able to have something satisfying :
Finally, we are close to the factory setting, but the behaviour is a bit weird. I guess maybe I didn't let the uBitx radio to heat up longer before making the calibration. I've red elsewhere that you need to turn on your radio at least 30 minutes before trying to calibrate. Maybe it's that.
However, I think this method is quite simple and easy to measure by the community of WSPR spotters. Another useful thing with WSPR !
After the conclusion of the calibration, I wanted to test if the heating of the electronics could have a noticable impact upon the deviation of the frequency.
So, I've tested my Ubitx with a WSPR spotting just after a cold boot (after at least one day without turning it on) and after a while, at least 30 minutes.
The results are in the Test folder, with the name WSPR deviation heating up.ods.
After a cold start :
After at least 30 minutes, without changing everything else :
The result is much better !!
Let's try after more than one hour :
OK, it's not better. Maybe there is not enough values to validate the results but it seems that there no improvements after 35 minutes.
The point is that you need after all to wait before making some calibration to be sure to have right values and not fluctuing measures.