For me calibration is about trust. If I measure something I want to trust the result to be as good as I think. To know that, I check (calibrate) my instruments or references against something else that I trust (that is know how accurate it is or said in another way know the measurement uncertainty). If I know someone else has promised how accurate it is and I trust the source I can use that. That is if I buy something from a manufacturer with a specification I can trust it if I trust the manufacturer. It will of course be a small risk my unit is bad for some reason. But this is also true if I calibrate something (by myself or external). Almost all external calibration use a confidence of 95% for that reason. That is, it is 95% chance it meets the uncertainty specification.
For the Isotemp 143 OCXO we have a datasheet from a reliable manufacturer but the problem is the source (used from eBay). Can we trust it?? If I bought it new I could be quite sure it met the spec of slightly better than 1ppm (1uHz/Hz). If that is enough for me I would trust it for further measurements like checking frequency counters. If I need better than the spec I must check it against a better source.
For me the best cheap source is a GPS module as the uBlox NEO6M, LEA-6T or similar and a 10MHz OCXO. As Gyro says in reply #2 you just need to divide down the 10MHz and compare it to the output of the GPS module with an oscilloscope. If you have just the 1PPS I would recommend a digital oscilloscope that you can trigger on the 1PPS. I have done this with just the PPS out from ublox modules and it works well if you divide down the 10MHz at least 10 times. Dividing down 16 times also works well (the output of the divider needs to be an integer of 1Hz).
Using radio or TV signals also worked for me but you need to be sure the transmitter really are locked to eg GPS. I learned this the hard way before I went to only GPS. Another problem I experienced is radio transmitter closing down in a quite short time. For the TV the change from analog to digital made the before long term stable signals almost useless.
After having used GPS to calibrate and adjust my 10MHz references I decided to build a GPSDO as I was tired of adjusting (and I had an Arduino I wanted to learn to program). With a GPSDO I don´t have to adjust or know the error. Of course a GPSDO is not perfect and you need to know the measurement uncertainty even for your GPSDO see for example:
Two links about GPSDO´s as frequency standards and traceable calibration:
http://tf.nist.gov/general/pdf/2289.pdf http://www.gps.gov/cgsic/meetings/2016/gust.pdfIf you need 1ppb (1E-9 or 0.001ppm) most GPSDO with OCXO´s if showing locked will be ok. But if you need more with good confidence you probably have to understand the specs, if they say enough ? or try to understand the TIC and DAC values to get a feel for the stability.
Note that ADEV is stability and not measurement uncertainty. But as the GPS is locked to a much more accurate reference ADEV can be used to estimate the uncertainty. My own rule of thumb is that the expanded uncertainty (95% confidence) is 2-10 times the worst ADEV at Taus longer than the gate time. To be conservative I normally use 10 times.
Lars
?