[time-nuts] Spectrum Analyzer Specifications

[timenut at metachaos.net]

Charles,

>>When using a spectrum analyzer to examine the output of a frequency 
>>standard, what are the minimum specification needed? Bandwidth, 
>>resolution, sensitivity, etc?

> It very much depends on what you hope to see with it.  That said, 
> virtually no SA will do anything useful in this application without a 
> fair amount of preprocessing.

> . . .

> I am all for having one or several SAs around the lab, but I fear
> that an SA may not be very useful for the particular tasks you are 
> hoping it will perform.

Being new to this, I don't actually know what I need to see with it. However,
I see charts that show the "spurs" on the output of an oscillator. I assume
that I will want to do similar tasks. As I understand it, those are the main
problem when multiplying the frequency by very much because they get
multiplied as well. Thus, one of the differences between a "good" OCXO and a
"bad" OCXO.

> Pay attention to ergonomics -- suffering with an SA that doesn't
> think like you do will drive you crazy.  For example, I prefer 
> HP/Agilent SAs over Tek SAs because the HPs allow you to specify a 
> start and stop frequency *or* a center frequency and span, while the 
> Teks only allow you to specify a center frequency and span.

As an example, I currently have a broken MTI 260 which I am repairing. I
currently have it cleanly disassembled so that it can be reassembled with no
damage. I know what needs to be done to fix it (at least the first pass) and
am waiting on a tool that I need. But, if I succeed in repairing it, what
test equipment do I need to determine the quality of output?

I can measure frequency to a point (7 digits, uncalibrated - 8 once I have a
frequency standard). That is not enough to determine very small deviations
from 10Mhz, but it will tell me that I have a sine wave at approximately the
correct frequency. The last one I (sort of) repaired, gave me a sine wave but
with a duty cycle that wasn't exactly 50-50. So there must be some distortion
present. Not surprising in that case because of the hacking that I did.

However, I understand that the quality of the OCXOs can vary widely. What
tools do I need to determine the quality of output? I assumed that a spectrum
analyzer would be a first step - to look for large spurs, overtones and
consistent noise (e.g. 120 kHz components) at unusual frequencies. Even jitter
and phase noise should show up that way - at least with sufficient resolution,
which might be unrealistic.


Mike