[time-nuts] Phase noise measurement with a scope

[Magnus Danielson]

On 06/13/2013 11:02 PM, Stefan Heinzmann wrote:
> Magnus Danielson wrote:
>> On 06/13/2013 04:26 PM, Stefan Heinzmann wrote:
>>> Azelio Boriani wrote:
>>>> The problem with sampling 'scopes is that you cannot get a continuos
>>>> samples stream. I think that the TimePod correlates continuously in
>>>> time.
>>> Does that matter for phase noise measurements? Doesn't that just make
>>> the measurement take correspondingly more time?
>>
>> It matters a lot, since the length of memory will limit the how close
>> in you can do it. You can naturally make multiple runs, and that's
>> what the TimePod do, but with decimations done in realtime by the
>> firmware.
>>
>> You can do it with high speed ADCs, but it won't bee cost efficient
>> and it will cost you in speed, as you need to do much in software
>> processing to get there.
>>
>> The TimePod is in that context a fairly well balanced design as in
>> bang for the buck. Another aspect I like is that it can do pretty neat
>> long-term measurements.
>
> That's of course true when you want to build or buy an instrument just
> for this job. I don't question the Timepod, on the contrary I think it
> is a very good instrument. My aim was rather to find another use for a
> scope that's already there (it isn't in my case yet, but will be).
>
> The R&S RTO seems to have a few unusual capabilities for a scope, which
> might help here. It does seem to do decimation and a number of other
> math functions in hardware and in real-time. Looking at their
> description of the I/Q option reminded me of the Timepod manual,
> specifically the block diagram in there, and brought me to the question
> I'm asking here. Have a look if you're interested:
> http://www.rohde-schwarz.de/file/1TD01_0e_RTO_IQ_Software_Interface.pdf
>
> While the scope may not be able to continuously acquire and
> cross-correlate, with no dead time, I would think it capable of taking
> fairly long shots by storing only the decimated data. There's a chance
> of it being suitable for phase noise down to 10 Hz from the carrier, I
> think. Perhaps closer than that. That would already be quite useful, I'd
> say.

Now, that would indeed be very handy.

It's not that you can use todays realtime sampling scopes and use two 
channels for cross-correlation, it will work, it will just take many 
runs to get anywhere. Cross-correlation is pretty cheap if done via FFT, 
and wrapping FFTW to do it only takes about 20-30 lines of code.

Having a scope able to do decimation up-front is quite similar to what 
the TimePod do, and when quickly browsing the manual it looks useful.

Cheers,
Magnus