[time-nuts] Some questions related to time-interval measurements and "modulation domain analysis"

[Bob Camp]

Hi

> On May 27, 2015, at 7:48 AM, Ole Petter Ronningen <opronningen at gmail.com> wrote:
> 
> Hello, all!
> 
> This will be a long post, my apologies..
> 
> A while back I scored a HP E1740A time interval analyser. This unit has a
> 48.8 pS resolution, which is great, but it can only measure up to 3.2uS
> with that resolution, which is a bit limiting. It can also store up to 500K
> readings internally, so it can measure pretty fast.
> 
> While thinking about interesting things to use the instrument for, I
> stumbled across "modulation domain analysis" - which this counter should be
> perfect for. I am just starting to scratch the surface of the topic, but I
> ran a couple of experiment that surprised me. That is, the results are far
> "better", or at least far more sensitive, than I would have expected. So I
> am looking for advice on what I may have done wrong..
> 
> The setup is basically this: a Rohde&Schwartz SMIQ-03B RF generator
> provides the signal. The generator has a decent OCXO reference. The signal
> was a 10Mhz sine at 7dBm, AM modulated at 10KHz, 0.5%(!) modulation. I
> verified the signal on an HP8563A, and observed a nice, low peak 10Khz from
> the carrier, as expected. Repeat on a Rigol DSA-815, same result.
> 
> The e1740A was configured to take 200k time interval measurements of the
> signal on channel 1, paced at 25 edges, triggerlevel 0v. I.e. I should get
> 200K measurements of how much time has elapsed from one rising edge, to the
> 25th following rising edge. These measurements are back-to-back, i.e. the
> stop-edge of one measurements is the start-edge of the next measurement. So
> the counter captured a gap-free time-record totalling 500mS.
> 
> I (slowly!) downloaded the samples to my computer, and basically did an
> FFT. The resulting spectrum showed a clear peak at 10KHz, growing and
> shrinking as the AM modulation index was varied. Only the first 1/5 or so
> of the spectrum gave much information, as the noise grew pretty quickly on
> higher fourier frequencies.
> 
> So here is my first question - since this signal is amplitude modulated,
> the only way this can (should) show up on the counter is if the triggering
> is not "perfect" - that is, it does not trigger at precisely 0v + whatever
> DC offset may or may not be on the signal. Is my undestandig here correct?
> A perfect zero-crossing detector would be immune to AM?

It should be, but nothing is ever perfect. Modulators and detectors both can have issues. 
You *might* have a phase slope on your setup. Anything that rotates one sideband relative
to the other will convert AM  <-> PM.

> 
> If so, this was a very valuable lesson in AM/PM conversion, I think. I
> would not have expected to so clearly see AM as little as 0.5%..
> 
> But there is a second conondrum in my experiment. I also saw an unexpected
> much bigger peak at 64KHz in the spectrum.
> 
> In order to verify my methodology, I repeated the measurement on an Agilent
> 53230A, continous timestamp at 100Khz samplerate, and FFT the result. The
> resulting spectrum was pretty much identical, including the large peak at
> 64KHz. This was also a little surprising, since the resolution of the
> 53230A is a lot better, but I suppose resolution is less important with so
> many datapoints.
> 
> There was also a much bigger peak at 50KHz which I suspect is originating
> inside the 53230A, since it does not show up on the e1740, nor on the
> 8563a. I plan to investigate this spur further.
> 
> I stuck the signal back on the spectrum analyser, to see if the 64Khz
> signal was in fact real, or an artifact of my measurement setup, which I
> was suspecting. Nothing on the DSA-815 nor on the 8563a. Not untill I have
> set RBW 30Hz, videoaveraging 80+ measurements on the 8563a do I see a peak
> at 64Khz from the carrier, at around -80dBc. So the signal seems to really
> be there.
> 
> I suspected a switching power-suppy might be coupling on to my cables, so i
> switched the RG-58 out for double shielded LMR-240, with no difference at
> all in the measurements. The peak was there both on the e1740A and the
> 53230A. So I am lead to believe the 64Khz spur is real, and originates in
> the signal generator, since is is visible on three separate instruments,
> and two separate measurement methods.
> 
> My question is this - how in the world can the counter pick up such a TINY
> signal?? Does this make sense at all?

Sure does. In a sense 1x10^-11 is 110 db down. Yes there is a little bit of 
math involved in the “down from what” part of that statement. 

> 
> I suppose it is possible that the spur originates in my lab somewhere, and
> is in fact quite strong, strong enough to couple into the counters with a
> more reasonable signal - and only appearing to be weak on the 8563a because
> it is better shielded?

It is not unusual to set this kind of thing up running on isolated power and in 
a shielded environment. It’s also fairly normal to spend lots of “quality time” 
switching all sorts of things on and off to figure out where a signal comes from. 

One basic problem is that skin depth essential guarantees that any coax you
can find will have a shield that is pretty poor at switching power supply frequencies. 

Lots of fun!

Bob

> 
> Thanks for any insight!
> 
> Ole
> _______________________________________________
> time-nuts mailing list -- time-nuts at febo.com
> To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> and follow the instructions there.