Displaying a pulsed RF signal on a Spectrum Analyser

[deephaven]

I have been trying out the Rigol DSA1030A-TG3 spectrum analyser. I tried displaying the signal from one of my products which generates a 10 ms burst of 434 MHz every 100 ms. The burst of RF happens to be phase modulated with data, but I think that is irrelevant to this particular discussion. I set the analyser to a relatively narrow 2 Mhz scan width to examine the signal. I find that the display beats with the pulse signal so that sometimes I see the signal and other times I don't. I couldn't find a setting which allowed me to see a consistent display. I could get a stable display by choosing an averaging option, but this is requires the results of several scans. Now, I was lucky in some ways that there was a 'beat' between the scan rate of the analyser and the rate of my bursts of signal. If the rates were exactly the same, I could miss seeing the signal altogether.

My question is would you expect this sort of behaviour from all spectrum analysers, or does the architecture of the higher end analysers provide a more reliable indication of a pulsing signal?

I could imagine the Tek MDO scope could well fair better in this test as you could trigger the spectrum display off the time domain display.

[KJDS]

You need a spectrum analyzer with gated measurement capability. Most of those from Agilent in the last 20 years, and other decent test equipment companies have this, though often as options.

[barnacle2k]

Many Spectrum analyzers offer Trigger / gate inputs that can be used.

If the Rigol doesn't offer this, you can try using the max-hold setting to get you the spectral outline of your signal. (requiring many sweeps)
By the way: Averaging is the WRONG option here, it gives you false readings since there will be some sweeps with no signal that will be averaged with the ones that do which makes your signal power reading lower then they are.

There is also a different kind of Spectrum analyzer: called "Real Time Spectrum Analyzer" that uses broadband FFT to show you the Spectrum vs. Time, aka waterfall diagrams, and also offer myriads of trigger options.
The Tek MDOs use that way for their Spectrum Analysis functions.

Edit: Seems like the Rigol does have an external trigger input on the back.

[G0HZU]

Because your OOK rate is so slow I think you may get the best performance by simply downconverting your 434MHz signal to about 20kHz and then looking at the signal on a PC soundcard using some decent spectrum analyser software.

You could then capture several seconds of pulsed and modulated signal as a wav file and then post process it as you wish.

But obviously this will only work for narrowband signals so a lot depends on the characteristics of your phase modulation.

[deephaven]

Thanks for all the comments. The main purpose for me getting an SA is for checking for unwanted radiated signals from an EMC perspective. I used my 434 MHz pulser as a test source to see how the SA would perform. I want to make sure that I don't inadvertently miss spuriously radiated signals purely due to my measurement method. Would a high end SA be just as likely to miss a pulsing signal due to the fundamental requirement to scan through a range of frequencies which obviously takes a finite time?

[nctnico]

I wouldn't be too much worried about pulsed emissions. During an EMC test the results are averaged over a certain time so pulsed emissions are dampened a lot in the result of the measurement. This is also why spread spectrum clocking 'helps' to pass an EMC test.

See:
http://www.adler-instrumentos.es/imagenes_web/productos/03_RF_emission_testing_e_2.pdf

[deephaven]

I wouldn't be too much worried about pulsed emissions. During an EMC test the results are averaged over a certain time so pulsed emissions are dampened a lot in the result of the measurement. This is also why spread spectrum clocking 'helps' to pass an EMC test.

See:
http://www.adler-instrumentos.es/imagenes_web/productos/03_RF_emission_testing_e_2.pdf

That's a very useful document, many thanks.

[barnacle2k]

Any traditional scanning SA will have the same limitations, no matter the price.
Even though higher scanning speeds yields higher probability of intercept.

Your instrument is perfectly capable of analyzing 433mhz transmitters.
The correct way here would be to use the max-hold function an scan for a very long time, that will reveal spurious emissions.
You don't have to use the trigger input for this but it would minimize the amount of false readings.
(other emissions getting into your test setup, which are not caused by your circuit)
To isolate these you can run the max-hold capture without the transmitter powered and save that spectrum for comparison.
By the way you should also check for harmonic emissions to least the 3rd (1.3ghz) better the 5th (2.1 ghz).

[deephaven]

That max-hold function sounds like a useful way for capturing those elusive pulses. In my particular setup, there is a beat between the SA scan rate and my pulses, so I do see the signal sometimes. I've only had a basic cheap SA up to now, so I guess I have some learning to do to get the most out of the Rigol.

[T3sl4co1l]

There are two quantities that matter: peak and quasi-peak.  Peak is simply peak, ever, and observes the worst case output of your system.  This is important to EMC, as peaky signals cause impulsive interference with radio systems, particularly AM type communications.  Quasi-peak is a strange one, but it has the same rationale: a fast rise time and a slow decay time, suggestive of the detector or AVC circuits in radios.  The decay time constant is something long like 250ms, so that a pulse repeating less than 250ms (i.e., >4Hz) will largely read as peak, even though its average value may be low.  The attack time (250us or 1ms, something around there -- I forget exactly) means that pulse durations substantially shorter than this will read as less than peak.

Tim

[deephaven]

Thanks for your comments, Tim. The Rigol has a quasi-peak function so I should be able to try that.

[Hugoneus]

Keep in mind that something like the Tek RSA USB spectrum analyzer has a 100% probably of intercept for a particular pulse width (100us) (Within its digital IF bandwidth). You can see this in my video review of it.

[deephaven]

Keep in mind that something like the Tek RSA USB spectrum analyzer has a 100% probably of intercept for a particular pulse width (Within its digital IF bandwidth).

Doesn't this have shortcomings in other areas like spurious responses?

[Hugoneus]

Keep in mind that something like the Tek RSA USB spectrum analyzer has a 100% probably of intercept for a particular pulse width (Within its digital IF bandwidth).

Doesn't this have shortcomings in other areas like spurious responses?

Yes, the scan rate is much slower, but within the IF bandwidth (40MHz) it is very fast, as I said, it can capture any event which is at least 100us in duration with 100% probably of intercept.

So if you are looking for spurs whose frequency you don't know, it would be harder to capture them without a gated spectrum analyzer.