looking for instrument that can do 2 channel phase comparison

Just a fast oscilloscope will do it.

Or two mixers with DC outputs, a 90 degree delay line, and a slow scope.

I do have a ghz scope and i can get a decent signal for higher level signals, but my signals are weaker and i am not getting a good result, however, maybe i can try amplifying the signals and see if my scope will work.

MiniCircuits makes some nice little mmic-in-a-box things, 20 dB or so gain.

What signal levels do you expect?

My power levels are about -20 dbm. I just put two channels of the scope at 1 ghz direct on rf sig gen and got levels up high enough for scope, but th ere is way too much phase error noise between two channels when using my sc ope. I took single shot measurement of both channels looking at exact same signal and phase error on each individual measurement varied by like 15 de grees. So i dont think direct measurement will work.

You need signal averaging, which is tricky given the pulse-modulated RF bursts. We need some goofy ideas here.

I assume that the 500 ns modulation pulse is incoherent to the RF. If it was coherent, it would be a nice scope trigger for averaging.

You could trigger off one channel and still do averaging, maybe.

An XY (lissajous) plot would be interesting, triggered or incoherent. You could do some eyeball estimation of the elipse angle.

You are correct thatthe pulse edge is not coherent. I cannot argue my way through averaging the noise out. I have to give these calibration procedures to others to use.

Sounds like you need splitters and mixers and a slow scope. Analyze the I/Q baseband pulses.

Radar people must have this issue all the time. I wonder what they do.

Our LeCroy 7 GHz scope does some radical post-acquisition processing, like synthetic data recovery PLLs and such. Maybe call them and see what they have.

How about comparing complex FFTs?

Did someone ask for goody ideas?

I assume that since you're talking about a signal generator that you have the originating signal. In that case, obtain that signal and its quadrature (which may or may not come out of the signal generator; if it doesn't then you need that 90 degree phase shifter).

Buffer, amplify, and whatnot the parent signal and it's quadrature version, and use it to drive a pair of quadrature mixers (MiniCircuits has these already packaged with connectors, so you can just rope them together. They may even have quadrature mixers so you don't have to mess with your own 90-degree shifter). Then look at the (four) outputs from the mixers at DC (well, at whatever counts for DC when your pulses are

Low-pass filter the mixer outputs as appropriate -- I'd start with a filter with a time constant of 500ns/3, but you'll want to adjust that up or down once you see how it works.

You'll probably have to account for DC offset, but it'll be right there on the scope. The phase at each mixer output will be the arctan of the ration of the DC shifts from "no signal" to "pulse end" (use whatever version of atan2 your number-crunching software provides). The phase _difference_ will be the difference between the two phases.

Keep an eye on coax lengths -- at 1GHz a two-meter length of coax is good for 360 degrees of phase shift; that works out to about 6mm per degree.

With the O-scope you probably can't do much better than 5 or 10 degrees. If you were willing to play games with samplers and maybe ADCs, and if you knew just when the pulse was happening, you could be limited by noise instead of by the scope.

If the absolute phases of the two inputs track the RF source accurately from pulse to pulse, this isn't too hard: use some Mini Circuits stuff to generate a LO and mix both of them down to some convenient IF. Triggering off another channel set up to generate the IF center frequency will do the trick.

If you don't have phase coherence like that (e.g. in an old-style radar where the oscillator built up from noise on each pulse), it's harder.

Cheers

Phil Hobbs

Thanks all for good suggestions. I got company to purchase 2k worth of min i circuit stuff and will do it that way. This thread got me to especially explore direct scope readings, and given that our scope is right on edge of 1 ghz it is too jittery in phase. Nobody could identify a specific commer cial insrrument. I saw some stuff at agilent that might work but at a mini mum it required a sw learning curve plus a longer lead time and very expens ive.

I dont think i will take it down to baseband i/q. I will mix to 10 mhz and see directly on two channel scope the phase difference. It does not need to be automatic, so at this stage i prefer seeing the two sine waves

This might work with your scope:

Use the scope to snapshot the RF pulse on the two channels. Export the samples to a computer.

Multiply/integrate the sets against one another, and plot the integrated product vs time shift between sets, in, say, 1 ps steps. Look for the zero crossing. Or do two product sets, one shifted 90 degrees (250 ps) and do the trig.

You could apply an identical 1 GHz signal to both channels to calibrate out any channel-channel phase shift.

What MiniCircuits setup do you plan to use? Heterodyne to some low freq, or baseband cross products?

That property could be used somehow. If nothing else it makes it easy to get a phase shift. Concievably a series of such cables giving delays could enhance the accuracy.

on

good

How many degrees for 2 m of coax did you say? 2000 or so? About 3 mm/degree in free space.

?-)

I thought the wavelength (360 degrees) of 1GHz in free space was C/f or about 300mm. That would be about .8mm, no?

The rule of thumb is 1 ns per foot. The time for 1 wavelength at 1 ghz is 1 ns, so wavelength is about 1 foot

So, how many mm/degree?

Sorry, I meant .8mm/degree. Yes/No?