About using a vector network analyzer (VNA)

[pgib8]

Hi Everyone.
I'm just starting to learn how to use the VNA but my colleague has been using it for a long time and I'm starting to think he's not doing it right.
He has it set to 1 port and then plugs it into the DUT which is a transmitter, which is also powered up.
So the transmitter (DUT) is sending RF energy into the VNA. Then he looks at the impedance to see if it looks good (while the thing is transmitting).

He didn't know that you had to calibrate the VNA for meaningful measurements, we don't even have a cal kit (I since ordered one). He's also likely exceeding the max power. The transmitter (DUT) has probably about 37dBm and VNA says max 26dBm. I told him he should probably use an attenuator.

The way I see it, with the 1 port setup, the VNA is supposed to send the energy and then monitor how much of it is reflected. Does it even make sense to have the DUT transmitting into the VNA?

[rfeecs]

Most VNAs are tuned receivers.  They mix down to an IF before detecting.  So it is possible to measure with another frequency applied, but it has to be far enough away so it doesn't fall into the IF bandwidth.

You might look up "Hot S22" measurement, where you try to measure S22 of a power amplifier with a large signal coming out of it.

Of course exceeding the maximum input power of the VNA is a bad idea.  Aside from blowing up the VNA, even if you don't exceed the maximum recommended power, too much power can compress the mixers or detectors and give you a bad measurement.

Not calibrating is also obviously a problem.

[xygor]

You should cancel the cal kit order.  The VNA is likely toast.

[JLNY]

Definitely not the right way to do it. As you say, for an S11 measurement, the VNA is acting as the signal source in this scenario. The VNA should really be the only active signal being sent. If you want to look at the output of a transmitter, you would be better off using a power meter or a spectrum analyzer or something. for safety, you might do as you say and include an attenuator in line with your measurement, but you will want to calibrate it out of course.

First off, under NO circumstances should you ever exceed the max input power of the VNA. Just to check, if you leave the VNA port open circuit and set it to measure S11 with correction turned off, you should see a squiggly line at around 0 dB. If the S11 line is closer to -10 or -20 dB, or has big nulls or other weirdness, it is possible that the input is blown. you could also try running the operators checks. you'll have to look up how to do them for your particular model of VNA.

In order to measure the S11, you should be measuring everything AFTER the transmitter, such as any cables, filters, and whatever the transmitter is going to (an antenna, some other part of the system, etc.)

Depending on the type of circuit, it might be possible to get some reasonable S11 data by trying to measure the transmitter with the power off. If there are any connectors, vias, or traces on the board that you suspect are creating high VSWR, you might be able to isolate the issue using TDR gating if your VNA has it.

[pgib8]

Wow, thank you guys so much, that is really helpful.
I checked the squiggly line by setting S11 to 'LOG MAG' and glad to report it's pretty much between 0dB and -2dB (with corrections off). I checked the DUT with a spectrum analyzer (and attenuator) and saw that the output power may only be as high as 28dBm, so only very slightly exceeding the limit of the VNA, which is probably why it still works.
Since he doesn't calibrate it, I guess he's probably just looking at the SWR but still that he plugs the VNA into the antenna port, thus looking towards the amplifier is a bit odd but maybe the readings he gets are somehow still meaningful to him.
Many thanks again for helping me better understand this piece of precision equipment. I'm also learning the Smith chart, very cool and interesting stuff

[LaserSteve]

That's almost as bad as the guy who thought he could cal  for open without a cal kit. He put on a ten foot Rg-58 coax and then held the floating  end in the air by hand.

Steve

[Someone]

That's almost as bad as the guy who thought he could cal  for open without a cal kit. He put on a ten foot Rg-58 coax and then held the floating  end in the air by hand.

Steve

If you're in a quiet RF environment it won't be that bad at all, when calibrating the reference plane to the ends of the cables the Fieldfox VNAs happily calibrate just on the open ends of the cable.

[skipjackrc4]

That's almost as bad as the guy who thought he could cal  for open without a cal kit. He put on a ten foot Rg-58 coax and then held the floating  end in the air by hand.

Steve

If you're in a quiet RF environment it won't be that bad at all, when calibrating the reference plane to the ends of the cables the Fieldfox VNAs happily calibrate just on the open ends of the cable.

The problem is that the connector geometry changes when it's mated to another connector.  The impedance of SMA-style connectors, and any other connector having a threaded jacket that flops around in the breeze when unconnected, will vary wildly with movement.  It needs to be locked down.  At frequencies where RG-58 is a reasonable choice, it may not matter that much, though.

[rfbroadband]

of course one needs to watch out for the max. power transmitted or send into the VNA. What you can do is to setup a sweep frequency range so that the analyzer sweeps very close to the frequency of the transmitter. You then select the internal bandwidth of the analyzer to be very narrow, depending on the analyzer it can be in Hz range. This is a common method to characterize the output impedance of oscillators for example. If the output power is not too high you can 'sweep over' the DUT frequency. The measurement results right at the DUT frequency will be invalid, but with a very narrow RX bandwidth of the analyzer you can get very close to the carrier measure the impedance accurately.

But yes, if you are not careful with the power levels in this setup, you can quickly damage the VNA.   

[rfbroadband]

attached a measurement result where I characterized the output impedance of an oscillator using a very narrow receiver bandwidth. The "invalid points" where frequencies overlap are obvious in the plot.