On 10/16/20 4:55 AM, Bob M. wrote:

If you put a 10dB pad on the CH0/Port1 and then try to calibrate it, the "open" won't be an open circuit, it'll be whatever parallal resistance the pad exhibits to ground. Similarly the "short" won't be a short circuit; it'll be whatever series resistance is in the pad. A 50 ohm termination should cause the pad to produce something that resembles 50 ohms.

Actually, not quite..

The pad is no different than losses *inside* the VNA - in the reflection bridge for instance. Taking the Open as an example, the calibration process "measures" the response when the connector is open. With the 10 dB pad, and a perfect bridge, the ratio between the stimulus and the reflected signal is -20dB. So that now becomes set to a calibrated value of 0dB (with the phase as measured).

The whole point of the calibration process is to take all these factors out mathematically.

Consider that reflection bridge on Port 1 - There's a stimulus signal, the UUT port, and a measurement port to a receiver that is (mostly) the reflected power from the UUT. And, either a power divider for the stimulus, or a forward power port on the bridge (if you're using directional couplers, for instance).

The bridge and power divider aren't ideal. The stimulus doesn't put out constant power. Not all the reflected power shows up at the measurement port. But you can measure the (complex) ratio between the forward signal and the reflected signal.

And then, you put some known UUTs out there - a 0 degree reflection, a 180 degree reflection, and a no reflection. You now can solve for the gain and leakage terms in your bridge/divider.

This is no different than calibrating with the standards at the end of a piece of coax. The loss and phase shift of the coax gets taken into account in the calibration.

When trying to measure the input return loss of the preamp, the signal coming out of the NanoVNA is reduced by 10dB. The reflected signal due to the RL of the preamp is also attenuated by 10dB on its way back to the NanoVNA. So you end up with a RL that's already 20dB, not the actual preamp's RL.

If you've calibrated the NanoVNA with the 10 dB pad in, that attenuation is already factored into the reading. In VNA speak, you've moved the "reference plane" of the measurement outside the attenuator. That's the beauty of VNA calibration.

Perhaps adding some attenuation to the outgoing NanoVNA signal could be compensated for by increasing the sensitivity to the incoming NanoVNA signal, to deal with the SNR issue. I didn't have such a problem using a SA/TG/RLB at -30dBm coming out of the TG.

The tracking generator spectrum analyzer approach (or a sweeper and oscilloscope with a detector) doesn't have the mathematical calibration process of the VNA. The key difference is that the typical sweeper scheme is that you don't get a phase measurement, so you can't fully calibrate the effects of the bridges and test set. There *are* amplitude only VNAs, but they require a bit more sophistication in the math and implementation - look up 6-port analyzers.

Once, though, you can make phase measurements, and do the underlying stepped or swept measurement automatically, and record the results, and then do the math, now you can do real Vector Network Analysis.

That was the huge change when VNAs were first invented - the math provided a systematic way to manage all those things, by creating a model of the losses and leakages, and then solving for them using measurements on known standards. It's really something that requires a computer to do (yes, you can do it by hand, but boy is it tedious - typically by hand, you'd do the measurements and calculations for one frequency.