# VNA calibration using cal kit vs PCB

So, its my understanding that the VNA (vector network analyzer) is designed to inject a known signal into an unknown device and analyze the reflected signal's amplitude and phase (both voltage and current). But it is also able to remove the unwanted reflections, delays, attenuations, etc of the test cable, connectors and so forth.

This is accomplished with the well-known calibration sequence in which the VNA samples what an open, load, short and through "look like" through the test cable, connectors, etc, so that it can effectively subtract the measurement errors they cause.

Am I right so far?

My question is, which is better, to use a precision cal kit (and then make electrical length adjustments), or make my own right on the PC board- actually 5 PC boards 4 for the SOLT cal, and the 5th of course for the actual measurements?

It seems to me a precision cal kit is both expensive and NOT what the test circuit will look like. I'm certain I'm missing something, but what?

This question was somewhat answered here, but what I don't understand is even if I get a pristine calibration with a good standard, I'm still going to have to connect to the circuit using some sort of pigtail that injects its own problems.

• Do you want to characterize a device that's mounted on the PCB, or do you want to characterize the whole assembly including the PCB and the devices on it? – The Photon Jan 31 '17 at 0:44

If you want to make your own cal standards on a PCB, consider using a "Through-Reflect-Line" (TRL) calibration rather than SOLT.

To achieve the kind of precision you're used to with SOLT, you not only need very good standards of each type, you also need to be able to characterize just how each standard differs from its ideal type. For example, how much excess capacitance is present in the Open standard, and how does it vary with frequency.

TRL cal is less dependent on the standards used being very close their ideal type. In particular, the Reflect standard just has to have a high magnitude of reflection; its phase doesn't have to be well known (as compared to the Open and Short standards in SOLT which must have very close to 0 and $\pi$ phase in their reflection coefficients).

A drawback of the TRL cal is that one set of standards is only valid over a limited frequency band (less than one octave, IIRC, but it's been a long time since I've worked with TRL cals).

The VNA has a sweep generator and two inputs. measurements are calculated for different setups that compare input or output reflection ratios to reference or output to input ratios. By performing calibration with an external test jig, using open, 50 Ohm and short circuit tests, then the device inserted in the fixture can be compared to the references with the fixture and external cable errors nulled out as much as possible..

The accuracy of resulting measurements is limited by the accuracy of the references. using pigtails will not likely give repeatable or excellent return loss. Precision calkits with 60 dB accuracy are like resistors with 1 ppm error not 1%. Also the machining of the coaxial ratios is done to the same precision so that it has this Zo=sqrt(L/C) ratio characteristic impedance due to the machining of inner to outer ratios. This is why they are so expensive with micro-machine precision for diameters and gaps.

You may not need this precision if your tester design uses pigtails.

You should explain or show more detail. You might want to consider using coaxial connectors or hardwired semi-rigid coax crimped SMA's on the fixture. More details on cal-kits