VNA : accuracy without a calibration kit

[imett]

kit de calibration vna pcb

Hi to everyone,
I know that for a vna (our is E5071 from agilent) calibration is required but sfortunatly i don't have a kit (agilent don't provide it with a vna!) , my measure is still good? What's the uncertainty?I've to buy absolutly a calibration kit ?
Thanks in advance

Matteo

 

[radiohead]

build your own vna

No, without calibration the measurement accuracy is lower, but still pretty good. It all depends on the performance of the VNA like port to port isolation and port return loss. You can check that yourself. Usually the datasheet also states both calibrated and uncalibrated performance. For most applications you don't need the calibrated accuracy.

But you can help yourself out by making a calibration kit yourself. Solder a short to a connector: your short, leave a connector open (without pins of course to prevent radiation): your open and use any 50 ohm termination for load. I know it is not perfect, and you don't have a good definition of your calkit, but at least you get rid of the cable losses and delay. Just use any SOLT calkit that resembles yours and your error will be reduced with another order of magnitude.

Helpful?

 

[kspalla]

homemade vna calibration

With out calibration the measurement is meaningless. just a trash

 

[Mazz]

Matteo

forget about accuracy without calibration, especially if freq is in the GHz range.
S parameter measurements are always referred to a measurement plane (that is defined by calibration).
I agree with radiohead only if you measure insertion or return loss.
You can built a home made cal kit, as suggested by radiohead to get some accuracy, but for low freq (below 1 GHz).

If your DUT is an SMD component, you can built a set of PCBs to make a TRL calibration with a good accuracy at very low cost (check agilent web pages for application notes related to TRL).

Hope it can help.

Mazz

 

[vfone]

As others mentioned, for low frequencies (~1GHz) you can make your own Short, Open, Through and Load, using standard SMA connectors.

The most limited in frequency from all these 4 elements is the Load element, because you do not have the special volume-type resistors (very low parasitic inductance) used by Agilent.

You can do a trick here to minimize the parasitic inductance of your own Load, placing three 150 ohms SMD resistors (0402) in parallel, and solder them carefully from the central pin of the SMA to the ground at 120 degrees. Using a clean soldering I managed to reach 2.5GHz equivalent calibration.

 

[Mazz]

It's becoming an interesting discussion.

The parallel SMD resistor is a good practice, but, to reach the 2.5GHz you need to consider also a raw model (C0 &L0 at least) for open and short, that will be difficult to have. Any suggestion on that?

A TRL (I have made few with very good results up to 3 GHz with standard PCB tecnology) is quite cheap although is narrow band and I don't know if it is supported by E5071. You move the difficulties from making an ACCURATE load/short/open to make SIMILAR TX lines.

Mazz

 

[wangyufn]

For TRL calibration, you can build different delay lines for different bands. It will be ok if you control the phase delay in the certain range, e.g. 30~150 deg. So TRL is not a narrow band technique.

TRL can be more accurate than SLOT, especially for high frequency. And the most important, it is less dependent on the accuracy of 50 Ohm load.

 

[Mazz]

wangyufn, you are right, TRL has its max accuracy at freq where line and thru are 90 deg.

for sure is narrowband respect to SOLT, especially if you want to go down in freq.

Mazz

 

[radiohead]

I agree with you guys that to have very good accuracy in the GHz range, you should have a nice and decent calibration kit.

* In coax this will get you absolutely accurate measurement up to the end facets of the cables. But how will you deal with the influence of the connectors?
* Therefore TRL calibration is better, this will enable you to make accurate measurement of SMD components in a through.

My point was that in many cases you don't need the 0.01 dB accuracy that you can get. For insertion loss this is often not relevant as the influence of connectors and other transitions by themself are one order of magnitude bigger. For return loss it will only change the phase of your return loss, but what interests you most is the magnitude right?

So yes I do agree that the performance is not perfect, but is good enough in most of the cases. You just have to know what you're doing and whether you need perfect measurements or not.