Is there anything particularly special about a "short" and an "open" in a VNA calibration kit?

Question

I would like to calibrate a VNA with the reference plane at the end of a male BNC feedline. Surprisingly, female SOL kits are hard to find. There are many BNC 50-ohm loads available but I'm having trouble finding a BNC open and a BNC short.

Obviously I can make one easily enough, but what about the "short" or "open" gives it high tolerance? For example, Fairview sells a 10GHz open and short for ~$1000 each, but my goodness, that is a lot for a high-tolerance short and open.

My target is only 440MHz, so I certainly don't need accuracy at 10GHz, but what should I consider when building my own short and open?

Answer 1

You're not paying the big bucks for the engineering, so much as the calibration certificates, and the tracability back to standards.

The reason you're having trouble finding BNC standards is that BNC is not a repeatable connector. It cannot be used for precision work. You might just as well use DIY standards, because you are not going to get high repeatability. N-type is far better, it's similar to BNC internally, but with a screw thread to hold it together. APC-7 is best. SMA is very common on VNAs.

With a short, the only thing you need to worry about is distance to the reflection plane. You can measure this physically, or calibrate it against a better standard. Dielectric will of course increase the distance.

A simple open radiates, making its |S11| < 1.000. You can fix this by making a shielded open. It also suffers from a capacitive fringing field, which you can't fix. There are tables to estimate how much this shifts the reflection plane from the physical end of the coax.

You would certainly get more repeatable results, and accurate results, and likely more usable results, if you did all your R&D in SMA, even if the final implementation was in BNC.

If you do want to make your own BNC standards, then probably the best route is to use connectors to 0.141" or 0.25" semi-rigid. As a hack, I've generally bought a ready made cable, and then cut it in half. Cut your short and open cables to the same length. Solder a short across the end of the cable for the short. This is your good length reference. Shield the open by soldering a copper foil thimble round the end. For shielding, it doesn't matter how far the shield is from the cable open end, however it will affect the estimate of end capacitance. Keep it three or more cable diameters away for negligible impact on the end capacitance, and then look up tables for the end correction. You could make a load, ideally using two 100 Ω resistors in parallel, but it would be easier and more reliable to buy a well-specified load (which wouldn't have much general use) or an attenuator (which will be useful elsewhere in the lab). An unterminated attenuator will have a return loss of twice its attenuation. A loaded attenuator will improve the performance of any rough load it's connected to by twice its attenuation. Generally you'll find the S11 of low value attenuators specified better than high value ones. I've sometimes used a 6dB + 20dB pair of pads, in that order, in a ghetto calibration set.

Before you put in a lot of work, try this experiment with your VNA (does it have N, SMA, or APC connectors?) Using a high reflection device (so short or open), break and make a connection several times, to observe the repeatability. Now put a BNC adaptor on it, and do the same with a BNC device. Compare. Can you live with the levels of repeatability you observe?

Answer 2

I haven't ever done any VNA work with BNC connectors, so I can't speak to the specific difficulties related to that connector type, but, ...

An open intended for VNA calibration is not simply "no connection". A good one will be carefully calibrated (at the factory) for its parasitic capacitance. Its outer conductor should extend a bit beyond the end of the mating connector to maintain the field pattern of the propagating mode. In olden times they sometimes had a bit of dielectric rod attached to tune the capacitance and to ensure the phase delay of the reflection it produced was correct for the reference plane of the connector being calibrated.

The short is simpler, but it must be carefully made to place the terminating surface at the reference plane of the connector it attaches to, so that the reflection it produces has the correct phase delay.

In both cases, the vendor will provide a set of correction factors that can be given to the VNA, so that the VNA software can adjust the calibration for slight non-idealities of the standards. The exact definition of these correction factors might be different for different VNA vendors and models.

Answer 3

Be practical. Calibrating at 440 MHz isn't nearly as tricky as 10 GHz. And then, what are you trying to achieve? If the system under test has a BNC connector, well, you're already committed to its limitations. You can use the VNA to see the effect of your termination: how much does the measurement change between calibration using SMA at the end versus through a BNC adapter?

Where do you want to know the impedance? Calibrate with open, short, and load as close as practical to there, through the actual hardware you intend to use.

And then, the ability of a VNA to measure system properties usually vastly exceeds the significance of the variation in properties. A mismatch factor of 2.6 corresponds to a loss of only 1 dB.

Answer 4

More than 20 years ago I used an HP 8753B network analyser with the amazing APC7 connectors on the front. The regular N-type calibration kit was broken - for the N-F open, HP supplied a brass slug outer with a hole in it, and a little open circuit inner socket which you'd push in after screwing in the brass, so as not to rotate the connector on the inner pin.
Image from Kirkby Microwave
The six (not four) tiny fingers were prone to breaking, and eventually it gave out.

In the back of the lab was a lovely 14x6x2 inch hardwood box with the HP 85054B kit in it - Opens, Shorts, Load and a sliding load which I think could be moved along to further improve the precision of the load.

The instrument didn't know about this special kit. Different kits had different distances from reference plane to short, and open, and different capacitance models. The standards were visibly different from the old ones, and it didn't give good results, even at 3 GHz.

You could type in the values for Offset length, Loss, and the coefficients of C0 C1 C2 C3 (polynominal in F), but we had no calibration coefficients. So I called Agilent in the USA; their support line was answered by a guy with a PhD in Electrical Engineering who promptly faxed or mailed me the appropriate coefficients for this 20-year-old kit.

Just look at these specifications (found again here):

• 48 dB return loss on the load (0-2 GHz)
• 42 dB return loss on the sliding load (2-18 GHz)
• Open deviation from the model: < 1.5 degrees at 18 GHz
• Short deviation < 1 degree at 18 GHz

---

That kit broke too. From then on, for many years, we used a simple kit I made with three flange-and-pin SMA connectors screwed to a bit of plastic. Open was just cut off flat. Short was soldered closed in the same plane. Load had two 100 Ohm 0805 resistors, one each side. All coefficients set to zero. This was quite good enough for (professional) 0-3 GHz antenna work.

So the answer is there's a time and place for $4k cal kits

and for $4 kits made from a few ebay SMA connectors.

I think this one is still is better than 25 dB return loss, 1.1 VSWR up to 3 GHz, good enough for making antennas.

Keep the open and short roughly in the same plane, remove any spare wire like the pin of the connector. Use two resistors for the load instead of one.

I recommend building your kit in SMA, and then doing a port extension to get to your desired reference plane. Add (or remove) adapters and connect your antenna (or a fresh connector of the same exact kind). Short circuit the connector with a scrap of copper tape or a knife, and adjust the port extension to get a nice short circuit on the Smith chart. Short is better than open because you can apply it in situ. See this answer for more detail and photos of that process.