23GHz SMA Calibration Kit

If you have ever had to measure any high frequency signals (say in collage or at a job) you most likely know that the Vector Network Analyzer (VNA) has to be calibrated to account for cables and other linear errors. This is done by connecting known standards - open, short, load and thru - to the device under test. Just like any good measurement equipment these standards can be incredible expensive; it's easy to spend a few thousand dollars on a testing kit.

While this certainly is important to high end measurements, it's often just not ecconomical for daily usage, especially if you need a few sets or don't want to risk pricey equipment when handing it out to students (like me). Sure, you can get cheap alternatives, but they often perform poorly at high frequencies.

In this instructable I'll show you how to make high quality SMA standards, suitable up to 23GHz, for about 6$ each. The proposed design is the result of several iterations, optimized for the best frequency response. Depending on you requirements you may also choose a variation for an even better low frequency response or higher power rating for a lower maximum frequency. The performance has been confirmed with a calibrated R&S ZVK 40GHz Vector Network Analyzer.

While this project is mostly done as a personal challange, it was issued and financed by my university, which has agreed to publish the documentation on my personal instructables account. They've also provided me with the VNA mentioned above and a proper calibration kit to compare my build against. Thank you!

Although this instructable is written mostly for advanced makers or academic use, I need to point out that my own experience with HF design is limited. Each iteration is based on experimenting with a few ideas and refining the one which worked best. Please excuse the lack of photos and measurements on the interim builds, initially I didn't plan on publishing this work.

This instructable also doubles as a official documentation to the project. If you have been given this calibration standard by my university, contact me at nqtronix@gmx.net with your university email address to get a copy of this as a pdf and all documents I have.

At high frequencies everything becomes an electrical component - an extra mm of wire, a droplet of solder, everything. So it's not surprising that careful component selection is just as important as the actual design. Below you find the details to each iteration step.

After some research I found this blog entry by Andrey E. Stoev, who used and recommended a 0603 high-frequency rated resistor. To say the least the measurements of the first prototype were disappointing (only about -3dB absolut at the target frequency of 18GHz).

A major improvement was archived by soldering the resistor vertically on the center of the conductor, which led me to assume a symetrical layout is critical. The result is still bad, but twice as good as it was before (-6dB absolut at 18GHz).

Several further experimental builds indicate that the size of the component has a large impact as well. For further testing I used generic 0805, 0603, 0402 and 0201 SMD resistors. In general reducing the component size (lenght) in half seems to double the usable frequency range. However smaller componets have several downsides, most important the overly complicated assembly. Without proper (and time consuming) measures it's almost impossible to use 0201 components as shorts are very likely.

I also investigated the effect of multiple parallel resistors as suggested by this article by Claudio Girardi. Most test builds indeed had a better frequency response - at least up to a certain point. For size 0402 three 150Ω resistors in parallel was better than 2x100Ω, 4x200Ω and 6x300Ω. The same test were also done with 0201 for up to three parts, but the single 49.9Ω resistor was better that 2x100Ω and 3x150Ω. A star shaped layout of 300Ω resistors (two vertically and four flat, on to each side) was also tested, but performed worse than any of the other builds.

The amount of solder coverage on back seems to also have an effect on the performance. For operation up to 18GHz it seems reasonable to coat the copper tape with an even layer of solder. Slightly improved performance at even higher frequencies (about -6dB relative at f > 25GHz) can be archived by covering all 4 sides with solder.

The kapton tape, which has to been added for manufacturing reasons, does not seem to influence performance in any significant way. It can be removed by soaking the connector isopropyl alcohol for a few seconds and gently pulling on the stripes with tweezers. It may take about half an hour until all solvent has evaporated and the measurement will stabilize. Due to the low impact you should rather leave them in place and add solder coverage on all sides as mentioned above.

To keep the cost of the iterations low, the connectors have been reused many times. One particular load showed a very poor result (10dB-20dB worse across all frequencies) despite identical construction. I did assume this is due to change within the dielectric when repeatedly exposed to high heat.

The experiment was repeated with new connectors, unfortunately results improved only slightly. I happend to notice that mecanical stress on the case changes the measurements. Squeezing "bad" parts yieled improvements of up to 8dB, but can also make things worse. I assume that heat induced mechanical stress during soldering is a main contributor for bad builds, but did not investigate this any further.

My university requested four load references which were all made according to the instructions below. Despite identical construction the measured frequency response varies between "good" parts of typically 5dB (f <10GHz) to 8dB (10GHz < f < 18GHz). This figure does not include failed builds with way worse performance. The 3^rd match (graph C) is not close as good, but still somewhat usable. The 4^th match failed (short) before I could take the final measurement. Most likely it's caused by internal mechanical stress, for regular use device reliability needs to be investigated further.

The diagram above was recorded with a calibrated R&S ZVK 40GHz Vector Network Analyzer (both are pictured in the next step). Obiviously calibration isn't perfect, the spike at 380Mhz should not be there, but I could not find the fault in the setup. The blue graph labeled "HP902C" is a measurement from a second match (which has been in the lab) and was added as a reference target value. Compared to Claudio Girardi's match (2^nd diagram) mine is slightly worse at frequencies <1GHz due to the worse accuracy of the resistors.

Tools:

• tweezers, sharp tip for general use
• tweezers, extra sharp tip (or cheap enough to make it extra sharp)
• tweezers, wide and dull tip
• vice
• file, various small
• jigsaw with a metal cutting sawblade
• saw or dremel to cut the tube
• soldering station, temperature regulated with at least 50W, fine tip (0.4mm)
• antistatic brush
• multimeter
• scissors or utility knife
• lighter (or hot air station if you want to go fancy)
• working microscope, but if you have great eyesight a bright light might be enough
• calibrated VNA

Consumables:

• solder, with lead (lower temperature, more reliable joints and better solder flow), 0.5mm thick
• flux, non-clean, I prefer the liquid type in a flux pen
• solder iron cleaner, dry brass/copper recommended, but wet solder sponge is ok, too
• kapton tape or any other thin, heat resistant (300°C) tape
• copper tape, I used 3mm wide tape and cut it slimmer later
• isopropyl alcohol (lab grade, 99% pure)
• paper towels
• sandpaper, grid 120-400

Materials:

• SMA connector, I used the Amphenol 901-9895-RFX rated up to 18GHz, do not skimp here
• SMD resistors, type depending on application, see next step
• brass tubing, 8mm internal diameter (I used copper, but ony because I had it on hand)
• heat shrink tubing, about 10mm diameter (flat 17mm wide), various colors (if you like)
• box for storage + foam padding (optional)

Depending on your requirements you may choose one of the following configurations:

Please note I did not get around to test the specified 50Ohm resistor in size 0402 (the star rating is based on a generic 0402 part), but if you do please conteact me, so I can add it to the list.

*estimate based on resistor tolerance

For this project we need to get rid of the center pin as it acts as a piece of unnecessary wire. Cut off most of the length with a small jigsaw, and file the rest flush to the dielectric. Do not even try to cut it with side cutters as it will likely damage them.

Afterwards clean the area thoroughly with the antistatic brush and some isopropyl alcohol as the dust will decrease performance.

Despite I got all of my tweezers super cheap (a pack of 8 for 8$) they work reasonably well for all components size 0402 or larger. To securely hold 0201 parts in place I had to "sharpen" the tip of one pair with a file and sandpaper. Sure, you can also simply buy a suitable pair instead.

When you plan on using multiple resistors as the load the easiest way ist to "bundle them up" before soldering them to the connector. Stick the resistors vertically in a row on a piece of kapton tape. Push hold them together with tweezers while you tap the freshly cleaned tip of a soldering iron on one end. When one side is done flip it upside down and repeats the process for the other side.

Unlike most connectors this one uses teflon as an insulator, which does not only have superior HF properties, but also a fairly high melting point of 326°C. To account for temperature loss from the heating element to the tip I've set the temperature of my iron to 330°C. The high temperature significantly improves solder flow and promotes better solder joints. For this it's also helpful to tin and clean the tip regularly.

Before you can place the resistor(s) you need to tin the pad first. Apply solder until a small dome forms. This ensures a sufficiently large surface area for a good heat transfer. Grab the resistor and place it vertically at the edge of the pad while holding the soldering iron in place. After a few seconds the solder should have wetted the contact of the resistor (you can actually see this when working under a microscope). Push the resistor into the center and remove the soldering iron in one motion. You may need a few tries to get this right.

Remove the remaining flux with an antistatic brush and some isopropyl alcohol. This doubles as a quality test, a good solder joint will not break even if you brush it forcefully.

The smaller the resistor is, the easier it is shorted when you solder the copper tape to it. To prevent this place about 1-1.5mm wide stripes of kapton tape on either side of the resistor as shown. For clean cuts you can stick the tape to a cutting mat and slice it with a sharp utility knife.

For a good performance you need to provide a solid ground connection to the second terminal. Copper tape is ideally suited as it can be soldered and easily applied.

Cut off a short piece and make it slim enough to fit it between the pins as shown in the picture above. Use the utility knive pokiest tweezers you have to poke a square hole in the middle. Peel off the paper back and align the hole with the resistor. Any excess tape should be cut off.

Solder the resistor to the tape with just a little bit of solder. Measure the resistance between this new solder joint and the center pin of the connector, if is not 50Ω ± 0.5Ω find the short or bad solder joint before continuing. Unfortunatly the heat can change the resistorand change its value, in which case you need to replace it.

When everything is ok, heat up on side of the connector with the largest contact area possible. Add plenty of solder, you can use the picture as a reference. Ensure the solder has covered all pins well. If you want you can add two short pieces of copper tape, one on each open side and cover them with solder afterwards.

Cut of a 1cm piece of the tube and remove all sharp edges. If it doesn't fit right over the four pins you remove a little material of the connector. Solder it only at the edges to the connector with some additional flux. The other end of the tube is intentionally left open to prevent moisture from accumulating. Finally add a piece of heat shrink tubing.

As mentioned in Step 2 mechanical stress has an influence on the measurements. This knowledge can be used to "tune" each device by gently squeezing it with pliers. Be warned, this method can very easily backfire and ruin the part. I'd suggest to only use it otherwise unusable parts.