Experimental Insights: Miniature filters I

For the last few weeks I have been designing and building new miniature low temperature filters. The idea of these little filter boxes is to be relatively good low pass filters (with no re-entrance of transmission at multi-GHz frequencies). The standard way to do this is to use a technology known as the powder filter. This idea was first used by Martinis et al. in the mid 1980s. They are pretty much essential for direct measurements on qubits and quantum phenomena in Josephson junctions, as they are a really good way of removing high frequency noise which causes decoherence and obscures the measurements.

Powder filters usually consist of a coil of wire with an inductance, L, and resistance, R, embedded in a powder of fine metal granules. A capacitance C can also be added to the system. The skin depth in the metal is similar to the grain size and therefore microwave frequency signals are absorbed well by the powder. Originally Copper powder was used, but stainless steel powder seems to give a better response (probably because the grains are more resistive). The result is a low pass LRC filter. LC style filters usually have problems stopping very high frequency signals, because at some frequency the inter-winding capacitance of the coil begins to look like a short, and the high frequencies jump straight across the filter. The powder helps to reduce this problem by absorbing such frequencies.

So how do you make them?

In a fully shielded system the filters must be housed in a metal enclosure. This also handily stops your powder ending up on the floor. I’m going to do a separate post on how to assemble the things. In the pictures below I haven’t put the lid on the filter yet, so you can see inside it. The filter must essentially be hermetically sealed. A rule of thumb: if it’s water-tight, it’s microwave tight 🙂 You can screw the thing together if the filter is machined, but seeing as it is handmade, and I didn’t happen to have a spare lifetime to look for all the sub-1mm screws that I would inevitably lose during this process, I soldered it together instead.

Here is the resulting filter…

The first picture below shows the little coils inside the enclosure. There are 3 filters in one here. The enclosure measures 20mm x 50mm, and the coil wire is 0.08mm thick. I don’t know exactly how many turns there are on each coil, but there’s over 100. The feedthrough capacitors are mounted on both ends of the filter coil to giv’ it some C.


Next the powder is added. This was a test run, so the powder was not permanently set in place, but was rather mixed with a solvent which allows the powder to flow (rather like cement) and then evaporates, leaving the grains tightly packed. Using this technique the powder can be removed again by adding more solvent. Here is a picture:


Once the filter has been tested, the coils can be permanently set in place. This is done with a mixture of the metal powder and an epoxy-resin known as STYCAST. Stycast sets really hard, and also does not crack at low temperatures (unlike almost every other type of glue) so it’s pretty much ideal. Here’s a picture of the finished filter.


What is the response like? Well it looks like this:


The frequency range is from 50MHz->40GHz and the reference line is -40dB with 10dB/div. Not bad for a miniature test filter. You usually chain about 3 or 4 of these in a row to get a good response. This test filter has MCX connectors to attach it to the VNA, which make it slightly larger than it should be. In the full system these will hopefully be replaced by some smaller, neater, shielded twisted pairs. I’ll write a little series of posts about other aspects of these filters, as there’s too much information to put it all in one.

5 thoughts on “Experimental Insights: Miniature filters I

  1. rrtucci says:

    Very Informative. Thanks. I don’t know how to make low pass filters but I’m a good cook 🙂

  2. physicsandcake says:

    Yes it is a bit like cooking sometimes, especially mixing up all the powders and then baking the things with a heat-gun 🙂

  3. andy says:

    I’m not sure I understand your power-spectrum picture. The transmitter power seems to be below the -40 dB line for a large part of the frequency range, but what is really transmitted? the peak at low frequency seems to be at about -35 db?
    we have a spectrum analyzer in our lab where the whole first div on the frequency axis is sort of unreliable data, yours also shows a dip down to -infinity at zero frequency.

  4. physicsandcake says:

    It’s because the transmission is below 50MHz due to the cut off frequency of the LC circuit (which is usually between 1 and 10MHz), so you can’t actually see it on this plot. The signal frequency used in the experiment is below 1KHz, but the experiment itself is very sensitive to microwave frequencies. The reason we use this VNA is to check that there’s nothing getting through at these higher frequencies. I probably should have explained that!

  5. […] 12, 2009 by physicsandcake I talked a little while ago about a prototype miniature low temperature LCR filter I made. Well I’m making the real […]

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s