Mechanical, non-magnetic RF switch

[mtwieg]

For my research I often have to compare the performance of different RF networks in a signal chain (for example, a LNA connected between an antenna and a mixer). Normally I have each stage as a separate PCB with BNC or SMC connectors, so I can manually arrange things how I want. But the problem with this is that the measurements are actually sensitive to the routing of cables and the position of the components, so manhandling the cables and connectors always disturbs things. I'd much rather have good RF switches to select the DUT without having to move anything. High isolation and very low loss (like >60dB and <0.3dB at 100MHz, respectively. Power level is very low, typically 10dBm max) are important, so mechanical switches would be ideal. The issue is that this has to be done in a powerful magnetic field, so no magnetic components are allowed, which seems to rule out all the mechanical switches I've seen, which all rely on relays or passive magnetism to work well.

So does anyone know of any good off the shelf nonmagnetic mechanical RF switches? Or how to DIY one? Even just a simple SPDT would be a great help.

 

[nxt_]

try solid state relays

 

[mtwieg]

I haven't been able to find any solid state switches that meet my needs (especially the isolation spec).

edit: however I will say that much higher insertion loss would be acceptable so long as the matching between the two paths was excellent (like within 0.1dB). However from what I've seen datasheets never actually specify the matching.

 

[nxt_]

i don't know the power rating of your system.. but here is 74HC4316 Quad Bilateral Switch

Features
Four Digitally Controlled Analog Switches with Common Enable Pin
Up to 6V Analog Inputs
Each Switch has a Separate Control Input
Diode Protection for Inputs and Outputs
Equivalent to CD4316

 

[chuckey]

Would buy some "Yaxely" type wafer switches, throw away any ferrous components and replace them with brass and or plastic. The only problematic area is the indexing of the knob to the switch position, just use (make?) knob with a long thin pointer.
If the measured performance of your devices is changing due to lead position, I would sort this out first , try better screening, else the selected best solution on your test bed might turn out to be inferior once in production.
Frank

 

[mtwieg]

Would buy some "Yaxely" type wafer switches, throw away any ferrous components and replace them with brass and or plastic. The only problematic area is the indexing of the knob to the switch position, just use (make?) knob with a long thin pointer.

There's no way rotary switches like those are actually suitable for RF though, at least not for switching between transmission lines. If they were then any old switch would work as well.

If the measured performance of your devices is changing due to lead position, I would sort this out first , try better screening, else the selected best solution on your test bed might turn out to be inferior once in production.
Frank

Basically I'm working with a flexible loop antenna, and so far I haven't found a decent way to prevent it from deforming while manipulating its cables. That's mainly why I want the switch.

 

[thylacine1975]

RichardsonRFPD offer a nice selection of wideband solid state switches that I've used to create automated RF test assemblies in the past.

At first glance, they don't seem to have anything that simultaneously meets your isolation and insertion loss specs though. For example, the PE4251 (https://www.richardsonrfpd.com/Pages/Product-Details.aspx?productId=937745) achieves 62 dB and ~0.6dB over 10 MHz-1GHz. With such a low insertion loss, it's highly probable that the matching between paths would be acceptable. Although they don't specify a lower bound on insertion loss, I've empirically observed the spread of insertion losess across devices/ports to be no greater than 0.2 dB at 2.4 GHz (without accounting for all of the sources of potential measurement uncertainty such as measurement calibration, connector repeatabilities, cable bending losses, variation between soldered joints* etc).

Have a peek and see if some device/combination of parts might work for you (I can vouch for the usefulness of their 4-port absorptive [internally terminated] devices

* At 2.4 GHz, we've found mounting a SOT23 package just slightly crooked on it's pads (so it's still a valid placement, just not perfectly aligned with the PCB) results in a clearly discernable phase difference (primarily due to length deltas) in the two paths of an SPDT switch.

 

[biff44]

I would use these. They are probably beryllium copper, and not magnetic, but check the data sheet to be sure.

https://www.digikey.com/us/en/ph/Murata/SWF_Connector.html

 

[mtwieg]

RichardsonRFPD offer a nice selection of wideband solid state switches that I've used to create automated RF test assemblies in the past.

Some of the Peregrine parts come pretty close, I think I will just have to try and see if the path matching is good enough. Fortunately phase difference doesn't matter to me, though.

And I did find a MEMs RF switch from OMRON that meets my specs, but naturally it's obsolete and it doesn't look like it's been replaced with anything...

I would use these. They are probably beryllium copper, and not magnetic, but check the data sheet to be sure.

https://www.digikey.com/us/en/ph/Murata/SWF_Connector.html

I've seen connector switches around, and they look like an option. The limited info I've found says they contain stainless steel (both in the connector and the test probe), which may be magnetic... might be worth sending murata a call though.

 

[chuckey]

As the wave length of 100MHZ is 3m, a couple of cms of wire in a 5 cms box will not upset the match by any measurable amount. You will have to try hard to wire up any solid state switch with less unmatched wire.
How about having a proper test test up, a couple of bulkhead connectors mounted on a bracket and weighted down (screwed?) to the bench, this is your measurement interface, keep your instruments and their leads constant on one side, your dangly PCBs and their leads to the other, now you can measure to and from the interface. Your measurements will then take into account the PCBs and their leads, which I guess are required in actual practise. If your final arrangements do not require the leads then wire the lot up and test as a system, which will give you comparative results
Frank