The Testing Fixture for 6-Lead QFN SPDT RF switch

[neilhao]

I purchased several SPDT RF switches which manufactured by a Chinese listed company. The chip’s footprint is 1.0mm x 1.0mm x 0.45mm 6-lead QFN package. The manufacturer did not provide S-parameters. Thus, I had to design the fixture and measure the S-parameters by myself.

The measurement data conform to the Insertion loss and Return loss specifications in datasheet. Unfortunately, the measurement data exceed the isolation specifications in datasheet. However, for $0.04/pcs, I should not complain to much...

This article could also be fetched from the project’s wiki if the screenshot of this article is not comfortable for reading. Project's Wiki:

https://uniteng.com/wiki/doku.php?id=rfmeasurement:testing_fixture_6_lead_qfn_spdt_rf_switch

[3isenhorn]

Hi

Good measurement!
I was wondering if the switch could be directional for insulation?   Since they call the pin Ant(enna), they could only consider isolation when power flows into that port.
So maybe it is worth repeating the first measurement when the switch for the S34 part is open sit? (Second RF port matched with 50ohm)

[RadioNerd]

Before putting down the switch as not compliant to the specifications, bear in mind that performing isolation measurements at several GHz can be quite tricky and pose high demands on the design of the test fixture.

I n general, good isolation is achieved by
1) providing lowest inductance between the IC GND and PCB ground (lots of vias under the GND pad of the IC)
2) miminizing direct RF coupling between the traces on the PCB.

I cannot see the details of the PCB layout but at least point 2) could be improved in your test fixture. It might help to take a look at the layout of evaluation boards of the most important RF switch IC manfacturers. Skyworks for example has several Switch IC with very similar characteristics. Often there is also a drawing of the corresponding evaluation PCB in the datasheet. As an example: https://www.skyworksinc.com/-/media/SkyWorks/Documents/Products/1901-2000/SKYA21012_203021D.pdf

[neilhao]

Make sense! Just similar as mmwave radar design, better GND plane design between TX and RX can improve isolation, sometimes a GND via between TX and RX can improve the isolation by several dB.
However, for the fixture design, it seems it is not easy to measure the isolation of the IC itself. No De-embedded algorithm can work with crosstalk problem..

[neilhao]

Before putting down the switch as not compliant to the specifications, bear in mind that performing isolation measurements at several GHz can be quite tricky and pose high demands on the design of the test fixture.

I n general, good isolation is achieved by
1) providing lowest inductance between the IC GND and PCB ground (lots of vias under the GND pad of the IC)
2) miminizing direct RF coupling between the traces on the PCB.

I cannot see the details of the PCB layout but at least point 2) could be improved in your test fixture. It might help to take a look at the layout of evaluation boards of the most important RF switch IC manfacturers. Skyworks for example has several Switch IC with very similar characteristics. Often there is also a drawing of the corresponding evaluation PCB in the datasheet. As an example: https://www.skyworksinc.com/-/media/SkyWorks/Documents/Products/1901-2000/SKYA21012_203021D.pdf

Update:
I did the EM simulation for isolation between the traces RF1 and RF2 on the PCB. The isolation is not perfect but still far better than the measurement.