Lossy in both terminated and through settings

I should have a way to check current draw for this one part on this board.  Thanks very much for the info.  Will report back.

There is a series 0402 jumper feeding this circuit.  I removed it and did the network analyzer measurement again. So in this configuration where we definitely know it is off, the S21 looks the same.  I carefully bridged the 0402 pads with a DMM in ammeter mode, fully expecting there to be no current draw, but it shows 2.75mA current draw.  The overall plot jumped up slightly when I bridged the 0402 pad with the DMM probes, but the overall shape stayed the same.  It rose maybe 1dB or 2dB?  I didn't have enough hands to capture a screenshot of powered vs unpowered.

So my S21 measurement lines up with the reference OFF state, but my device draws 2.75mA current. If it was somehow shorted under the package body I would expect it to draw more than 2.75mA

With the series component removed, and board powered down, DMM DCR measurement from this node (pin-1) to ground is 2.32kohms.  The only thing included in this measurement is the IC and a few decoupling caps.  

I repeated this on my other assembled pcb and It behaves the same way on both boards.

I had a tech replace this switch and now it measures better on S21.  Current draw is now 1.4mA.   It's possible they are both damaged on both boards.  I did not expect that.  Will report my findings.

S parameter measurements are better but still not right.  I'm trying to figure out the big impedance dip at the switch approximately 100mm in using time domain mode.  Like I showed earlier with the switch removed, the impedance rises quickly as expected at this point.  In the below measurement, I have removed the output coupling capacitor to separate it from the output SMA connector pad.  I was worried the SMA pad geometry and reference plane overlap was bad on the SMA connector and that this was the result.  The plot still shows it going low impedance at the switch even with the output open.  This makes it look like a problem with the switch or the switch footprint itself. 

With output cap back so I can test S21, here it is:

Here is a sanity-check through path with my analyzer setup, same scale...

Switch removed again.  Input and output terminated with 0402 50 ohm resistors soldered to RF switch pads as marked.  S11 signal path is input SMA, coupling cap, input switch HMC1118LP3DETR, coupling cap, ~70mm trace, 50 ohm termination.  S22 path is output SMA connector, coupling cap, 50 ohm termination.  

The EN pin is tied low, right? The nomenclature on this pin is poor, that is, one would expect that a high on EN would turn on the part. However, it's the other way around. EN needs to be low to turn on the part.  - Eamon

I broke out the other two boards from their panel.  The two new ones look the same on S-parameters.  I checked control voltages like you suggested.  The board that I thought was marginally better?  The CTRL line voltage looked a little low like it was drawing current.  It was 2.3V instead of 3.3V.  I desoldered some parts and isolated CTRL pin from the MCU and filters -- turns out it's shorted to RFC and RF1 and RF2. I think I loaded RFC 50ohms to ground on the inside of the coupling cap on this one board attempting a test.  So if you can ruin the switch that way, maybe that's what caused it?

I did verify correctly intended control settings for these.  CTRL 3.3V and EN 0V. I checked with DMM and scope to make sure I wasn't doing something dumb like accidentally PWM-ing them.  So our desired state appears correct.

I opened a conventional tech support case for this as well, but haven't heard back.  If we can get that going offline I can send full schematics, gerbers etc to try and figure out what's going on.  I can also share info about our other boards that work well using the same switch.  

This is resolved.  We've been concentrating on the path connected to RF1.  On RF2 there was a component connected with a resistive path to ground.  I removed that and bypassed it with a series jumper.  It's all good.  Current draw is back down to 1.2mA as mentioned.  S parameter measurements look much better.  Thanks for the comments and suggestions.  

So to summarize the failure mode where S21 looks like it's off but the switch draws excessive current -- look for a current path to ground on RF2 (Maybe RF1 or RFC too?).  We missed a blocking cap on our schematic when the signal path changed from one sheet to another.  Bummed to get tripped up by something so basic, but happy for things to start making sense.  

Glad to hear that it's resolved.