SMIQ06B - Low Power Over 1/2 of Band and Freq. Unstable

[energy_hunter]

I posted this back in Nov. 2019 but am now back on the repair horse for my SMIQ06B (300kHz...6.4GHz Signal Generator) and will more proactively work on this repair and post the results.  I already replaced the EMI glass for $450 last summer so I don't want to lose this nice instrument.

The unit at first puts out full power in a frequency sweep. After a few minutes the SMIQ06B puts out 20dB lower power up to 3GHz and the frequency begins to move around and be a little unstable (~20-40MHz wandering around when centered at 915MHz).

I've started documenting the symptoms and repair on google docs here.  Any advice welcome. Could be a switch that is failing. I've read that the Hitite switches may have been overbiased and seem to fail sooner than other components. This would explain a loss of 20dB since that is about the isolation of a switch that may be off when it should be on.  However this doesn't explain the frequency jitter so it may be two problems in one.

https://docs.google.com/document/d/1K95gHN0iE7RXR8iz1eGwksqNJzK_yJJNtmcQg3tauM0/edit?usp=sharing

Thanks to DaJMasta, smgvbest and hendorog for giving their advice earlier in 2019. I will document this repair as best I can for folks to help in the future.

[DaJMasta]

It really strikes me as two issues, because I can't for the life of me figure out how the frequency would be effected by the level with this kind of generation - maybe there's something strange about the LO board or the lower band board that I just haven't dug into, but I can't think of something.  If, for example, you had an LO underpowered that was feeding a mixer, you just would get varying/low power output at the expected frequency, not strange frequency variation.  It's worth finding out, though, whether this frequency offset is repeatable or is linked to the frequency of interest or is just a fixed offset.  That could point to a number of potential things, but if, for example, it was a specific LO that was under frequency for whatever reason, when you turn on a path with a doubling, you'd expect to see its offset double as well.  If the variation is more random, then maybe it's from a VCO fault of some sort (though you'd think you'd see PLL errors with that), but I don't know of what could give you a wrong frequency offset in just one case.  Is a modulation enabled and the modulation data just garbage?

For the level, yes, it definitely could be a switch.  It would be worth just disconnecting the coax from the main synthesizer board that goes into the RF module with the switches on the 6GHz extension to measure to see if it's actually a fault in the output of the lower band board (though if the detector isn't seeing it and giving you an unlevel error, it's less likely).  If not that, then I'd also check the drive voltages for the switches - just trying to eliminate the much easier to fix/replace options than trying to find new dies or a whole new 6GHz output module...

[energy_hunter]

Thank you for your post. I agree it is most likely two issues.  I did have a corroded green EMI glass in this instrument and bought the replacement glass from R&S for $450 and replaced it. I wonder if there isn't a great EMI seal somewhere if that would be causing the frequency instability by allowing signals to leak around in the instrument. Anyway, working on this in dribs and drabs. If anyone has an SMIQ06B that they want to sell then maybe having two instruments could help me fix this.

[energy_hunter]

We have traced the problem to the 6GHz extension card to the "A1 Power Unit", pg 47 of the pdf: Rohde_Schwarz_SMIQ6GHZ_SMIQ_6GHZ_Extension_Board.pdf.  The schematic shows a M/A-COM power amplifier (MAAM2600-B1) that goes to a Wilkinson power divider and then two switches (HMC986A GaAs SPDT switch die).  The amplifier I believe to be working since the power for the upper band is fine but the lower band is not. Since this amplifier is common to the full band this points to one of the switches. I seem to remember a post about the switches being biased too hard and failing on one repair. I probably will need to send this to an assembly house to fix this part with a new die and then wirebond it.  If anyone is willing to sell their 6GHz extension card please let me know.  The switches cost $110each on Mouser.com and then I have to pay an assembly house to remove and reattach new die and wirebond so the total repair will probably cost $700 I'm guessing.

[DaJMasta]

It is an R&S custom part, so the module itself is probably hard to come by, though I have occasionally seen SMIQ boards for sale.  Cheapest option may actually be a for parts SMIQ, though honestly they're not that cheap in the 6GHz flavor.


An option if you don't mind trying a hack: find an encapsulated switch with suitable bandwidth and characteristics, then adhere it to the ceramic substrate (or on top of the dead switch) and jump wires to it from the wire bond pads (they are soldered to for some parts, so it should be manageable, and without throwing the impedance too out of whack).  It's a hack, but since you have schematics you should be able to get the drive required to use it, and since 6GHz isn't crazy bandwidth these days, you can probably find an off the shelf part that will do the job fine, so it could save the wire bonding fees, if nothing else.

The module is fully removable if you desolder the pass through connections, so it is replaceable on its own, but I think it's going to be quite difficult to find outside of a surplus board, in which case it may be easier to just swap it in.