Per the seller, it had a bad "Solid State Transfer Switch". But the price was right -- working 85046A units typically run about $1000 on eBay, so I thought I'd take a chance with it, given my previous positive experience modifying an 85046B (click here for details).
A quick check -- S11 and S22 (with both ports open (unloaded)) were way off (e.g. -30 dB, where they should have been about 0 dB). Hopefully this was the fault of the bad switch and nothing else would be wrong with the unit.
But only one way to know for sure...replace the switch!
Here's the solid-state Transfer Switch (p/n 5086-7539), in situ.
Equivalent replacements are usually quite expensive (per eBay), but there's a simple, inexpensive solution: replace the solid-state switch with a much less expensive mechanical coaxial switch.
HP originally designed these S-Parameter boxes with a mechanical coaxial switch to perform the Transfer Switch function. So, by making this mod, I'd just be returning it to an earlier version.
Per the HP manual for the 85046A/B, the mechanical coaxial switch's part number is 5081-8178. It's rated from DC to 18 GHz, 1 watt max., and has 24V coils.
Well, I didn't have one of these in my junkbox, but I did have an HP 33311B and an HP 33311C coaxial switch. The 33311B also has 24V coils and is spec'd from DC to 18 GHz, 1 watt max, so it looks to me like it's an equivalent replacement for the 5081-8178. (The 33311C is spec'd similarly, but with a wider frequency range to 26.5 GHz).
Replacing the bad solid-state switch with the 33311B switch is straight-forward:
- Disconnect the 3 rigid coax cables from switch's SMA connectors
- Unscrew the two screws holding the switch to the chassis (mine were spline-heads)
- And unsolder the three wires from the terminal on the back of the case (noting which color wire went to which terminal).
And then to install the mechanical coaxial switch, just reverse these three steps.
Here's the wiring on the original solid-state Transfer Switch:
And here's the newly wired 33311B (ready to be installed on the chassis).
Logic Board Modifications:
But there's one more step. The "Logic Board" that drives the Transfer Switch also needs to be modified.
The Logic Board used to drive the solid-state transfer switch is p/n 85046-60016. But, per the manual, the p/n of the Logic Board which drives a mechanical coaxial switch is 85046-60051.
Fortunately, the modifications to make a -60016 board into a -60051 are simple -- there's only a slight difference in component values between the two boards. Here's a summary of those differences:
Solid-State Switch Mechanical Switch
R7 39 ohms 56 ohms (2 watts)
C2 4,700uF, 35V 10,000uF, 35v
R25 0 ohms unstuffed
R5 0 ohms unstuffed
I've highlighted these four components on the PCB layout diagram below (click on image to enlarge).
Now, I could have replaced all four parts, but being lazy I decided to take the easier route and just remove R5 and R25 and see how the unit worked with just those two parts removed. I left R7 and C2 at their original values.
And to make it easy to back-out my mods, I just lifted one leg on R5 and one leg on R25. Here's a picture.
An Unintended Step to the Left:
After I finished the mods and reconnected the cables between the 8753A and the 85046A, I powered up the system for a quick test.
Hmmm...S11 still wasn't working correctly -- in fact, it was behaving very strangely. Oh oh -- could I have a bad Directional Bridge?
I poked around and noticed that when I was measuring S11, the voltage at the coaxial switch's "Common" terminal, which should have been around 20 VDC, had dropped to 17.5 VDC. Yet, for S22 measurements, this pin was at the correct 20 volts.
Could my coaxial switch be flakey? I removed it from the 85046A and tested it with my lab supply. Still flakey -- one of the coils would continuously draw current when it was selected. In normal operation the switch should flip and the coil stop drawing current.
After taking the case off the switch and watching its operation while I powered one coil or the other, I noticed that the switch wasn't completely flipping to one side and instead it would hover midway and continuously consume current (thus the drop to 17.5 volts).
The fact that this failure also occurred while I was testing with my lab bench supply implies that the relay is bad (I was hypothesizing that perhaps I should have also increased C2 to 10,000 uF).
So I replaced the 33311B with the other coaxial switch that I had, the 33311C, and I left C2 unchanged. The installation procedure for the 33311C is exactly the same as I described above for the 33311B.
After its installation I powered up the system and...success, it works!!!
Although I left C2 unchanged, you might want to increase the value of C2 from 4700 uF to 10,000 uF.
Dick Benson, W1QG, has also performed this modification with success.
Some network analyzer versions might require the solid-state Transfer Switch. For example, if the network analyzer requires that the four S-parameters (S11, S21, S22, S12) be updated continuously (i.e. continuous switching), then you need the solid-state transfer switch. Refer to "Solid-state Switch Operation Considerations" in the Operating and Service Manual (p/n 85046-90001, April 1999), page 24 (of 78) in the PDF, or page 4 of section 3 (Operation) of the printed version.
So before you make this modification, verify that the relay version is adequate for your purposes.
Obviously, if you make this modification, you should have the necessary skills, knowledge, and experience to undertake this task.
Also, I might have made a mistake in my equations, assumptions, drawings, or interpretations. If you see anything you believe to be in error or if anything is confusing, please feel free to contact me or comment below.
And so I should add -- this design and any associated information is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.