Variable inductors in HP 3577B (and 3577A) IF filter stage

[precaud]

I'll try not to make this too long-winded. A couple years ago I bought three "known bad" HP 3577A's at a local auction. Previous owner was Raytheon (Phoenix). I know from past experience to be careful buying their surplus, these guys are very hard on their gear. Some of it modified/customized internally, or even outright sabotaged. (I recall a large Sorensen power supply with bus wire wrapped around the bank of capacitors inside, so it blew fuses at power-on. Some department head wanted a new supply, and that is how he got it.)

So I expected issues with the 3577A's and was not disappointed. One unit in particular stood out from the others. Late-serial-number mainframe, transitional from A to B, with some boards from each. Powered up fine, weak crt, but all three input boards were bad. Two were "B" boards, one "A". Not just bad, but strategic parts removed to make sure they tested bad, passed no signal, and would be deemed unrepairable. Unless you took one apart and knew what to look for, you'd never know.

On all three boards, they removed one or both of the variable inductors in the 1st IF filter stage. See pics (yellow arrows) and schematic, parts L3 and L5. They even did a hasty job of it, leaving some of the pins in place.

At first I thought it might be a customization. But no. I modeled the filter, and removing these parts turns the bandpass into a band reject at the IF frequency (250kHz), passing no signal and making it fail every self-test.

So I need to find five of these tunable inductors to bring these boards back to life. Part number 03577-60329. Spec'ed as variable 100uH to 109uH, shielded, DCR is about 250mOhm. They have to be variable; adjusting them is part of the cal procedure for meeting phase specs.

Keysight parts ID says its obsolete. Lots of co's list them on their website but noone has any.

Once upon a time, variable inductors like this could be bought fairly easily. Now it appears they're as rare as hen's teeth.

I looked for equivalents in other HP instruments with IF filters, whose manuals I have on hand, and only found one that uses a 100uH variable. The HP 3585A uses one. But its ten times the DCR and twice the diameter, so it won't fit.

Any suggestions? Am I gonna have to make them?

[KE5FX]

I think I'd just replace both LC components with a standard fixed inductor and a parallel combination of an NP0 ceramic cap and a trimmer.  Any shielded inductor that's suited for use in modern switched-mode power supplies would be fine.

[precaud]

Interesting idea. Made me wonder why they didn't do that in the first place. And I think the answer is: temperature stability. That's why they would use the big mica caps and a variable inductor to tune the notches/peaks.

This is what HP says the spectrum of the filter with 50kHz either side of the center freq should look like. My modeling shows it to be very sensitive to tuned L values. I didn't find a combo of the three L's that matched this shape. They all had the notch closer to the peak, which I doubt is desirable.

[chris_leyson]

It looks like the pin spacing is 0.2" so an EPCOS/TDK P9x5 pot core will probably fit.
https://www.tdk-electronics.tdk.com/inf/80/db/fer/p_9_5.pdf
4-pin mounting is B65518D2001X000 and Mouser Digikey have them in stock.
Ferroxcube P9x5 pot cores will probably fit the same mount and that will give you a few more options in terms of materials, Al value, air gap size and tuning slugs.

[precaud]

Thanks Chris, I may be headed in that direction. Their coil formers don't appear to be threaded for tuning slugs.

[chris_leyson]

Hi Precaud, the thread for the tuning slug is moulded into the mounting base and the external metal clamp holds everything together. I threw out a box of EPCOS P9x5 cores and bobbins years ago and have regretted it ever since.

[precaud]

LOL, yeah, I know the feeling. Unfortunately most of us don't have room to hang on to everything...

I'm thinking a 9x5 will be too small to get the need inductance and low DCR. The stockers look like 14x8.

I guess its time to take a deeper dive through the pcb's in the junk pile...

[KE5FX]

Interesting idea. Made me wonder why they didn't do that in the first place. And I think the answer is: temperature stability. That's why they would use the big mica caps and a variable inductor to tune the notches/peaks.

This is what HP says the spectrum of the filter with 50kHz either side of the center freq should look like. My modeling shows it to be very sensitive to tuned L values. I didn't find a combo of the three L's that matched this shape. They all had the notch closer to the peak, which I doubt is desirable.

Another thought might be to replace it with an active RC filter.  Opamps are much better than they were in the early 1980s when the 3577 was designed.  It's vanishingly unlikely that someone undertaking a similar design today would do it with old-school ferrite inductors.  (They'd most likely do it digitally, but aside from that, an active analog filter would be next on the list.)

[precaud]

Another thought might be to replace it with an active RC filter.  Opamps are much better than they were in the early 1980s when the 3577 was designed.

Yeah, I've seen that in another VNA designed around the same time, though its IF freq is lower, around 90kHz IIRC. NE5532/5534's (and selected varieties from Analog Systems) were around back then; there's one on this board. You can't beat the passives for noise and stability.

[precaud]

Success!

Turns out these inductors are Litz-wound. Seeing that tweaked the ol' memory that, once upon a time, I had parted out a "universal modulation analyzer" which was chock-full of boards with tuned circuits with small variable pot-core inductors. Did a deep dig (say that 10 times real fast...) in the garage, found the box, and sure enuf, 7 boards each one with 16 of them, all 8x14 cores of various permeabilities. Was able to make four of them the right L tuning range and DCR, and stuffed them into the shielded holders stolen from the 3585A boards. Installed two in one of the bad boards, found and fixed a couple other issues (someone had overdriven the 50 Ohm input), did the cal procedure on the board, and voila It works fine and passes cal.

This does not qualify as an easy fix. But certainly a satisfying one.