Interfacing MMIC prescaler to ECL (to replace an Agilent MMIC)

[fenugrec]

Hi,
I'm in the process of replacing the notorious 1GC1-4210 prescaler (custom HP/A/K part) that has bricked countless ESG* signal generators and ESA44* spectrum analyzers such as the E4402B. Quick summary for those not familiar : the 1GC1-4210 divides the YIG signal (3-7GHz) by 1,2,4,8, or 16 inside a PLL loop, depending on 3 mode-select pins. On the spectrum analyzers, only the ratios 8 and 16 are used :


One viable repair strategy is a small (tiny !) pcb with two chained HMC862A dividers (1,2,4,8 ratios). I thought this was a bit wasteful (the HMC862A is quite expensive) since only 2 ratios are needed,  so after a bit of research I'm hoping one HMC862A wired to divide by 4 or 8, followed by an ECL divide-by-2 (MC100EP32), might work :


However I have doubts on the compatibility of the input/output structures (see datasheet extracts below) : the HMC862 has apparently a 50ohm output, probably similar to the original 1GC1-4210, and meant to be AC-coupled to the next stage; I'm not sure how well it would play if connected directly to the ECL divider ?



And (possibly more problematic), this ECL divider would end up driving the original AC-coupled attenuator, and I think this would probably require a 50R on both outputs ?

Really hoping I don't need ac-coupling and termination between the HMC and ECL divider, as this pcb is extremely tiny and needing 5-6 extra passives would make this solution much less attractive. I've attached a tentative layout to give an idea of scale - 9.2x6.3mm really doesn't leave much space at all :

[minsik]

Why not https://www.leobodnar.com/shop/index.php?main_page=product_info&cPath=125_127&products_id=324&zenid=822d6aca34a60019cb77d6535385c9bc from Leo Bodnar. Professional built and supplied. Owning the precise GPS reference, which is the pinnacle of GPS frequency accuracy in my workshop I have no hesitation is suggesting this would be a way to go.
Minsik.   

[fenugrec]

Professional built and supplied.

Yes I am aware of the Leo Bodnar PCB, and it's great to have such a thing available "off the shelf". However it is expensive, as I mentioned partly due to the BOM cost. I was planning on maybe building a few, where it made more sense to build myself at a third of the cost.

Since I started this thread, I built a dual-HMC862A board (in the style of Bodnar's) to repair my ESA first, but now I'm not sure if I'm going to proceed with the simplified version, as I am not really equipped to validate RF performance of the mixed MMIC + ECL stage.

Planning on publishing more details of my build, including design files, in the near future.

[fenugrec]

Finally got around to documenting this.

I opted for the more prudent 2x HMC862 option; I had the feeling that using the MC100EP32 for the second stage division was more of a risk and would require 4-5 more passives, making the PCB even more challenging.

I posted details and photos to https://www.qcte.ca/testeq/e440x_sa/
Additions and corrections welcome !