TSP

In this episode Shahriar investigates a failed Keysight MXG RF generator. The instrument shows UNLEVEL for frequencies above 10GHz. The measurements confirm that the instrument fails to achieve the desired output power. The block diagram of the synthesizer is examined in details and potential failure points are identified.
The teardown of the instrument and various components are also presented. A detailed description of the multiplier unit is shown along with the PCB X-Ray. Measurements confirm the failure of an Agilent-branded component (doubler + amplifier module). The repair has to be paused for now until a replacement component is sourced.
www.TheSignalPath.com
/ thesignalpath​​​
www.Patreon.com/TheSignalPath

[@w2aew]

Really great analysis technique - hopefully you can find the part or a replacement module.

[@funtechu]

Great teardown and debugging as always. It's always a shame when the path ends at a custom part that is hard or impossible to replace.

[@rapsod1911]

I suspect that part 2 is going to be: because I could not find part I designed my own ;)

[@electroboy7422]

Really great job

[@carlosramirez6639]

Love your videos, I have a similar unit with a different problem. Mine works for 10 minutes and after that I get a error, 508, frac n loop unlocked. I will open and take a look inside whenever I have some time.

[@t1d100]

Good Job

[@dtiydr]

11:04 All in on via stitching.

[@pahom2]

Detective part tracking down the issue is always interesting

[@danielm3711]

It seems they knew about the electro-migration/aging issues in their high-speed switches and put a lot of thermal pads underneath them. Not surprised the failed chip had the highest amount of thermal pad. If you ever found the chips, replace the other chips with thermal pads underneath them.

[@linuxaos]

In another few years we will be sold a brick with some controls and some connectors. We are almost there. Shahriar had to use an x-ray machine to debug this. Who else has one of those ? Nice job BTW ! I love your videos!

[@trickyrat483]

We are already there. Remember, this is an old analogue piece of kit. Latest version is probably all digital and integrated into a handful of FPGAs and/or ASICs - with bugger all chance of even the OEM being able to "repair" it; simply board swap or throw away.

[@linuxaos]

@@trickyrat483 I agree. And even if you can repair it you need a fortune in test equipment. The most you can do is maybe (maybe) check PS voltages. Yet, the price is still astronomical. So, spend 40-50K and add to that an expensive maintenance contract all so in 4-5 years you have to junk it. Ok, fine, 5-10 years. I have 3 TEK spectrum analyzers and they work perfectly and they are about 30 years old. Thank the test equipment Gods!

[@WizardTim]

Great analysis as always! Shame it stopped at the custom module, however have you tried pressing down on the module or the old freeze spray? With all that thermal interface material I wouldn't be surprised if this failure is a result of thermal cycling and it could just be a failed solder joint or bond wire? The fact there is still an output just with excessive loss makes me think that could be the case as if it were catastrophic damage to the active junction it would probably result in no output. If it's a bad connection somewhere repairing that may be more feasible than sourcing a replacement for less than the cost of an entire replacement PCB.

[@faust-cr3jk]

Out of sheer curiosity, can you please decap the faulty chip and check it under microscope please? I wonder what exactly is broken. The doubler clearly works, so my guess is that the problem is related to some amplification stage within the problematic chip.

[@Meow-hw5wi]

Maybe you can try to build your own hybrid (if it is worth doing so...)? I believe you can get some off-the-shelf components to build up the internal circuits :p

[@arrbam02]

Very interesting debugging, almost glad it wasn't a broken connector :) I hope you find that replacement part so we will get a part 2!

[@rjordans]

Nice round of debugging again, too bad for the custom part. Seeing that there are some open source PDKs and free tapeout options nowadays if you open source the design. How big of a project would it be to make a replacement part?

[@gacherumburu9958]

👍👍

[@kfixit]

Since it doesn't work anyway, try reflowing the chip.

[@KeritechElectronics]

Microwave tech, "the blackest of the black arts in analog electronics" as CuriousMarc wonderfully stated it. Damn fine piece of kit - way too complicated for me to understand.

[@Rehbet]

2:30 That is an ancient USB drive

[@askild_eide]

Could you de-solder the custom chip and x-ray it by it self? does your x-ray have enough resolution so we could see how the IC is built up?

[@bumbochinelo4796]

Dont you miss old Tek and HP tools? manual with full schematics like my old 545 or 200ab.😀

[@mekatronikorg]

Can the entire module be done using discrete RF parts?

[@halo41293]

Can you explain the BPF architecture embedded into the board? How did you know that structure was the BP? and what does that structure look like? From top down it looks like these features are all on the same layer and are simply microstrip traces close to one another?

[@adrianwolmarans]

Hi there, I've recently discovered your excellent channel, well done keep it up. I also have a query. I am trying to repair a R&S SFU that has excessive quadrature phase error. R&S no longer supports this equipment and are unable/unwilling to supply me with schematics. Is there an online forum for RF test instrument techs where I would be able to source schematics, ask questions or alternatively source spares? And of-course, if you have experience with this generator then I would welcome your opinion too!

[@TheBackyardChemist]

The doubler seems to work but the level is too low. What if you bodged in an amplifier in the faulty path?

[@drdeconstruct9341]

Hello Shahriar, where do you get these schematics from? I am a college student who is working with a few instruments from Keysight this summer during an internship and want to inspect the architecture to use them more effectively. I tried searching online, but found nothing

[@bansci]

They're typically in the user or service manual. I've not double checked, but here's the service manual for this range www.keysight.com/es/en/assets/9018-03964/service-manuals/9018-03964.pdf

[@JaredKaragen]

I have a bag of cards/boards from some disassembled gear. I’ll see if any of them have that IC for you. Will send for free if I have it.

[@Thesignalpath]

Thank you!

[@TheDefpom]

@The Signal Path Did you check the 10.5GHz input to that IC you think is bad to be sure that it isn't before that, you seemed to make assumptions about the path before it being good, with that IC being so hard to find I think checking before it as well is a good thing to do, I would be checking directly at the output from that IC switch @16:16 be be sure that IC isn't the one giving trouble instead.

[@Thesignalpath]

Everything before that module is fine. That is why we get the sub-10GHz signals to work fine.

[@TheDefpom]

@@Thesignalpath Ah, right, you are correct, I just went back and looked at the block diagram, I was thinking of the 2nd switch and bandpass filter section, which is obviously what that IC you suspect is doing. Another possibility is a switch after the final bandpass filter being stuck on one of the other filter paths, resulting in an effective back feed into a different filter.

[@pahom2]

Rohde & Schwarz product placement )

[@mahuspro]

This component like box. You can find small components inside it if carefully take out upper side. There is switches, amp, resistors. Be careful to open, but you can find solution without replacing

[@mylesl2890]

did you try reaching out to manuf. for a replacement part/chip??

[@Thesignalpath]

I plan to do that.

[@PapasDino]

@@Thesignalpath Hope you have better luck with Keysight than I did with R&S; even Dr. Rhode couldn't shake loose schematics from the company for my FS300 spectrum analyzer when I asked him (he's a ham operator so I had a connection, but unfortunately no joy).

[@vincei4252]

@@PapasDino That is depressing.

[@fourierv2]

National Test Equipment claim to have this part in stock, listed under 'PRG-1GM1-4226'.

[@Thesignalpath]

I wrote them already. They don't sell the part. They are only willing to "repair" my unit for $4500.

[@vincei4252]

@@Thesignalpath yikes

[@andymouse]

Wouldn't keysight provide one as you promote them alot ?....cheers.

[@isoProxanol]

@Waterbeach Electronics Ltd That's the point. Agilent/Keysight is a test gear manufacturer, not a chip manufacturer. Because they have such specific requirements for amplifiers or other components they need to design their own ICs. They're good enough to work quite reliable but eventually they're just not as good as what is available for everybody on the market.

[@bzuidgeest]

After all the videos on their stuff you think keysight could make an exception and just sell you, or better even, just send you the part you need.

[@Thesignalpath]

To be fair, I have not actually asked them yet.

[@bzuidgeest]

@@Thesignalpathdoesn't hurt to ask, though to be honest i don't expect them to help. But a nice surprise if they did.

[@artiem5262]

Chances are very good that someone who works on this instrument will see this video over the weekend. Maybe you'll get an e-mail Monday morning...(I supported the HP/Agilent Santa Rosa site for a number of years, where this one might have been born.)

[@Jonas_Meyer]

If they do they get probably flooded with e-mail from people asking if they can get replacements IC.

[@bzuidgeest]

@@Jonas_Meyer i doubt that. This kind of repair takes more skill then most have. And even is it did, they build and sold it with a custom part so they should provide spares. I am a firm believer in the right to repair. Preventing waste by throwing away repairable stuff is part of that.

[@jimmy_jamesjams_a_lot4171]

Would it be a worthwhile endeavor, to throw down some Hardware Description Language, and see if you can’t configure an FPGA to do the same tasks that your broken IC should be doing? THANKS AGAIN, as always, for sharing so so much!

[@preste3]

I'd like to know the most likely reason these high performance chips fail. Heat drives what mechanism? The cooling solution here looks pretty amateurish to me.

[@great__success]

My guess would be a loose gold wire that bonds to the actual wafer inside.... what could have happened due to the heat

[@rjordans]

@@great__success if that's the case, might be worth seeing if these chips can get decapped somehow and try to re-bond?

[@karlbesser1696]

You could design the IC yourself and have it manufactured in China.😇

[@qwertyaasdfg8210]

Hi! We have 2 such generators, and this chip also failed. It is impossible to buy it. It would be nice to replace it with a similar build. I think it helped a lot of people in the repair. I hope it will be possible to make an assembly that is not original.

[@wuukaa9079]

Haha your synthesizer can do 20Ghz?.. mine can only go up to 20kHz 🎹😅

[@Thesignalpath]

Ha!

[@SaabFAN86]

How diffcult, if at all possible, would it be to create a replacement-module for the chip? It looks quite big, so with tiny enough parts the necessary components might fit on some high quality PCB with castellated holes that can be soldered onto the board. It would not have the performance of the original part, but I think, the worst inaccuracies can be compensated for with different calibration coefficients. Would save the rest of the instrument from the scapper or spare parts salvager.

[@vaualbus]

you need custom silicon for this sorta of thing so yeah no discrete component hack, at the point probably is better to find a replacement for the entire module nit just the IC. 🙃

[@alikhosravi7579]

please slow down talkig

[@Thesignalpath]

Have you tried listening at 0.75x?

[@Kilohercas]

Why not make it yourself ? It should be easy to do

[@trickyrat483]

Yep, wood, screws and a bit of glue. Easy.

[@kevina.4036]

sure he just has a fab next to the dining room for one-off MMIC designs...

[@Kilohercas]

@@kevina.4036 If you go on digikey, you can find LNA and RF switches with same performance, or maybe even better one. And where are special pcb material for low loss operation. So before any saying things like that maybe you should try to think a bit. Also he is quite capable engineer, i would not comment if he couldn't do it.