Hewlett Packard 35665A Power Supply Repair (+ Memory Upgrade)

[Smoky]

Having no time for any electronics projects in the past six months, I wanted to take some time to open up the power supply to an HP 35665A DSA project because I've read that Rifa capacitors and leaking electrolytics may be lurking inside of this thing.



The first thing to do is to get the covers off of the power supply. One stubborn little screw-mounted circuit board support makes it difficult to separate the case so I decided to cut it using a small pair of side cutters to avoid cracking the PC board. The replacement part number for this support is TCBS-4-01 and it's made by Richco. They're available at Digikey but I found them much cheaper on Ebay (20 for $3):

https://www.digikey.com/en/products/detail/essentra-components/TCBS-4-01/395438?s=N4IgTCBcDaIC4GMBGBnAtAFjQBgIwgF0BfIA

All you have to do is un-screw what's left of the old support from the stud on the cover and then screw the new one in place and it's ready to be put back together:



The DSA passes all of its self-tests and is in great physical shape although it doesn't have any of the available options. George at GLK Instruments helped me out with that by mailing-off two EEPROM's loaded with options



Many thanks to Louiebluey for the link to the full HP 35665A Service manual. It's going to help this project out immensely

https://drive.google.com/file/d/1foUXt5DUGoiw3OGIxmDoXEJ-oenD2k0c/view?usp=sharing

...and thanks to Sören Marodören for his threads on how he successfully added memory to his 35665A:

https://www.eevblog.com/forum/testgear/how-to-upgrade-the-memory-of-your-hp-35665a-dynamic-signal-analyzer/msg3314476/#msg3314476

https://www.eevblog.com/forum/testgear/hp-35665a-dynamic-signal-analyzer-memory-upgrade-hack/msg2744506/#msg2744506

I believe that I have enough of the FPM DRAM to bring my analyzer to at least the 4mb level (1C1 option).

I have a new battery to install on the CPU board and a floppy disk drive to open up and clean too:



If I find anything dreadful in this SMPS, I'll be doing this all over again in a 35660A and 3588A.

[Smoky]

The large group of electrolytics include eight 2200uf/16v Nippon Chemi-con LXF caps.

It didn't take a second to know if these caps were leaking just by that familiar smell when de-soldering these things!





I read an article by Shaun Merrigan regarding his HP equipment repairs:

https://shaunmerrigan.info/electronics/hp-3245a-capacitor-replacement/?msclkid=8828afd9ba1811ecb624ed1cc309c20f

He mentions that the LXF capacitors that he found in his machines that were leaking electrolyte still passed all of the capacitance tests. That goes the same for the 2200uf caps that I removed from this power supply:



When the electrolyte conducts between the two leads, bad things start to happen:



After a thorough cleaning, thank goodness I see no damage to the PC board

[Smoky]

I replaced all of the electrolytic and Rifa capacitors with new:







I measured every old capacitor and made sure that the new replacement measured better/higher. For example, the original Nippon Chemi-con KMH bulk 1000uf/200v measured 887uf on the Sencore Z-meter. The new Epcos B43644 1200uf/200v caps measured 1040uf.

As for the Rifa capacitors, they were all cracked and crazed.



Anywho, the service manual lists all of the test points where the voltages are measured. Here is the table that shows what's coming out of the power supply through two cables, J1 and J2:



From here, I moved on to the floppy drive. I removed the CPU board and detached the floppy drive. It's made by Epson:



Without any screws to unwind, the cover just pops off. Boy, this drive is clean



I don't believe the head seen much use at all. Look at the horizontal metal area without a speck of dust. I just lightly swiped the head with a cotton swab and a touch of Windex. Even the original lubricating grease was soft and in place. This analyzer must have spent most of its life in an office:



I put the floppy drive back together and mounted it back on the CPU board. I then swapped out the old socketed IC with the new one from GLK Instruments to open the door to the analyzer's options. And now the CPU board is back in the analyzer:

[Tomorokoshi]

I updated Sören's thread with the correct memory available from 1C1.

With the service manual available now I should double-check my memory board.

Here is additional information from my unit.

[Smoky]

The power supply is back together and the analyzer is back up and running







I used a DMM to measure the voltages and ripple levels of the rails I can find/get access to. The service manual doesn't state how much ripple is too much.

Here's the +5v rail with ~2.3mV of ripple:

     

And the +12v rail with ~15.5mV of ripple:

     

And on the Digital Processor Board (A1), here is the +15v rail with under 200uV of ripple:

     

Both the +15v and -15v showed a very-low measurement of ripple so I used a TDS420 oscilloscope to look at it:



That's pretty quiet!

Anywho, I couldn't find a test point for the +18v and -18v because when the voltage makes it to the input boards, it's regulated down to +15v and -15v.

I followed the power-up of the analyzer with an alignment of the CRT using the service manual's procedures. The display looks oodles better than what my camera can capture:



...and checking the options menu, there you go, GLK Instruments upgrade EEPROM does what it should:



What's next is the memory upgrade. I hopefully found the correct DRAM...

[m3vuv]

Looks like you caught those leaking caps just in time,had same in my hp dso psu,took longer repairing tracks i very nearly gave the job up,was 30+caps,smelt like a stray tomcat had been using it as a toilet,i wonder why most rifa caps co nuclear,bad production runs, under speced etc,anyone know?

[Smoky]

Well, it's on to the memory upgrade attempt.

Looking at all of the pictures that I can find of 35665A's which have the "1C1" or "ANA" memory upgrade option installed use 4MB Mitsubishi M5M44400AL-7 Fast Page Mode (1048576-word by 4-Bit) Dynamic Ram in the 20-pin Zip package. Well, this Mitsubishi stuff seems to be hard to find at an affordable price. But I guess looking for "new" 30 year old stuff can be. The data sheet for this DRAM is on the bottom of this reply.

On RaccoonRider's gaming website, are charts listing a bunch of different memory by type, size, and bit. Here is one of the charts that I found the M5M44400 listed on. I put a red dot next to it:



He identifies the memory if it is the FPM or the Static Column type after the part number. I don't know much about this but it allows me to whittle-down the list when searching for options since the Mitsubishi memory mentioned above is the FPM type.

I began searching for the data sheets of all of the part numbers in that column (1Mx4). After looking for a while I noticed that a few of the part numbers on the list weren't available in a 20-pin Zip package, so the list got smaller.

The "L" in the part number seems to represent a "battery backup operation" version, but I'm not absolutely sure. If this version is a requirement for the 35665A, then that reduces the remaining part numbers to 3 manufactures. Mitsubishi, Hitachi, and Toshiba.

Using my best guess, I went in it alone and bought some memory before getting professional advice from the EEVBOG community

As we all know, that can be a costly mistake

[Smoky]

I bought 20-pin Zip sockets to install since I might be swapping memory out of the analyzer if it's not correct or if it's bad/phony.

I checked the clearance of the memory board to the under-side of the power supply to make sure that there is enough breathing room when adding the sockets. There is. I see a solid 3/16" gap.

I bought 20 pieces of Toshiba TC514400AZL-70 DRAM. It's dated 1994. The data sheet is at the bottom of this reply.



The data sheet on this stuff has"preliminary" written at the top of it and I'm not even sure if there is an updated copy available.

I find data sheets hard to read when it comes to the specs especially when they are not in any particular order.

If this memory isn't a correct replacement, I'll save my pennies for the Mitsubishi stuff, maybe even just eight of them, so to get me to the "1C1" level of memory.

Anywho, here's the old memory that needs to be replaced:



First thing I did was purchase a new tip for my de-soldering gun. What a game-changer!



Even with the the pads being so small in diameter, the heat was contained with no visible damage after the memory removal

Here's a shot of the bottom and the top of the memory board after de-soldering the 320 holes





I then proceeded to install the 20-pin Zip sockets. They fit perfectly between the capacitors too!





Next, I removed the 1M RAM resistor (R101) from the PC board. Btw, the "1M" marking makes sense since the memory chip that is used is abbreviated as "1Mx4" type:



Installing the memory followed by the new lithium battery was next:



Sliding the memory board back into the 35665A analyzer is all that's left to be done. Then it's power-up time

[Smoky]

The memory modification worked





This table is an example of how the amount of memory installed in the analyzer affects some programs:



One thing about the memory that's used. That letter "L" in the part number represents a standby current of ~200uA. Any other DRAM seems to have a standby current ~3mA. If you plan to do this modification, the part numbers that I found that fit this in a 1Mx4 20-pin Zip package is the Mitsubishi M5M44400AL/BL, the Hitachi HM514400BLZ/CLZ, and the Toshiba TC514400ALZ/BLZ.

I guess this low standby current allows the internal lithium battery to hold the current information when the analyzer is turned off and then back on again after lunch



Anywho, the next project coming up is the power supply in an HP 3588A spectrum analyzer

[Cubdriver]

GREAT write-up!  Thank you.  I have one of these in my repair queue; this will encourage me to dig it out and look at it.  Much appreciated!  The manual link will be a big help, too - lack of any kind of service info on it was a large part of my disinclination to work on it.

-Pat

[slbender]

That really is a well done report.  I do think that such exotic memory zips may have only been produced by one or two mfg’s.  So some might be relabeled and the standby currents could be a maximum value, while the smaller value a “typical” value.  Zip 20’s are rarely seen, but I believe my old Commodore Amiga 80386 machine has a few dozen 1Mx4’s, too bad that set always crashed from day one until it was relegated to an under chair position over a decade ago.  I will be more likely to spend for an HP 35665A now than I would have considered otherwise.  Many thanks for the info!!!

[Reprobyte]

Hello and thank you so much for the amazing write up you have done here, this is so helpful for people going forward, like my own situation!

I notice from your posts I have read recently that you have owned both the 35660A and the 35665A.

I am an owner of a perfectly working 35660A, but I just ordered a broken 35665A on ebay - https://www.ebay.com/itm/275773441226 as I would love swept sine and the other extras

Looking at the description I think it may be a PSU issue. I am looking between part numbers and it appears the CRT and the PSU of my 35660A are the same part numbers in the 35665A.

Hoping I can just swap over the PSU from my unit as a quick fix, before I get to work on the old PSU, if that is all it is I will be very lucky.

The unit is only missing 2 options from what I can see 1D0 Computer Order Tracking and 1D3 Curve Fit/Synthesis - I need to have a read if these will be useful, and if so I will follow your info on upgrading the EEPROM too.

I think its almost impossible to resist following your guide to double my memory also, thanks for the part numbers and explanation, that will help me loads.

The unit on the ebay pictures says "ACS Program" in marker which I hope to mean American Chemical Society Program, and not an error!

Thanks again!

[Reprobyte]

Also had a couple questions for you, did you upgrade the firmware in your restoration? And can the floppy drive be swapped if it is broken or do we need to repair the unit it came with?

Edit: Notice yours is an Epsom SMD-1300 so those are very cheap if it's possible to swap, maybe mine works I'm getting ahead of myself.

Thanks again!

[Smoky]

Nope, I didn't upgrade the firmware in the 35660A or the 35665A.

As for the floppy drive, I would clean and lubricate it before you use it. The one in my 35660A came right to life after I refurbished it. I haven't tried the one in the 35665A yet.

I rebuilt the power supplies in my 35660A, 35665A, and the 3588A Spectrum Analyzer. They all share the same power supply part number but were made by different companies over time. Here are the labels from on top of the other two power supplies:





None of the three power supplies looked the same inside.

I'm too busy in my construction job right now but next Fall I'm going to upgrade the memory, change out some relays, and install a new battery in the 3588A Spectrum Analyzer. It throws a random "calibration error" message every now and then.

Good luck with your project

[Reprobyte]

Thank you for your reply, that’s great to know as I can do a quick test with my working supply before I fix the assumed broken one. having the spare part made me much more confident I can at least boot it up and then see what else is wrong, hope I get lucky.

Strangely my psu from my 35660A is dated 9111 and 2.1 revision, I would have thought mine would be earlier but I must have a later model while they were sold side by side.

Thanks again and I hadn’t actually checked the 3588A out before, but with that frequency range now that’s another thing I need!

[Reprobyte]

Great news, swapped the PSU and it works! Perfectly from what I can see functional tests and did a quick sine sweep on some junk EQ here, got the first memory upgrade option and almost every other option, apart from Order analysis, which I don't know if I need, but it seems like a waste to buy the upgrade chip for a single option rather than the full package. Also gave it an extreme clean when tear down, remove marker pen and dirt etc as that bothers me more than anything.

Both PSUs appear to be the same batch, the 35660A is newer though remarkably. Now I need to follow what you did to your PSU to try and fix my other one.

Got lucky here, offered so cheap it cost double that to ship to the UK glad the risk worked out, and thanks again for your info here, I am going to need it to continue the project!

[Reprobyte]

Installing the memory followed by the new lithium battery was next:



Sliding the memory board back into the 35665A analyzer is all that's left to be done. Then it's power-up time

Sorry another question, does the board just pull out with the ring-pull on the battery/memory board? And did you re-solder the leads to the new battery, or can you get them like yours?

Thanks again!

[Smoky]

Yep, just tug on the memory board and it'll slide out.

I purchased the batteries with leads from Digi-Key:

[Reprobyte]

Ahhh, brilliant, thank you once again!