Repairing and Upgrading an old PM3320A - Including adding FFT!

[SaabFAN]

As you might remember from the Test Equipment-Subforum, I bought a PM3320A about 2 weeks ago.



When I turned it on, it worked and showed no sign of the defects the seller mentioned.
I thought: Well, maybe nothing wrong with it afterall/transport-knocks have solved the problem.

As we say in Germany: "Pustekuchen!"
Today I finished the first part of a self-made "Almost All-in-One Bench-Tester" (Will contain Signal-Gen, LCR-Meter, Multimeter, 10W Amp to simulate Generators, etc.) that grew out of the desire to build something to sort of calibrate my PM3350, and wanted to test it with the PM3320A.
On a hunch I checked the Battery Compartment and yep: The batteries (Ni-Cd Accumulators) were done for! 300mV on the first one, 600 on the second one. And the contacts were corroded. Some time in the past, batteries must have leaked in there.
After cleaning the contacts and replacing the batteries, I turned the scope on and thats when the trouble started: I had previously observed some pretty rough calibration-waveforms when I first turned the scope on, but now that the RAM was wiped, the scope didn't start with the stored parameters, but instead started a "Auto Set"-Routine set everything up. And it hangs up during that routine.

So I opened the case and started measurements with the 5V-Rail.
The result was kinda expected: Highly contaminated with random noise: In analog-mode the trace is just a 2 divisions wide blurred green band. Digital Mode shows it jumping all over the place.
(Yes, the new LCD for the PM3350 arrived. Thx again to JohnnyBerg. I just didn't have the time to install it )

The upper trace is Channel A, which measures 5V on the Display Memory-Board, the lower one is Channel B, which measures 5V on the Processor-Board.

Next I measured the Calibration-Output: The 1Vpp 2Khz Square-Wave was horribly garbled and also modulating around something else. I suspect Mains-Frequency, but didn't measure it. Maybe I do tomorrow.
You can see how both look like in this video: https://youtu.be/oz7jPVhGgyU


At the time I have three culprits:
- Reference-Voltage (5V-Rail is regulated by a LM358 that uses the Reference to compare the output-voltage):


- 5V Regulator (Something might be wrong here, hopefully not the LM358, because its not socketed and I hate desoldering DIP-Packages^^):


- Capacitors inside the Power Supply:


As you can see, one of the Caps has what appears to be a scorch-mark. I don't have a IR-Thermometer and I won't go anywhere near that PSU when the scope is powered up with my fingers or a thermocouple, but I guess that cap deserves a treatment at the recycling plant.
I think, I'm going to replace all the Electrolytic capacitors with "Old New Stock" my dad got from Philips when he worked there 20 years ago. They are the same style and should fit without problems. He even has 2 of these big fat "Mother-of-Caps" beasts with several hundred µF and rated for over 200V


Do you guys have any more ideas where to look for sources for that kind of problem?

[tautech]

A few thoughts, no particular order:

Why on earth would you be scared of the PSU, don't you have a DMM?  Measure the voltage potential to Earth and then you know if it is dangerous or not.
Little on the secondary side should bite.

Well you can't see or conclude much @ 20 ms/div.
It seems to be a repeditive waveform, so at what frequency?
SMPS frequency would be my guess, ~40 KHz.
Smells like a cap or diode not doing it's job.


2 KHz Cal waveform ?? most unusual.

Removing DIP's.
Snip leads, remove package, desolder pins, install new package.

Scorch mark, more likely a soldering iron burn mark.
Measure the voltages across some of those caps, if not high, don't be scared to touch them and with your other hand behind your back. (no voltage path risk across your body)

I'd be reluctant to replace those fine Philips caps without reason.

IMO forget everything except getting the PSU to Service manual spec at the test points specified.
Then worry about other problems if there are any.

Keep up the good work and post your fixes. 
I've got a 3070 (100 MHz) to fix up in the next few months and all your experiences will be a welcome help, meanwhile I keep watching. 

[PA0PBZ]

Scorch mark, more likely a soldering iron burn mark.

That's what I was thinking, but then look at the lead, it's blackened. Almost as if it spilled the guts and corroded the lead.

[dom0]

Those blue axial Philips caps are notorious for failing. I'd replace them all.

[SaabFAN]

Scorch mark, more likely a soldering iron burn mark.

That's what I was thinking, but then look at the lead, it's blackened. Almost as if it spilled the guts and corroded the lead.

Yep! It literally spilled its guts. (See Picture). That cap is done!
Now I have to get the board out, which isn't so easy^^

On another note: The board n the rear plate shows no sign of corrosion. So the battery-fluid hasn't reached it.
At the moment I'm thinking that, while I'm in there anyway, it might actually be a good idea to replace the Voltage Reference Diodes and upgrade them in the process. The ones used are the C-Version of the BZX79 and the BZV12. They are both about 2% accurate, which surely can be improved with today's available silicon, right?

[Wim_L]

I think, I'm going to replace all the Electrolytic capacitors with "Old New Stock" my dad got from Philips when he worked there 20 years ago. They are the same style and should fit without problems. He even has 2 of these big fat "Mother-of-Caps" beasts with several hundred µF and rated for over 200V

Be careful, electrolytic capacitors that were left unused for many years can become very leaky. They may require reforming before they're good again. At least check them before you install them.

[SaabFAN]

Well, the cap is out and not surprisingly, it had no remaining capacity whatsoever. Instead it had resistance: Fluke stabilized at 2,5kiloOhms after about 30seconds.

So I went to the parts-box and searched: 680µF at 10V. Guess what: Not in there!  And Conrad doesn't have them in stock either.
So I took a Radial one instead. Doesn't look pretty, but should do the job. Its a Yageo (don't know if that brand is any good, didn't have anything better available) 1000µF/16V, Datecode: 11/07.
I don't know if I'm going to order a batch of caps from Farnell, or just go with the radial option for the rest of the caps. There is enough space inside the case to house the radial ones, but it won't look so pretty anymore

Soldering the board is really easy by the way. I have the iron (Weller WECP-20) at about 275°C and the leads almost come out/go in on their own. 


@Wim_L: What's the correct procedure for reforming them? Haven't heard of that before.
I have tested a few of the 1000µ/6,3V Caps with a Component-Tester from ebay (don't know how much I can trust it, but it's based on the circuit of the AVR Transistortester from the german Forum Microkontroller.net: https://www.mikrocontroller.net/articles/AVR_Transistortester). They show a bit high capacitance and 0,2Ohms ESR. Interestingly the Fluke shows OL in Capacitance Mode...

EDIT: What I just noticed: Many of the caps have copper-colored spots where the lead meets the can. Normal or sign of imminent faillure?

[tautech]

Soldering the board is really easy by the way. I have the iron (Weller WECP-20) at about 275°C and the leads almost come out/go in on their own. 

The joy of leaded solder when it comes to rework. 
Nice tidy job with the radial, you could sleeve the leads and make it better.

If you have the correct value but a higher voltage AND it fits....use it.

And because you have the PCB out AND have a cap tester, pull them all and check them.

Google "reforming capacitors"

[SaabFAN]

Update on the caps: I decided to look for replacement-parts and there are basically 2 brands: Vishay (which cost at least 1,50€ per piece) and Multicomp, which cost 0,23€, but don't fit the footprint. So it is either a quite big investment, using radial caps, or make a adapter-board out of the approx. 2 square-meters of prototyping-board and build arrays with some of the 1000 47µ/10V SMD Tantals I have from my granddad
Does anyone know a cheaper source? So far I checked Conrad and Farnell. I don't have a Mouser or Digikey account and don't know if I can get them to sell me stuff as a private customer.

I also measured some more caps: They all check out fine so far: Capacity within Spec (nominal +50%, -10%) and ESR about 0,2Ohms. About the value: I didn't have the exact Value, but the caps are all specified as "+50%, -10%", so a modern 10% 1000µF cap as a replacement for a 680µ shouldn't move the circuit out of spec. Don't know much about flyback-converters, but they seem to rely a lot more on the actual values of the components than the DC-DC Converters I am used to (Step Up, Step Down and SEPIC).

By the way: What kind of trace-material did they use? I guess copper, but its strange looking through the solder-mask: All matte black. All the PCBs I had in my hands so far had rather bright, shiny looking traces. Did they use some kind of special isolation back in the 80s?

Sidenote: While hunting for more bad caps, I found a chip on the memory-board that was made in 1978, according to the datecode. Its an Intel-Chip. Didn't google the number yet though.

[SaabFAN]

And another update:
After replacing the cap, I put it back together (had 1 screw left over  ) and powered it up after verifying that all connectors were correct and searching for the place where the one screw had to go for about 10 minutes.
It worked! The trigger had a few issues, as the seller told me, but it displayed a waveform and ran quite fine actually.
Then I pressed the "Auto" Button and it locked up in Calibration-Mode.
Cycling the power brought me back to where I started: Stuck in Auto Set-Mode and with all kinds of crazy stuff on the waveform-display.

I then checked the voltages I could reach: 7V-Rail showed 7,45V. The negative one showed -7,44V. The 5V-Rail showed 5,1V. The rest I couldn't reach with the probes of the multimeter.

With my little DS203 I then measured the 5V-Rail on the memory-board and was able to capture what looked to be some kind of 250khz-signal in AC-Mode. The analog-part of that small pocket-scope is not the best, but maybe just because of that, I was able to see that 250khz component. It was rather low peak to peak: Just 6mV, but again: The analog-part of the DS203 is pretty crappy so I don't trust measured height of the amplitudes. They claim 8Mhz, but as early as 2,5Mhz, a square-wave is becoming heavily distorted and getting more and more sine-like. At 6Mhz it also drops in amplitude.

In conclusion: The fault probably lies a lot deeper than just inside the PSU. I just hope the damaged cap, which was sitting on the negative 7V-Line, didn't drag any of the ASICs with it to the grave. That would probably mean the end of any repair-attempts as those parts are most likely out of production for 2 decades now.
What I also noticed was the temperature the digital components reached pretty quickly. Especially the ceramic-package ROMs were getting quite warm, as did the processors (socketed 68k CPUs in ceramic package with metal cover over the die). I find it a bit odd, because the CPU in my Sega Mega Drive-II is also a 68k, running at a little more than half the clock-speed and I can barely feel it warming up at all.
While I'm on the topic of temperature: The Tantal-Caps on the processor-board and the IEEE-Board were also getting a bit warm. Are those (they are drop-shaped) caps prone to loosing capacity over time too?

Attached:
5V Ripple - Measured with the DS203 on the Memory-Module

[dom0]

drop-like tantals are also very prone to failure. If they get hot they are probably shorted internally ; a common failure mode for them.

They claim 8Mhz, but as early as 2,5Mhz, a square-wave is becoming heavily distorted and getting more and more sine-like. At 6Mhz it also drops in amplitude.

That's the meaning of 8 MHz (-3 dB) bandwidth

[tautech]

drop-like tantals are also very prone to failure. If they get hot they are probably shorted internally ; a common failure mode for them.

Bead tantalums are reknown as problematic, especially older ones like yours. Either they are having to cope with excessive ripple on the rail, or are failing.
Check the rail for ripple at the leads of the known hot units to be sure.
Then remove and test for piece of mind.
Substitute temporarily if necessary for comparison.

It may be that failing tants are stressing the PSU and the main smoothing caps are struggling to manage the increased ripple from a higher than "designed" PSU currents. This may explain the high ripple on the PSU in the OP.
Detective work required. 

If it where mine I'd replace all tants with modern manufacture Bead Tants.
Watch for polarity.
Try to get a voltage rating 2x the voltage they are exposed to for long service reliability.
AFAIK They are only available to 50V rating.

[SaabFAN]

Okay, I did more measurements: The scope did work for a while btw. (about 5minutes before it stopped updating the traces) It seems that the RAM has to truly forget everything to come out of that boot-lock. But once the scope tries to go into calibration-mode, it locks up.
I attached a few pictures for you to see what it measured. Probeswere in x10-Mode during measuring.
The supply-rails are still noticeably contaminated with noise, but what I also discovered: The 10V Reference-Rail is 0,5V too high and also full of noise - About 300mV. Of course the Adjustment-Pot is right in the middle of all that high voltage-stuff -.-
The other voltages I could measure so far:
7V: 7,45V
-7V: -7,44V
5VDigital: 5,1V
10V Reference: 10,48V
I guess, this is the reason why the scope stopped updating the traces: Some part overheated. Probably the ADC.

I guess, that scope has had quite a lot of hours logged on it, so I'll probably have no other choice than to replace ALL the caps.
As Tantalums are way too expensive, I'm going to use 100µF/25V caps to replace them all. The axial Caps in the PSU will be replaced with radial ones and kept from vibrating with hot glue.
There is a problem replacing the small axial Caps though: Radial Caps won't fit and axial caps with the same specs aren't available. So I came up with a small adapter-PCB that fits the footprint and can be populated with SMD components.

Btw. the cheapest tantalum I could find with the right specs was 78cents
The cheapest Aluminium Cap I found: 4cents!

[SaabFAN]

Okay, I adjusted the 10V Reference-Voltage and the scope is much more reliable now. Although the 5V-Rail on the digital-boards is now down to 4,83 and 4,91V.
But it is kinda running now for about 4 minutes. Then the trace-updates become erratic. Sometimes it recovers, but after a few times, there are no trace-updates anymore. So I guess there is a problem with the ADC or the associated circuits, which means I have to get into the bottomside of the scope to take measurements.
Also the ADC-Chip on the A11-Unit is getting pretty hot. I can just barely touch it and the longer I leave the scope off, the longer it updates the traces.


I was able to get the Service-Mode running (Push the top and the bottom soft-button on the scope at the same time)
Can anyone with a PM3320A (Software-Version: 4.2 L. v. 0s 89-12-01) verify these Checksums?
D1724:84FC
D1726:A75C
D1727:715F
D1728:7DC5

The Software-Tests also indicate a faulty Display-RAM board - When I select the Display RAM Test, all the leds start blinking.
So there are at least 2 parts faulty or at least behaving erratically: Display-RAM and ADC.

Edit: When its cold, the Display-Memory tests "OK". I'm tempted to give it the "PC-Treatment": Add heatsinks!

And about the SMD-Adapters: Would anyone here in the Forum be interested in them too? When I order a batch at ShenZhen2U, I'll probably have a few hundred of them and I probably won't need all of them.
There will be 2 versions: Large electrolytic Caps with more than 7mm diameter and small ones that fit in the places of the smaller caps. They can be populated with 0805 ceramic caps and electrolytic caps and will have at least 2, maybe even 3 different lengths of pins available to fit as many footprints as possible. If they are too long for a footprint, the uneccessary parts of the PCB can be broken off easily.

[SaabFAN]

Okay, I think I'm going to call it on this one. I'm out of ideas.
I managed to get the excessive Ripple under control, readjust the voltages (10V Reference was 0,5V too high), recapped and added caps on the digital boards and most of the power-supply, which also brought down the heat generated in the chips, but still no luck. It works for a few minutes, then the traces fail to update.
Diagnostics-Software shows Display Memory Error, but the display-tests (which draw all kinds of patterns on the display) show no problems.
On top of that, the Service-Manual seems to be for a slightly different model: Mine has 2 Coax-Cables going into the ADC-/T&H-Board and also has 2 adjustment-pots, whereas the Manual only shows one cable and no adjustment-points.

Does anyone of you have any more ideas where the fault could be? Otherwise I'm going to sell this thing. It would be nice to get this scope working again, but I have another project I have to finish and also got to do something for the university so I don't have the time to check every little component.
The CRT is still in pretty good condition and puts out a crisp and bright picture. The front-panel has no damages and the PSU, as well as the Modules A4, A5, A6, A8, A9 and A11 have been equipped with new capacitors where feasible. So I'd say, 70€ should be a fair price.

Edit: WTF?!
I turned it upside down to reach the sandwiched boards on the bottom-side and decided to turn it on.
It now ran for about 10 Minutes and didn't stop updating the traces. I'm using my DS203 and the PM3350 as Signal-Generators and both traces display nicely and I can do measurements and everything. Display Memory-Test still fails though.
Blows my mind to be honest. I was ready to give up^^
I'll let it run for about 30 minutes now to check if there is still hope for this thing. I did spray a good amount of Kontakt Gold 200 oil onto the Euro-Card connectors btw.
There are some parts on the bottom-side that are too hot to touch by the way. The quadratic heatsink right behind the Power-Knob for example.
Because I now had a lot more time to test the scope, I also discovered that the Ground-Level receives an offset of almost +2 divisions when the timebase switches from 200ms/div to 500ms/div.

[macboy]

You should be able to calibrate out the offset. In the service manual there are different adjustments for different timebases. This is because of the way the CCD sampler works, as it samples the ground level every other sample in order to digitally compute the signal level w.r.t. ground after the ADC pulls the samples out of the CCD. I guess this is due to some drift while the samples sit waiting to be shifted out of the CCD. At slower timebases, the ADC operates directly on the input signal, bypassing the CCD.

Mine has about 0.5 division offset when switching to the slower timebase. It bothers me but I haven't gotten around to trying to calibrate it yet.

[SaabFAN]

Ah, thanks for the insight. Once I get it to actually run, I'll do that.

While poking around on the bottom of the powered scope, I managed to come across 12,58V where 14V should be.
So I ordered 8 10.000µF/16V and a few 63V and 100V Caps to replace the last few caps in the PSU. I hope that will bring the voltages back to spec and then it will be put on ebay. Found another one, which, according to the Number, has the FFT-Option installed! (I'll of course only sell it if I can't repair it or get the other one )

One thing I don't understand though: Why is the damn Display Memory failing the Memory-Test after about 30seconds after a cold start, but the display shows no graphics-glitches?

[dom0]

Display memory might not mean the frame buffer, but for example the memory of the display processor. In that case corruption can occur without any visible artifacts.

[SaabFAN]

Display memory might not mean the frame buffer, but for example the memory of the display processor. In that case corruption can occur without any visible artifacts.

You might be right about that. It would also explain, why it sometimes restarts updating the traces when I play around with the settings.
If the fault lies on one of the DPU-Boards, I have a serious problem: On the DPU Control Unit (Module A8), there is a programmable logic device calles PLS105N. It is part of the Start/Stop-Logic and generates control signals for the DPU. It also is the only chip that has a maximum Rating on the inputs of 5,5V, close to the voltages I measured when I first poked around with the multimeter.
The chip is socketed, so I can get it out and a new one in (they are available on ebay), but how do I get the programming out of that chip and into a new one?
There is also a sticker on that chip, reading: "D1457  8719.1"

[dom0]

PLS are essentially a PAL with a few bits of feedback data storage. Or, with other words, they can directly implement a state machine.

The hard part might not be to find a working chip to copy from, but rather to find a programmer capable to do so.

[smjcuk]

This one sounds like an 'FMH' job (F*** Me Harder). I've got a Tek 453 that is doing this to me. I've come across a rationale to manage effort on such things. I draw five lines and £50. I write down the problem on the first line and fix it then I write the next problem down and so forth. If I run out of budget, it gets stripped for parts and disposed of. I'm on problem 5 and £47 on the 453 now. Its hanging by a thread. Currently waiting on obscure FETs to arrive from the US. Failing that, it's a source of quality pots and carbon resistors (low parasitic inductance ones unlike modern carbon film ones).

As a side note, I've had nothing but trouble from Tek scopes. They really dont deserve the reputation they have. Might start a thread on this. And as for philips, the power supplies are total trouble as well. I worked in the TMM department at a large company and there was a stack of dead PM kit always. Usually PSU problems.

As for your problems I reckon one of the boards is probably knackered if its fixed by orientation. Not sure if they're multilayer but if they are its probably a write off.

Good luck though.

[SaabFAN]

Yeah, especially as now a functioning PM3320A/41 (the one's with FFT) and a PM3382A popped up on ebay.

The idea with the lines and the max. budget is a good one.
So far I have nailed down the problem to two places: PSU and the digital boards, probably the DPU. Because the rest of the scope seems to work fine, although the signal has a larger than unsual noise-ground, which is PSU-Related. Even when the Display-Memory-Test fails, I can select the display-adjustment-menu and it displays the different test-waveforms (with which one can adjust the huge amounts of Pots on the DAC and Amp-Board) without any problems. I'm giving this thing one more chance with the 10.000µF-Caps I ordered and will remove the old electrolytic caps (they are in blue plastic enclosures on the digital boards) to reduce possible leaking-currents - This brought down the heat-generation on the Processor and Processor-RAM boards significantly. Now the ROMs produce almost no heat at all. If it's not running, I'm selling it. I bought it for 70 and invested about 30€ in Caps by now.

[smjcuk]

Not a bad investment even if it doesn't work. At that price you can shift it for what you paid for it

[SaabFAN]

While waiting for the caps to arrive, I did some research on the PLS105-Chip I suspect to be fishy.

It seems as if that type of chip was the first or at least one of the first FPLAs and was called 82S105 for a time and produced by Signetics, as well as Texas Instruments.
The datasheet states that it is compatible with Jedec Design Software and therefore I think it should be programmable with a universal programmer.
Unfortunately, the only compatible programmer I found were the really expensive ones for 800€ and higher. Does anyone here have a programmer that could read and write that chip, or knows how to make the much more affordable TL866 compatible with that chip?

[guido]

Those PLS things loose their content.... Similar to "ROM-rot" with old Mostek ROMs.

I can read/program them (acient All-07 from hilo). I can even "assemble/disassemble" the contents. In case it is (partly) empty, that only works if you know wat needs to go in (if it is simple or well described, maybe the logic rules could be figured out). However, if that is clear you could also replace it with a more generic GAL or so.

They are also used a lot in Tek 27xx spectrum analysers (in SMD variants). But for those machines, at least somebody has dumped the contents. Maybe somebody has done that also for this one, try google...