Introduction

6 years ago I acquired an HP 8640B that needed work (see HP 8640B Part 1: initial state).

Over the last few years, I managed to clean up the unit cosmetically, but not dive in to it any deeper. Until now.

I recently found some time to work on it again and started looking at the power rails. I knew that (at least) the -5.2V rail had intermittent issues, and had noticed the caps on the AC input needed to be replaced as well.

RIFA caps

AC Line powered devices usually have some filtering on the AC input. As part of their required EMI/RFI suppression requirements there might be a capacitor between line and neutral, and another between line and ground. Since these caps are connected to line voltage, they need specific safety certifications.

Class X capacitors are used between line and neutral, whereas class Y capacitors are used between line/neutral and ground(PE). IEC 60384-14 has different classes with different peak voltage and peak impulse voltage ratings. The capacitors that meet the requirements for industrial equipment are rated as X1 (between line and neutral) and Y1 (line/neutral to ground). Common household appliances have lower requirements, and can use the class X2 and Y2.

As Y capacitors are tested to fail open instead of shorting out, Y class capacitors can replace X class capacitors but certainly not the other way around.As X capacitors are connected between the line and neutral, they are allowed to short as that would simply cause the upstream circuit breaker to trip.

Some good additional information can be found in the document “Capacitors for RFI suppression of the AC line: basic facts, released by EVOX/RIFA in 1996.

Over the years of its production, several input filter modules have been used on the 8640B.

Line filtering AC input filtering in the 8640 with the filter (middle), power switch and indicator light (top), voltage selector board (bottom) going to the transformer (right).

Mine had an external RIFA branded class X capacitor between line and neutral:

RIFA cap The soon-to-be-replaced x-rated cap between line and neutral.

These capacitors can often be found on older equipment. Some of these are known to develop tiny cracks over time. As moisture finds its way in through the cracks, these capacitors become dangerous. Plenty of horror stories can be found easily. I replaced mine with another with better specifications.

Mains filter The mains filtering and voltage selection module. RIFA cap left top.

Heat damage

The A12 rectifier board is sandwiched between the potted transformer and the big filter caps:

A12 Rectifier board The A12 rectifier board (1) sandwiched between the transformer (2) and the filter caps (3).

A quick look at the board shows that two power rails had been drawing too much current for the rectifier board to thermally sustain:

Burned rectifier board back side Burned rectifier board back side

Burned rectifier board front side Front side with some diodes removed.

The scorched bridge rectifiers were those for the +5.2V (CR9-CR12, bottom) and -5.2V and the rear fan driver (CR17-CR20, top) power rails. According to the manual, the +5.2V is rated for 2.25A and the -5.2V for 1.75A. Several of the diodes had not survived the heat, as expected. Replacing all of them is the only sensible option.

The back side needed some work as well. Several traces were scorched and needed repair:

Damaged traces Damaged traces on the backside

Remarkably, all of the “big” electrolytic caps were still in spec, even after all the thermal abuse.

Maybe we can patch this… Temporarily?

At first, my reasoning was that I would temporarily patch the rectifier board so that I could continue and assess damage further down the power path. I had some suitable diodes in stock that could be used to replace the damaged can diodes. The legs of those were too big to fit the holes left from the previous diodes, so I had to drill the holes slightly larger:

One of the diodes replaced by a suitable recent component One of the diodes replaced by a suitable modern version

Clearly this was going to be a bodge job, possibly even making further troubleshooting more difficult by causing vague issues. If you can’t start troubleshooting from a known good state, you’ll never know what you’re looking at and might see the effects of another issue further upstream…

I also took another look at the backside of the board. This made me doubt it was worth to put effort in “temporarily” fixing the board.

…Or maybe we shouldn’t ?

The rectifier board consists of 4 bridge rectifiers and an input crowbar circuit. That last circuit is between the +44.6V rectifier and the regulator. When the output of the rectifier exceeds 75V, the SCR enables Q1, shorting the rectifier. The resulting current draw blows the primary fuse.

Since this is a pretty basic board, it made sense to lay out a replacement board using modern components. The steps are simple:

  • Take measurements of the board itself and the edge connector on the bottom, carefully copying the dimensions to the pcb layout
  • Find suitable replacement components
  • Draw schematic
  • Layout board

In honour of the original boards, no soldermask was used. For the prototype board I went for HASL. The final version will be ordered with an ENIG finish.

I came up with this as a first attempt:

First attempt First version of the replacement A12 rectifier board.

Even though the backside is stuck very close to the transformer (there is about 2mm of clearance), I added large copper pours at the back, hoping it would provide a little bit of additional heatsinking.

A colleague of mine challenged me to try to add a light behind the eye of the fish/logo. Why not?

After populating the board, it is ready for testing.

Damaged traces (again)

With the board populated, I started testing functionality. Given I don’t generally just trust my own work, the board was tested out of the HP first. I used an old variac to generate the correct AC voltage levels for each of the power rails and used an active differential probe to measure the rectified output without a load.

Testing with a variac as input. Using a variac to replace voltages from the transformer.

After connecting the +44.6V input voltage and powering on, a pop could be heard and there was no output on the edge connector. Looking on the back side, I quickly realized I had not given enough thought to track clearances. With an input of 51Vrms, the peak voltage (Vp) on the tracks on the input to the +44.6V AC rectifier is up to +/- 72V. This could create a spark between tracks, briefly ionizing the air around it into a conductive plasma and vaporize the trace.

Vaporized trace The vaporized trace between the via and the copper pour.

Besides not having used enough clearance, not having a solder resist does not help.

I found an old, expired can of conformal coating in my box of chemicals. The coating had become too thick to apply with anything besides a spatula, but by mixing in some synthetic thinner as a solvent, it became liquid enough to use an old brush to get an (uneven) layer of coating on the problematic traces of a new board.

Checking under a suitable UV light source shows the messy application. As a conformal coating, this certainly would not be acceptable, but to proceed with testing it is good enough.

Conformal coating applied Conformal coating under UV. Not pretty but enough for testing.

With the conformal coating cured, I tested the +44.6V rectifier again. This time, there were no sparks and the board was ready for the instrument.

I first tested the power backplane (board A17 “power supply mother board”) for shorts, removed the power regulator boards (A18, A20 and A22) and, using the supplied connector extender, checked the power going in to the regulator boards. After adding them one by one and checking all outputs, the 8640 was ready for testing…

Initial testing

First of all I wanted to check if the YAG oscillator was still working. If that part had issues, I would have to reconsider where I wanted to go with this restoration.

The best reference I have available is my TTi TG1006 which generates waveforms up to 10MHz and has a counter that works up to 120MHz. I used that to test the generator as well as the counter functionality.

Testing the counter at 5MHz Testing the counter at 5MHz.

Testing the generator at 20MHz Testing the generator at 20MHz.

This looks good enough as a start. I’ll have to test at other frequencies, but for now, other issues are more pressing…

Rectifier board overheating

After some testing, and even though the instrument top cover was removed, I noticed that the rectifier board had become quite warm. Taking it out after a while and checking it confirms that there is something wrong:

Thermal view Out of focus, but clearly rather warm…

As could be expected, these are the +5.2V and -5.2V rectifiers that were burned in the original board as well. The manual has some clear steps on how to proceed and check the regulator boards, which will be the next step.

Doing this and figuring out what causes the overheating will be for next time…

References and footnotes