Hewlett-Packard 3488A Switch/Control Unit
Posted: Tue Aug 19, 2025 3:36 pm
This was a fiver bargain buy at the Flight Refuelling rally but it needed a wash and brush up as well as a new cap for the mains power switch.
Who was I kidding?
As soon as the cleaning was complete I tried out the now easy-to-press power switch a few times. And you know what is coming. It stopped powering on.
I sighed and decided to examine the PSU, which is really easy to extract. It's a secondary switcher with 30V DC to work with and 5V at 5A to output. It also hosts the buzzer and sends a couple of status conditions to the instrument's computer using TTL, which we'll return to.
First observation was that the switching transistor isn't a transistor, it's a PIC601 a TO-66 IC with 4 leads and an isolated case. This is actually a combination device containing a PNP, NPN and diode, all on my favourite beryllium oxide substrate apparently. Lovely. Bound to be broken, right? It's always the switching element that fails isn't it?
This was easy to extract using my Engineer solder sucker, which I recommend very strongly indeed if you haven't got a hakko hot pump thing. Anyway, I found a 12V 2W M.E.S. light bulb and used the PIC601 to turn it on and off with a 470Ω driving the switching input. It was fine.
Maybe the output inductor or the current sense resistor was open? Nope.
Maybe the capacitors are bad? Nope.
PIC601 was driven by a 2N2219A in a sort of cascode to the internal transistors in a semi-ubiquitous SG3526J PWM control IC, with a comparator controlling the 2N2219A for reasons of startup sequencing or similar. The 2N2219A was fine.
The SG3526J was getting warm but not running, which was odd, the VREF pin should have been 5V but it was 4.3V. This seems peculiar so I decided to help it get to 5V with a carefully current limited external 5V supply. This caused the faulty supply to spring into life and deliver a healthy 5V to the 39Ω resistor that HP helpfully include as a dummy load with a link that you can bridge to use it (yes, really!).
So… the SG3526J works, but the VREF is low. Looking at the datasheet for this venerable old chip, apparently they sort of encouraged designers to make use of the generous 20mA of VREF current and even showed a way to boost it to 1A, so you could run your power supply health monitoring and power-on sequencing from the VREF before the supply comes up. HP loved this idea and there are plenty of voltage dividers hanging off it, driving LM339s which are powered by a separate LM317 hanging off the 30V rail and providing 12V before the SMPSU starts. I changed the LM317 just in case it was the problem. It was not the problem.
One of the things hanging off the 5VREF line was a CMOS dual monostable that works in combination with the LM339s to produce healthy supply flags for the computer. Old CMOS can do strange things and after the 1Hz oscillator running at tens of MHz that I found in that RACAL 9084 generator I decided it was probably best if I changed it. No, the MC14538BE was innocent and got chopped out for nothing.
The other chip that was literally being powered by the 5VREF line was a 74LS10, which combines the health messages from the various LM339 outputs and sends them to the computer (despite it having no power at this point, go figure). I found a 3.3Ω resistor that allowed me to monitor the current that all the 5VREF stuff was drawing, it was over 50mA which is way over the limit, so not surprising that the PWM chip was getting warm and not starting. LIfting the vcc power pin on the 74LS10 and powering it from an external supply was interesting, sometimes it would operate normally and draw the proper quiescent current of a few mA but other times it'd draw 56mA. Oddly there was no sign of oscillation on the outputs, it was just drawing way too much current.
This logic chip whose outputs aren't even used in the power supply's control loops and only serve to tell the instrument's main computer about the state of the supply was causing the entire supply to fail to start. Replacing the 74LS10 returned the supply to normal operation. I'm generally very wary of fixing switched mode power supplies, even if they're only low voltage ones, as the control loops can be very destructive in all sorts of fault conditions.
Later versions of this power supply do not use the SG3526J+PIC601 combination and are a little simpler as a result - they also run the LM317 supply at 5V output, as there's no need for 12V, but they seem to have a different stock fault and were subject to warranty repairs.
My remaining problem with the instrument is the silly flat ribbon cable that connects the PSU to the main backplane - it's one of those failed experiment HP ribbons that turns into a self-destructing sticky mess if you so much as look at it. For now I'm ignoring it and pretending that having multiple bare conductors about 0.1mm apart is perfectly normal. It's only bad if they touch and they're inside the box, it'll be fiiiiiine!
Who was I kidding?
As soon as the cleaning was complete I tried out the now easy-to-press power switch a few times. And you know what is coming. It stopped powering on.
I sighed and decided to examine the PSU, which is really easy to extract. It's a secondary switcher with 30V DC to work with and 5V at 5A to output. It also hosts the buzzer and sends a couple of status conditions to the instrument's computer using TTL, which we'll return to.
First observation was that the switching transistor isn't a transistor, it's a PIC601 a TO-66 IC with 4 leads and an isolated case. This is actually a combination device containing a PNP, NPN and diode, all on my favourite beryllium oxide substrate apparently. Lovely. Bound to be broken, right? It's always the switching element that fails isn't it?
This was easy to extract using my Engineer solder sucker, which I recommend very strongly indeed if you haven't got a hakko hot pump thing. Anyway, I found a 12V 2W M.E.S. light bulb and used the PIC601 to turn it on and off with a 470Ω driving the switching input. It was fine.
Maybe the output inductor or the current sense resistor was open? Nope.
Maybe the capacitors are bad? Nope.
PIC601 was driven by a 2N2219A in a sort of cascode to the internal transistors in a semi-ubiquitous SG3526J PWM control IC, with a comparator controlling the 2N2219A for reasons of startup sequencing or similar. The 2N2219A was fine.
The SG3526J was getting warm but not running, which was odd, the VREF pin should have been 5V but it was 4.3V. This seems peculiar so I decided to help it get to 5V with a carefully current limited external 5V supply. This caused the faulty supply to spring into life and deliver a healthy 5V to the 39Ω resistor that HP helpfully include as a dummy load with a link that you can bridge to use it (yes, really!).
So… the SG3526J works, but the VREF is low. Looking at the datasheet for this venerable old chip, apparently they sort of encouraged designers to make use of the generous 20mA of VREF current and even showed a way to boost it to 1A, so you could run your power supply health monitoring and power-on sequencing from the VREF before the supply comes up. HP loved this idea and there are plenty of voltage dividers hanging off it, driving LM339s which are powered by a separate LM317 hanging off the 30V rail and providing 12V before the SMPSU starts. I changed the LM317 just in case it was the problem. It was not the problem.
One of the things hanging off the 5VREF line was a CMOS dual monostable that works in combination with the LM339s to produce healthy supply flags for the computer. Old CMOS can do strange things and after the 1Hz oscillator running at tens of MHz that I found in that RACAL 9084 generator I decided it was probably best if I changed it. No, the MC14538BE was innocent and got chopped out for nothing.
The other chip that was literally being powered by the 5VREF line was a 74LS10, which combines the health messages from the various LM339 outputs and sends them to the computer (despite it having no power at this point, go figure). I found a 3.3Ω resistor that allowed me to monitor the current that all the 5VREF stuff was drawing, it was over 50mA which is way over the limit, so not surprising that the PWM chip was getting warm and not starting. LIfting the vcc power pin on the 74LS10 and powering it from an external supply was interesting, sometimes it would operate normally and draw the proper quiescent current of a few mA but other times it'd draw 56mA. Oddly there was no sign of oscillation on the outputs, it was just drawing way too much current.
This logic chip whose outputs aren't even used in the power supply's control loops and only serve to tell the instrument's main computer about the state of the supply was causing the entire supply to fail to start. Replacing the 74LS10 returned the supply to normal operation. I'm generally very wary of fixing switched mode power supplies, even if they're only low voltage ones, as the control loops can be very destructive in all sorts of fault conditions.
Later versions of this power supply do not use the SG3526J+PIC601 combination and are a little simpler as a result - they also run the LM317 supply at 5V output, as there's no need for 12V, but they seem to have a different stock fault and were subject to warranty repairs.
My remaining problem with the instrument is the silly flat ribbon cable that connects the PSU to the main backplane - it's one of those failed experiment HP ribbons that turns into a self-destructing sticky mess if you so much as look at it. For now I'm ignoring it and pretending that having multiple bare conductors about 0.1mm apart is perfectly normal. It's only bad if they touch and they're inside the box, it'll be fiiiiiine!