Solid as a rock – Fixing an old AC stabilizer

Sometimes, a piece if equipment finds its way to me for repairs. This time, it was an old Devere voltage stabilizer, apparently used in conjunction with a 5108 enlarger. Very nice piece of kit. With one problem – it didn’t work. But I think we got ‘er fixed up again.

These things are like sudokus to me, really. The sane thing in a case like this is probably to not even bother with a repair. One option (which I in fact suggested to the owner as a temporary workaround) is to simply forego the stabilizer altogether. He prints B&W and it’s unlikely he’ll ever notice any effect of grid fluctuations on his prints. Another option is to simply purchase a new stabilizer; they’re still being made, and especially if it doesn’t need to be insanely powerful, they can be quite affordable.

Here’s our BBB – Big Beige Box. Actually photographed after repairs – note the dial that’s nicely where it should be.

Then again – this unit is rated at 1.5kW, and a quick search for cheap replacements were more like 600W or thereabouts. And there’s the whole business with solar panels, which means that passing clouds can and *will* create voltage fluctuations on local grids. Heck, not just fluctuations – our nominal 230V may very well run around 250V or so at times. Incandescent/halogen enlarger bulbs are often run a little hot to begin with, at the cost of a reduced lifetime. Run them at higher voltage, and I wouldn’t be surprised if their lifetime is cut short quite dramatically. And besides, if you have one of these nifty stabilizer units, I imagine you want it to actually work, right?

So I set to work with the screwdriver and the multimeter.

But first, what was the problem with this one? It was brought in with the generic problem description “enlarger bulb doesn’t come in, checked all fuses but they’re OK – please help.” Powering up the actual power supply and timer (not pictured) directly from a wall socket suggested those are just fine.

The output from the stabilizer, however, wasn’t fine. It was reading way low – 150V or so. Hmm. What’s more, when plugging it in, the unit would make a brief humming noise, the dial would drop a little and the voltage would settle a little lower, still. I tried adjusting the little screw thingy on top of the dial, which turned out to be the voltage adjustment (in fact, it turned out to be an ordinary 10k potentiometer). Adjusting it down seemed to work well – something buzzes inside, voltage goes down. Up, however…not a chance. Not good, but hey, it does something at least.

Alright. Let’s have a look inside, then. This is what I found when I took away the pretty front cover:

Behind the front panel: a big variable autotransformer, driven by a small motor, and some control electronics hidden inside another box.

Up to this point, all stabilizers for enlargers I had worked on were Durst units – basically phase-control thyristor circuits. This is something else! A little Googling tells me that this approach is actually still used today, even though it looks like something that would pair well with mercury valve rectifiers and TV-sets covered in wood veneer.

The concept is simple: at the bottom is a variable autotransformer (a.k.a. Variac) with an output range of something like 10 – 250VAC. Mounted above this is a little AC motor, coupled with a gear to the variable transformer. A magic electronics box above controls the motor: it monitors the output voltage, and turns the transformer down if the voltage is too high, and up if the output is too low. Simple, effective – and quite efficient, because the autotransformer is essentially lossless. And it’s a nicely constructed and engineered unit, to boot.

Another view, with the top electronics box moved out of the way. The AC motor is the black box at the top of the picture. It drives the central shaft of the autotransformer through a gear linkage. Two limit switches prevent the motor from stalling if tries to turn out of the variac’s range.

Not only nicely constructed – also in very good nick given its 35+ years of age. No corrosion, gears nicely lubricated, everything running smoothly…and yet, it doesn’t work. Sounds like an electronics problem. So I dug into the little box that says “Rilton Electronics”. And yes, I tried googling the type number, and it provided exactly zero useful hits. Drat.

The underside of the control box. I always take a snapshot of this so I know what goes where when the times comes to put it all back together again. The left five contacts go to the high-voltage electronics: black is neutral, green is live regulated output, red is live unregulated input. The other bundle goes to the voltage adjustment pot meter in the front panel; yellow/green turned out to be virtual ground for the DC circuit inside, white is the wiper and yellow is the top, which should be at 12VDC.

Inside the box, there’s a sandwich of PCB’s and some electronics that make me go all woozy and nostalgic:

Note the red terminal block at the top where the connections from the outside world come in. The top PCB shown here is a single-sided job with mostly discrete through-hole components on one side and a couple of leads sticking out on both top and bottom side.

Underneath that PCB is a metal frame, and at the bottom of that is another PCB with two transformers and two relays mounted on it. Sorry, I didn’t snap a pic of the relays, nor did I jot down a type number, but they were Omron 24VDC relays. The motor that drives the variable transformer shaft is bidirectional, and one relay is used to drive each of the two coils in the motor. So switch one relay and the motor turns left, switch the other and it turns right. Easy-peasy.

Here’s a look from the long side with the top PCB lifted away (seen in the background). Note the metal frame and the bottom PCB with thick traces and two small transformers and one of the relays visible in the bottom right corner.

Oh, and did you note that inspection sign? 4-2-‘ 87! I assume that would be 2 April 1987, which makes it almost exactly 37 years since this was put together. No signs of anyone having been in here since the device was assembled, which means that this thing must have been operating problem-free for close to 4 decades. Not bad, Rilton people.

At this point, I couldn’t resist. A nice collection of discrete components. Nothing too complicated, everything fairly well accessible – I felt I should attempt to figure out what’s wrong, and hopefully fix it. How hard could it be, after all!?

I started number and identifying components just to get a feeling for what’s going on.

There’s a bridge rectifier top left, and two big electrolytic capacitors. So some kind of power supply section. Then a collection of PNP and NPN transistors; BC557, 558, 547 and 548. All run-off-the-mill, common types. Some diodes that color-decoded into 1N4148 – nothing fancy. A couple dozen resistors and a handful of tantalum capacitors. Two diodes that I figured had to be Zener types and I could just make out something along the lines of BZX55-15, which would make them a common 15V type. A lonely NEC 78L12 12VDC linear regulator in the middle of the board. And the only two IC’s are easily identifiable as 741’s – the archetypical opamp.

All considered, it looks like the engineer made this cannibalized his 12-year old son’s “Get started with exciting electronics projects – everything you need to get started included instruction booklet!” box. Which I like, because it means that nothing outlandish is happening here and even at my pay grade, I should be able to figure this one out.

I numbered the parts according to no particular logic. The numbering certainly didn’t follow any schematic, because, well…I had none. So that was the next step. I started tracing connections and after some poking around with a continuity meter, I came up with this Monday morning sketch by a hungover Piet Mondriaan:

Don’t try to read it. It’s useless. I’ve got a cleaned up version, but be ware, it does not follow the same component numbering in the photo above. So if you happen to work on one of these units and that’s why you’re reading this blog (I couldn’t imagine any other plausible reason), please keep this in mind.

Reconstructed schematic of the control unit

While the first sketch seemed rather chaotic, the second version immediately above shows a clear logic. And indeed, things aren’t quite as intimidating as they seemed before.

Top left is the power supply section. It’s a simple affair with an AC voltage being rectified. A +24V (more like 28V in reality) is used to feed the relay coils. For the control electronics, a symmetrical +/-15VDC supply is made with two zener shunt regulators, using a BC547 NPN BJT as a regulator for the positive supply and the complementary BC557 PNP for the negative side. The 78L12 turns out to be used only as a voltage reference used in the error amplifier section.

This error amplifier consists of two mostly identical parts around two 741 opamps. An voltage is derived from the second AC transformer, half-wave rectified and filtered. This is fed into the opamps, and compared to a reference voltage that’s set with the front-panel 10k pot meter together with a ‘ ‘sensitivity’ 500 Ohm pot meter that’s only accessible after removing the front panel. The sensitivity setting controls the hysteresis and offers a way to compromise between tight output voltage control (but relays that chatter pretty much constantly and a motor that’s hopping back and forth) or somewhat more relaxed output control (and a much quieter work environment).

The error amplifier outputs feed a relay driver circuit for each relay. The top right one controls the relay that turns the variable transformer down – so it activates whenever the output voltage surpasses the set voltage. The bottom driver circuit drives the transformer up whenever the output voltages sags below the control value. Save for an additional inverter stage in the top circuit, both drivers are nearly identical. A 2.2uF capacitor creates a delay that prevents the relays from chattering.

Annotated photograph showing the sections of the circuitry. The problem section on this board was the lower right section (‘voltage up’ control). Note that the component numbers don’t match the schematics!

And since these are all discrete components, and I was drawing schematics anyway, I figured I might as well play with it a little, too. So I put the cleaned-up schematic (without the power supply section) back into LTSpice and gave it a go. This proved immediately useful, since it helped me to remove one or two transcription errors from my schematic. Here’s the version that I simulated:

And it also showed that in principle, the circuit should work quite nicely:

In this simulation, I fed the input with a slowly fluctuating signal, which should alternatingly trigger the increase resp. decrease voltage control subsystem. And indeed, it does – I set the hysteresis to the maximum value for this plot, which shows how nicely the error amplifier circuit (traces Comp1 and Comp2) prevent both relays switching at the same time. The voltage at the bottom end of the relay coils (VRL1 and VRL2) switches in accordance with the error amplifier outputs, but also follows the delays programmed with the two 2.2uF capacitors.

Pretty neat. But in reality, it still didn’t work at this point. That is to say, in an experimental setup, I could see that one side of the circuit worked just fine: the downward regulation. But the upward regulation was as dead as a doornail. The error amplifier outputs both looked OK, but no relay switching on the upward regulation part.

Bench-top “don’t try this at home” test setup

I did the obvious things. I measured the relay coils, and even disconnected the leads between the control PCB and the relays to trigger them with a bench power supply. Both worked just fine. I probed the transistors in the problematic section with a diode tester – and found no anomalies. This was odd, because I was really expecting to see one of the PNP’s fried. In fact, I was so sure that this would be the problem that I desoldered the transistors from the affected section and tested their Hfe. They all tested perfectly fine. And of course, I measured voltages throughout the section as well, and everything looked OK, and at the same time, not at all.

The one thing that was really not right was this bit here; the first stage after the comparator/error amplifier output:

The voltages with the idiotic number of decimals are what the simulation comes up with. The single-decimal voltages are my own measurements. As you can see, around the emitter of the PNP (in reality a BC557, but I didn’t have one handy in LTSpice), things aren’t right. The simulation suggests the voltage around R14 should drop to around 4.5V. My measurements put it pretty close to the 15V supply voltage. Odd, because the base of Q1 is fed with the (high) output from the 741.

The numbers just didn’t add up. So I went downstairs for a cup of coffee.

And then it dawned upon me – as it actually should have done pretty much right away when I opened up this electronics box. The capacitor.

C3 in the image above is a nice little 2.2uF tantalum capacitor. It looked absolutely fine, just like the other identical ones, and there was no indication that anything was amiss. Except that if this capacitor leaked, it would explain perfectly well why Q1 didn’t appear to do doodly-squat, while in reality it tested just fine. Q1 could do whatever it pleased, but if C3 passed current, all bets would be off anyway.

So I plugged in a virtual “leakage resistor” in parallel with C3 and simulated the whole thing again:

Lo and behold – the simulation all of a sudden matched the actual situation nearly perfectly! Alright, I did fiddle with “Rleak’s” value a bit to find the best fit, and found it would have to be around 4-5kOhms.

This at least gave me a very concrete thing to check, so I whipped out the soldering iron and took out C3. How a leaky capacitor behaves in a circuit is a different matter from how it measures with a DMM, but still, I couldn’t resist. Guess what – taken out of the circuit, it actually measured almost exactly 4k5!

With confidence, I put in a replacement 2.2uF cap, powered up the whole thing and lo and behold – it worked absolutely fine. Evidently, I replaced the other capacitors as well as they were the same or similar types as the defective one. No use sending this unit back to its owner only to have it returned a few months down the line for a similar (or worse) problem.

I reassembled the unit and had it run for a couple of hours on my test bench, which it spent merrily clicking and buzzing away whenever the grid voltage fluctuated slightly (solar panels!)

Of course, I could have simply started replacing things left and right and I might have finished quicker. But where’s the fun in that? Besides, I might not have gotten lucky and would have been none the wiser after a lot of work.

But admittedly, when I say those tantalum capacitors, I might have jumped at them a bit quicker than I did. I’ll be mindful of those, next time.

Leave a Reply

Your email address will not be published. Required fields are marked *