Ouch! Watch out for power supply memory buttons

What is it about power supplies?  They are so very boring, until suddenly they bring excitement. It’s always the wrong kind of excitement, too…

It was a quiet day at work. I was using a GW Instek PST-3201 bench power supply, a nice unit with three 30 volt/1 amp outputs. It was purring along at 12V and a couple dozen milliamps, powering a rare and expensive prototype while I did firmware development. Then, when I reached over to grab something, I bumped the “RECALL ↑” button. Bang! The supply’s relays clacked, and suddenly all three outputs were configured for maximum output. The supply pushed a full 30 V, with a 1 A limit, into my circuit.

Needless to say, I switched the output off as fast as I realized what happened. Thank heavens for the clatter of those relays, which told me something unexpected had happened. With the output off, it was time to check for damage.

To my surprise, there was none. The power supply wasn’t plugged directly into the prototype — instead it went through a test board that interfaced a PC to the device under test. By very fortunate coincidence, the wimpiest chips on that board were rated for a hefty 36V, and some were rated for as high as 42V. (Those “wimpy” chips weren’t wimpy in any real sense — they were some LT1970 high-power op amps, more than capable of hefty output of their own.)  The interface board was undamaged. Better yet, none of the high voltage made it to that precious prototype.

As I breathed a sigh of relief, I looked for the cause of the accident. It turns out this supply came from the box with half of its memories set for 30 V / 1 A and the other half set for 0 V / 0 A. Hitting “RECALL ↑” took it to one of the 30 V memories.

This could have been much worse. It was sheer luck that all of the parts I picked for that interface board had maximum ratings higher than 30 V.

Needless to say, if you happen to use GW Instek PST-3201 supplies or the closely related PST-3202, it would be wise to check the memories and set them all to zero volts. That’s what I did as soon as I realized I was not going to have to tell my boss I had blown the prototype.

Updated 3/24/12: Corrected the name of the button at fault. It is “RECALL ↑”, not “MEMORY ↑”.

Cool chip: LT1970 power op amp with current limit

LT1970 power op amp configured for a gain of two and a variable current limit inputThis week I want to share a new favorite chip of mine, the LT1970 power op amp from Linear Technology. This nifty device runs from a 36 V supply and can source or sink 500 mA. That’s not too unusual, but how about programmable current limiting, outputs to indicate over-temperature and over-current, and more? Think of it as a four-quadrant power supply on a chip, or as the source part of a source measurement unit.

The first nice feature of the chip is the programmable current limit. The chip has two inputs, one for sourced current and one for sunk current, that accept a voltage from 0 V to 5 V. It compares these voltages against the current measured across an external sense resistor and limits its output when the measured current exceeds 1/10 the voltage on the relevant control pin. Differential input pins, independent of the output pin, are provided for the sense resistor.

Another great feature is that the output stage has its own power pins. One of the neater “trick” circuits out there is to boost the current capability of an op amp by using the power supply pins to drive external transistors. An ordinary op amp without separate power pins can also be used this way, but the bias current of the external transistors will be proportional to the op amp’s quiescent current, which means one will probably need a low quiescent current model. By pinning out the output stage’s power pins, the LT1970 makes it easier.

This circuit ends up running the output transistors in class AB. The bias current in the transistors is controlled, in part, by the quiescent current of the op amp. When the op amp tries to increase the voltage of its output, it will deliver more current to the output, pulling that current from its power supply. That, in turn, changes the voltage on the base of the upper external MOSFET, causing it to source more current as well. The equivalent happens when the op amp needs to sink current.

Another advantage of choosing the LT1970 for this configuration, instead of an ordinary op amp, is that all stages except the output stage will see a stable power supply, keeping the PSRR up where it belongs. As for the last stage, when combined with the external transistors, it resembles two Sziklai pairs (also called “complementary Darlingtons”) biased into class AB.

The separate power pins also make it easy to use a lower-noise supply for the input stages and a high-current supply for the output. I used them for another purpose: My application needed a wide input range but drove an input sensitive to overvoltage. Running the input stages off of 24V and the output stage from a lower voltage ensures the output will not hit an unsafe value. Watch out, though, because V+ and V- must be within the VCC and VEE supplies to the op amp’s input stages.

Another neat feature is three open-collector outputs that indicate that the op amp has gone into thermal limiting, into current source limiting, or into current sink limiting. These pins can drive LEDs for an easy status display.

The biggest drawback of the LT1970 is its package, a 20-pin TSSOP with a thermal pad on the bottom. That thermal pad means you pretty much have to have either a hot air or reflow setup to solder it.

As you might expect for a part like this, there are ways to use it beyond the obvious. The current-limiting flag outputs can be used for snap-back current limiting, where the output current is sharply reduced after it enters limiting. The current sense pins can be used as voltage-limit inputs, enabling circuits like a symmetric, voltage-controlled limiter. These ideas are from the data sheet; I haven’t tried my creativity at coming up with other ways to (ab)use the chip.

No, I used the chip for a boring old power supply. It was a two-quadrant supply, though, meaning (in this case) that I needed a positive-output power supply that would either source or sink current to maintain its output voltage. The voltage was controlled by an analog input. What I needed was basically just an op amp with current limits and a few hundred milliamps output. The LT1970 was perfect. I fitted it with potentiometers to adjust the current limits, a set of LEDs for the status outputs, and as much PCB copper as I could manage for a heat spreader. I also included some compensation for capacitive loads. It worked great. I haven’t had any issues with oscillation, the LEDs have been handy to tell me when I mess up, and the current limit has saved my bacon at least once. Sure, I could have built the same supply from separate parts for a lower cost, but it would not have been nearly as easy or as quick. This is another case where it was worth spending a little money on a chip in exchange for saving a little engineering time.

The feature I miss most from the LT1970 is a current amplifier output. It does current measurement internally, for the current limit feature, but a voltage proportional to current is not brought out. For my power supply, I had to add an external difference amplifier to measure the output current. The difference amp used the same current sense resistor as the LT1970, but it would have been nice to have that function built in.

The quantity 1 price for the LT1970 is $9.88 from Digi-Key or $6.05 directly from Linear Technology.

Figures: Copyright 2002 Linear Technology Corporation. Used by permission.

Crazy PCB Layout from the Big Hair Decade

Check out the bizarre PCB layout on this power supply.  A Pulse Instruments PI-702, I bought it for a few dollars at a hamfest. It was made in the mid-80’s and plugs into Tektronix TM500-series mainframes. When I powered it up, BANG!, after which it had no output.  Here is what I found when I opened it up for a look:

The curvy, hand-taped traces are typical for the period, but look at how few components are in the two-thirds closest to the front panel compared to the number of pads! Plenty of those pads look like they have an 0.3″ DIP pattern, but have either discretes or nothing soldered to them.  The layout is also full of dead-ends and traces that don’t go anywhere, and there is no silkscreen.  The back third is neat and tidy — it is all a bit Dr. Jekyl and Mr. Hyde.

All of this leaves me wondering what the crazy layout is for! If it is meant to dissuade reverse-engineering, it might work (it worked for me, so far…), but who would want to protect something as simple as a linear power supply? It is even stranger that the digital-to-analog conversion circuitry near the edge connector gets so little of the board, and is laid out quite cleanly compared to the power supply.  I suppose that this might be a case of multiple models using a single PCB, but what a devious mind it would take to merge multiple schematics into something that looks like this!

The problem itself was easy enough to find. One electrolytic capacitor dried out and blew up. You can see it at the lower-left of the second photo. It will be easy to fix, assuming it didn’t take any other components with it.

The device itself is interesting.  It has three outputs, two of which are bipolar, each covering -25 V to +25 V continuously. They can be independently set with front panel controls, track an external input, or be controlled digitally. The bipolar outputs are limited to a wimpy 25 mA each, which is undoubtedly why it is called a bias supply. It also has a fixed +5 V output at up to 0.5 A.

Now I know that the 80’s were not only the decade of MTV and big hair, but at least one rather strange PCB layout.