My Book power supply, part two

21.09.2010 12:40

Some time ago I was writing about a MyBook power supply that went bad. It turned out that bad soldering wasn't the culprit there because the external hard drive kept causing problems. It would turn off at random times, sometime restarting immediately and sometimes staying off for good. For the sake of posterity I'm posting a description of my second (failed) attempt at repairing it.

I took the power brick apart again and started poking around. Indeed I could see with an oscilloscope that the switcher would sometimes turn off and the output voltage would fall to zero. Sometimes it started again after a few moments and sometimes it would stay dead for several minutes. Shaking and stressing the circuit board didn't reproduce the problem and neither did heating and cooling. So I considered it unlikely that it is a mechanical defect.

The topology of the circuit is consistent with a flyback converter. The controller IC is in a 6-pin SMD package marked 63813. Although I can't find the specific chip, the circuit around it looks very much like the typical application for the SG6858, which is also available in 6-pin SSOT-6 (identification of the IC was wrong in my previous post). For reference, here's my (incomplete) sketch of the circuit:

MyBook power supply partial schematic

I started on the secondary side since its easier and safer to measure, but I soon discovered that everything seemed to be in order. Most importantly, the current through the feedback optocoupler decreased after the output voltage already dropped when the bug occurred. This pointed to a problem on the primary side.

On the primary side I first focused on the Vdd supply for the controller IC (pin 3). This voltage is supplied by 1 MΩ resistors at start-up and later by an auxiliary winding of the transformer. When the PSU wasn't working the supply voltage went between 10 and 15 V in a saw-tooth. This is consistent with the hysteresis of the under-voltage protection of the controller IC, which means the IC was turning on, depleting the charge in the 10 μF capacitor and turning off until the capacitor again got enough charge through the 1 MΩ resistors. So the signal must have got blocked somewhere between the control circuit in the IC and the gate of the MOSFET.

Both the 100 Ω resistor and the diode D9 measured fine and didn't fail when heated, cooled or mechanically stressed. However it did appear that doing the same to the 1 kΩ resistor in the current-sense network (which was nearby) increased the likelihood of the bug.

At this point however something must have failed completely. After I replaced the 1 kΩ resistor the controller chip went dead for good and I can get no response from the circuit at all. It's possible I shorted something while replacing the tiny SMD resistor although I find that unlikely - I closely inspected everything after the fact and could see no solder bridges.

So, this circuit is going into the spare-parts bin and the external disk will get a new generic power supply from the nearest shop.

Posted by Tomaž | Categories: Analog | Comments »


18.09.2010 23:11

Half of the country is fighting floods and I just couldn't sit in my (mostly dry) apartment. With many roads closed I went by foot to the closest hydroelectric power plant to see how it's coping with the torrential rain.

Hidroelektrarna Fužine on Ljubljanica

This is Hidroelektrarna Fužine, a small power plant on Ljubljanica, downstream from Ljubljana. It was built in 1897 for the purposes of a local paper mill and was in fact the first power plant to supply alternating current in the country.

Hidroelektrarna Fužine on Ljubljanica

The usual sight here is a dry spillway and except for a small fish ladder all of the river's flow goes through one or two turbines on riverbanks. Obviously today's sight wasn't usual. I do not remember seeing water this high - neither here nor in the center of Ljubljana. Not to mention so many trunks of uprooted trees in the water.

By the looks of it at least one turbine was working, but I guess all this mud in the water wasn't doing miracles to its longevity. Also it was quite apparent that this abundance of water actually decreases available power because the high water level below the dam significantly decreases the drop.

Posted by Tomaž | Categories: Life | Comments »

Ars Electronica 2010

06.09.2010 21:48

I spent the weekend in Linz visiting this year's Ars Electronica festival. I hitched a ride with a group from Kiberpipa that has been visiting the event for several years.

The festival is huge and two days we had in Linz was just barely enough time to walk though all the exhibitions. It's happening all over the city and it's surprising how much effort the city as a whole puts into modern art - Ars Electronica Center and most other venues host exhibitions all year, not just during the festival.

The exhibitions themselves ranged from what looked like completely commercial exhibits of modern technology, science-museum-like presentations to installations and abstract art involving computer graphics, electronics and mechanics.

"Hylozoic Grove" installation

This machine, for instance, mimicked a living creature. It sensed your presence with what looked like IR sensors and responded to it by contracting muscles made out of memory alloy.

"Measuring Angst" installation

This one replayed the break-up of a glass bottle in slow motion, going forward and backward in time.

I also saw the results of the toaster project, cigarette flinging robotic hands, electric drill powered monowheels, giant rabbits and a whole bunch of robots that fell well into the uncanny valley.

So a lot of interesting ideas. Imagine something like the Art & Beauty place in CCC multiplied by several orders of magnitude. What did caught my attention however was how often these ideas were lacking in implementation. All too often some robot or installation wasn't working because "it takes 6 hours to charge it and it only runs 5 minutes on batteries". Some of the most interesting things you could only see on a video recording (the rain mirror for example).

Some parts of the exhibition I also couldn't quite connect to Ars (homeopathic medicine demonstrations for instance) or Electronica (performances where people hang themselves on meat hooks).

I should mention that this year the official motto is repair and I can't agree more with the Repair Manifesto. There was a whole floor dedicated to workshops on how to repair everyday things like clothes or shoes (again, not really Electronica).

Linzer Klangwolken 2010

Last, but not least, I was blown away by Klangwolken "Baby Jet" event on Danube. Kind of a science fiction action movie combined with a musical played live in front of you. Helicopters, ships, trains, lasers, explosions and a story involving a secret invention by Konrad Zuse. I hope there will be a video available somewhere so I can watch it again and hopefully understand a bit more about what was going on.

Posted by Tomaž | Categories: Life | Comments »

Noise of a straining transformer

02.09.2010 17:38

My lab power supply is now almost complete. It passes most tests I throw at it with flying colors except for one. When the output is even moderately loaded the toroid mains transformer is giving off a loud 100 Hz hum.

This is usually a sign that the windings are badly overloaded. Another warning sign that pointed in that direction were the thyristors in the preregulation stage that were dissipating more heat than I anticipated. This called for further investigation.

This is how the voltage on transformer secondary (blue trace) and capacitor (yellow) look like when the power supply is in short-circuit with current limit set to maximum. The gray trace shows transformer voltage when the power supply is idle for comparison.

Transformer and capacitor voltage oscillogram

The preregulator is working correctly here - it fires the thyristors in the correct moment so that the sine wave charges the 4700 μF tank capacitor just enough to replace charge lost due to load current during the previous cycle.

But that voltage drop on the transformer looks worrisome. Adjusting my SPICE simulation so it matches these measurements it gives peak transformer current at 17 A and RMS of almost 7 A. This is quite a bit above the 4 A RMS specification of the transformer.

Although I don't trust these simulation results completely they do confirm that components are getting overloaded. This is much more current than I accounted for in the design - I was counting on the stray inductance of the transformer and capacitor ESR to dampen the response a bit. With such current spikes it's no surprise thyristors are heating up.

The solution I'm looking into right now is to put a couple of power inductors between thyristors and the tank capacitor. They should provide enough reactance to smooth out the current.

Unfortunately to lower the RMS current all the way down to 4 A I would need a pretty big inductor. Bigger in fact than I'm willing to invest into right now, considering these things cost and arm and a leg to ship to this end of the world. For a start I'm ordering a pair of cheap bobbin-types from eBay. Although they shouldn't be enough according to SPICE at least I will be able to more accurately estimate the correct value.

Here's the simulation with the 680 μH inductor in place:

Simulated transformer and capacitor voltages

Green - voltage on transformer secondary. Blue - voltage on the tank capacitor.

Simulated transformer current spike

Current through the transformer secondary.

So, a few lessons learned: with a thyristor regulator RMS current can be even 3.5 times higher than the rectified DC current (compare that to the 1.6 derating factor you usually see on transformer datasheets). To get that down to a reasonable level you need expensive inductors, which eliminates one of the advantages over switch-mode regulators.

Posted by Tomaž | Categories: Analog | Comments »