Teardown and Repair of a 850W Power Supply from the '80s

Qualidyne 10094, 5V, 120A + 3 Auxiliary Outputs

By Dave Erickson

Youtube Video Part 1

and Part 2

qual1

Intro

I was about to toss this broken power supply into the E-waste bin. It was from a Datacube MaxBox 20 slot VME chassis. When I powered it up, the +5V output voltage was +4.2V. I found a loose component inside it, re-installed it, and messed around a bit, but the unit went from almost working, to fully dead. I pulled this beast out and replaced it with a spare PC power supply. I put the old power supply aside, waiting for E-waste day.

But I could not just toss it. I thought that some useful knowledge and possibly some useful parts could result from a tear-down. After removing the cover I saw that it was not easy to access the components on the main board. The different assemblies inside were all hard-wired together. A teardown would require significant un-soldering of wires. As I have absolutely no use for a 1980's, 5V, 120A, 850W, four output power supply, I didn't feel bad about scrapping it for science. Maybe someone will learn something.

The supply was made by Qualidyne, its date code is "8849", so late 1988. One year younger than my daughter. I was the designer of the MaxBox at Datacube, and specified this power supply for the design about 1987. We built about 100 MaxBoxes. Later, Martin-Lockheed requested a militarized version of MaxBox for their Lantirn IR targeting system for the F15 and F16. They had designed about 8 MaxVideo-10 VME boards into the Lantirn test and calibration system. These test systems were required in each air base that deployed Lantirn. We designed a new ruggedized MaxBox chassis to their specs and sold about another 100 systems to Martin. This cash cow helped pay our 80+ employees salaries for several years in the 90's. Each MaxBox and Martin system had one of these Qualidyne Power supplies. It's an old friend.

The power supply specs are:
 Unlike most modern multiple output power supplies today, all 4 outputs fully regulated and individually current limited. And all isolated from each other. This is the classy and more expensive way to build a power supply. I think we paid about $800 for these supplies, about a buck-a-watt.

Here is a MaxBox 20 slot VME chassis. The power supply lived in the back.
 
 maxbox

Repair? 

I removed the covers and soon discovered that the supply was quite difficult to troubleshoot without a schematic and parts placement. With the main +5V not working, the next thing to do on a typical power supply is to check the main DC capacitors for +320VDC. That was fine. That indicates that the AC fuse, input filter, rectifier and main caps are all OK.

Next is to check the main controller. Typically one would check that it is getting DC power and that its reference voltage and oscillator are working. Then look at the DC inputs and the pulse outputs. That all assumes that you know what IC it is and you have access to its pins. Unfortunately I only had access to the bottom of the board. No access at all to the component side: the giant output diode heat sink and output caps completely covered it. Unfortunately all this circuitry is buried directly under the main diode heat sink, and is all screwed together. From the solder side, I could see a single DIP IC, most likely the main controller. I didn't know what IC it was, but an intelligent guess would have been either a TL494 or a SG3525, popular1980's PWM controllers. That would have gotten me very close. It's actually a SG3527, a SG3525 with inverted outputs.

Here is the top level schematic I drew, once I got the diode heat sink and main transformer off. Due to the 4 switching transistors and the nearly 1KW output, I assumed a full H bridge. I was confused by the base drive transformers. I could clearly see only one high-current secondary (base drive) winding on each transformer.  How to control 4 transistors with only 2 base drive windings? I later solved this mystery. It's a half bridge, not a full bridge. The high side and the low side each consist of two transistors connected in parallel. That also explains the large film capacitor in parallel with a beefy 200 ohm resistor. They connect the transformer primary and the heat sink to the neutral / common of the power supply.

top sch
And the main +5V 120A output stage. The current transformer actually senses 2 parallel diode paths.

The electronic load board provides the minimum load required by the 5V supply.
sch2
This is the original configuration, with covers removed. The bottom board contains the AC in, Filter, main caps (black), H-Bridge (right) and all the main controls. The top board is the three Aux supplies.

qual1
Here it is with the top board removed. The rear heat sink is the diode bridge. Big blue caps are the output filter, transformer is upper right. The Half H Bridge is on the right heat sink.

 qual3
Interesting magnetics. The top one is the main transformer. The bottom one is the main 120A output inductor. The 3 wires, R/Bk/Wh (disconnected here) are the transformer primary, the top wires are the windings that go to the Aux board. A total of 6 isolated secondaries.
qual4
Here is the main output assembly and 120A path. From left to right: Transformer atop inductor, Diode heat sink, output caps, output terminals with filter board.
The yellow toroid is the output current sense transformer. 
qual5
4 beefy 60A diodes, stud mounted to the heat sink, which is the common for the 4 cathodes, and the + output of the supply. The 4 blue resistors and mostly hidden orange cap are snubbers.

qual6
The main 120A winding of the transformer and the inductor are copper buss bars. If you look close behind the white wire, you can see a black blob. It's heat-shrink over the the connection from the transformer CT buss bar and the and the inductor buss bar.
qual7
This shows one side of the main 120A secondary winding.
qual8
The main board. From bottom left, CCW: AC in, fuse and filter, input diode bridge and inrush limiter, over he 6 main caps, H-Bridge heat sink,  and a hole where the transformer goes.
qual12
Main controller area with SG3527A controller.
qual13
The half-H bridge and its toroid Base drive transformers and driver transistors.
 qual9
Close up of the base drive circuit.
qual10
The back of the main board. AC in and 6 main capacitors on top, H bridge driving the transformer primary on the bottom. The rest is control circuitry.
qual14
AC In, rectifier, inrush limiter and 6 black filter caps. Blue output caps and diode heat sink block the view.
ac in
Inrush limiter Triac, resistor,  and AC-DC Diode Bridge
 qual14
The Aux board contains three isolated buck controllers for the +12, and -12,  10A and +24V, 4A power supplies. Here is the power path schematic for each of the of the Aux channels.
qual16
Auxilliary 3 power supply board. Main AC Transformer windings come in on the bottom.  DC out is the (hidden) barrier strip on the top left.
qual15
The back side of Aux Board
 qual16
Here are the power devices for one Aux channel. The small toroid L2A was the part that was loose in the case when I first bought it.
quala17

Interesting Stuff. At least I think so. 

H-Bridge,  Base Drive Toroids

The H-bridge is old school, using four RJH6678 NPN transistors. These are 400V, 15A, TO-218. The isolated base drive is from 2 toroidal transformers, driven from two DH44H10 NPN transistors. These in turn are driven from the main SG3527A controller IC. The toroids need to provide about 1A of fast switched base drive to the H-Bridge transistors. I couldn't figured out how 2 toroids with one secondary each drive 4 high-voltage transistors. Another mystery of these toroids is the one-turn current-sense-looking wire running through each one.

Later, I discovered that it is not a full H-bridge, it is a half H-Bridge. The four transistors are actually two pairs, each pair in parallel to increase the current capability. The other side of the transformer is driven by the common of the +/- 160V DC, via a big film capacitor in parallel with a 200W power resistor.

Each base drive transformer drives two transistors. Mystery solved.

Iron Core Inductors, not Ferrite

The main 120A inductor,  and the three Aux filter inductors are laminated iron core, not ferrites. After examining them closely, the laminations are thinner than typical AC magnetics. Apparently thin enough to minimize eddy currents at the 26KHz switching frequencies. The main inductors are wound with copper buss bars. As is the main 120A secondary of the main transformer.

The AC common-mode choke is also iron core.

AC Inrush Current Limiter

An AC inrush current limiter is used to charge the six big DC capacitors without blowing the AC fuse. It consists of a 12.5 ohm power resistor in parallel with a big power triac. When AC power is first applied, the triac is off, and the ac current is limited by the 12.5 ohm resistor. Then when the main transformer starts up, an isolated winding drives the triac gate and turns ON the triac, allowing full AC current to flow to the diode bridge.

On other switching power supplies, there are several other strategies used to address this problem:

PFC (power factor correction) is primarily to reduce AC current harmonics, but can also provide controlled inrush current limiting PFC is generally required in Europe for power supplies of > 80W. PFC reduces the current spikes and therefore AC waveform current distortion of the power grid. The regulatory spec is "IEC 61000-4-7 for harmonics and inter-harmonics measurements". It specifies the maximum amount for the harmonic content of the AC current waveform. The US and other regions are not so restrictive. But if you intend your high-power product to be sold in Europe, PFC is a must. For any product that draws over 80W, you should have PFC. Not sure how cheap 200-500W PC power supplies get around this.

Aux Channels

Each Aux power supply channel is a buck (step-down) regulator. It gets its input from an isolated winding of the main transformer. I haven't figured out if the buck converters are synchronized to the main transformer. I suspect not. The controller for  each channel is an old school Motorola (now On-SemI) MC34060.

There is a small PCB that connects between the Aux and Main boards. It has four TL431 references, 4 trimpots, and two ILCT6 dual opto isolators. Pretty sure it is the over-voltage detector for the auxilliary channels. 

I was able to power up and test the two 12V Aux channels by simply applying +24V to their input transformer connections. I applied + 40V to the other +24V supply but no output. I didn't do much testing, just no-load. Gotta save something for later...

Low Voltage AC transformer

A small AC transformer was mounted to the front panel, near the fan. This has a single center tapped primary,  it can be wired to either 120VAC or 240VAC. The 120/240 jumpering takes care of this. The secondary is about 15VAC, so ~20V peak. This transformer provides low voltage to the controller, the fan and the other control circuits. The 24V fan gets about +20V so does not run at full speed.

After the main transformer starts up, an isolated "boost" winding and a separate diode bridge take over and provide the low voltage DC power. It's probably a bit more efficient that way.

Minimum Current Load Option Board

When we first bought these power supplies, there was a minimum load requirement of a few amps on the +5 supply. Without some load, the main transformer did not have enough oomph to drive the various Aux and other functions. Early PC power supplies also had this requirement. Later, the design was changed to remove this requirement. This small board, wired to a large power FET provides about 5A of load when the +5V load current is below the minimum. Here is a guess at the simplified circuit. The power FET is mounted to the Diode heat sink. This makes sense for 2 reasons:
  1. The drain of the TO-218 N-FET is screwed directly to the heat sink, which is the +5V output. No heavy wire or insulation required.
  2. The circuit only operates when the power supply and thus the output diodes, are at low current. So the extra 25W or so that the FET dissipates does not cause extra temperature rise on the heat sink. 
The additional wires are likely there to turn the load on and off.

load

Moment of temporary insanity

As I looked at this dissected pile of 1980s power electronics, I couldn't help but think: "Hey, I can fix this!"
I set about returning the main section 5V 120A power supply to semi-operational condition. I had cut a lot of wires to separate the assemblies, but fortunately the colored wire stubs were still there. The PCB is labeled with many of the wire functions. These were a big help Big help. Some of the wires that were cut or bypassed were:

With no load, the unit would charge the output caps to 5V and then shut down. It would recover after a while, and then do the same thing. I added the 1 Ohm (5A, 25W) load, and it did pretty much the same thing, but wold not recover until all power was completely removed.

Understanding that there are many conditions that could cause a shutdown, I set about isolating which one(s) were causing it.

SG3527 pin 8 is the SOFT START pin. Pulling it low causes shutdown. Pin 10 is the SHUTDOWN pin. Applying +5V causes shutdown. Pin 10 was the one causing the shutdown.  I traced the circuitry via a PNP transistor to an unknown part. This TO92 package was labeled SIY 88F. Couldn't find it on Google. Normally one pin was 10V and the others low voltage. Kind of like an NPN transistor. When it shut down,  the input were still 0, but the output went low. Huh? Turns out it is a small SCR that provided the overvoltage Crowbar function. An SCR is used here because it stays latched and shuts down the system until power is removed. I disabled the output of it, and the supply went into a bizzare screeching mode, with the 5V output only 2.5V. I reconnected it quickly.

Here is the shutdown circuit.

shutdown

"When you get strange DC results, turn on the oscilloscope."  I triggered the 'scope on the +5V rising. Strangely, the output would ramp up to about 5V, then increase to +7V. The output had about 6V p-p of sawtooth at the switching frequency,  for about 200mS. Then it would shut down due to triggering the over-voltage circuit crowbar.  Here is the scope shot of the +5V coming up. 6V p-p of 26KHz crud!  This is not a happy power supply.

scope1

For about 3 seconds, I wondered how those giant 46,000uF caps could have so much high frequency across them. Duh, both capacitors were dead! I jumpered a 1,000uF 35V cap across the output terminals, and the output quieted. I removed them, and on my LCR meter they both measured under 1uF. Fortunately my junk bin had some large caps. I installed a 12,000uF  25V and 18,000uF 25V, and the supply came to life, outputting nice stable +5V. 30,000uF is not 92.000uF, but it's way better than 2uF. New caps are about $20-30 each, and I ain't paying that. I'll see how far these mismatched old ones get.

The 5V output power-on with new caps installed. Sweet!

scope2
So you're probably asking "Why didn't this dummy suspect the output caps at first?". Well I did, but they looked good, no bulges or leakage, and seemed to hold a charge. Also there are 2 in parallel, so they'd both have to be real bad to cause the power supply to not work at all. Yep, they were. They could not be measured in-circuit easily with an LCR meter. Since they were after diodes, I could have charged them with an external power supply or my trusty DIY-SMU, to see if they held a charge or were excessively leaky. SMU at 10mA range should take T = C * V / I, so T = 90,000e-6 * 5V / 10mA = 45 seconds to charge or discharge 5V. Easy test, hindsight is 20/20. A 5V power supply through  a 500 ohm resistor would also do it. 

I have no excuse for not 'scoping the output at first. Better late than never. Hopefully others will learn from my mistake.

Frankenstein power supply. It's Alive!!

Here are the main board, main transformer and main output path set up to try and get it working. Note the mismatched output caps on the left.
open

High(er) Current Test

I operated the supply with a 5A minimum load and a 10 to 20A load from my DIY E-load. 20A plus the 5A minimum load is a 25A, so a 125 watt load. Upon increasing the load to 25A, the power supply drew about 200W of AC power. The 12.5 ohm resistor in series with the AC line became very hot. This is because the Triac had not been turned on. I connected the Triac to the Triac Gate winding of the transformer, and the 12.5 ohm resistor operated at a cool temperature. But after an AC power cycle, the 2A 20mm fuses in the AC line blew. I replaced these with 6.3A fuses. These were the highest value 20mm fuses at my local hardware store. To prevent blowing more $4 fuses, I temporarily operated the supply from a variac until I was sure it could start up and operate at 25A. If I want to continue using this at high currents, I would use fuse(s) larger than 20mm, about 10-15A, maybe slow-blow
Interestingly,  when it does operate with the inrush triac, there is a 120Hz buzzing noise from the AC input section. I traced this noise to the AC filter common-mode choke. The varnish that holds the choke's iron core laminations together may have dried, causing the buzzing. Or the noise may have always been there, but was masked by the board being in a box that was inside another box, and the fan noise from the power supply and system fans.

Main Transformer

This is a detailed schematic for the main switching transformer. The primary has 2 taps,  the white wire tap is unused. There are 6 secondaries. You can calculate the turns ratios from the inductances if you are interested. N2/N1 = sqrt(L2/L1).

transformer

What to do with this thing

As I've stated, I have no use at all for a 5V 120A power supply. However I do have a use for a system that can test transformers, inductors, other switching power supply components, etc.

If you stand back and squint, you'll see a similar architecture used in old PC power supplies. The old ones use a similar Half-bridge and are quite hackable. My advice. If you have an old half-bridge PC supply, particularly one that uses a TL494, hang onto it.

This supply has many things in common with old PC supplies. Here is a typical old-school PC 200W Power supply. It is kind-of an old-school PC power supply on steroids:
 Note that modern PC power supplies (> ~y2000) are smaller, and most use more advanced configurations. These are not be as hacker-friendly as the older ones. Fancier transformers, more integrated. Here is a schematic of a more modern supply. https://www.smpspowersupply.com/atx-power-supply.html They also use a saturable reactor to regulate the 3.3V. Old technology in a new application.

Analysis of half-bridge PC power supply

I discovered an interesting winding on halfbridge power supplies. It's on most PC power supplies, but in addition I found it on the 1980s Qualidine power supply that I did a YouTube video on. In that case it was very distinct. It's a heavy wire that's run through the center of the toroid base drive transformers as if it were a current sense. But why would a current sense be on a base Drive transformer? I've seen a few theories about what this does. One theory is that it self oscillates the Transformer but that doesn't seem to be the case.

Here are the two base drive toroids on the Qualidyne. There are three windings on each one:

baase

And here is a PC power supply schematic similar to the one I have. Note that the base drive transformer T2 secondary has an extra tap on one of the windings, pins 1 and 6. It is wired in series with the main transformer primary.  Similar to what is happening on the Qualidyne.

 I'm guessing that it's about a one turn winding due to the very low inductance. And that it serves the same purpose as the heavy wire run through the center of the Qualidyne base drive toroids. On the PC supply, I measured the base drive voltage on the primary, while varying the load on the power supply, and the base drive voltage varies considerably as a function of the power supply current load. At low currents the drive is low and it high currents the drive is high.I measured the inductance of one PC transformer, similar to the schematic below.

So if the current tap is 1T (turn), there are about 14T on each secondary and 50T on each primary side.

7502

Here is the base drive waveform on one side of the primary. The lower white trace is with a 12V, 1A load (12W) and the yellow trace is at 7A (84W). It increases from about 13V to about 25V. The significant increase in drive with load is due to the 1T current sense winding. The center tap is driven by a 1.5K resistor, so is is not driven hard. This allows the current sense winding to change the drive voltage. Wish I had a good current probe to measure the actual primary.

base

Analysis for the base drive.

The output pulses of the controller are open drain or collector. Both are normally off, and pulse low. They are pulled high to about +15V by the 1K/1K pull-up resistors. (my power supply uses 2.7K's to +15V. These active low pulses drive the NPN transistor bases. The common-emitters are biased positive with ether one or two diodes to ground,  plus a bypass capacitor. The schematic above shows one diode, my power supply has 2 diodes. This is probably so that when the controller outputs pull low (to GND) the transistor bases are pulled below the emitters, for faster turn-off.  You might think that the two half's of the transformer are driven low alternately, but no! The transistors are both ON normally and are turned OFF by the pulses. So the transformer pulses are positive: common emitter stages invert. The center tap is driven by a 1.5K resistor pulled up to +15V. So without pulses, and with both transistors ON and saturated, the center tap (CT) is pulled low, close to the emitter voltage, 1-2 diode drops. There is about 10mA bias current flowing through the CT from the 1.5K resistor. Then when each side pulses, its transistor turns OFF due to the 10mA bias current,  applies a positive 10mA current pulse to the transformer. The voltage is limited by the Output transistor base load on the secondary So the transformer is basically in current mode. Its 4:1 turns ratio applies about 10mA * 4 = 40mA to the base drive.  DC balance is maintained by the other side doing the same to its winding.

The fact that it is i current mode allows the transformer to have the additional current sense winding: 1 turn that senses the main transformer primary current. At low currents, it does little. At higher currents, it causes the increase in base drive.

BTW, this extra winding is only used with Bipolar power transistors. Modern power supplies with FETs do not have this mystery winding. And Gate-drive transformers are typically about 1:1 turns ratio. The drive circuit typically outputs 12V pulses, and the FET gates want to see 12V pulses. so 1:1 transformers.



Back to Dave's Home Page
Last updated 1/6/2024