Rectifiers in parallel

[ppppenguin]

Do rectifiers in parallel share nicely? These three bridges didn't. They are from the Phlips PM5544 electronic testcard at the Broadcast Engineering Museum. The symptom was massive hum on the picture. See photo from 'scope. It was actually the main +5V rail feeding all the TTL that was in trouble. Exactly how all the logic kept working we don't understand. For strange reasons known only to the Philips designers they decided to use three bridges in parallel to provide about 3A for the logic. They were also mounted tight up against each other. Now replaced by a single chunky bridge bolted to the back panel.

It's not the first time that the PSU in this unit has fried. These rectifiers were actually RS branded replacements. These repairs must have been done when it was in use at BBC Pebble Mill. The bridge bolted to the back panel that's in one of the photos had probably been fitted at the same time. It supplies one of the analogue power rails.

I don't think much of the way the BBC fitted their favourite XLR-LNE mains inlet either.

By chance I've also been fixing a Philips LDK4210 SPG recently. The mains transformer had overheated and had shorted turns. A similar bridge had an open circuit diode. Halfwave rectification puts DC though the winding and saturates the transformer. The problem is that if there's enough margin and/or the mains voltage is high enough the unit carries on working despite the rectifier being halfwave.

[Kalee20]

I think the photos answered your question!

The trouble is, the forward voltage has a negative temperature coefficient... if one of the three runs slightly hotter, it's forward voltage drops a bit, it takes more than its 1/3 share of current, and gets hotter still, till that rectifier is carrying nearly all the current.

The reduction in power dissipated due to the voltage drop decreasing, isn't enough to cancel out the increase in power due to it hogging more current.

The only safe way to parallel rectifiers is to have current-sharing resistors. If transformer-fed, winding resistances can do this - but you need a separate secondary winding for each rectifier.

[Mike Watterson]

Rectifiers do work in parallel if same construction / part (if current rises, so does volts, so they share), but a bridge has effectively two in series and that is more complex. I've never paralleled entire bridges, though should work if the same diode part.

A capacitor factory I worked in (1970s) sent all the worst stuff to RS if short of production quota because RS never had made returns and wasn't known to even sample test. RS bridges with same RS part might not be from same supplier, thus current sharing might not work properly.

Mounting tight actually makes sense, re TC. as Kalee20 says.

[Mike Watterson]

I've also paralleled the ~ & ~ to have a higher current + or - rectifier, or both for a simple doubler.

[BusyBee]

I've never really thought of bridges before but I know 'current hogging' is a big factor to be aware of with paralleled transistors. Hence the emitter resistor being useful but maybe not a practical thought with bridges?

[ppppenguin]

As Kalee20 pointed out, you can use series resistance to balance the currents. Exactly how much resistance you need is another question which may not have a simple answer. You also have to allow for the voltage drop and wasted power.  Might be possible to do it with suitable lengths of wire.

Far better to use a single part that can handle the whole current, though if it's 100s of amps this may not be possible.

My mistake with this Philips unit was to leave the paralleled rectifiers instead of just replacing with a single bridge as a matter of good practice. Here are my notes from July 2024 when I overhauled the unit:

This has suffered a serious failure in the past. There is heavy scorching on both the PCB and some wiring. GR1702-4 (3x BY164 bridges in parallel for +5V supply) have been replaced. There is also a heavy duty bridge bolted to the back panel. The 1250uF main reservoir capacitors have been replaced with 2200uF 50V parts. C1702, the 2500uF reservoir for the +5V supply, has been replaced with 2x 2200uF in parallel. All this work was done at the BBC. Fig XIII-4 in the scanned manual (internal page 58) shows a single external bridge for the +5V supply. The PCB is renamed as MA8555, Unit 19.

Philips' own manual shows that they had second thoughts about using paralleled bridges.

[boater sam]

I have had similar problems with load sharing using 78xx voltage regulators in parallel. There is always one that runs hotter and eventually fails first.

[Mike Watterson]

I have had similar problems with load sharing using 78xx voltage regulators in parallel. There is always one that runs hotter and eventually fails first.

That doesn't work. But there is a way to add external transistor(s) to an LM317. It is very sensitive to layout and like the higher power LM317 it can "sing" from audio to RF, depending on layout. Getting 20A on an LM317 + transistor circuit was challenging.

You can also add an external transistor to an 78xx, but the regulation is poor compared to the LM317 which has a feedback pin from a divider on the output instead of the 78xx earth/common.

[Refugee]

I found a data sheet for the 317 with an example circuit with the extra transistor as an emitter follower and the feedback potential divider on the emitter of the added transistor.

[Kalee20]

This is how to get more current from an LM317. Above about half an amp the external PNP transistor starts to pass current and bypasses the '317.

You do lose the current-limit feature of the LM317.

It should be possible to add several PNP transistors if the collectors are strapped, the bases are strapped, and each emitter has its own current-sharing resistor (0.1 ohm or so). But, stability might not be guaranteed...