Upcoming product - Replacement power supplies for the DECserver 700-8 and -16

[Terri Kennedy]

[Note - I'm not selling these, just pointing out something that's coming out soon.]

Adam (another volunteer at the Large Scale Systems Museum) is known as the Power Supply Guru. Last August, I asked him if he could look at a pair of defective H7881-AA DS700 power supplies. One had been re-capped (by someone else) without success, and the other would start up but then go dead after anywhere from a few seconds to a few minutes.

He dug into them and decided they were old, complex, and subject to future failures even if repaired. So he started a project to create a whole new power supply board that would fit into the existing case. I have a pair of pre-production units and they work perfectly. It uses an off-the-shelf power supply as well as a little extra logic to implement the combination "power supply is unhappy / mainboard is unhappy" status LED. It goes into the same metal power supply case as the existing PSU board. These run cooler and should contribute to a much longer life for DS700s - you don't have to risk shattering the case by opening it repeatedly, for one thing.

My understanding is that these will be a "send in your old supply, get it back with a new board inside" product, as it takes 3 different screwdrivers and a fair amount of coercing just to get the metal case open. A DIY kit of just the board might be available - that involves re-using some parts from the original supply (IEC input connector, output harness - although a new grommet is provided, LED, fan connector, etc.)

 

[czunit]

That is really neat. One request for him: Please include the schematics for whatever power supply he's using to replace the original. Reason being in 20 years that thing will blow up and trying to fix it without a schematic or documentation will be.... even more annoying.

The older power supplies can be weird, but at least there is some documentation. I recently repaired three Sun 386i power supplies for my two units (plus one spare): One of them was totally shot so I "mellon balled" in a modern XT supply and ignored the -15 supply (only used by a few early ISA cards). The other two I committed to fixing and found that the +12 volt "snooper" circuits that power up the rest of the supply were bad so I wired in a small 120-240 to 12v supply to provide the startup voltages when plugged in. Left the rest of the supply for someone else to figure out in 10-20 years if needed.

But these systems are starting to become like antique clocks and I think we should learn the lessons from that era: People would "repair" clocks by drilling in screw in bearings (which destroys the back plate in 20 years) or junking watch movements for cheap quartz movements (which look like hell or die in 30-40 years). 140 year old watches can be made to run fine with time and effort, and are usually documented by the factory.

I'm not for being 100% luddite (my VT52 has a -15 supply that is a simple 7915 wired in around a broken transistor that I can't seem to find) but I am for being 100% reversable if someone better comes along and really wants to run the supply with a true transistor step down regulator that you have to adjust with a 60 year old pot....

We made these mistakes with clocks and watches in the 1950-1980's, Let's not make the same mistakes going forward with antique computers.

 

[Terri Kennedy]

That is really neat. One request for him: Please include the schematics for whatever power supply he's using to replace the original. Reason being in 20 years that thing will blow up and trying to fix it without a schematic or documentation will be.... even more annoying.

It's a stock 3-voltage (+5, +12, -12 supply). I doubt the vendor provides a schematic, although I'm certain that equal-sized (or smaller) ones will be available in the future. For the "I'm OK/You're OK" LED circuitry this seems reasonable.

The older power supplies can be weird, but at least there is some documentation. I recently repaired three Sun 386i power supplies for my two units (plus one spare): One of them was totally shot so I "mellon balled" in a modern XT supply and ignored the -15 supply (only used by a few early ISA cards). The other two I committed to fixing and found that the +12 volt "snooper" circuits that power up the rest of the supply were bad so I wired in a small 120-240 to 12v supply to provide the startup voltages when plugged in. Left the rest of the supply for someone else to figure out in 10-20 years if needed.

There is no schematic for the H7881. "Replacement" original units go for $600 and up (that's why there's so many "naked" DS700 boards on eBay - the power supplies died and the cases shattered when opened without exercising extreme care regarding their very brittle plastic.

We made these mistakes with clocks and watches in the 1950-1980's, Let's not make the same mistakes going forward with antique computers.

You'll get your original PCB back (if requested) with the upgraded power supply, so you always have the option of trying to repair it. Adam is a switcher expert (who fixed a Data General 6-voltage disk drive power supply he'd never seen before in < 30 minutes, other than the time spent re-forming the caps) who spent the better part of a week on each of my two before deciding it would be faster / easier to just design a new one with modern components. You get the upgraded power supply back in the original case, so it's visually indistinguishable from the original DEC unit (aside from a peel-off sticker saying it was rebuilt, and the date, for warranty purposes).

These newer DEC switching supplies were vendor-supplied and it's doubtful even DEC had schematics for them.

As another example, Island Computers will sell you a genuine HPaqital DS10 power supply for $159 with a 30-day warranty. And Island can only do that because they got a HUGE load (more than several hundred) of DS10's in a few years ago. Island commissioned a new design in the same form factor which is $249 for a unit with a 1-year warranty, up to $348 for a 3-year warranty. That's how certain they are that the HPaqital supplies are going to go "poof" sometime after 30 days.

 

[Terri Kennedy]

That is really neat. One request for him: Please include the schematics for whatever power supply he's using to replace the original. Reason being in 20 years that thing will blow up and trying to fix it without a schematic or documentation will be.... even more annoying.

Since I'm outside the edit window, I'll add that the actual switcher is a MEAN WELL RPT-60B which is a "Medical Grade" supply with a stated MTBF of 677,800 hours (per Bellcore SR-332) and a 3 year warranty. Although I'd take any MTBF number that high with more than a grain of salt - that kind of number can only be extrapolated from individual component specs and aggregated.

Datasheet (for those who are curious): https://www.meanwellusa.com/upload/pdf/RPT-60/RPT-60-spec.pdf

 

[czunit]

Reasonable. I'm pretty good at fixing these older supplies (got the 8/E supply working most recently, it's not that bad a unit. Ok, it is bad but it is original) and I get torn between getting things running for the next 10 years vs the next 100 or so.

 

[Terri Kennedy]

Reasonable. I'm pretty good at fixing these older supplies (got the 8/E supply working most recently, it's not that bad a unit. Ok, it is bad but it is original) and I get torn between getting things running for the next 10 years vs the next 100 or so.

I'm mostly familiar with PDP-11 supplies. Up through the 11/70, these tended to use the H74x modular regulators which were 30VAC in, various DC voltages out, simple switchers. Schematics and service documentation were provided by DEC.

Then the 11/44 came long with its extremely bizarre power supply. While schematics and board layouts were provided as part of the 11/44 print set, you have to try something, discharge the 400VDC bus (which, for some strange reason DEC FS tended to clip the bleeder resistors off of), and try something else. Other than probing on the PSU backplane pins, there's no visibility into what's going on. To give you some idea of how bizarre this supply is, the 3 fans are 35V 70 Hz square wave drive only fans, just to get rid of something the power supply makes in an intermediate step and needs to get rid of. And a $2 optoisolator is the only thing between 120/240 mains power and the LTC signal on the CPU backplane.

By the era of the VAX 4000 and the DECserver 700, DEC was using absurdly complicated switchers purchased as complete assemblies from companies like Astec. We've already covered the H7881 in the DS700. The H7874 in the VAX 4000 is another disaster to work on because it's a fiendishly complicated fold-out one that comes in several utterly different designs (rev B and C units are completely different inside). Those tend to have obvious capacitor failures, although de-soldering them is a PITA. I had a good Rev C in a badly bent case and a bad Rev B in a good case, and figured "I'll just swap the boards". Nope. Everything is different, even the front panel LED / connector board. I managed to get everything moved, although it involved the drilling out of rivets and many harsh words,

 

[Adam.S]

Thanks for the kind words Terri. I'm still working on pricing but yeah I plan on selling the replacement board itself or offering the swap service. I reuse the old wiring harness so doing it yourself requires some soldering of the old harness and fan connector. An add on option is replacing the noisy fan with a new quiet model. So stay tuned everyone.

 

[Terri Kennedy]

Since people have been clamoring (not really, more like 2 people asking ), here's pictures of the completed H7881 supply with the new PCB installed. This PSU also has a new fan installed as most DS700 fans have turned into howlers or growlers. I'm including a picture of the whole PSU as well as a close-up of the label area. Note that this label is my design, and actual rebuilt units may have a different label installed.