• Please review our updated Terms and Rules here

Repair/replace IBM 5110 power supply board (or seeking anyone experienced in similar work)

For reference, the PSU cover on the 5110 also has the Vdc ranges (and which wires on the Y1 cable belong to what)

1651556068624.png


There is another label on this same PSU enclosure, does anyone know what this sticker below is referring to?

1651556133579.png
 
Also I was able to remove the "lodged" Y1 connector end that was stuck in the A1 board.

After taking the first photo, I realized the problem is that "black rail" is clipped over the top of the connector! That rail is just two screws - I removed that, and the connector then came right out with some pliers. I'm not sure what that black rail is for - it doesn't rest on anything when the A1 board is flipped back over (and some of the paint is coming off on the one for my original 5110). I guess it's just an extra guard, or possibly something to hold while flipping the A1 board.


1651556394753.png


Rescued connector end below: At least something easier to measure and reference for replicating as some STL 3D model for duplicating later.

1651556672632.png
 
For reference, the PSU cover on the 5110 also has the Vdc ranges (and which wires on the Y1 cable belong to what)

Ah yes, these tables and accompanying details in the manual were much used when I was developing Voltmitten and my IBM SLT connector substitute.

There is another label on this same PSU enclosure, does anyone know what this sticker below is referring to?

This must be a diagram for the naming scheme used for the backplane pins.

Take a look at the PDF file I shared in this post. Look at the letter labels inside the pin matrix. For full-width boards, the left pins on row 2 are prefixed with D, the right ones with B. On row 3, it's J and G respectively. Row 4, P and M. Row 5, U and S.

D B
J G
P M
U S

just like on the sticker. Half-width cards, meanwhile, just get

D B
J G

I don't know what that circled roman numeral III means.

I'm not sure what that black rail is for

I think that just retains the plugs that go into connectors Y1, Y2, Y3, and Y4. Keep in mind that those connectors are upside-down when the case is closed: moreover, they could easily be wrenched out by the rather thick ribbon cables that go off to the power supply or other parts of the computer.
 
I think that just retains the plugs that go into connectors Y1, Y2, Y3, and Y4.

Fair point, I thought the rail was inadvertantly clipping the edge of those connectors, but it is probably intentional for precisely the reason you said. A fairly strong person could accidentally rip those pins out, if they didn't take care to notice the rail securing them in place.


The power sticker has Roman numeral "VI" on it, so I'd assume those circled numerals (like "III") are maybe just reference for the assembly folks who are placing the stickers? " 'VI' goes here and is oriented like this ", etc? I'm not sure if enough 5100's/5110's were being made to justify a more automated production or line. I saw a YT video once on the Commodore C64 production line over in Europe -- early 80s and fairly automated.
 
I spent a lot of time wondering about the gauge of those wires. My guess is that it's 20 gauge; the 18-gauge ribbon that I used for my Voltmitten installation feels thicker.

I've never measured current draw for sure, but based on how hot the Voltmitten 5V solid-state switch gets (and in comparison to how less warm it gets for a 3A draw), my very vague guess is that it's drawing about 4A-5A on that rail. The system appears to draw much less current on the other rails.
 
No volunteers yet to do a component checkout on the original intact PSU (many folks on travel).

However, a friend pointed out that there are these inexpensive "buck" DC converters.

So I'm testing some out. I'm not sure yet how I'll do the inverted voltages (-12, -5), but those lines seem to just run to the tape-unit (and external IO) which I could just ignore for now.

These takes any up-to-30Vdc input, and down-convert to whatever DCv you have them set to output. And they're small enough, I could fit 5x of them on a board the size of the 5110 PSU board.

Mulling it over. The units get a little warm (they are passive and indicate 33degC).

buck_test.jpg
 
I would not ignore the negative voltages. Take note of the table on PDF page 120 in the 5110 MIM: -5V is used for "tape control card, storage cards, BSCA cards, and common ROS". -12V you might get away with omitting: it's for "tape unit, 5114 R1 relay, BSCA cards, and asynchronous communications/serial I/O card". Still, I'd be quite reluctant to tempt fate in my own machine.

The computer might work for a little while without -5V in particular, but there's a considerable risk that this could damage the ICs in your machine. In the 1970s, many ICs (especially DRAMs) required a reverse bias voltage in order to completely switch certain signals within the chip. Without the bias voltage, some signals might not turn all the way off when they should, and ultimately having voltage where voltage should not be can lead to damage. Ken Shirriff did some good writeups about this stuff in the context of early microprocessor development, and I did some DRAM datasheet dives when I was working on Voltmitten. Links:


We don't know if those datasheet warnings or Ken's explanations apply to IBM hardware, but I would be surprised if it were otherwise.

Besides gathering a small flock of buck converters, one other popular option is to see if a quality AC->DC multi-output power supply from Mean Well or a similar manufacturer might meet your needs. From the Mean Well catalogue, the RQ-50 and MPQ-200 may be promising avenues (although you will need to consider all the specs in detail before you make a choice). Neither option has +8.5V, and for that a buck converter drawing power from the +12V rail (or even perhaps a linear regular, properly heat-sinked, if ripple is a concern) could be useful.
 
Ok, did some reading, and I think I have the -5 and -12V sorted out !!

I'll be testing this in the "3rd spare" 5110, that is basically for parts anyway. Still bench testing and mulling over what I've missed - for example, I've set this up while no-load, maybe I should bump the voltages up a tad (.2 - .5 V?) ? Should I go ahead and put small inline fuses (maybe like 5A small automotive styles ones)? The manual on these bucks say they have an undervoltage shutoff at 4.7V

Will be running them again one more night (mostly checking temps). But I think aside from running 1 more chassis ground cable, I'm about ready to give this a go in the actual system tomorrow. Which if it works, then I'll give more wiring-up details, sure.


One thing I've noticed is on startup - the voltages are a tad high, for a very brief moment. For example, the 12V might start at 15V. I've noticed this both on the LCD of the pucks and measuring the pins on the multimeter. It might actually "start" at the input voltage (19V) and work its way down -- but it is doing DC/DC conversion during that time it is walking down to the target output voltage. It's quick - maybe around 1 second to settle - but just something I've noticed. Maybe that's normal?

Image below is just with the Y1 wiring harness, haven't connected the external harness yet.

5100_psu_beta.jpg
 
Last edited:
Wanted to show how the Voltages of the -5 and -12 buck are elevated above the input voltage (of 19Vdc in this case).

19+12 = 31 (Vdc)
19+5 = 24 (Vdc)

And I assume this extra V is why those two units run a little warmer than the others.

1651995799911.png
 
I'm a bit puzzled by how this is working. Are these "buck converters" possibly also "boost converters", in that they can convert an input DC voltage to a higher DC output voltage? Buck converters typically only reduce voltages.

Whatever the case, there may be a few design objectives that you want to uphold here before you plug things in. A good one that I can think of is making certain that what you call ground in your PSU is basically the same thing as what most of the other stuff around your 5110 also thinks of as ground.

I mention this because there could be certain arrangements of power converters that establish a kind of "virtual ground" at a voltage that is higher than the general idea of ground in your lab (or whatever). You could just choose ground to be +20V relative to mains earth, say, and then supply +8V for the -12V rail, +15V for the -5V rail, +25V for the +5V rail, and so on. Everything is relative and in order inside the computer, so it's fine...

...until it's not, and that happens on the day when you attach some other piece of equipment that isn't following the same convention. Say you plug a wall-powered CRT monitor into the video out port. Now you're shorting your +20V virtual ground to +0V mains earth, and that will probably go poorly.

And actually, now that I think about it: the 5110 has some AC appliances inside already: how are these powered in this setup? The ventilation fan is an AC fan, and the tape drive motor (if you have one in this #3 machine) is also an AC motor. The presence of these AC devices suggests the potential for other ties between the ground that the logic uses and mains earth.

If the +0V in your setup is basically the same is mains earth, then perhaps this is no factor.

One advantage of using an engineered PSU replacement like the Mean Well is that it will do a pretty good job of keeping all the grounds on level ground.

I note that the 51[012]0 documentation (plus stickers on the power cord) are pretty insistent about good grounding, although I don't know their particular motivation for all of that warning.
 
Going out for a bit (Mother's Day!), but some quick notes:

(1)
Here are the notes I used to get the buck converters to invert voltage...
"MAKING A VOLTAGE INVERTER FROM A BUCK (STEP-DOWN) DC-DC CONVERTER" from maximumintegrated.com

(2)
The main fan next to the PSU and the fan of the internal tape have their "own" power. Or in other words, even with the entire PSU removed, if you "power on" the 5110, the main fan and internal-cassette-fan still spin up. In the "AC BOX" below the CRT there are J1,J2,J3 AC lines. J2 is that system fan, J3 is for the tape unit. J1 goes over to the PSU for everything else. (and the System Logic Manual also shows this, will try to get a shot of it later)


EDIT: Oh, and yes, I've ordered one of those MeanWell PSUs - they aren't that expensive as I thought. I may do a 2nd build involving that and a mix of 2-3 of these bucks. Or in general, I'm looking at more options on what they offer. I still may use this first alternate-PSU, as I don't know what works in this "3rd 5110 spare" (and if the CRT doesn't work, I am hoping the external video out does)
 

Attachments

  • 1652030400670.png
    1652030400670.png
    81.2 KB · Views: 3
Last edited:
The Maxim note looks reasonable to me, although I can't be trusted to know whether even a correctly-designed power supply is a safe design. Not my area of specialty :)

Oh, and yes, I've ordered one of those MeanWell PSUs - they aren't that expensive as I thought. I may do a 2nd build involving that and a mix of 2-3 of these bucks.

Note that some of the quad output PSU options like certain RQ-50 and MPQ-200 models already provide +5V, +12V, -5V, and -12V.

Provided those outputs met your needs, you would then only need one buck converter on the +12V line to get down to +8.5V.
 
Alright - good news: it works! I was able to boot the Executive ROS using my "home-brew" buck PSU. The internal CRT didn't work (yet) but the external (BNC) video at the back did. Very exciting, everyone literally hid behind cover while I prepared to power it on the first time (in system).

However, there are some issues still to be worked out...


The main issue is that the 5V buck was reading above 4.3A, which is beyond its spec max of 4Amps. It got super-hot fairly quickly, shutting itself down at 99degC.

These bucks have CC and CV. So briefly stated: If the load uses more Current than your CC setting, a red light comes on and it starts reducing Voltage (CV) to compensate. Fortunately, the 5110 A1 board and cards were quite forgiving about this (I initially had the bucks set to 1.0A, so various voltages were dropping below the setting as I discovered these Amps). Hopefully nobody else has to go through that again. I'll say this, "funny things" ended up on the display as I got closer to the expected Voltages (first sync/scroll issues, then various grid of same or random characters showed up -- so I was on the right track, and eventually then the "ABCDEFGH" startup sequence showed up!).

Here's what the little bucks reported:

+5V --> ~4.3A (over 90degC, shutdown at 99)
+8.5V --> 2.4A (~42.7degC)
+12V --> 2.4A (~35.9degC)

-5V --> 0.05A (I only booted the Executive ROS, didn't transition yet to the Common ROS -- this is the older Executive ROS that halts on the 013 D80 no disk drive error, which is ok for now)
-12V --> 0A (as expected, no tape in this system, not using anything that uses the -12V line)
(these negative lines stayed around 30-33degC)

Some general notes:

- The 5110 executive ROS (and common/language ROS) will boot with no keyboard attached to the system.

- The 5110 will boot (and external video BNC works) with both the -5V and -12V disconnected / OFF [ I went ahead and did a -12V line just for completeness, it may apply more when I get back to exploring the external IO; the -5V may come into play when I insert a newer Executive ROS that can bypass the no-disk error -- but point is: lack of the -5V doesn't hinder the BNC external display from working ]

- When the +5V buck overheated and turned itself off, the BNC-connected screen went blank, the audio-buzzer behind the built-in CRT issue a mono-tone sound, and the red light at the front panel came on. Let that cool down a bit, and everything powered back up as before.

- The 19V/4A power brick I'm using does work fine (taking 120V AC main and converting to DC; the ground on that main is tethered into the PSU ground that is also used by all the bucks). I switched over to a variable DC power supply - if you're already in a fairly cool air-conditioned room (or have some other external fan), I've confirmed you can just run the A1 board cards and the external BNC display with just DC-power. 2 Amps didn't work, did have to bump it up to 4A (going higher than 4A -- I tried 6A -- didn't seem to have any effect)


One might be able to use multiple 5V bucks instead of one single one for all 5x lines on the Y1 cable (3x internal to the A1 board, 2x the others for external IO). Hopefully I have the new buck and MeanWell by next weekend, to give those a try - but as a proof of concept, I'm glad to have "spare3" going, this setup can be used to "benchtest" a 5110.
 
Last edited:
Found a 12A version of the same DC-DC buck converter (the ones I have are 4A versions):
Should be a drop in replacement, verify that's the only issue and how warm it gets (passive).


For a follow up build, I'll try the MeanWell ‎RQ-125B‎
AC/DC 5V 12V -5V -12V (with the 5V port handling over 10A)
It's a bit longer than I'd like, but as suggested I should just need 1x DC-DC puck for the 8.5V (and won't have to hang a "power brick" off it, so it should be a more concise setup to fit in the PSU bay).
 
Last edited:
For replacing/substituting the stock PSU in the IBM 5110, here may be an option....

I used "spade connectors" to be plug-compatible with the original IBM wiring.

Use "screw" or "twist" connectors to avoid needing any soldering. I used 18 gauge wiring.

On the Y1 wiring harness from IBM, they're all the same color (yellow) so mark them somehow as you're unplugging them from the original PSU.

Configure the buck converters first before connecting outputs (there is an "A" amp probe on the Fluke multimeter to help know approximately where the CC is set at).


IBM5110_alt_power.jpg
 
Last edited:
The RQ-125B sounds like a solid choice, and as you note, the only puzzle you will have to solve is where you will get the +8.5V from.

Looking at the datasheet, two things stand out to me. At +5V, the RQ-125B has a rated current of 11A. At +12V, the rated current is 4.5A.

Your current measurements for the power rails is really valuable --- that's new information. I'm personally somewhat pleased to see that my estimate of the +5V current was pretty accurate :) But anyway, note the following:

You are already consuming a measured 2.8A on +12V, and that leaves 4.5A - 2.4A = 2.1A left over in your current budget.

Do you know how much +12V current your +8.5V buck converter would consume when it's serving out 2.4A? It could be less than the 2.1A you have left. But there may be some loads on +12V that you may not have tested yet:

magnets.png

Those tape select magnets are electromagnets that turn on whenever the tape drive wants to move the tape:

tape.png

I have not measured how much power they use, but I would assume that they use a fair amount of current in order to press the jackshaft rollers against the spindle. I'm concerned that you may overrun the RQ-125B's +12V current budget of 4.5A!

There may be another option for an all-in-one solution though: instead of getting +8.5V by converting down +12V, you might try upconverting from +5V with a boost converter. Let's round your measurement for the +5V current, 4.3A, to 5A to add some safety margin: you now have 6A remaining out of the RQ-125B's rated 11A to dedicate to making +8.5V. Power = current*voltage, so 6A * 5V = 30W. Now, 30W / 8.5V = 3.5A: I suspect whilst making no guarantees that as long as you get an efficient boost converter (since it will consume some current of its own), then this will give you the power you need with some margin. This is the approach I would likely prefer, although more study is probably advisable.

More notes in a follow-up post.
 
Last edited:
Miscellaneous additional notes:

1. +90C-plus is a toasty time. Voltmitten's +5V solid-state switch gets to somewhere over +60C with passive cooling and no heat sink but a big copper fill area for radiating heat. The switch I use is an automotive-grade component and is operating comfortably within its operating spec, so I don't worry about it too much, although the old rule of thumb is that 10C extra cuts your component lifespan in half. Who knows.

2. I wouldn't run multiple DC-DC converters (particularly fancy switching ones like buck or boost converters) in parallel unless they were designed for that kind of usage. One failure mode you might discover is this: owing to small variations in the devices, one unit winds up trying to deliver all the current at once and fails due to overload --- now it's offline and one of the remaining units will be the next to get overloaded and perish, and then the next, and so on. Classic cascading failure: it happens in enormous power distribution networks and it can happen on your benchtop too.

3. For someone who has spent a whole lot of time and effort ensuring that my 5100's A1 board etc. gets only exactly the voltages specified by IBM, your experiments make me a bit dizzy to read about! :) I am glad (and also personally reassured) that you are discovering that the IBM silicon is somewhat robust to your trials --- and I note that this echoes an offhand comment by YouTuber Jerry Walker, who did some work on a 5120 in one of his videos. This said, I really feel like it is tempting fate to omit the -5V voltage under any circumstances. A lot of silicon from the '70s really needed that inverse bias voltage for the reasons I described earlier, and while we can't easily know what 1970s IBM would think about not supplying that rail (perhaps it's in a Bitsavers doc somewhere), the manufacturers of analogous commercially-available components like DRAMs very clearly warned you against leaving that rail unpowered. I can only repeat my recommendation to have a look at the datasheet excerpts I've collected in my Voltmitten sequencing documentation --- I promise I'm not pointing anyone that way for the website traffic!

PS: Don't be fooled by the fact that the -5V rail does not consume very much power. When it's operating as a bias voltage for the logic I've been describing, I'm pretty sure that it's only going to sip milliamps. It's still essential.

4. If you absolutely cannot resist omitting the -5V rail, please, please tie it to your ground rail --- don't leave it floating. If it were to float somewhere above ground --- chips really hate that!
 
Ah, I did forget to put a "not for the faint of heart" disclaimer, or the old "do not try this on your homeworld." I was ok with potentially sacrificing "spare3" for science - but thankfully no sparks or odors. Just some concerning strobes on the monitor connected to the BNC output, until everything got dialed in (at one point I swore I saw some blue text-characters on the screen, reminding me of some tricks in the old 8088mph demo - a vision of "color output on a 5110!?" flashed in my mind).


The other "spare2", however, does have an internal tape unit and is in overall pristine condition - except for its PSU as well. Good call on not accounting for whatever Amps the tape unit will draw - indeed, a boost off the +5V of the RQ-125B sounds like a good plan. But I do want to keep this all passive - no extra fans. Will read up about those.


The little manual on these 4A bucks state, "Protections":

- Input reverse connection protection
- Output anti-backflow protection
- Short-circuit: Yes. Can directly connect to the short circuit on the output.
- Over-temperature protection. (100degC, confirmed!)
- Over-power protection (60W)
- Over-current protection (6A)

Not sure if any of that addresses your "in parallel" concerns? The 2nd button on these devices disables the output. And that's how I inadvertently ran without the -5V: I accidentally bumped its "output off" button. Which is one nit I'd give these bucks: no way to "lock settings" (like if the knobs for CC and CV could be removed, to avoid them inadvertently getting adjusted). No intention to keep that off, but just noted the observation that the system stayed up (even through doing a reset). Since that went so well, then I just intentionally turned off the -12V also (after observing it had no Amp draw anyway). When I swap over to the newer Executive ROS and do a full to-a-prompt bootup, the -5V will be there for sure.



Oh, before I forget: I have one of those "Kill-A-Watt" devices to measure power consumption of AC devices. My functional 5110 (sans tape) reads only 1.6A out of the 120V AC socket. Just thought that was interesting. The internal vents seem to push the air over to the A1 cards. I wonder: Are the fans in the 5110 intended more to just keep dust out of the system, more than for actually cooling anything? Practical, for a system designed to be used outside.

Also to note, I'd say it took a minute or so for that 5V to heat up to over 90degC (from room temp) -- so not immediately, but a bit gradually. And I could sure start to feel it on the plastic. I wondered if the plastic enclosure of those bucks really help or hinder (for cooling)- they remain open on the ends.
 
Last edited:
Back
Top