• Please review our updated Terms and Rules here

HP 2116B rescued - two boards missing :-(

MattisLind

MattisLind

Veteran Member
Joined
Sep 30, 2013
Messages
1,441
Location
Stockholm, Sweden
A couple of weeks ago I was contacted by an employee at a big Swedish company. They had a storage building which was going to be demolished within weeks. In this building there were among others sitting a HP 2116B computer. He had himself worked with servicing HP21xx systems and felt a bit sad seeing this machine going to scrap. He asked if I was interested in taking care of it but it had to be picked up rather urgently so last Monday I met up and picked this machine up together with some other HP 21xx equipment.

When examining the machine I quickly understood that two boards were missing. Unfortunately they are not even optional. They are required for the machine to operate. Without the Power Fail Board the machine will not even power up.

Those are:
  • 02116-6175 or 12588-6001 power fail board
  • 02116-6300 or 02116-6274 Memory Module Decoder
I fully understand that there is just a slim chance to actually find those boards, hence I am reaching out wide to try to find them! Anyone sitting on some old HP2116 board that happens to have these boards?

The boards are not very complex so it should be possible to recreate them if there weren't the fact that they use the Fairchild CTuL technology instead of more standard TTL circuits. CTuL is almost unobtanium. The only source is scrapping other HP cards which is not what I want to do.

Unless something pops up I will try to recreate these boards with modern technology. CTuL used -0.5V to 2.5V logic levels with 1.0V threshold. I will test a hypothesis that a HCMOS chip with GND at 0.5V and VCC at 2.5V could potentially be used to interface with these chips. I will at least try mocking something up to experiment a bit.



IMG_5929.jpeg

IMG_5927.jpg
 
Sigh. More boards missing. When I pulled out the card cage to start working on the PSU I immediately saw something more missing. Both the PCBs for the control circuitry for the regulator are missing ☹️

IMG_6041.jpeg

Someone also removed all the fuses. I wonder why. Perhaps they had blown?

Anyway. Not giving up. These boards can also be recreated somehow. Still need to check pass transistor if they are fine.
 
MattisLind said:
Someone also removed all the fuses. I wonder why. Perhaps they had blown?

Anyway. Not giving up. These boards can also be recreated somehow. Still need to check pass transistor if they are fine.
Click to expand...
My guess is the system was used for parts for others.
If I were taking parts from a system I would pull the fuses to prevent someone from powering it up.

I hope you can get it running again, HP stuff was always so nice looking.
 
Good. With the thread moved into a better place I can make this into a restoration thread.

jlang said:
My guess is the system was used for parts for others.
If I were taking parts from a system I would pull the fuses to prevent someone from powering it up.

I hope you can get it running again, HP stuff was always so nice looking.
Click to expand...

I fully agree that this poor machine became a parts machine when decomissioned. In my opinions the thing with HP is maybe not the exterior but the interior. All the carefully designed small details that makes them amazing. The entire card cage is sliding out without a hitch to give access to the PSU from the front. If it had been rackmounted you could still do the service from the front.

In total 16 HP 2116 based systems were ordered from HP in 1969 to do avionics testing on the Viggen aircraft. These systems were probably upgraded with never processors and more modern disks.

My plan is approximately as follows:
0. Label (if necessary) and remove all cards.
1. Open up the big capacitor box ( where the connectors for the missing cards are in the picture above ) and reform the capacitors. There are four 100000 uF capacitors in there among others. Hopefully they still are in good shape.
2. Reinstall the missing 30A fuses. Power it up using a variac. Without regulator cards the linear pass transistors will not conduct. Check for leaky germanium power transistors. I just wonder how much leakiness that can be allowed? Ge is notorious for leaking. I have had people reporting that they have seen this in 2116. What if I need to replace transistors? Getting those Ge power transistors will likely be very hard. Perhaps redesign using power PMOS if that happens?
3. Using a bench supply to the driver of the pass transistors to see that it is possible to get some output.
4. Add some load on the output. I just wonder how I can load the -2V properly. 22.5A max means 0.09 ohm. Maybe just use 15 meters or so of 2.5 mm2 cable?
5. If the all the pass transistors are fine I am going to make modern replicas of the regulator boards. All the schematics are available. Again they made frequent use of Ge transistors. I need to fully understand what implications that have.
6. The good thing is that I have a lead on the power failure board so hopefully I will not need to recreate that one.
7. I am thinking of recreating the memory decoder using a modern GAL-chip and a 74LS30. The output driver will be simple emitter followers. Some kind of pretty fast NPN transistor should be enough. 2N2369 (MMBT2369AL) ? Some design used 2N2222.


Well. And then I need to adapt along the road…
 
Good news for once! I managed to purchase a power fail board for the machine and hopefully it will arrive in a couple of weeks. It is the more advanced power fail board that also has auto restart, 12588-6001.

So we are down to three boards that need repliacs to be made. I also found out that there are NOS 2N2156 high power Ge transistors to be purchased if one of them are bad in the machine so hopefully I can stay with original design.

This means that I need to design those replica regulator boards. The original design made frequent use of the fact that a Ge-transistor has a 0.2V forward drop over the BE junction. For example the current limiting function is using one single Ge-transistor that has base and emitter connected over the current sense resistor. Whenever the current is so high that the voltage drop exceed 0.2V the transistor would start to conduct and force down the output. I am not entirely sure it is a good idea to design the replica using old Ge transistor even though it seems to be possible to find NOS on Ebay.

The other choice is to redesign it using modern op-amps which I think should be doable.

Screenshot 2025-11-07 at 09.58.55.png



This is the schematic of the logic power supply regulator that creates +4.5V, -2V and +12V. I tried to annotate the schematic with info on what kind of transistor the HP house numbers corresponded to. In one case there simply were no match and I just wrote what it said: "SI AF Amp NPN". Basically a low frequency silicon NPN medium power transistor, I think.

Any opinions of what strategy to choose? Exact replica or redesign with new components?
 
Last edited:
Nice find!

For the capacitors, maybe rig up a computer controlled PSU and run for example the software that Shelby at Tech Tangents (formerly "AkBKukU") uses?

Unless the PSUs already have crow bar circuits or some other form of disaster protection, maybe add that too while designing replacement boards?

Re the removed fuses: In addition to it being a good idea to remove them when boards were removed, I would also say that those were not commonly available everywhere if those used the size that's common in North America (6,3x32mm) rather than the in Europe common 5x20mm fuses. Thus they might had been used as spare parts over the years.
 
MattisLind said:
Any opinions of what strategy to choose? Exact replica or redesign with new components?
Click to expand...
I'm relatively pragmatic when approaching questions like this. Is the original design unique or ground-breaking in some manner that makes it worthy of exact replication ( are you expecting to show off the equipement by opening it up and pointing to the high-quality replica)? Can a functional replacement be designed that meets the same physical constraints and electrical/electronic requirements? Could the functional replacement include useful enhancements that aren't entirely inconsistent with the original design (e.g., crowbar)? Can the redesign be easily swapped out should the original implementation later become available?

In this case I don't think that the original design is especially "noteworthy" compared to the digital aspects of the HP2116; the power supply is simply a necessary precondition to a functional system. Substituting a regulator design with more capable components less likely to fail (and probably with a greater safety factor) than the original design & components seems like a good insurance policy against power-induced damage elsewhere in the system. That it would be a simple board-swap should an original regulator board appear is definitely nice.

I'd redesign the regulators (undoubtedly less well than you :-}), replacing as much of the original circuit design as desirable and retaining original circuitry where it's already "safe and effective" (and the parts are still reasonably readily available).
 
Thanks for the input!

I have been thinking a bit more about this. Recreating an exact replica has the problem of sourcing exact the same components. Sometime I need live with the fact that I have to choose something close enough. For example certain values for zener diodes might not be readily available. Creating a functional replica with modern components adds the complexity that I need to FULLY understand the current design in ALL details so that I can recreate it with modern parts. Now this is analogue stuff and even though I know bits and pieces of analogue design I am nowhere that skilled that I could say that I fully understand the current design. For example it has a capacitor and resistor in series between base and collector in every regulator. Apparently this give some negative feedback from the collector into the base which increases with the frequency. So a low pass design if I understand it correctly (yes, I had to ask ChatGPT). There are a lot of those small details that will force me to simulate the full design in Spice to understand it.

Given I think that making an as exact replica as possible of the regulator board would be the easiest way forward. Then it is just the money and hassle of sourcing the right components.

The other option is to replaced the entire PSU with a set of Meanwell switchers but I don't like the idea very much. Don't throw away something that is not broken.

This is HP stuff. They already thought of crowbars so there are a crowbar circuit on every output.
 
Fully agree. That would be the last last last thing I try (probably should hand over the machine to someone else rather than going to that extreme). Now I am mostly thinking of recreating the missing boards and that will be done by re-using the existing design but not to the point where the circuit board look old with same hand drawn layout. No, this will look like an new PCB but have NOS and new components it. Then eventually the actual board will popup on Ebay or elsewhere and then I buy them.

Didn't LCM replace all power supplies in the KA10? Never really understod why. A PSU consisting of a ferro-resonant transformer, some rectifier bridges and a whole lot of capacitors just don't fail in way that can't be fixed. Just replace a blown rectifier or bad capacitor. Or were they intending to run the machine 24/7 so the lower energy consumption would actually matter?
 
Last edited:
I'd say that a period correct thing to do would be to replace components with other components of similar vintage. Like if it's hard to find the North American germanium transistors I'd say that it would be fine to use European OC/AC transistors. Not sure if that would help at all, but still. (IIRC I have a few NOS OC71 if that would help).

If you were to do a slight redesign using "modern" silicon semiconductors I'd gladly take a look and maybe come up with suggestions. From what I see on the schematic you posted it seems to be linear regulation? If so it should be easy to follow the circuit.

In general anything between base and collector are as you already found out a local negative feedback, It's in general only used with capacitors, as the more common flat frequency (or even only at DC and lower frequencies) way of having local negative feedback is a resistor in series with the emitter.

If we count max current consumption at the correct output voltage as a resistive load, then smoothing capacitors across the outputs in combination with the resistance of the load should roughly estimate at what max frequency the regulation need to work up to.
For power supplies that aren't lab supplies or for other special usage it's usually way easier to have a large enough smoothing capacitor on the output and make the regulation relatively slow, to avoid the risk of oscillation. (Somewhat related anecdote tangent: Back in the 80's, when I still was in primary school, I made an attempt at building my own lab power supply (IIRC using a TL084 OP amp) and I didn't know anything about frequency response and whatnot. At the first test it oscillated so badly that I could hear the oscillating as a sound emitting from the wiring or whatnot 😁 :eek: ).

Btw for the germanium transistors for the current sense circuits it seems like a good idea to use OP amps (that can withstand any theoretically possible over voltage) with some simple reference voltage (I.E. voltage drop over a regular diode + voltage divider or whatnot) for each of the current sense circuits. If you want it to look "cool old electronics" style, you could use round metal can silicon OP amps.
 
Screenshot 2025-11-12 at 15.02.22.png


This is the part of the PSU that do the logic voltage. The transformer is not seen as the crowbars and sense resistors which are on another page. Below this part is the memory supply.

If you come up with a design that is handling all these voltage and does it with the same functionality in terms of sequencing the outputs when power it up I am all for it.

As you can see the PSU is actually a -6.5V supply and a -4.5V volt supply which in sort of grounded in the positive output.

I think reading up in the manual is also good to understand all the sequencing. http://bitsavers.informatik.uni-stu...tallation_and_Maintenance_Volume_2_197010.pdf
 
MattisLind said:
Creating a functional replica with modern components adds the complexity that I need to FULLY understand the current design in ALL details so that I can recreate it with modern parts.
Click to expand...
So this is where I think that you misunderstand my perspective, which was that "a functional replacement be designed that meets the same physical constraints and electrical/electronic requirements". So In-accordance-with (IAW) the specified voltages, currents, allowable noise & other limits, timings, control signals, etc. that HP documented, or perhaps improved (e.g., higher current, lower noise, faster crowbar). From my perspective the objective wouldn't be to replicate the consequences of design decisions that made sense *then* given the available technology, components, and manufacturing costs, but which no longer apply. One would design a "black box" (what happens inside to produce the externally observed behavior is immaterial) that drops into place that meets the specs, so while it's functionally equivalent it's not necessarily absolutely indistinguishable when evaluated side-by-side with the original at the scope-trace level. It doesn't need to be; just "good enough" -- which is IMO good practical engineering practice :-}.

That said, it's fair to point out that our modern expectations regarding TTL requirements when it comes to power supply may not apply well to the CTuL that the 2116B principally employs. But there are surviving design sheets for a number of CTuL devices to serve as a source of (some) information on the subject.
 
I think I see what you mean. I think you mean the system level. The PSU as one single unit which (hopefully) has a detailed spec. One can maybe workout a spec on the basis of the functional description and some curves that show sequencing and tables of voltage tolerances. Perhaps.

My intention from the beginning was to replace as little as possible. Just the regulator boards which control the series pass transistors. The transformer, the bulk capacitors, the output smoothing capacitors, the series pass transistors and crowbar are left as is. Just the regulator module is going to be replaced somehow. If I get hold of the regulator boards that are missing I just put them in place and go. Thus my interface with the world is what surround these two cards. And then a lot of the requirements come from various design decision they made back then. For example: we have Ge-transistors (Q33), lets make them the current limit sense circuitry by having a 0.2V drop over the sense resistor when they start to conduct or use a Ge -transistor (Q34) with some diodes to create the power sequencing that doesn't allow +4.5V until -2V has reached correct level.

Since I am not intending to replace the current sense resistor or the Ge pass transistors at at this point I need create something that correctly interfaces all those things. I don't think I can find a HP spec for that on the unit level. Perhaps they have a detailed system level spec (not sure about that either - a lot is just implied by the schematic).

On the other hand if I do start to replace various parts exterior to the regulator board, like replacing all the pass transistors with modern PMOS parts (which I gave some thought initially - before I concluded that it was possible to find 2N2156) , I will be on the slippery slope towards replace everything with Chinese switchers.

And actually I sort of would like to open it up and show the quality of the HP design and engineering which is something special compared to DEC or a flimsy DG.
 
There were this seller on Ebay that happened to list boards that came from a 2116B. Luckily I got them all for on averge 8.5 USD per board. They include four ALU cards and a time base generator card for the CPU and sense amplifier and driver boards for the core memory. Very good to have these spare parts. But even nicer was that the set included the EAU, Extended Arithmetic Unit.

I have no immediate need for the rest of the boards. But perhaps someone else have? At least they were saved from the gold extractor gang.


1763021613182.jpeg
 
You must log in or register to reply here.
Back
Top