Isn't your power supply a buck converter? Your "pass switch" transistor should be turning on and off at some 10's of kHz frequency. I'd disable the crowbar and the overcurrent protection (just remove the SCR and the input at the the current protection transistor). Then feed the thing with some current-limited power supply at F1 and ramp it up slowly. That should allow you to debug it with more safety and less ElectroBOOM potential... At some point the pass transistor should start chopping, then you can work on the regulation loop. You can even temporarily open the regulation loop to untangle things. Shooting from the hip here, I have not looked at the doc, so take this with a big grain of salt.
thats what i thought! the pass switch, "free wheeling" diode and inductor look like a standard buck converter to me. although the feedback and regulator circuit look very different from what I'd expect in a modern switching supply 🤔
@@juliankandlhofer7553That is what I thought. Everything is working as designed, but there is no load. I would want a headlight bulb on the output to see.
The best kind of parents are the ones who have no problem stopping to throw a couple hundred pounds of minicomputer into their trunk on their way to visiting you.
I've done it a few times - the biggest being a 10 ohm 50-watt resistor (I can still remember the foot-tall flame that shot out of it - 30+ years later! The things that happen as a rookie tech are easily seared into one's mind).
That style of switch mode power supply was quite common in 1970s computer and digital processing gear. They knew the efficiencies of a switch mode power supply which is almost mandatory at the current levels needed, like over 20 amps. But at the time we didn't have reliable high voltage AND high current transistors or MOSFETS. There were TV horizontal output transistors but that was not much current. Triacs and SCR were not suitable at the ultrasonic frequencies needed for a direct line operated supply. So the solution was a standard 50/60 hz transformer/rectifier based unregulated power supply feeding a switch mode regulator. That drive circuit is actually an oscillator which means a feedback loop. Very difficult to troubleshoot. As the component count is reasonably low, I would just test all the semiconductors. Even in circuit you can get a pretty good idea if the part is shorted or open.
I see you have an 11/60 on your inventory list! I designed the Writable Control Store in it along with the clock and other odds and ends. I also wrote the microcode that implemented the instruction set, boot up, and all the other control stuff. Neat to see that one is still alive. At the time it did the fastest register to register operations of any of the PDP-11s... to bad that was not an important metric to customers though.
I really wanted an 11/60 back in the day, including when I was trying to start a computer school in the mid-80s. I had a lot of equipment for it, including SMP KL-10 pair from Intel and a 2065 from Columbia University. Then I developed inflammatory bowel disease, couldn't work, and lost everything in storage.
If the 11/60 wasn't limited to 18-bit addresses but had the 22-bit extensions from the 11/70 I think it would likely have been far more successful than it was. The microcode also could do all the floating point instructions if the FP hardware wasn't present. Over 40 years ago when I was working for DEC we investigated using the WCS option to create special instructions to handle some operating system functions unfortunately the performance gain was insufficient to take it into any product. Unfortunately I no longer have the microcode listings although I still have all the WCS manuals.
The 11/60 was, in effect, considered to be an enhanced 11/40 and the 11/70 was an enhanced 11/45. The 11/40 and 11/45 were really different classes of machine. In fact the head designer on the 11/60 also was the head designer the 11/40 (and the original 11/(20)). So the feature space we worked in was more limited than the one the 11/70 design team worked with. That’s one of the reasons for the difference in address space size. A microcode implementation of floating point might be faster than a regular software implementation but I doubt that it would come close to the hardware implementation performance. The FPU for the 11/60 ran asynchronously so you could be doing integer arithmetic (or anything else) at the same time the FP instruction was running. A microcode implementation would run inline with the program. Also there are cases in floating where the result is, in effect, unspecified. So any floating point implementation would have to be sure to produce the same unspecified results and the 11/40 did or it could end up breaking process control systems that used the PDP11. There wasn’t a formal instruction set for the PDP/11s, though at one point I think someone tried to document one. The key goal in implementing the 11/60 instruction set was to make sure that the 11/40 diagnostics would run on it (as best the could)and the 11/40 version of the operating systems ran on it. The 11/70 team had similar goals for 11/45 diagnostics. One big area of differences were in the setting of status bits that are set at the completion of an instruction. The diagnostics and operating system had dependencies on undocumented results in the status registers…each model set those bits differently (surprise, surprise). In some cases they used them to determine what model machine they were on. In some cases customer software made use of undocumented status (and other) results too. I had to insure that the 11/60 set the same undocumented values in the status registers as the 11/40 did… did I mention that those settings were undocumented :-). The 11/60 had a one-way associative cache and the 11/70 had a two-way associative cache. In practice I don’t think the 11/70 ended up with a better hit ratio than the 11/60. The 11/60 and 11/70 were the first cache machines for DEC. Needless to say the cache did a number on the existing memory diagnostics… so they had to be updated. Oh hum, things I haven’t thought about in five decades.@@jeremybarker7577
There are tons of PDP equipment in the Rhode Island Computer Museum. They have PDP-8, PDP-9, PDP-11 in perfect working condition. Tons of spare parts as well. There is a person there who specializes in PDP restorations of all kinds
28:19 Yeah, I laughed. You can have lots of spare parts, but it doesn't help if they're all the wrong part. Adrian had power supply problems in his episode yesterday, but at least he had the right kind of junk to scavenge from. Even then, he still ended up using a HOT from a CRT as a pass transistor.
Basic switching supply theory is not that difficult. Diagnosing one however can be a real headache. I think you have been incredibly lucky so far in most videos. A video like this was bound to happen. I'm surprised it doesn't happen more often.
Many moons ago I had the job of testing 5V 200A power supplies. They would never start up until I realised that they required a small initial load to get them going (rtfm). Very similar symptoms to what is going on with the PDP psu.
What my eye went to immediately I saw that circuit is the resistor network on the control pin of the voltage regulator. This network is used to boost the voltage over the 5V which the regulator is set to by default, in order to compensate for the voltage loss in the load transistor. Since a crowbar circuit responds to an over-voltage condition and adding greater resistance to the control pin of the regulator brings the voltage up, I'd be looking for a dud resistor on that control pin. There are two resistors (one is the potentiometer) which could cause an over-voltage if they went high, and one (seems to be connected to -15V) which could cause the same problem if it went low. But the first thing I'd do is attempt to adjust the potentiometer and see if it's possible to adjust the voltage down until the crowbar no longer triggers. Less resistance should lower the voltage in this case. The whole circuit is complicated by having more than one voltage output. If it's possible to isolate all other voltages then I'd do that in order to confirm that the fault is only on a single supply output and not on multiple outputs.
10:40 I've had to do that exact fix (I refuse to call it a bodge) on numerous older systems where something got stomped after years of being stored. The clean socket is a massive help for testing and checking continuity. I've even gone do far as to set up a little tester, just a breadboard which the socket sits in and I can test continuity with a hand-held meter no problem; excellent to see. And the cable fix is spot on!
That was a great haul of DEC hardware and very good of you to distribute it to other DEC hobbyists. I used a small, desk sized PEP 11 in a realtime programming class, circa 1980. Our lab had a dozen of them. I am familiar with Bomen. I worked for a company called Digilab 1978-79. Digilab was the first company to build and sell FT-IR spectrometers. I did the final test and troubleshooting of 2 or 3 spectrometers every month. They used Data General machines. I think Bomen got into that business in the '80s.
Thank you! That old Weller is awesome! Everything I'm doing could totally be done on a much cheaper, older analog scope too. I just use the digital job when filming so I can take scope shots mostly.
If you go into a professional electronics repair workshop, you'd probably be shocked just how old and apparently low-spec the test gear is! Quite often, when hobbyists get into electronics, it turns into a Pokemon collect-all-the-gear hobby, where people are convinced they need a 4 channel, 10GSPS smart scope and a 7.5 digit meter.
The base emitter junction will act like a diode with the collector disconnected. That's not "leakage" as you said. It's normal behavior. Dnnn transistors are 2SDnnnn. Just a Japanese number.
The red and green transistors are ancient General Electric parts. D44xx were quite popular many years ago. That is the full part number. They are not 2SDxxx Japanese parts. The D44C8 is still being made by ON Semi.
I have a nicely working PDP-11/23 that has a SCSI q-bus card and can boot from a semi-modern hard disk if you'd be interested in it, along with a variety of spare parts (RAM, RL02 interface, RL02 drive, etc).
I have seen similar principles in power supplies. They are designed to power high-load consumers and sometimes they don't work correctly without a load. I would advise connecting some kind of safe load to this power supply. I usually used a couple of powerful 12v x 21-watt car lamps as a load.
I have replaced maybe 100 of these power supplies during my field service life. I have never tried to use one that did not have a load in the backplane. Put a proper load on that power supply. It was not designed to operate without a normal load. Dec Field Service Engineer badge number 83903.
Many years ago, in the early 1990ies I helped a friend bring a PDP 11/23+ to life. He had got it from his uncle who had had the intention to use it in his company. We got it together with two spare CPU-chassis. We didn't have that much information on it, and there was no internet to query. We turned it on, and it run for maybe a few seconds, and then a large pop and it went silent. Turns out we had put it into an unearthed outlet by mistake. Since we had no knowledge on how to repair the PSU, we took the PSU from one of the other chassis and put it into this. We had to modify the chassis a bit to make the other PSU fit, but it worked perfectly. We played around with the machine for some time, making very sure we always plugged it into a grounded outlet. But one day when we should turn it on, we heard that pop-sound again! We just couldn't understand why, it was plugged into an grounded extension cord, as usual. But on closer inspection someone had connected that grounded extension cord to an extension cord that was NOT grounded, but had been modified to be able to plug that one into a grounded outlet. So we were down to one working PSU. THAT, if noting else, taught me to ALWAYS double check your connections.
Hi, I just tried to wrap my brain around it: The "pass transistor" is actually the switching transistor. While the 12V output is at 0V and the base of Q15 is not pulled to 0V via the 12V shutdown signal, Q15's base will be pulled to Vin via R47 and R75. This feeds Vin to the input of the 7812. The current also flows through R75 (the 10R resistor you blew). Normally when all parts are in place the voltage drop over R75 will reach 0.7V at a current of about 70mA through R75, Q15 and the 7812. This will turn on Q18 and the switching transistor Q17. The voltage will ramp up until the current through R75, Q15 and the 7812 drops below 70mA and then the regulation by switching Q17 takes place. Your light emitting resistor R75 let me suggest that you have some severe load on the 12V rail (some shorted cap or similar). As without Q17 turning on, the current will never get high enough to trigger the over current protection. R75 just burns out (nearly the full Vin drops on it with Q17 off). So the next step is to find the cause of the low resistance of 12V to GND, then R75 should no longer burn out and you can put the other parts back into place... Let me know if my thoughts are right...
I admire your discipline and principles! If I would run old computers (I don't), the first thing I would do is to replace the power supply with modern replacement kits and emulate disk drives whereever possible. Your persistence in trying to keep these things as original as can be is amazing and motivates me in my daily life to never take cheap shortcuts that degrade perfection.
If you put in a modern PSU and emulate the drives, you may as well emulate the whole computer with a Pi. It takes all the fun out of vintage computers. 👍
I don't think anyone should get in to this hobby if they aren't willing to restore and use original equipment wherever possible. Emulated drives are good for hobby reproduction systems... or if you can't find the right drive/media. Neither of the things you mentioned would be unreversible though... so not too bad.
@@maskddingo1779the original drives are aging and in far more limited supply than the actual systems they serve. Emulated storage isn't just convenient, it also saves you from repeatedly cycling an aging drive while diagnosing for example. I think a restorer with emulated drives is just smart. If you can keep an original working that's great, but using emulation saves it from wear and tear.
@@maskddingo1779also power supplies in the past did not always have the greatest protection circuits. If they had them at all. Using a modern well protected power supply, might be less original, but it protects your machine from being blown up because of power supply failure. And failure becomes ever more likely with age. Even curious Marc recently had a video where he added a crowbar circuit to a vintage power supply that had broken and killed the computer attached because it did not have a crowbar circuit. Something HP added to later models. Original no, safe yes.
@@bzuidgeest Has UE said what drives were in / were avalible to this system at new (late 1980 to me isnt that long ago)
Před 8 měsíci+1
I love these types of troubleshooting episodes. They are both entertaining and educational! It matters very little to me personally if you manage to wrap up and fix it in a single episode, I'm here for the journey 😅
You know I don't really know too much about electronics, but I really like the logical breakdowns. I think if I keep watching I may learn something. Thanks!
You need to put a small load on the 5 and 12 volt outputs. Say a tenth of the normal load. Without a load the series regulators are running outside their lower current spec. Also a variac is your friend. As are a few of those $9.95 DVM's from HArbor Freight. Slowly turn up the AC line voltage on the variac while monitoring the outputs, under a light load. That way you'll blow fewer transistors and resistors. If it helps, I've been stumped by several Tek 485 scope power supplies. They are extremely complex, with a bunch of overvoltage and overcurrent shutdown circuits. If any power supply is sensed to be out of spec, everything shuts down. Tek even published a separate troubleshooting guide just for the power supply, it's so convoluted. The scope itself is very high tech, but I've ended up junking two scopes due to at least for me, unfixable power supplies.
He had high watt load resistors on the power supply at one point in the video. The LM7805 is a linear type regulator, used for the 5v rail and it doesn't require a minimum load as the internal draw of the IC (2-3ma from memory) is sufficient. If I recall the 485 has a switching style power supply not a linear system.
@@user-uz1yv2oc9v The 5 volt supply is also a buck converter, just using a three-terminal linear regulator pressed into duty as the controller for switching.
Enjoy your 11’s once you solve this power issue (which you will!) They are great computers to use and program on (I did for nearly a decade!) I also build up a home collection a few decades ago; however, I miss owning PDP11’s now. I had a 11/05, 11/34, 11/44, RL02’s, and a ton of peripherals. Then I realized I needed to reduce my hobbies and they all went to another collector where they live happily to this day. But I kept the PDP11 books so I can be nostalgic anytime I need too. Oh … and I had a PDP8/E that I restored (a much simpler power supply fix - just throw in new caps!) Funny how things changed and got more complex from the 8/E to the 11/03; however, they also got a lot lighter!!!
Interesting stuff again. It really made me laugh when you said that something went over your head because all the things you do, I just usually don't understand, heheh. Cheers from Finland!
The base-emitter junction of Q15 SHOULD pass voltage (although it wouldn't pass much current before you fried it) Unfortunately, since ALL of the current must now pass thru that B/E junction, it's quite probably been transformed into a 3-way piece of wire... I know there's no 'external' load on the PSU, but there IS an internal load being the 560uF output capacitor (C20) that has to be charged up! (More on that later). The fact that you are getting over 13V on the output is worrisome to me... The 7812 SHOULD keep it very close to 12.0V, so MAYBE this too is 'fried'? However, this circuit allows you to 'adjust' the +12V rail a bit (using the potentiometer based voltage divider on the output. I'd confirm the EXACT voltage between pins 2 and 3 of the 7812 before I called it dead! (Pin 2 is NOT directly connected to the overall circuit ground!) Next comes the LER (Light Emitting Resistor)... You REMOVED Q18 driver and therefore ALL of the current on the 12V rail cannot come from the series pass transistor (Q17) since it no longer has any drive. This means ALL of the 12 rail current MUST now pass thru Q15!!! Let's ASSUME for a second that there is a direct short across between the emitter of Q15 and ground. The incoming +24V will turn ON Q15 by way of the base bias current (R75 + R47) and thus you now have 24V across a small (1/4W?) 10 ohm resistor (R74)... 24V/10ohm = 2.4 AMP! WAY more than it could ever hope to handle, so it becomes a FUSE instead and gives off the light show in the process. (R74 has 'left the chat' and is now an infinite ohm resistor) So now the ONLY remaining current path is thru R75 + R47 + base-emitter of Q15 into the 7812. With R47 being 330 ohm, there's simply not enough current flow to create a new light show. My _GUESS_ is that the 7812 is also NFG but that's yet to be determined! If's possible that the adjustment is turned up so far that it causes the crowbar to kick in, but, without the 10 ohm resistor, there is enough of a voltage drop across the R74 + R47 base-bias that the crowbar doesn't get a chance to trigger. 1: Check C20 for leakage / short 2: Disable the crowbar and retest (remove D30, R62 and Q14)
Switch Mode power supplies can be a real bear to diagnose. Lots of ways to fail and interacting subsystems to throw you off. I see you have fixed it now (in a new episode). Congrats
No.1 before powering any old power supply is just replace ALL electrolytic caps with good quality low ESR types. Then check all diodes & transistors with a meter (in circuit). Look for dry joints too, before cranking up slowly with a variac, preferably with a lightbulb in series in case something does short out.
try testing power supplies by powering up their outputs, using bench supplies. this is how i test PSUs that i repair, as well as new off-line designs - by powering the primary side SMPS controller from a LV supply but no AC input, you can check that the output stage can reach its setpoint, and also check that the feedback network is working correctly
0:52 If that's the same Dave McGuire I'm thinking of, I met him once at his house ~20 years ago (in Tampa, FL or thereabouts). He had a supposedly working full-rack PDP/11 right in his living room, next to an original radioactive-decay tube-based time standard, and had converted his two-car garage to a data center. There were 2 Crays in there which he was using to calculate Pi digits for some kind of contest. He bragged once about his ~$800 power bill (in ~2004 dollars). I would hope that all of those machines are now part of the exhibit in his museum, as they were in great condition back then, and he was pretty fastidious about such things.
It's often faster to just desolder the power transistors and check them all and also replace any electrolytics that seem a bit duff. You can check a few other components with a meter in circuit like low ohm resistors. Bipolars often fail via direct shorts between B-E or B-C.
To be honest, I wouldn’t apply mains to something this old without replacing electrolytics outright. Call me overcautious but there were two, ahem, “excursions” in this video and there is simply no future in going forward with 40-year-old electrolytics even if they appear to hold back DC.
@@hubbsllc Yes there is a good argument for that, especially considering how much it could damage if that 5V turned into a 12v supply. However I suspect the old Western electrolytes are longer lasting than the new ones. With the old ones you would only have a problem if they were in close proximity to the places where they could get hot, so pull out some of the best suspects and test them to get an idea what the rest are like. The thing is, with a fuse like that I'd see it more likely one of the power transistors has failed, and sometime that takes out a driver transistor. Sometimes they can half fail so you have a leakage across the junctions. It's better to spend 20m checking all this before applying power as well.
@@Andrew-rc3vh Yes - a blown fuse (especially if it had been checked from the get-go) is a good indication that something already went awry so you know there has already been an overcurrent event. Those electrolytics can appear to hold a voltage but in the case of filter caps they're expected to experience AC current - not exactly the conditions of that test. And on top of that they may appear to hold voltage *now* but what about after 15 minutes in an energized circuit? This does imply that if you're going to entertain getting a machine like this running and keep it running, there's a certain cost/effort to sink into it before you even give it mains which is why you need to be clear about what your goal with a piece is and how committed you are before you even put it on the bench. I restored a 1954 tape recorder and started by rebuilding the whole power supply onto a perf board; a dual-element power resistor had one leg that was open-circuit and there were two can caps that I removed outright. I didn't even plug the power cord in until that much was done.
Hmm, as you replaced that 180 µf cap, the bigger electrolytic, and eventually pulled that D44C9, I kept thinking, "Yeah, I'm sure I've got something in the basement that will work." (Exact replacement for the smaller cap, smaller form factor replacement for the big one in case you want to hide it in the original case, and "D44C9 - (checks data sheet) 60-70 volt collector voltage, don't need anything close to that here, pretty wimpy beta, 30 watt power dissipation, something like an MJE15028 ought to work fine - its fT is a little slower but I bet the switcher isn't clocking that fast, and the pinouts (E, C, B) are even the same." But my basement is in New York and your shop is in Texas, and I'm sure you have parts on order already. In any case, it looks as if you have a buck converter on your hands. They don't regulate well with no load, and that's probably tripping the crowbar. Try loading the 5V rail with a 5 ohm 10 watt resistor and maybe a 20 ohm on each 12 volt rail, and see how it does. (I'm guessing that you haven't poked into the guts of a switcher before, because you're approaching this thing as if it's a linear power supply.) By the way, an SCR is a kind of thyristor.
Your calculation of frequency from the scope trace is not correct. The period is ~250ms, or 4 Hz. That is a compact power supply - good luck. I enjoy you working to resurrect these vintage pieces.
Yes, definitely come see us at the Large Scale Systems Museum! One day I'll get Dave to be willing to talk on camera long enough to catch all his stories :)
Would it not have been simpler to just replace the connector on the ribbon cable? Looks like a fairly standard DIP/DIL IDC to ribbon cable connector to me that are still widely available.
Absolutely! Except, I didn't have a spare, so instead of waiting on parts to arrive, I figured I'd take a stab at repairing it so I could continue on with the build. I'll order up some replacements and fix it properly once I know for sure the PSU is going to come back to life.
Great video! I've got a power supply issue on an Apple Lisa (also a big 'pop' after replacing the fuse and found a trace entirely vaporised). Watching you do this gives me the understanding and confidence to forge ahead. Cheers!
I would love if the PDP11 from The Weather Channel ever surfaces, they used one to run all the STAR units in the field over satellite, unearth that long lost 4000 software.
I suspect the PS had a catastrophoc event that cascaded one problem into several. Switching PS are a feedback control system which make isolating the problem extra difficult. You are getting close, though.
Will the Power supply Grimlins be defeated find out on the next exciting episode on Usagi Ball Z. I cant wait to see what happens next. this is about the only way ill get to mess with these old computers. Thank you for all the amazing content, I appreciate all the hard work that goes unseen. I know this kind of stuff isn't always easy.
It seems a lot of people are unaware of the fact that the common equations for switchmode converters don't apply at light load where the current in the inductors goes into the discontinuous region, dropping fully to zero each switching cycle. With buck converters, as used here, the usual expression is _Vout = Vin x duty cycle_ but that *only* applies with continuous inductor current. In the discontinuous region the duty cycle must be reduced and at very light loads it must go very close to zero. Lots of older control circuits couldn't manage zero duty cycle very well. Sometimes they were OK except you'd get quite a bit of ripple because the "squeg" - run for a few cycles then turn off completely. The circuits here use three terminal regulators to actually do the control of the switching. I get very frustrated trying to find the brief view of the relevant schematic bits in videos like this, so I haven't looked closely at the design. That slow rise of the 5 volt rail very strongly suggests to me that the circuit can't bring the no-load duty cycle down far enough to maintain regulation I would test these converters with at least 10% of nominal full-power loading. That won't likely be enough to assure continuous inductor current, but may be OK anyway (it is common to design to place the discontinuous/continuous boundary at something in the range of 20 to 40% of full rated load).
In general, the presence of an LM7805 and LM7812 implies they are being used as voltage references for the pass transistors that are handling larger currents. Since the 5V and 12V outputs are both showing high, I'd suggest pulling both regulators and testing them in a reference circuit on a breadboard. They're common jelly bean parts and are likely easier to source replacements for than the pass transistors, etc.
The other culprit could be the 7805 or 7812 regulator. It looks like the circuit is using them as a kind of current switch. When the Vout is going above the 7805/12 regulator voltage the linear regulator will limit the current and turn off. This will stop all current through the control transistor and R27 (5v circuit) which will have the effect of switching off the driver transister and pass switch. The Vout should drop as the inductor loses energy and the cycle will start again. If the regulator is no longer regulating then it might behave the way you are seeing. Worth doing the same checks on 12V and testing the crowbar circuits for both outputs, I think you could just hook your bench PSU up to it with the current limit set to something sensible and bring up the voltage on your bench PSU to see where they are tripping. Saves you having to switch it on and risking other components.
I agree with your take on how the circuit functions. One hook with using a bench supply, which in general I agree is a much safer approach in terms of avoiding doing more damage, is that sometimes these sorts of supplies have some additional low-current supplies that are necessary. In this circuit a -15V supply is used in the setpoint networks for the 3-terminal regulators. All it does is provide some bias current for diode used as a crude negative voltage reference. In things like bench power supplies a negative supply may have a more critical role. It certainly isn't an insoluble problem, but something relative novices may not have seen or may not understand..
Your « failures » are as interesting as your successes. For a viewer, it’s the process which is interesting more than the result. And it is probably just a temporary failure so, no regrets.
Any time you are confronted with circuitry with adjustable parts that has been mucked about with by someone else, you *must* assume the mucking about has included fiddling the twiddlables. In fact if an amateur was as it, is is very likely said amateur started with such twiddling. This can make logical troubleshooting substantially more difficult. It can be difficult to determine if something is wrong because something has failed or because some knob turned a knob.
I'm by no means an electrical engineer but that is a bizarre way of implementing overload protection. Maybe it's commonplace and somebody will correct me, but usually most circuits just kill power downright. Definitely a complex looking protection system on this thing. At least it was still doing its job all these decades later, so the designers definitely got something right!
FTIR is a common spectroscopy technique to identify or otherwise get structural info on molecules. It'd be super cool if you could snag an old spectrometer. Old science instruments are really cool.
Interesting circuit, completely linear voltage regular that uses a 7812 as voltage reference and error amplifier at the same time. The reason R74 burn up when Q18 is removed is because the B-E junction of Q18 is keeping R74 at 0.7V. Without Q18, the current going through Q18 is essentially all the current going though the 7812. Q15 is just a simple common base amplifier configuration, it is there as a switch for the shutdown cutoff, it is unlikely to be broken. Leakage isn't what you are looking for either, as B-E junction of Q18 is 0.7V, so the R75, R47 is passing current though to the B-E junction to power the 7812. As for why the voltage output is creeping up, my guess is the 7812 is drawing more quiescent current than the circuit is initially designed for, so the extra current is causing the Q15 to pass more current to the output, even if the 7812 is in full cutoff state. Add a resistor from 24V source to ping 1 of 7812 might help, experiment is necessary to find the correct resistance. General description, Q15 is the cutoff switch. Q18 is a current gain stage and also to change the direct of the current flow and the Q17 is the main power transistor.
I don't think power supply repair is your speciality 😁 I never would just switch on such an old PS like you did. What I also don't understand is that you remove transistors without testing it. Most transistors and diodes can be tested in circuit with just an ordinary DMM.
Absolutely. Where is his supply protection? Where is the ability to ramp up the supply voltage gradually (variac), rather than just subjecting any gear to a power input surge after years of sitting idle? It was the same with the G15 at first switch on .... He hadn't even separately tested the fans or the drum motor. He just threw the switch. There he was lucky, but it shouldn't have to rely on luck.
Signetics 82S100 PLA chip on that CPU board. Same as the infamous PLA in the Commodore 64. The original 82S100 was pretty reliable, despite running hot as hell. It was the MOS clones that frequently failed.
I think it would be awesome if someone got one of these up an running and showed the Fourier analysis described. Seems like something more than just doing a listing of directories could be done on these machines. A lot of time and effort put into just getting to a command prompt and listing directories and files.
If it were me, I'd just pick up a 5 & 12V switching power supply from the getting spot and be done with it, but I can understand wanting to keep this as original as possible. Certainly makes the job a lot more difficult, but that just means you'll have learned more when you're done.
Maybe I am wrong, but remember that a BJT (Q13) is essentially two diodes, in this case, from base-emmiter and base-collector. When you put your input voltage from the rectifier, is normal to pass through R75+R47=890OHM, the B-E junction to be regulated by the 7812 and so you should have those 12V at the 12V output power rail, albeit at a very low current. Without R74 (10Ohm) the behaviour of Q13 is normal, it is not leaking anything. Actually, you shoud check the voltage between the ground pin (the middle pin) of 7812 and ground and add to it 12V; this you should have at the output. If is not ok, your 7812 may be shot. 7812 is not regulating to 12V in refference to the ground but in refference to the voltage set by the resistor network R77, R86, R87 and this voltage can be regulated by the R87 potentiometer. Try to check/clean those potentiometers (R37/R87) and any others you can find, connect the transistors after you checked them and readjust the output voltage by tunning them. You may (in fact should) have to disable the crowbar circuits to ignore the overvoltage protection to do this easyer. Actually, that 10 Ohm R74 may have burned because the crowbar circuit, if nothing else is drawing current (check if 7812, D32, C24(?)=0.22 are ok).
I know the feeling! I was given a Zenith 486 laptop yesterday. Power supply was dead. Found a shorted diode and replaced it with one from a scrap board. Laptop POSTed once and then went in to a power cycling loop. Pulled apart the machine and found loads of leaky surface mount electrolyitics... And I have none of those on hand. 😔
Very helpful background info regarding these supplies, I have a poorly LSI11/24 with a dead H777 PSU which I'd love to get going at some time. Many thanks!
I run my franken-11 off a salvaged ATX supply and generate the power-on signals and LTC with one of those $3 ESP-8266 modules. If that thing has you flummoxed, you don't *have* to use an expensive and complicated DEC supply.
I cannot even describe how many devices I seen been killed by old capacitors by now, They leak electrolyte and corrode traces, they short out, they leak electrically and causes massive fluctuations in voltage ripple. it is very common when a cap short out it takes many other components with it such as diodes, resistors and transistors that suffer over current. you should never ever trust old caps to be good. it is a good thing to make resistance measures to check for shorts and if caps are leaking before powering stuff and it can save a lot of trouble down the line. I have the feeling those shorted caps can be one reason why the supply have died this much, when things literally light up from shorted stuff you know there been a good BBQ party in the PCB house
Ah, happy/not happy days of 3 engineers in our workshop also working on an 115V 11/03 PSU. We were defeated too, so I hope there's success in the next video. Too late for our machine, this was 15 years ago. Fingers crossed!
Got an idea for you .. Why not take a 100 watt incandescent light bulb and put it in series with your power cord to limit current going to a unit when you plug it in for the first time. Could use maybe 2 of them .... Keeps fuse from blowing , and if the lamp lights up bright you know you got a problem.. If it lights bright and dims.... then it may be OK ..... I started doing that back in the 70's and it saved me a lot when replacing high power audio transistors.. Just an idea..
Isn't your power supply a buck converter? Your "pass switch" transistor should be turning on and off at some 10's of kHz frequency. I'd disable the crowbar and the overcurrent protection (just remove the SCR and the input at the the current protection transistor). Then feed the thing with some current-limited power supply at F1 and ramp it up slowly. That should allow you to debug it with more safety and less ElectroBOOM potential... At some point the pass transistor should start chopping, then you can work on the regulation loop. You can even temporarily open the regulation loop to untangle things. Shooting from the hip here, I have not looked at the doc, so take this with a big grain of salt.
It's conversations and comments such as this that make me feel like an eager student at the Lyceum.😊
thats what i thought! the pass switch, "free wheeling" diode and inductor look like a standard buck converter to me.
although the feedback and regulator circuit look very different from what I'd expect in a modern switching supply 🤔
I heard elevator music in the background while reading this :)
also could be that the unloaded output is actually whats causing the crowbar circuit to kick in
@@juliankandlhofer7553That is what I thought. Everything is working as designed, but there is no load. I would want a headlight bulb on the output to see.
The best kind of parents are the ones who have no problem stopping to throw a couple hundred pounds of minicomputer into their trunk on their way to visiting you.
A light-emitting resistor. I've got to remember that one.
I'm just glad he had a camera pointing at it to capture the event.
every electrical component can emit light (and smoke), *At Least Once*
Goes with smoke-emitting diodes.
I've done it a few times - the biggest being a 10 ohm 50-watt resistor (I can still remember the foot-tall flame that shot out of it - 30+ years later! The things that happen as a rookie tech are easily seared into one's mind).
That style of switch mode power supply was quite common in 1970s computer and digital processing gear. They knew the efficiencies of a switch mode power supply which is almost mandatory at the current levels needed, like over 20 amps. But at the time we didn't have reliable high voltage AND high current transistors or MOSFETS. There were TV horizontal output transistors but that was not much current. Triacs and SCR were not suitable at the ultrasonic frequencies needed for a direct line operated supply. So the solution was a standard 50/60 hz transformer/rectifier based unregulated power supply feeding a switch mode regulator. That drive circuit is actually an oscillator which means a feedback loop. Very difficult to troubleshoot. As the component count is reasonably low, I would just test all the semiconductors. Even in circuit you can get a pretty good idea if the part is shorted or open.
I repaired a lot of PDP power supplies two decades ago... Zip-ties are exactly the correct way to secure the capacitor.
I see you have an 11/60 on your inventory list! I designed the Writable Control Store in it along with the clock and other odds and ends. I also wrote the microcode that implemented the instruction set, boot up, and all the other control stuff. Neat to see that one is still alive. At the time it did the fastest register to register operations of any of the PDP-11s... to bad that was not an important metric to customers though.
I really wanted an 11/60 back in the day, including when I was trying to start a computer school in the mid-80s. I had a lot of equipment for it, including SMP KL-10 pair from Intel and a 2065 from Columbia University. Then I developed inflammatory bowel disease, couldn't work, and lost everything in storage.
If the 11/60 wasn't limited to 18-bit addresses but had the 22-bit extensions from the 11/70 I think it would likely have been far more successful than it was. The microcode also could do all the floating point instructions if the FP hardware wasn't present.
Over 40 years ago when I was working for DEC we investigated using the WCS option to create special instructions to handle some operating system functions unfortunately the performance gain was insufficient to take it into any product. Unfortunately I no longer have the microcode listings although I still have all the WCS manuals.
The 11/60 was, in effect, considered to be an enhanced 11/40 and the 11/70 was an enhanced 11/45. The 11/40 and 11/45 were really different classes of machine. In fact the head designer on the 11/60 also was the head designer the 11/40 (and the original 11/(20)). So the feature space we worked in was more limited than the one the 11/70 design team worked with. That’s one of the reasons for the difference in address space size.
A microcode implementation of floating point might be faster than a regular software implementation but I doubt that it would come close to the hardware implementation performance. The FPU for the 11/60 ran asynchronously so you could be doing integer arithmetic (or anything else) at the same time the FP instruction was running. A microcode implementation would run inline with the program. Also there are cases in floating where the result is, in effect, unspecified. So any floating point implementation would have to be sure to produce the same unspecified results and the 11/40 did or it could end up breaking process control systems that used the PDP11.
There wasn’t a formal instruction set for the PDP/11s, though at one point I think someone tried to document one. The key goal in implementing the 11/60 instruction set was to make sure that the 11/40 diagnostics would run on it (as best the could)and the 11/40 version of the operating systems ran on it. The 11/70 team had similar goals for 11/45 diagnostics.
One big area of differences were in the setting of status bits that are set at the completion of an instruction. The diagnostics and operating system had dependencies on undocumented results in the status registers…each model set those bits differently (surprise, surprise). In some cases they used them to determine what model machine they were on. In some cases customer software made use of undocumented status (and other) results too. I had to insure that the 11/60 set the same undocumented values in the status registers as the 11/40 did… did I mention that those settings were undocumented :-).
The 11/60 had a one-way associative cache and the 11/70 had a two-way associative cache. In practice I don’t think the 11/70 ended up with a better hit ratio than the 11/60. The 11/60 and 11/70 were the first cache machines for DEC. Needless to say the cache did a number on the existing memory diagnostics… so they had to be updated.
Oh hum, things I haven’t thought about in five decades.@@jeremybarker7577
@@dansullivan6787 Thanks for that additional information Dan. The work I did dates back to 1978 so like you it's a long time ago to remember.
There are tons of PDP equipment in the Rhode Island Computer Museum. They have PDP-8, PDP-9, PDP-11 in perfect working condition. Tons of spare parts as well.
There is a person there who specializes in PDP restorations of all kinds
If you want a 'portable' DEC system, what you need to look for is a VT103. That's a VT100 terminal with a built-in four slot backplane.
28:19 Yeah, I laughed. You can have lots of spare parts, but it doesn't help if they're all the wrong part. Adrian had power supply problems in his episode yesterday, but at least he had the right kind of junk to scavenge from. Even then, he still ended up using a HOT from a CRT as a pass transistor.
I was just watching that video this morning!
Pure coincidence that we're both battling power supply problems, haha.
Basic switching supply theory is not that difficult. Diagnosing one however can be a real headache.
I think you have been incredibly lucky so far in most videos. A video like this was bound to happen. I'm surprised it doesn't happen more often.
Many moons ago I had the job of testing 5V 200A power supplies. They would never start up until I realised that they required a small initial load to get them going (rtfm). Very similar symptoms to what is going on with the PDP psu.
I think it needs a load too.
An empty mechanical lead pencil with metal top are the perfect tool for straightening bent pins.
A wire wrap tool also works.
Well, I guess you’ll have to design a +5/+12 VDC power supply out of vacuum tubes. That should use up some 6AU6s. 😅
Switch mode vacuum tube power supply, that'd be "cool" 😂
What my eye went to immediately I saw that circuit is the resistor network on the control pin of the voltage regulator. This network is used to boost the voltage over the 5V which the regulator is set to by default, in order to compensate for the voltage loss in the load transistor. Since a crowbar circuit responds to an over-voltage condition and adding greater resistance to the control pin of the regulator brings the voltage up, I'd be looking for a dud resistor on that control pin. There are two resistors (one is the potentiometer) which could cause an over-voltage if they went high, and one (seems to be connected to -15V) which could cause the same problem if it went low. But the first thing I'd do is attempt to adjust the potentiometer and see if it's possible to adjust the voltage down until the crowbar no longer triggers. Less resistance should lower the voltage in this case.
The whole circuit is complicated by having more than one voltage output. If it's possible to isolate all other voltages then I'd do that in order to confirm that the fault is only on a single supply output and not on multiple outputs.
Love your videos. The only “complaint” I have is they come only once a week. 😀🤷🏼♂️
the Thyratron sounds like something that Suzanne Somers would have done a infomercial for back in the 80's
Squueze it, squeeeeeze it! Tighten those buns ladies!
10:40 I've had to do that exact fix (I refuse to call it a bodge) on numerous older systems where something got stomped after years of being stored. The clean socket is a massive help for testing and checking continuity. I've even gone do far as to set up a little tester, just a breadboard which the socket sits in and I can test continuity with a hand-held meter no problem; excellent to see. And the cable fix is spot on!
Please don’t apologize! You have no idea how much I get out of watching you troubleshoot the issues and work through the possible solutions.
That was a great haul of DEC hardware and very good of you to distribute it to other DEC hobbyists. I used a small, desk sized PEP 11 in a realtime programming class, circa 1980. Our lab had a dozen of them. I am familiar with Bomen. I worked for a company called Digilab 1978-79. Digilab was the first company to build and sell FT-IR spectrometers. I did the final test and troubleshooting of 2 or 3 spectrometers every month. They used Data General machines. I think Bomen got into that business in the '80s.
I immensely appreciate that you went to the trouble of shooting an insert of you plugging the power cord into the wall. 😄
An official DEC power cord, no less! Subtle flex. 😂
What at awesome video. I love that you are doing it all with an old weller soldering iron and an inexpensive o-scope.
Thank you!
That old Weller is awesome!
Everything I'm doing could totally be done on a much cheaper, older analog scope too. I just use the digital job when filming so I can take scope shots mostly.
If you go into a professional electronics repair workshop, you'd probably be shocked just how old and apparently low-spec the test gear is! Quite often, when hobbyists get into electronics, it turns into a Pokemon collect-all-the-gear hobby, where people are convinced they need a 4 channel, 10GSPS smart scope and a 7.5 digit meter.
I think this weekend I'll binge watch your channel and everytime you say "close enough" I have to drink a shot. Could be a fun day.
The base emitter junction will act like a diode with the collector disconnected. That's not "leakage" as you said. It's normal behavior. Dnnn transistors are 2SDnnnn. Just a Japanese number.
The red and green transistors are ancient General Electric parts. D44xx were quite popular many years ago. That is the full part number. They are not 2SDxxx Japanese parts. The D44C8 is still being made by ON Semi.
12:49 I always suspected you might have a screw loose but now I know for certain 😝 Good luck with the transistor replacement!
Nothing a screwdriver can't fix! At least that's what I told the doctor, but he just said "Nah man, you crazy."
I have a nicely working PDP-11/23 that has a SCSI q-bus card and can boot from a semi-modern hard disk if you'd be interested in it, along with a variety of spare parts (RAM, RL02 interface, RL02 drive, etc).
This channel is gem on youtube.
I have seen similar principles in power supplies. They are designed to power high-load consumers and sometimes they don't work correctly without a load.
I would advise connecting some kind of safe load to this power supply. I usually used a couple of powerful 12v x 21-watt car lamps as a load.
Lamps or ceramic resistors are a great test load. Both keep the bench warm and cozy.
I have replaced maybe 100 of these power supplies during my field service life. I have never tried to use one that did not have a load in the backplane. Put a proper load on that power supply. It was not designed to operate without a normal load. Dec Field Service Engineer badge number 83903.
YAY New Zealand. I have a PDP11/23 in my garage with software and manuals
Many years ago, in the early 1990ies I helped a friend bring a PDP 11/23+ to life. He had got it from his uncle who had had the intention to use it in his company. We got it together with two spare CPU-chassis.
We didn't have that much information on it, and there was no internet to query. We turned it on, and it run for maybe a few seconds, and then a large pop and it went silent. Turns out we had put it into an unearthed outlet by mistake. Since we had no knowledge on how to repair the PSU, we took the PSU from one of the other chassis and put it into this. We had to modify the chassis a bit to make the other PSU fit, but it worked perfectly. We played around with the machine for some time, making very sure we always plugged it into a grounded outlet. But one day when we should turn it on, we heard that pop-sound again! We just couldn't understand why, it was plugged into an grounded extension cord, as usual. But on closer inspection someone had connected that grounded extension cord to an extension cord that was NOT grounded, but had been modified to be able to plug that one into a grounded outlet. So we were down to one working PSU.
THAT, if noting else, taught me to ALWAYS double check your connections.
Hi,
I just tried to wrap my brain around it:
The "pass transistor" is actually the switching transistor. While the 12V output is at 0V and the base of Q15 is not pulled to 0V via the 12V shutdown signal, Q15's base will be pulled to Vin via R47 and R75. This feeds Vin to the input of the 7812. The current also flows through R75 (the 10R resistor you blew).
Normally when all parts are in place the voltage drop over R75 will reach 0.7V at a current of about 70mA through R75, Q15 and the 7812. This will turn on Q18 and the switching transistor Q17. The voltage will ramp up until the current through R75, Q15 and the 7812 drops below 70mA and then the regulation by switching Q17 takes place.
Your light emitting resistor R75 let me suggest that you have some severe load on the 12V rail (some shorted cap or similar). As without Q17 turning on, the current will never get high enough to trigger the over current protection. R75 just burns out (nearly the full Vin drops on it with Q17 off).
So the next step is to find the cause of the low resistance of 12V to GND, then R75 should no longer burn out and you can put the other parts back into place...
Let me know if my thoughts are right...
I admire your discipline and principles! If I would run old computers (I don't), the first thing I would do is to replace the power supply with modern replacement kits and emulate disk drives whereever possible. Your persistence in trying to keep these things as original as can be is amazing and motivates me in my daily life to never take cheap shortcuts that degrade perfection.
If you put in a modern PSU and emulate the drives, you may as well emulate the whole computer with a Pi. It takes all the fun out of vintage computers. 👍
I don't think anyone should get in to this hobby if they aren't willing to restore and use original equipment wherever possible. Emulated drives are good for hobby reproduction systems... or if you can't find the right drive/media. Neither of the things you mentioned would be unreversible though... so not too bad.
@@maskddingo1779the original drives are aging and in far more limited supply than the actual systems they serve. Emulated storage isn't just convenient, it also saves you from repeatedly cycling an aging drive while diagnosing for example.
I think a restorer with emulated drives is just smart. If you can keep an original working that's great, but using emulation saves it from wear and tear.
@@maskddingo1779also power supplies in the past did not always have the greatest protection circuits. If they had them at all.
Using a modern well protected power supply, might be less original, but it protects your machine from being blown up because of power supply failure. And failure becomes ever more likely with age.
Even curious Marc recently had a video where he added a crowbar circuit to a vintage power supply that had broken and killed the computer attached because it did not have a crowbar circuit. Something HP added to later models. Original no, safe yes.
@@bzuidgeest Has UE said what drives were in / were avalible to this system at new (late 1980 to me isnt that long ago)
I love these types of troubleshooting episodes. They are both entertaining and educational! It matters very little to me personally if you manage to wrap up and fix it in a single episode, I'm here for the journey 😅
You know I don't really know too much about electronics, but I really like the logical breakdowns. I think if I keep watching I may learn something. Thanks!
You need to put a small load on the 5 and 12 volt outputs. Say a tenth of the normal load. Without a load the series regulators are running outside their lower current spec.
Also a variac is your friend. As are a few of those $9.95 DVM's from HArbor Freight. Slowly turn up the AC line voltage on the variac while monitoring the outputs, under a light load. That way you'll blow fewer transistors and resistors.
If it helps, I've been stumped by several Tek 485 scope power supplies. They are extremely complex, with a bunch of overvoltage and overcurrent shutdown circuits. If any power supply is sensed to be out of spec, everything shuts down. Tek even published a separate troubleshooting guide just for the power supply, it's so convoluted. The scope itself is very high tech, but I've ended up junking two scopes due to at least for me, unfixable power supplies.
He had high watt load resistors on the power supply at one point in the video. The LM7805 is a linear type regulator, used for the 5v rail and it doesn't require a minimum load as the internal draw of the IC (2-3ma from memory) is sufficient. If I recall the 485 has a switching style power supply not a linear system.
@@user-uz1yv2oc9v
The 5 volt supply is also a buck converter, just using a three-terminal linear regulator pressed into duty as the controller for switching.
I was a FORTRAN programmer on 11/23 back in 1985 and moved quickly over to Vax 11/750 and 8600 about a year later. Fond memories.
Enjoy your 11’s once you solve this power issue (which you will!) They are great computers to use and program on (I did for nearly a decade!) I also build up a home collection a few decades ago; however, I miss owning PDP11’s now. I had a 11/05, 11/34, 11/44, RL02’s, and a ton of peripherals. Then I realized I needed to reduce my hobbies and they all went to another collector where they live happily to this day. But I kept the PDP11 books so I can be nostalgic anytime I need too. Oh … and I had a PDP8/E that I restored (a much simpler power supply fix - just throw in new caps!) Funny how things changed and got more complex from the 8/E to the 11/03; however, they also got a lot lighter!!!
I'm not a fan of mysteries, but I sure love the diagnostic investigation and troubleshooting work!
Interesting stuff again. It really made me laugh when you said that something went over your head because all the things you do, I just usually don't understand, heheh. Cheers from Finland!
I absolutely love your channel. It is so fascinating watching you debug hardware. Thank you for what you do.
My first thought was, 'Give it a moderate load. The lack of load could be allowing voltage to ramp up and then trigger the over-voltage protection.'
Usually these powersupplies with a blown fuse, you should always check the transistors and mosfets. Good starting place
odd that once fuse replaced everything sort of worked as it should (I assume circuit and drive draws finally blew the old one)
that video looks like a Curiousmarc episode... fantastic 😎
The base-emitter junction of Q15 SHOULD pass voltage (although it wouldn't pass much current before you fried it)
Unfortunately, since ALL of the current must now pass thru that B/E junction, it's quite probably been transformed into a 3-way piece of wire...
I know there's no 'external' load on the PSU, but there IS an internal load being the 560uF output capacitor (C20) that has to be charged up! (More on that later).
The fact that you are getting over 13V on the output is worrisome to me... The 7812 SHOULD keep it very close to 12.0V, so MAYBE this too is 'fried'? However, this circuit allows you to 'adjust' the +12V rail a bit (using the potentiometer based voltage divider on the output. I'd confirm the EXACT voltage between pins 2 and 3 of the 7812 before I called it dead! (Pin 2 is NOT directly connected to the overall circuit ground!)
Next comes the LER (Light Emitting Resistor)...
You REMOVED Q18 driver and therefore ALL of the current on the 12V rail cannot come from the series pass transistor (Q17) since it no longer has any drive. This means ALL of the 12 rail current MUST now pass thru Q15!!! Let's ASSUME for a second that there is a direct short across between the emitter of Q15 and ground. The incoming +24V will turn ON Q15 by way of the base bias current (R75 + R47) and thus you now have 24V across a small (1/4W?) 10 ohm resistor (R74)... 24V/10ohm = 2.4 AMP! WAY more than it could ever hope to handle, so it becomes a FUSE instead and gives off the light show in the process. (R74 has 'left the chat' and is now an infinite ohm resistor)
So now the ONLY remaining current path is thru R75 + R47 + base-emitter of Q15 into the 7812. With R47 being 330 ohm, there's simply not enough current flow to create a new light show. My _GUESS_ is that the 7812 is also NFG but that's yet to be determined! If's possible that the adjustment is turned up so far that it causes the crowbar to kick in, but, without the 10 ohm resistor, there is enough of a voltage drop across the R74 + R47 base-bias that the crowbar doesn't get a chance to trigger.
1: Check C20 for leakage / short
2: Disable the crowbar and retest (remove D30, R62 and Q14)
Switch Mode power supplies can be a real bear to diagnose. Lots of ways to fail and interacting subsystems to throw you off. I see you have fixed it now (in a new episode). Congrats
Neat to see ink on the dip switches (5:45) from the techs that just used their pens.
"light emitting resistor" you had me at that one haha!
No.1 before powering any old power supply is just replace ALL electrolytic caps with good quality low ESR types. Then check all diodes & transistors with a meter (in circuit). Look for dry joints too, before cranking up slowly with a variac, preferably with a lightbulb in series in case something does short out.
Just a minor setback on your journey to successful restoration. I firmly believe in you!
try testing power supplies by powering up their outputs, using bench supplies. this is how i test PSUs that i repair, as well as new off-line designs - by powering the primary side SMPS controller from a LV supply but no AC input, you can check that the output stage can reach its setpoint, and also check that the feedback network is working correctly
0:52 If that's the same Dave McGuire I'm thinking of, I met him once at his house ~20 years ago (in Tampa, FL or thereabouts). He had a supposedly working full-rack PDP/11 right in his living room, next to an original radioactive-decay tube-based time standard, and had converted his two-car garage to a data center. There were 2 Crays in there which he was using to calculate Pi digits for some kind of contest. He bragged once about his ~$800 power bill (in ~2004 dollars). I would hope that all of those machines are now part of the exhibit in his museum, as they were in great condition back then, and he was pretty fastidious about such things.
It's often faster to just desolder the power transistors and check them all and also replace any electrolytics that seem a bit duff. You can check a few other components with a meter in circuit like low ohm resistors. Bipolars often fail via direct shorts between B-E or B-C.
To be honest, I wouldn’t apply mains to something this old without replacing electrolytics outright. Call me overcautious but there were two, ahem, “excursions” in this video and there is simply no future in going forward with 40-year-old electrolytics even if they appear to hold back DC.
@@hubbsllc Yes there is a good argument for that, especially considering how much it could damage if that 5V turned into a 12v supply. However I suspect the old Western electrolytes are longer lasting than the new ones. With the old ones you would only have a problem if they were in close proximity to the places where they could get hot, so pull out some of the best suspects and test them to get an idea what the rest are like. The thing is, with a fuse like that I'd see it more likely one of the power transistors has failed, and sometime that takes out a driver transistor. Sometimes they can half fail so you have a leakage across the junctions. It's better to spend 20m checking all this before applying power as well.
@@Andrew-rc3vh Yes - a blown fuse (especially if it had been checked from the get-go) is a good indication that something already went awry so you know there has already been an overcurrent event. Those electrolytics can appear to hold a voltage but in the case of filter caps they're expected to experience AC current - not exactly the conditions of that test. And on top of that they may appear to hold voltage *now* but what about after 15 minutes in an energized circuit? This does imply that if you're going to entertain getting a machine like this running and keep it running, there's a certain cost/effort to sink into it before you even give it mains which is why you need to be clear about what your goal with a piece is and how committed you are before you even put it on the bench. I restored a 1954 tape recorder and started by rebuilding the whole power supply onto a perf board; a dual-element power resistor had one leg that was open-circuit and there were two can caps that I removed outright. I didn't even plug the power cord in until that much was done.
Lovely repair you did for that ribbon cable plug!
Hmm, as you replaced that 180 µf cap, the bigger electrolytic, and eventually pulled that D44C9, I kept thinking, "Yeah, I'm sure I've got something in the basement that will work." (Exact replacement for the smaller cap, smaller form factor replacement for the big one in case you want to hide it in the original case, and "D44C9 - (checks data sheet) 60-70 volt collector voltage, don't need anything close to that here, pretty wimpy beta, 30 watt power dissipation, something like an MJE15028 ought to work fine - its fT is a little slower but I bet the switcher isn't clocking that fast, and the pinouts (E, C, B) are even the same." But my basement is in New York and your shop is in Texas, and I'm sure you have parts on order already.
In any case, it looks as if you have a buck converter on your hands. They don't regulate well with no load, and that's probably tripping the crowbar. Try loading the 5V rail with a 5 ohm 10 watt resistor and maybe a 20 ohm on each 12 volt rail, and see how it does. (I'm guessing that you haven't poked into the guts of a switcher before, because you're approaching this thing as if it's a linear power supply.)
By the way, an SCR is a kind of thyristor.
Your calculation of frequency from the scope trace is not correct. The period is ~250ms, or 4 Hz. That is a compact power supply - good luck. I enjoy you working to resurrect these vintage pieces.
Yes, definitely come see us at the Large Scale Systems Museum! One day I'll get Dave to be willing to talk on camera long enough to catch all his stories :)
Would it not have been simpler to just replace the connector on the ribbon cable? Looks like a fairly standard DIP/DIL IDC to ribbon cable connector to me that are still widely available.
Absolutely!
Except, I didn't have a spare, so instead of waiting on parts to arrive, I figured I'd take a stab at repairing it so I could continue on with the build. I'll order up some replacements and fix it properly once I know for sure the PSU is going to come back to life.
It was a very clever solution to an immediate problem. And if it works, it works.
Thats obvious of course it would have been simpler I bet he wishes you were there to tell him that....oh wait a minute he didn't have one.
Don't mind me I'm grumpy :)@tradde11
Great video! I've got a power supply issue on an Apple Lisa (also a big 'pop' after replacing the fuse and found a trace entirely vaporised). Watching you do this gives me the understanding and confidence to forge ahead. Cheers!
I would love if the PDP11 from The Weather Channel ever surfaces, they used one to run all the STAR units in the field over satellite, unearth that long lost 4000 software.
A fantastic look at a power supply! I remember you saying you didn't know much about power supplies, so this is a great start!!
Might be a good idea to test all components for shorts.
I suspect the PS had a catastrophoc event that cascaded one problem into several. Switching PS are a feedback control system which make isolating the problem extra difficult. You are getting close, though.
Will the Power supply Grimlins be defeated find out on the next exciting episode on Usagi Ball Z.
I cant wait to see what happens next. this is about the only way ill get to mess with these old computers. Thank you for all the amazing content, I appreciate all the hard work that goes unseen. I know this kind of stuff isn't always easy.
Hmmm Heathkit once sold PDP-11/23 half size cards and I once say a TELCOM company use a VAX as a long distance telephone switch.
It seems a lot of people are unaware of the fact that the common equations for switchmode converters don't apply at light load where the current in the inductors goes into the discontinuous region, dropping fully to zero each switching cycle.
With buck converters, as used here, the usual expression is
_Vout = Vin x duty cycle_ but that *only* applies with continuous inductor current. In the discontinuous region the duty cycle must be reduced and at very light loads it must go very close to zero. Lots of older control circuits couldn't manage zero duty cycle very well. Sometimes they were OK except you'd get quite a bit of ripple because the "squeg" - run for a few cycles then turn off completely.
The circuits here use three terminal regulators to actually do the control of the switching. I get very frustrated trying to find the brief view of the relevant schematic bits in videos like this, so I haven't looked closely at the design.
That slow rise of the 5 volt rail very strongly suggests to me that the circuit can't bring the no-load duty cycle down far enough to maintain regulation
I would test these converters with at least 10% of nominal full-power loading. That won't likely be enough to assure continuous inductor current, but may be OK anyway (it is common to design to place the discontinuous/continuous boundary at something in the range of 20 to 40% of full rated load).
In general, the presence of an LM7805 and LM7812 implies they are being used as voltage references for the pass transistors that are handling larger currents. Since the 5V and 12V outputs are both showing high, I'd suggest pulling both regulators and testing them in a reference circuit on a breadboard. They're common jelly bean parts and are likely easier to source replacements for than the pass transistors, etc.
The other culprit could be the 7805 or 7812 regulator. It looks like the circuit is using them as a kind of current switch. When the Vout is going above the 7805/12 regulator voltage the linear regulator will limit the current and turn off. This will stop all current through the control transistor and R27 (5v circuit) which will have the effect of switching off the driver transister and pass switch. The Vout should drop as the inductor loses energy and the cycle will start again. If the regulator is no longer regulating then it might behave the way you are seeing. Worth doing the same checks on 12V and testing the crowbar circuits for both outputs, I think you could just hook your bench PSU up to it with the current limit set to something sensible and bring up the voltage on your bench PSU to see where they are tripping. Saves you having to switch it on and risking other components.
I agree with your take on how the circuit functions.
One hook with using a bench supply, which in general I agree is a much safer approach in terms of avoiding doing more damage, is that sometimes these sorts of supplies have some additional low-current supplies that are necessary. In this circuit a -15V supply is used in the setpoint networks for the 3-terminal regulators. All it does is provide some bias current for diode used as a crude negative voltage reference. In things like bench power supplies a negative supply may have a more critical role. It certainly isn't an insoluble problem, but something relative novices may not have seen or may not understand..
That power supply chassis looks very familiar to me (some minor differences), I had one when I was a wee lad. but I now know what it was from
"Never give up, never surrender!"
Your « failures » are as interesting as your successes. For a viewer, it’s the process which is interesting more than the result.
And it is probably just a temporary failure so, no regrets.
Any time you are confronted with circuitry with adjustable parts that has been mucked about with by someone else, you *must* assume the mucking about has included fiddling the twiddlables. In fact if an amateur was as it, is is very likely said amateur started with such twiddling. This can make logical troubleshooting substantially more difficult. It can be difficult to determine if something is wrong because something has failed or because some knob turned a knob.
Ask for 'the Repair Man' at the OLG, if he still works there.
I'm by no means an electrical engineer but that is a bizarre way of implementing overload protection. Maybe it's commonplace and somebody will correct me, but usually most circuits just kill power downright. Definitely a complex looking protection system on this thing. At least it was still doing its job all these decades later, so the designers definitely got something right!
power supply explanations are great, and thanks for showing the cable repair I am always curious how you guys do those.
FTIR is a common spectroscopy technique to identify or otherwise get structural info on molecules. It'd be super cool if you could snag an old spectrometer. Old science instruments are really cool.
Watch out for carbon film resistors. They like to drift a lot over time.
... Especially once they've been used as light emitting resistors... LOL
Interesting circuit, completely linear voltage regular that uses a 7812 as voltage reference and error amplifier at the same time. The reason R74 burn up when Q18 is removed is because the B-E junction of Q18 is keeping R74 at 0.7V. Without Q18, the current going through Q18 is essentially all the current going though the 7812. Q15 is just a simple common base amplifier configuration, it is there as a switch for the shutdown cutoff, it is unlikely to be broken. Leakage isn't what you are looking for either, as B-E junction of Q18 is 0.7V, so the R75, R47 is passing current though to the B-E junction to power the 7812. As for why the voltage output is creeping up, my guess is the 7812 is drawing more quiescent current than the circuit is initially designed for, so the extra current is causing the Q15 to pass more current to the output, even if the 7812 is in full cutoff state. Add a resistor from 24V source to ping 1 of 7812 might help, experiment is necessary to find the correct resistance. General description, Q15 is the cutoff switch. Q18 is a current gain stage and also to change the direct of the current flow and the Q17 is the main power transistor.
Holy Crap! Thats a Freaking Truck Load!
Nice Pull! Nice sharing it Out too!!!
I don't think power supply repair is your speciality 😁
I never would just switch on such an old PS like you did. What I also don't understand is that you remove transistors without testing it. Most transistors and diodes can be tested in circuit with just an ordinary DMM.
Absolutely. Where is his supply protection? Where is the ability to ramp up the supply voltage gradually (variac), rather than just subjecting any gear to a power input surge after years of sitting idle? It was the same with the G15 at first switch on .... He hadn't even separately tested the fans or the drum motor. He just threw the switch. There he was lucky, but it shouldn't have to rely on luck.
Signetics 82S100 PLA chip on that CPU board. Same as the infamous PLA in the Commodore 64.
The original 82S100 was pretty reliable, despite running hot as hell. It was the MOS clones that frequently failed.
I think it would be awesome if someone got one of these up an running and showed the Fourier analysis described. Seems like something more than just doing a listing of directories could be done on these machines. A lot of time and effort put into just getting to a command prompt and listing directories and files.
If it were me, I'd just pick up a 5 & 12V switching power supply from the getting spot and be done with it, but I can understand wanting to keep this as original as possible. Certainly makes the job a lot more difficult, but that just means you'll have learned more when you're done.
Maybe I am wrong, but remember that a BJT (Q13) is essentially two diodes, in this case, from base-emmiter and base-collector. When you put your input voltage from the rectifier, is normal to pass through R75+R47=890OHM, the B-E junction to be regulated by the 7812 and so you should have those 12V at the 12V output power rail, albeit at a very low current. Without R74 (10Ohm) the behaviour of Q13 is normal, it is not leaking anything.
Actually, you shoud check the voltage between the ground pin (the middle pin) of 7812 and ground and add to it 12V; this you should have at the output. If is not ok, your 7812 may be shot.
7812 is not regulating to 12V in refference to the ground but in refference to the voltage set by the resistor network R77, R86, R87 and this voltage can be regulated by the R87 potentiometer. Try to check/clean those potentiometers (R37/R87) and any others you can find, connect the transistors after you checked them and readjust the output voltage by tunning them. You may (in fact should) have to disable the crowbar circuits to ignore the overvoltage protection to do this easyer. Actually, that 10 Ohm R74 may have burned because the crowbar circuit, if nothing else is drawing current (check if 7812, D32, C24(?)=0.22 are ok).
I know the feeling!
I was given a Zenith 486 laptop yesterday. Power supply was dead. Found a shorted diode and replaced it with one from a scrap board. Laptop POSTed once and then went in to a power cycling loop. Pulled apart the machine and found loads of leaky surface mount electrolyitics... And I have none of those on hand. 😔
Very helpful background info regarding these supplies, I have a poorly LSI11/24 with a dead H777 PSU which I'd love to get going at some time. Many thanks!
There's a whole bunch of old caps in there that remind me strongly of those which go pop.
I run my franken-11 off a salvaged ATX supply and generate the power-on signals and LTC with one of those $3 ESP-8266 modules. If that thing has you flummoxed, you don't *have* to use an expensive and complicated DEC supply.
My franken-11 uses that same backplane too. You can wire wrap it to 22 bits and run one of those 11-73 cards in it.
also replacing the 7812 and 7805. The old ones are known to become faulty.
But these are easy to get.
Man I always wanted a full PDP-11 stack for myself but now this is making me kind of reconsider.
10:44 get a piece of new ribbon cable, crimp a new connector. These terminations are still being made afaik.
Light-emitting resistor, lol
Great video! We learn more from our struggles than we do from our success.
You should use an isolation transformer when playing with high voltage supply stuff.
I Love if it's not working :D I can see workflow of diagnostic of other people ;)
This video was featured on Hackaday, and some knowledgeable folks have made some comments there as well.
I am looking into my year end project for school and I am going to be looking for a pdp 11 over the next year.
I cannot even describe how many devices I seen been killed by old capacitors by now, They leak electrolyte and corrode traces, they short out, they leak electrically and causes massive fluctuations in voltage ripple. it is very common when a cap short out it takes many other components with it such as diodes, resistors and transistors that suffer over current. you should never ever trust old caps to be good. it is a good thing to make resistance measures to check for shorts and if caps are leaking before powering stuff and it can save a lot of trouble down the line. I have the feeling those shorted caps can be one reason why the supply have died this much, when things literally light up from shorted stuff you know there been a good BBQ party in the PCB house
Ah, happy/not happy days of 3 engineers in our workshop also working on an 115V 11/03 PSU. We were defeated too, so I hope there's success in the next video. Too late for our machine, this was 15 years ago. Fingers crossed!
Got an idea for you .. Why not take a 100 watt incandescent light bulb and put it in series with your power cord to limit current
going to a unit when you plug it in for the first time. Could use maybe 2 of them .... Keeps fuse from blowing , and if the lamp lights up bright you know you got a problem.. If it lights bright and dims.... then it may be OK ..... I started doing that back in the 70's and it saved me a lot when replacing high power audio transistors.. Just an idea..
Power supply design wasn't one of DEC's strongest points. But they loved vertical integration.