00:39 You've defined a new class of vintage computers. There was the mainframe, the mini, and the luggable computer. Now there is the huggable computer. Bonus points if it's all warm when you hug it.
Make sure to remove the batteries in the diode tester before it goes back to the shelve :) Also what I do when I reversed engineer something, I put the print-out into the device itself, so that the next person working on the device in 50 years will get the necessary headstart by having a fantastic documentation to start.
In high school we had a G15 sitting in the back of the Algebra classroom that they had removed from service in the mid 70's. I spent many hours as a freshman starring at that front panel, confused why you had to adjust the AC voltage on it, or what in the world the giant switch at the bottom was for. This was the late 70's so I was used to seeing IBM's & PDP's, not this weird 50's relic. I took all the electronics courses the school had to offer and I remember seeing that diode tester sitting on a shelf, I never put it together that it was for the G15. I think junior year I found a dusty South West Technical Products terminal kit laying around and since I was totally fascinated by computers I asked the teacher if I could build it. He said sure! He was kinda excited to see that he was creating a passion in at least one of his students and happy that I was taking an interest. He purchased a cordless Isotip soldering iron for me to use building this kit. At the time, all we had were unregulated 110v soldering irons in the lab which weren't suitable for that sort of work. After building it, at some point I was joking with him about dragging the G15 over so we had something to connect it to. We didn't really have the space in the lab for it and since my buddy and I were the only 2 students interested in the project, he couldn't justify it. Much to my surprise, a few weeks later he told me that the school wanted to get rid of the G15 and that I could purchase it from the HS for $20! Sadly, my parent's wouldn't go with the program and let me buy it. I did spend a bunch of time reading the manuals for it being particularly interested in the drum memory unit. At the time I wondered how many hours of audio you could store on a drum like that. You'd see surplus drum memory units pop up at surplus places and electronics magazines back in the day and would have been a fun experiment. My electronics teacher then gave me a KIM-1 that had been donated that nobody had been interested in to play with. I thought it was the coolest thing ever! This helped inspire the teacher to start a Digital Electronics course the next year. I built a Sinclair ZX-80 kit out of the back of popular electronics I got for X-mas and have been banging keys ever since. After watching this series, I'm so glad my parents were smart enough to say no. It would have been wayyyyy too far over my head and skill set at that time to have resurrected that machine, even with the 2 foot pile of manuals and tapes that were with it. The original Bendix terminal was there too as well as a 2nd, smaller Bendix cabinet that I have no idea what it was for.
In reading your post, I kept asking myself if I had written it - at least for the first few sentences. At my high school, we had a G-15, also in back of the algebra / geometry class. I would have played with it around 1968 - 1969. Is there any chance that your high school was Skyline, in Oakland, CA ?
@Usagi Electric The 12AV7 tube is NOT a substitute for the 5965 (E180CC). The 5965 and other tubes specially made for flip-flop circutry, has to withstand being biased into sharp cut-off for extended periods, hence the more rugged construction, a lot more heat has to be dissipated in cut-off. Unfortunately the 5965 and the E180CC are wrongly considered an "audiophile" substitute for the 12AU7 - ECC82 types, so prices are stupidly high for tube wich is prone to hum, noise and microophonics. For anyone who doubts my claims, please read the Philips datasheet for the E180CC. Much ❤
Just a picky point: how does cutoff increase dissipation compared to conduction? I thought I'd read it was a problem with cathodes becoming coated with foreign matter from other tube components (like silicon from heaters) during cutoff for computer tubes, hurting their performance when they must conduct. The "Hot cathode" article in Wikipedia seems to corroborate this, calling the phenomenon "cathode poisoning." Perhaps for short term testing the 12AV7 would work, but the computer tube will be needed for best reliability.
Just looking at the 12A?7 family, a 12AV7 isn't an exact electrical replacement for a 5965, and isn't close on the ruggedness chart, and isn't designed _specifically_ for extended cutoff operation. But given that 5965's are now essentially unobtanium, and that this machine isn't going to be sitting powered on for hundreds of hours, a 12AV7 is probably a perfectly fine substitute for occasional operation. And it is available, which is better than no tube at all. Given the specific logic conditions in use here, it would be quite feasible to engineer a solid-state replacement for the 5965. I'm sure it has been done in the past, probably more than once, but the parts for those versions are likely obsolete by now too. Since we aren't interested in audio distortion, just a good curve matching a nominal 5965, it wouldn't be too hard to make something with some power MOSFETs or the like and a few other parts to do the job. (Don't forget the heater load resistor!) However, that would be kinda defeating the purpose of getting a vacuum tube machine working, if it didn't use vacuum tubes. 12AV7's are probably the much better available choice.
Audiophile nuts have created supply shortages snd crazy prices for a number of vintage items. Not long ago I ran across someone trying to sell Western Electric manufactured wire which had been pulled out of a central office main distribution frame. He was asking a few US Dollars per foot per strand, which is a crazy price for old used wire. I have quite a few feet of Western Electric 25 pair gray plastic sheathed cable and I can attest to the fact that is quite average plastic coated wire. The wire itself appears to be tin plated copper or just copper depending on the cable. I’m pretty sure it is tin plating because it does not tarnish as silver would.
It is amazing that you were able to get an actual Bendix test instrument for this G15. It can't be understated how amazing this is to have this tool. You're getting closer every time you work on this. Thanks for sharing your journey with us!
@@wtmayhew Knowing HOW it works, makes it possible to build more modern semiautomated replacements in future. Personally I'd run it off a USB power supply rather than rely on potentially leaky dry/alkaline cells
@@miscbits6399 Thank you for replying. There really isn’t much to the tester conceptually. It is a switch matrix to select a particular diode from the card under test. There’s a cross connected DPDT switch to forward or reverse bias the diode under test. The micro ammeter is wired in series with the battery to monitor the current. The secret sauce is in that rare card edge connector and the wiring of the wafer switch to make it easy to select a diode. The diode logic in the G-15 is probably composed of diodes similar to the venerable 1N60 germanium junction diode. The 1N60 has a peak reverse voltage rating of 50 Volts. The test box originally had a 35 Volt battery which is probably enough to spot a leaky diode. Something like a modem multimeter or a vintage VTVM usually on biases test components with about 3 Volts. The test box lets the operator flip the rotary switch through all the diode positions on a card in a few seconds - one rotation for forward bias, flip direction switch, then a second rotation. That sure beats probing individually with a meter - which doesn’t put out enough voltage to catch a reverse bias breakdown.
Two steps forward, one step back. It's the cycle of life. Considering these diodes are the same age as my father, the fact that such a small number only have failed is astounding. Also Lloyd, if you are reading this, you are a GOD among men. Never forget it. You and Dave are making magic here, but he'd be going at 1/10th the pace or even slower if not for your help. Thank you.
I agree. My experience of germanium in the 1980s wasn't pleasant and I spent a lot of time cobbling things up in cases when silicon wasn't a simple drop in answer (Why? Working in a SW comms station commissioned in 1951 and considered well past end of life even in 1980 let alone the end of the 80s, No money was put into replacing anything as it was already expected to be turned off (and was, in 1992), so things like 45.5 and 50baud tty circuits had to be kept running. Adding to the pain was the fact that large portions of the building had been declared off limits due to asbestos comtamination and a realisation that berryllium dust had been accumulating in nooks and crannies everywhere (a lot of old high power tubes use it for the heat conductivity qualities, and a bunch used thoriated cathodes with 2-3kW heater elements to enhance electron emissions. Let's not even go into the hard xray emissions from the mercury arc "rectum-fryers" or needing copius quantities of distilled water because the cooling loops were directly cooling anodes running at 50-75kVDC)
I was taught in Electronics school that you always adjust an analog meter to ZERO with power OFF. You adjusted with power on, and not to ZERO. I can hear the voice of my teacher in my head, screaming, for doing that. LOL 😂
I didn't show it directly on camera, but I wanted to see if when adjusted such that Lloyd's specs are on point, does the needle drop down to zero. This was a way of confirming just how far off base we might be. It did indeed drop back down to zero, and once powered off, I confirmed, the needle was still at zero. In the video, it seems like I blazed through it, but in reality, I spent about 15 minutes confirming and double checking the needle position and cross referencing against any and all notes and documentation we had.
@@UsagiElectric I'm just giving you a hard time! 😉 I have NO DOUBT that you are more than competent and capable in the field of Electronics Technology. The Bendix G15 is literally my dream computer. I was given a Bendix volt meter off one as a kid and always loved them ever since. I still have the meter even today. I absolutely LOVE watching the G15 slowly roar back to life! Keep up the great work!
The thing I love about this channel is the in-depth stuff, but not too complex, the really old systems (not just 8-bit) and the community. Reverse engineering stuff is amazing, people helping out with coding etc. Love it.
It's like Quantum Leap - you don't get to disappear until you're completed the task you were set out to do. I assume that Al is talking to Ziggy just off camera. Keep up the amazing work, this series has been fascinating, thanks for sharing it.
He says "You're still here?!" but like we're always here, staring at the blank screen waiting for the next one to start. I thought that was understood.
The more of these episodes I watch, the more I'm impressed by the G15's industrial design. The swing-out access doors, the keyed and coded tube carriers, the metal finishes and colors -- they really wanted to make sure people felt they were getting their $50,000* worth! Even in this little tester it was as if somebody said, "Okay, we've got a way to test the diodes, but we can't just cobble this thing up on a breadboard -- it has to match the G15's aesthetics!" Thanks for taking the time to show all the nuances. *For a way to talk about prices without getting into currency exchange, you could say something like: "In 1956 this computer cost $50,000 -- at today's prices, that's the equivalent of how much profit Amazon makes every 56 seconds..."
It amazes me that I have meters, component testers including a curve tracer but the Bendix engineers knowing others would be having this exact problem solved it with basically a battery, a couple resistors, some rotor switches and a meter. Small and rugged for it's time compared to carrying a curve tracer and a scope considering scopes back then were very heavy vacuum tube jobs that required their own cart and were very fragile. I am having to repair my RF analyzer now because of a bad screen when being transported even in a special anvil case and it's within the last 20 years modern.
They yield on early semiconductors and diodes wasn't real great so the parts were well checked before they left the factory. Infant mortality was also more of a problem so parts failing a few weeks or month into the field was also an issue. I would expect a well equipped service man to have a tube tester, diode tester, multimeter and scope when they visited a customer site because that many parts will produce a failure sooner than latter.
@@denawiltsie4412 The vacuum tube machines I was familiar with (admittedly larger than this machine) always had an on-site customer service engineer that would some on shift around 6 AM, power up the machine, run diagnostic, and fix anything he found broken or marginal, hopefully turning it over to the customer by 8 AM. Then the machines would usually run till around 6 or 7 PM, when they would either be switched off or put into a low-power standby state until the next morning. It was rare to not need to replace a few tubes every day or so.
@@lwilton The oldest systems I used where transistor so I really haven't had hands on experience with vacuum tube computers. I was lucky enough to see one of the Air Forces SAGE (Semi-Automatic Ground Environment) installations where they had two very large vacuum tube computers with core and drum memories. One system backed up the other but I had the impression that the system was reasonably reliable. At least enough that they allowed a regular stream of visitors to view the system in operation. They trusted the system enough that they would run demos on the idle system.
People like this are a national treasure... I dont think this country realizes how badly we truly need people that can, and will, do this type of stuff... 🎉
I remember a TV commercial from decades ago, for, I think, a motor oil, the machanic states "You can pay me now, or you can pay me later." BY ALL MEANS check all the diode cards now. Not doing so is going to cause you grief tracking down which card is causing the fault. You will end up spending more time in stressful panic debugginf than you will take leisurly and calmly checking them now.
Loving the lacing cord on the back of the panel. Reminds me of my first job to pay my way through university, working for a company making instruments for the North Sea oil rigs.
No idea how you get one of these videos made every week. Major highlight of my weekend! I mean, you do SO MUCH, not only al the work on the machines, all the trial and error, but you also record AND EDIT 20-30 minute videos EVERY WEEK?! Madness!
I am older than this system and got to play with one in high school in the early part of the 1960s. Pure magic watching this come to life. Tic Tac Toe next!
That screw on the front is the Mechanical Zero. When the device is OFF and the meter is reading nothing, you use that to adjust the meter to read exactly Zero. Any other adjustments should be electrical calibrations. What you did may work, but it is not the correct use of the mechanical zero and it may affect the accuracy across the dial. Yeah I know, I am being pedantic. lol. I do love your videos, and this tester will definitely save you a heap of time.
Just a thought but 4x 9v batteries would be 36v if you wanted to eliminate any active electronics and phantom power draw while not in use. Also if 36v is too high just add a few diodes in series with the batteries as a voltage drop.
In this case, adding diodes in series with batteries would cause an issue if a test was measuring something less then the forward current of the series diodes.
I was the MIL-STAR. MIL STD 1750. FEU (functional equivalent unit) repair man. Many a trip to Trouble W in Redondo Beach and LMSC in San Jose. Ha Ha. A simple Rassberrie Pi project, A DAC & ADC pulse it and read back and check reverse bias test. Looks like a Rube Gold Berg set-up.
The G-15 series is my favourite of the ones you're currently running. I'd never heard of it till I found your channel, but I'm completely invested in this thing. I'm excited to see this one work... soon!
Given how rare these machines are now, I am tempted to say that the value (at least to those who collect machines like this) of a working G15 today is even higher. Of course, commercially not so much, as you can get orders of magnitude more computing "oomph" for orders of magnitude less "ka-ching". Your "see you in the next episode" didn't work for a simple reason. Everyone knows that your videos end with a good portion of your "furry family members". So unless you "roll the fluff", we know the video isn't over. 😄
A working G15, I can't say for sure, but a few have changed hands over the past year. A G15 with MTA2 tape drive, two typewriters, and some other small stuff sold at RR Auctions about a year ago for $60,000. More recently, a G15 with a typewriter, and an MTA-2 tape drive from the LCM sold on Christies Auctions for around $20,000. So, it seems the value of these machines is still pretty impressive, but still well under $100,000 today. I was wondering how many people I would catch out with the fake ending, but I also know y'all are all really here for the fuzzy family at the end!
I enjoyed your video. The tube failure you experienced, is the perfect example of why, at Bletchley Park during WWII, when the British turned on the Colossus machines, from the time Tommy Flowers, and his team, turned on the first Colossus machine, the machines stayed on continuously until they were turned off and destroyed at the end of the war. Leaving the machines on, was the key to improving the reliability of the vacuum tubes. PS: Each mark II machine, had 2500 tubes
I have a 70's Elka Concorde 802 organ with many, many diodes also doing a lot of signal control and whatnot. It al runs with similar voltages too: 0V to -27V. So this video is excellent info in debugging some (future) issues!
Think about the 1956 factory technician who had to test all the boards needed for a full scale production run. They were looking for the forward / reverse reding for hundreds of boards a week. Fast forward 20 years and you could test each board in 30 seconds, all diodes and all resistors.
I would bet quite a few dollars the factory had automated board testers. I know I had them at Burroughs in the late 1960s. You could build one as a special peripheral to be run by a G15. You could make a simpler version with some stepping switches and limit-indicating voltmeters and ammeters, which were fairly common in those days.
This is the beauty of discrete logic: No special chips that are now obsolete and impossible to find, just standard parts that you can get everywhere. That test box is nice! But for testing large numbers of things (parts, modules, whatever) a simple "pass/fail" indicatzor (red light or green light) would have made testing much faster and simpler. Maybe a comparator that measures the "meter current" with a shunt resistor and compares to a fixed voltage depending on the position of the forward/revers switch? Probably not worth designing a custom test box, but I probably would have done it. (Automating something is way more fun then actually doing the work ...)
SUGGESTION: With a fine sharpie, number all the remaining packs with some type of notation such as rack, vertical position, horizontal position- e.g. 3-5-16 May be useful once you get the machine running as you will definitely have failing packs/tubes. Put the info in a spreadsheet for the next guy that is going to service the beast. 😊
There are only about 460 vacuum tubes, so it's a much more feasible task to tackle that problem one by one as issues crop up during machine testing. I'm working under the assumption that the majority of them are okay, and we're just going to troubleshoot the machine slowly and surely. I do have a tube tester, but it's not a very good one, so it won't highlight issues with emissions and balance on the flip flops or things like that. So, if we find a flop isn't behaving and the tube filaments look fine, we'll get into some more in depth testing at that point!
@@UsagiElectric You might look around and see if you can find a Hickok Cardmatic in working order, along with the test cards. They are pretty much idea for this sort of batch testing. The Model 123A had a set of test cards for the 5965 specifically.
Magnificent work, especially on reverse-engineering and documenting the tester. Also, i noticed you were using Australian Central Time in your scheduling tool, which I can totally understand as a guess at the best average timezone to use for the country, but in actuality, over 80% of the population lives on the east coast, so it would probably make more sense to use AEST unless a disproportionate number of your Australian viewers are tuning in from WA, SA, or NT. And, yes, technically Queensland is also in a different timezone because they don't do Daylight Savings Time, but we all just make fun of them for that, so I wouldn't factor it into your considerations...
In Usagi's case, I think that might be because Japan uses the same zone as central Australia. So he defaults to that, when choosing time zones for international comparison purposes.
not marking which slot they came out of is craziness to me. that's my first instinct. especially on a legacy system, I'd keep the configuration as close to original as possible just in case the assembly order is meaningful
As volvo09 said, the slots are all keyed and all the cards are labeled and color coded by function. Every card should be completely interchangeable with other cards of the same type. I actually specifically chose not retain the same configuration as it's a good way of highlighting failures we may not have caught yet. It can be a little frustrating and time consuming to track down though, haha.
@@volvo09 oh I have no doubt that it was designed to be interchangeable. but with something that old, I would think maybe some board warpage or some corrosion patterns or something might mean that returning something to its original slot could help make better contact or something that wasn't immediately apparent
Surprisingly, I think I've seen more testers than actual G15s! I have one, Bob still has two or three, Lloyd has two, the G15 that sold at RR Auctions a while back had one with it, and so on.
I think the carbon composition resistors should be taken into account; I have seen one vintage radio / TV restoration site where the person said he checks them and replaces anything that has drifted out of tolerance (gold band = 5%, silver band = 10%, no fourth band = 20%). That might have to be the next step after the diode test if the computer still acts like it hasn't had its morning coffee.
I'd suggest probably testing all the tubes first, they are more likely to have hard failures that will have an impact. The resistors without digging deeper you won't know if it being out of spec matters. Dud tube and you're going to have a bad day (as they found out(
Yeah nice jig, I once worked in a repurposed lab and in the back as a similar jig that was used to test disk contoller cards for I think DEC mini comuters, no panel metter just a two banks 32 leds and sbunch os switches to set up the card trst
From the very start of this project when you first gave us an overview of the machine and all those diode cards, I was wondering how in heck anyone was going to be able to check all that. We're getting there. That is, you're getting there. We're looking over you shoulder and trying to keep out of the light.
I had a similar problem once back in the late 1970s or so. I built a test fixture I could plug into my AIM-65 computer, and wrote a Basic program to run it. It could to an almost complete functional test on about 4 different types of cards, and let me test hundreds of them a lot faster than by hand or a manual tester would have let me do.
Wow, nice work and that test tool sure made that quick(ish) work. I bet a good percentage of those carbon composition resistors are bad too. They can go both low and high resistance. Looks like it's full of the little critters. And then, what kinds of capacitors does it use? Electrolytics are suspect but even more so any paper based non-polarized types.
The point contact germanium diode failure is caused by carrier diffusion drift. The dopant elements are attracted to the point of contact causing the diode to have a highly doped region in contact with a normal doped region and a depleted region between. These bad diodes won't switch right in logic ciruits due to shutoff oscillation but they have a secret. They can replace tunnel diodes in most circuits. ❤
About the heat, have there been any mention of how much power usage is? Cant say I’ve watched every video so far, but can’t remember it being mentioned in the ones I’ve watched yet.
I grew up a few miles from the Large Scale Systems Museum and I still live in the area. I considered stopping by for a visit and seeing if my dad would be interested as well since he worked with computer mainframes starting in the 70s. I gotta say though $100 per person to attend is a little much. I genuinely hope they raise a lot of money but feel like they could get more by having a lower cost of entry.
I may be late to the game - but with the Testofon (aka Contitest), you can quickly test any electronic component, especially if you have more than one of them. It works by applying a voltage and transforming the current that flows into a tone for your ears to hear and for your brain to decipher (if I test capacitors, I get a whole chart with the time/frequency axis and a scatter plot on what's good, what's tolerable and what's bad - and it'll fill it in in realtime). If you got all the diodes next to each other, testing 8 of them both directions will take around 5 seconds. It can also test transistors. Not great, but I found wonky ones that still worked in circuit most of the time, but either fail when cold or fail when warm. Like the multimeter, it won't find any that fail the forward voltage test above 4V, but it'll find shorted or reverse-leaky diodes. You might think of it as a glorified continuity beeper, but to me, the Testofon was a game changer. In a pinch, it could even replace an oscilloscope or logic probe.
I remember reading that tube computers (or anything with a sufficiently large number of tubes) couldn't theoretically work because the MTBF is such that there will always be at least one bad tube; the reason they do is because MTBF isn't the whole story, due to production variations some will fail early and some not, if you start with a whole new batch and replace them as they fail eventually you end up with all the better ones. The machine then works reliably until the tubes get into the end of life section of the bathtub curve at which point it is necessary to start again with a new batch. The other thing was never to switch the machine off as it is switching on which puts the most strain on the tubes, obviously not practical for hobby use. All this used to be observable with incandescent light bulbs in any building with enough of them.
The G15's MTBF was actually quite a bit longer than other tube computers of the era, often going for weeks of 24/7 operation without a failure. There's a few things playing into this. Only having 460 tubes brings the numbers game to a much more reasonable level. Also, the G15 is primarily using indirectly heated Noval tubes, which had much, much better life and reliability than the older octal or directly heated tubes used in machines like ENIAC. Also, filament warm up and start up is the hardest time on a tube, but the G15 has a clever soft start that sequences in, slowly increasing the voltage to the filaments over the course of 60 seconds. That means that all the filaments last much, much longer.
They used a couple of design tricks to extend the tube life. They ran the filament cooler either by a lower voltage or by designing the tube to handle a higher voltage than what was applied to it. Tubes normally ran much higher plate voltages between 45 and 90 volts. It surprised me a bit when I found they were running around 24 volts. Mean time between failure is a worst case calculation and often you will find that parts far exceed the numbers it produces after you get a few early failures out of the way. Running anything below specs will extend its life. I have a chandelier that uses 12 - 40 watt bulbs and was purchased in 1986. I have only replaced on bulb and that was because it broke while cleaning it. The trick is on high it produces so much light and heat that it's always on a dimmer running at perhaps half its full output.
MTBF is a mil-spec calculation based on component type, component count and stress applied to each component. You could have a poor design that never overstresses any parts but fails reliably far sooner than the calculated MTBF. And you could have good designs that run for longer than the MTBF.
@@UsagiElectric I was actually going to ask if you would consider using a soft tube starter. I'm glad to hear they included one, Bendix really thought of EVERYTHING!
@@denawiltsie4412 Incandescent lamp life is inversely proportional to the 13th power of the applied voltage. Running lamps at 50% voltage will extend the life by about 8000 times normal. Since the dimmer is probably a phase control unit, and you probably turn it on and off now and then, you probably won't get this. But if you have 1000 hour bulbs, it is still going to last for a whole lot of years! BTW, commercial and industrial lamps in the US were designed to run at 130V. Since most power is between 115 and 125, this made them last considerably longer than consumer lamps, but they put out about 10 to 15% less light.
Lloyd sure is a treasure to the community and to history. That Bendix tester is sure nifty. I sense a future video will be testing all the vacuum valves. 🤪 Cheers, Adriel
I see a nice gem there (The multimeter) If you have budget, try to get the UT210E-PRO version. That one has Frequency as an extra, and is more sensetive in the MiliAmpere range through the clamp. Nice habit of them is that they can be reprogrammed to start in DC mode on all ranges, that the backlight can be set to a longer value, or even to continious untill shut off again, and that the Auto shutoff of the meter can be omitted. All that by reflashing the Eeprom with an ESP32 flasher.
Just looking at the construction, this appears to be DTL logic instead of earlier RTL logic (diodes instead resistors) and was much more reliable. I worked for IBM in 1966 and their SLT flip-chips were much much smaller (approx 1/2"x1/2"), and one-chip could effectively replace one of those large PWB boards. The SLT chips had diff-amps available and thus could have myriads of other functions added directly on the hardware, such as flip-flops, monostable delay gates, direct NANDs, and multiple output Interface Drivers. The later IBM70+ systems pretty much killed any of these older PWB based single-card systems. That newer technology was actually driven by the development of solder-flow processes, whereby a huge large board could be wave-soldered in one pass. I lived through this whole cycle, and one of my sons now works on 4-to-7-nanometer scale device technology, and there maybe be yet a chance to get down to about 2-nanometers, before Moore's Law is broken. Now requires below ambient cooling, but the Spintronics people are working in that zone (can be done with CO2 cooling).
While about 400 G-15s were produced, you have to think that the number of diode testers was fewer, and that an arcane piece of diagnostic equipment for a long-obsolete computer would likely be discarded in decades past as its purpose and utility would be meaningless outside the hands of a very small contingent of specialists. What I'm getting at is that there are likely very Bendix diode testers still in existence, and this one in your workshop/lab/compound (pick your poison) may be the only surviving example for all we know. As such it will be highly sought-after by those working to restore the Bendix lineup of computers, who may shortly be pounding fists against your door to get hold of the unit. Here I want to recognize the immense value of what Lloyd has accomplished. By reverse-engineering and documenting the tester, he has made it possible to functionally recreate it. Those won't be virtual museum exhibits like yours, but the functionality is the point. A huge round of applause goes to Lloyd for his magnificent work. The one difficulty may be the slot connectors for the Bendix cards. If nothing commercially available can be made to work, there is the option of 3-d printing, thankfully. It will be a headache but something can be cobbled together. Thanks for another fascinating installment.
Great! 🤔 But how many more elderly germanium diodes will have heart failure in the next few years? Some times the "scattergun" repair method saves you far more future heartache! Especially when you have fairly bomb-proof Schottky's available. (But good job Lloyd! 😃)
Think about it like this: There were 60 bad diodes out of 3,000. That's just a 2% failure rate. If you were repairing a more modern machine that had 100 ICs on it, and two of them were bad, would you replace every single IC just to be safe? A 2% failure rate is incredibly small. I'm sure we'll come across maybe 10 more that fail over time, but we'll tackle those one by one when they happen. The rest of them I expect to soldier on without issue!
@@UsagiElectric Yep! I WAS playing "devil's advocate", but I suspect that the failure rate will increase dramatically as time goes on. Are these diodes a type susceptible to "tin-whisker" failure? This is the main killer of old germanium devices. Yes, it's a BIG job but as an experienced electronics engineer, I would from now on, probably spread the load by changing ALL the diodes on each board as they subsequently fail.
You may want to install a larger battery or a plug for a wall wort. I suspect that the tester with a meter, inverter board, and others will eat the AA batteries pretty quickly. 6V to 33V is about 6 times the voltage which requires like 6 times the amperage to convert, not counting the no-load current draw of the inverter board. Also I think the meter mechanism takes some power too to move.
The diode checker meter circuit. When you have a meter with a series resistor it is easier to see it as a voltmeter. That meter is indicating forward and reverse voltage drop across the diode. You can consider the meter resistance and series resistance in series as if it were a resistor in parallel with the dioade with an ideal voltmeter (no current flow) in parallel. The meter is probably 50 microamps at full scale
One way you could test those germanium diodes is with an Oscilloscope that has a built-in component tester, my old Dick Smith Electronics DSE Q-1804 20MHz dual-trace analog scope has a built-in component tester that temporarily puts the scope into X/Y mode when you press the component test button, it is great for testing transistors and diodes.
Better way if his scope has no such capabiilty is to build a 5-component Octopus IV tracer 6V transformer, two resistors to bring it down to 1V pkpk, two banana plugs and that is all! Same functionality, in a pinch
I reckon both of these are probably pointless, given that the normal method of checking the parts made them look om. Did you watch the full video? These diodes are seeing reverse voltages of 24v. Neither of these two suggestions would offer significant enough voltage to show the reverse leakage issues discovered using the official tester. Similarly for the forward voltage drop when they're seeing 30v forward applied.
@@CollinBaillie Of course the simplest modification would be to bump up the test voltage to 30 Volts Peak to peak using a suitable transformer, that's easily done. My own Octopus has adjustable voltage and current limits to check Zeners and diode leakages
UsagiElectric - It's getting exciting that Bendix G15 is getting closer to run the codes through it's ALU circuits... Yay! Still needs a few works, of course, something I am looking forward to when you get back from the trip. I am glad that Lloyd gave you a hand fixing up that absolutely ancient computer as former Bendix technicians are as rare as hen's teeth nowadays. (BTW, for weird reasons, UA-cam refuses to let me tag you.)
00:39 You've defined a new class of vintage computers. There was the mainframe, the mini, and the luggable computer. Now there is the huggable computer. Bonus points if it's all warm when you hug it.
Hmm... 1956 hey, That's the same year I was built. Maybe I should get my diodes checked.
Check your capacitors also, hope they aren't leaking
The G15 is tough as nails and refuses to quit, so I'm sure you're just as tough and tenacious!
Same here. Hello.
Yeah, all mine have turned into zener diodes either random voltage drops. 😆
@AlchemyAmiga-C64since198Mine too! (1959 model) 😢
Make sure to remove the batteries in the diode tester before it goes back to the shelve :) Also what I do when I reversed engineer something, I put the print-out into the device itself, so that the next person working on the device in 50 years will get the necessary headstart by having a fantastic documentation to start.
In high school we had a G15 sitting in the back of the Algebra classroom that they had removed from service in the mid 70's. I spent many hours as a freshman starring at that front panel, confused why you had to adjust the AC voltage on it, or what in the world the giant switch at the bottom was for. This was the late 70's so I was used to seeing IBM's & PDP's, not this weird 50's relic. I took all the electronics courses the school had to offer and I remember seeing that diode tester sitting on a shelf, I never put it together that it was for the G15. I think junior year I found a dusty South West Technical Products terminal kit laying around and since I was totally fascinated by computers I asked the teacher if I could build it. He said sure! He was kinda excited to see that he was creating a passion in at least one of his students and happy that I was taking an interest. He purchased a cordless Isotip soldering iron for me to use building this kit. At the time, all we had were unregulated 110v soldering irons in the lab which weren't suitable for that sort of work. After building it, at some point I was joking with him about dragging the G15 over so we had something to connect it to. We didn't really have the space in the lab for it and since my buddy and I were the only 2 students interested in the project, he couldn't justify it. Much to my surprise, a few weeks later he told me that the school wanted to get rid of the G15 and that I could purchase it from the HS for $20! Sadly, my parent's wouldn't go with the program and let me buy it. I did spend a bunch of time reading the manuals for it being particularly interested in the drum memory unit. At the time I wondered how many hours of audio you could store on a drum like that. You'd see surplus drum memory units pop up at surplus places and electronics magazines back in the day and would have been a fun experiment. My electronics teacher then gave me a KIM-1 that had been donated that nobody had been interested in to play with. I thought it was the coolest thing ever! This helped inspire the teacher to start a Digital Electronics course the next year. I built a Sinclair ZX-80 kit out of the back of popular electronics I got for X-mas and have been banging keys ever since.
After watching this series, I'm so glad my parents were smart enough to say no. It would have been wayyyyy too far over my head and skill set at that time to have resurrected that machine, even with the 2 foot pile of manuals and tapes that were with it. The original Bendix terminal was there too as well as a 2nd, smaller Bendix cabinet that I have no idea what it was for.
Do you still have the KIM-1?
In reading your post, I kept asking myself if I had written it - at least for the first few sentences. At my high school, we had a G-15, also in back of the algebra / geometry class. I would have played with it around 1968 - 1969. Is there any chance that your high school was Skyline, in Oakland, CA ?
@@chrisp1653 high school reunion in YT comments thanks to Bendix G15?
The wiring in that diode tester, well, the G15 itself too really, is a work of art.
I came to comment on the wrapping of the wiring bundles. Tried to do it many moons ago, and it did not look that good.
Wire lacing is an art unto itself.
@Usagi Electric
The 12AV7 tube is NOT a substitute for the 5965 (E180CC). The 5965 and other tubes specially made for flip-flop circutry, has to withstand being biased into sharp cut-off for extended periods, hence the more rugged construction, a lot more heat has to be dissipated in cut-off.
Unfortunately the 5965 and the E180CC are wrongly considered an "audiophile" substitute for the 12AU7 - ECC82 types, so prices are stupidly high for tube wich is prone to hum, noise and microophonics. For anyone who doubts my claims, please read the Philips datasheet for the E180CC.
Much ❤
Just a picky point: how does cutoff increase dissipation compared to conduction? I thought I'd read it was a problem with cathodes becoming coated with foreign matter from other tube components (like silicon from heaters) during cutoff for computer tubes, hurting their performance when they must conduct.
The "Hot cathode" article in Wikipedia seems to corroborate this, calling the phenomenon "cathode poisoning."
Perhaps for short term testing the 12AV7 would work, but the computer tube will be needed for best reliability.
Just looking at the 12A?7 family, a 12AV7 isn't an exact electrical replacement for a 5965, and isn't close on the ruggedness chart, and isn't designed _specifically_ for extended cutoff operation. But given that 5965's are now essentially unobtanium, and that this machine isn't going to be sitting powered on for hundreds of hours, a 12AV7 is probably a perfectly fine substitute for occasional operation. And it is available, which is better than no tube at all.
Given the specific logic conditions in use here, it would be quite feasible to engineer a solid-state replacement for the 5965. I'm sure it has been done in the past, probably more than once, but the parts for those versions are likely obsolete by now too.
Since we aren't interested in audio distortion, just a good curve matching a nominal 5965, it wouldn't be too hard to make something with some power MOSFETs or the like and a few other parts to do the job. (Don't forget the heater load resistor!) However, that would be kinda defeating the purpose of getting a vacuum tube machine working, if it didn't use vacuum tubes. 12AV7's are probably the much better available choice.
high hum, noise and microphonics is how you can tell that it made a difference to the sound and was totally worth the money!
Audiophile nuts have created supply shortages snd crazy prices for a number of vintage items. Not long ago I ran across someone trying to sell Western Electric manufactured wire which had been pulled out of a central office main distribution frame. He was asking a few US Dollars per foot per strand, which is a crazy price for old used wire. I have quite a few feet of Western Electric 25 pair gray plastic sheathed cable and I can attest to the fact that is quite average plastic coated wire. The wire itself appears to be tin plated copper or just copper depending on the cable. I’m pretty sure it is tin plating because it does not tarnish as silver would.
The 407A tube good example of a way over priced tube, just a simple oddball triode but no way it is worth 50$
It is amazing that you were able to get an actual Bendix test instrument for this G15. It can't be understated how amazing this is to have this tool.
You're getting closer every time you work on this. Thanks for sharing your journey with us!
@@davida1hiwaaynet I was thinking the same thing, that the Bendix card tester is the proverbial hen’s teeth which are some rare.
@@wtmayhew Knowing HOW it works, makes it possible to build more modern semiautomated replacements in future. Personally I'd run it off a USB power supply rather than rely on potentially leaky dry/alkaline cells
@@miscbits6399 Thank you for replying. There really isn’t much to the tester conceptually. It is a switch matrix to select a particular diode from the card under test. There’s a cross connected DPDT switch to forward or reverse bias the diode under test. The micro ammeter is wired in series with the battery to monitor the current. The secret sauce is in that rare card edge connector and the wiring of the wafer switch to make it easy to select a diode.
The diode logic in the G-15 is probably composed of diodes similar to the venerable 1N60 germanium junction diode. The 1N60 has a peak reverse voltage rating of 50 Volts. The test box originally had a 35 Volt battery which is probably enough to spot a leaky diode. Something like a modem multimeter or a vintage VTVM usually on biases test components with about 3 Volts. The test box lets the operator flip the rotary switch through all the diode positions on a card in a few seconds - one rotation for forward bias, flip direction switch, then a second rotation. That sure beats probing individually with a meter - which doesn’t put out enough voltage to catch a reverse bias breakdown.
Two steps forward, one step back. It's the cycle of life. Considering these diodes are the same age as my father, the fact that such a small number only have failed is astounding. Also Lloyd, if you are reading this, you are a GOD among men. Never forget it. You and Dave are making magic here, but he'd be going at 1/10th the pace or even slower if not for your help. Thank you.
I agree. My experience of germanium in the 1980s wasn't pleasant and I spent a lot of time cobbling things up in cases when silicon wasn't a simple drop in answer
(Why? Working in a SW comms station commissioned in 1951 and considered well past end of life even in 1980 let alone the end of the 80s, No money was put into replacing anything as it was already expected to be turned off (and was, in 1992), so things like 45.5 and 50baud tty circuits had to be kept running. Adding to the pain was the fact that large portions of the building had been declared off limits due to asbestos comtamination and a realisation that berryllium dust had been accumulating in nooks and crannies everywhere (a lot of old high power tubes use it for the heat conductivity qualities, and a bunch used thoriated cathodes with 2-3kW heater elements to enhance electron emissions. Let's not even go into the hard xray emissions from the mercury arc "rectum-fryers" or needing copius quantities of distilled water because the cooling loops were directly cooling anodes running at 50-75kVDC)
I was taught in Electronics school that you always adjust an analog meter to ZERO with power OFF. You adjusted with power on, and not to ZERO. I can hear the voice of my teacher in my head, screaming, for doing that. LOL 😂
I didn't show it directly on camera, but I wanted to see if when adjusted such that Lloyd's specs are on point, does the needle drop down to zero. This was a way of confirming just how far off base we might be. It did indeed drop back down to zero, and once powered off, I confirmed, the needle was still at zero. In the video, it seems like I blazed through it, but in reality, I spent about 15 minutes confirming and double checking the needle position and cross referencing against any and all notes and documentation we had.
@@UsagiElectric I'm just giving you a hard time! 😉 I have NO DOUBT that you are more than competent and capable in the field of Electronics Technology. The Bendix G15 is literally my dream computer. I was given a Bendix volt meter off one as a kid and always loved them ever since. I still have the meter even today. I absolutely LOVE watching the G15 slowly roar back to life! Keep up the great work!
12:35 mark, it’s the only way - adjust the meter set screw to zero with zero current / voltage depending on the type of analog meter.
I found myself screaming at the TV about this.
@@UsagiElectric You are forgiven. I, too, crimged at that first adjustment. but later I saw the meter at zero and concluded you had fied it.
You should see the diodes you pulled for a fundraiser for System Source - that way people can buy a piece of Bendix for their own collection !
The thing I love about this channel is the in-depth stuff, but not too complex, the really old systems (not just 8-bit) and the community. Reverse engineering stuff is amazing, people helping out with coding etc. Love it.
It's like Quantum Leap - you don't get to disappear until you're completed the task you were set out to do. I assume that Al is talking to Ziggy just off camera. Keep up the amazing work, this series has been fascinating, thanks for sharing it.
He says "You're still here?!" but like we're always here, staring at the blank screen waiting for the next one to start. I thought that was understood.
Long Live Lloyd
FOR REAL. Bro just whips up schematics like its no big deal. I'm like..... is he a wizard?
The more of these episodes I watch, the more I'm impressed by the G15's industrial design. The swing-out access doors, the keyed and coded tube carriers, the metal finishes and colors -- they really wanted to make sure people felt they were getting their $50,000* worth! Even in this little tester it was as if somebody said, "Okay, we've got a way to test the diodes, but we can't just cobble this thing up on a breadboard -- it has to match the G15's aesthetics!" Thanks for taking the time to show all the nuances.
*For a way to talk about prices without getting into currency exchange, you could say something like: "In 1956 this computer cost $50,000 -- at today's prices, that's the equivalent of how much profit Amazon makes every 56 seconds..."
Bendix made a LOT of military flight computers and they DEMANDED good maintainability
It amazes me that I have meters, component testers including a curve tracer but the Bendix engineers knowing others would be having this exact problem solved it with basically a battery, a couple resistors, some rotor switches and a meter. Small and rugged for it's time compared to carrying a curve tracer and a scope considering scopes back then were very heavy vacuum tube jobs that required their own cart and were very fragile. I am having to repair my RF analyzer now because of a bad screen when being transported even in a special anvil case and it's within the last 20 years modern.
They yield on early semiconductors and diodes wasn't real great so the parts were well checked before they left the factory. Infant mortality was also more of a problem so parts failing a few weeks or month into the field was also an issue. I would expect a well equipped service man to have a tube tester, diode tester, multimeter and scope when they visited a customer site because that many parts will produce a failure sooner than latter.
@@denawiltsie4412 The vacuum tube machines I was familiar with (admittedly larger than this machine) always had an on-site customer service engineer that would some on shift around 6 AM, power up the machine, run diagnostic, and fix anything he found broken or marginal, hopefully turning it over to the customer by 8 AM. Then the machines would usually run till around 6 or 7 PM, when they would either be switched off or put into a low-power standby state until the next morning. It was rare to not need to replace a few tubes every day or so.
Reminiscent of shop built test equipment we used in production flight test when I worked at Cessna and at Learjet.
@@lwilton The oldest systems I used where transistor so I really haven't had hands on experience with vacuum tube computers. I was lucky enough to see one of the Air Forces SAGE (Semi-Automatic Ground Environment) installations where they had two very large vacuum tube computers with core and drum memories. One system backed up the other but I had the impression that the system was reasonably reliable. At least enough that they allowed a regular stream of visitors to view the system in operation. They trusted the system enough that they would run demos on the idle system.
Love my Sunday morning coffee watching Usagi!
I've got this terrible pain in all the diodes down my left hand side - Marvin (The paranoid android)
In the final scene of all, is the bowl pushed aside as a "Thank you for that" or a "Now refill, please."
It's a "I want more banana!"
Buns always want more nana 🙂
People like this are a national treasure... I dont think this country realizes how badly we truly need people that can, and will, do this type of stuff... 🎉
I remember a TV commercial from decades ago, for, I think, a motor oil, the machanic states "You can pay me now, or you can pay me later." BY ALL MEANS check all the diode cards now. Not doing so is going to cause you grief tracking down which card is causing the fault. You will end up spending more time in stressful panic debugginf than you will take leisurly and calmly checking them now.
Loving the lacing cord on the back of the panel. Reminds me of my first job to pay my way through university, working for a company making instruments for the North Sea oil rigs.
No idea how you get one of these videos made every week. Major highlight of my weekend! I mean, you do SO MUCH, not only al the work on the machines, all the trial and error, but you also record AND EDIT 20-30 minute videos EVERY WEEK?! Madness!
I am older than this system and got to play with one in high school in the early part of the 1960s. Pure magic watching this come to life. Tic Tac Toe next!
Great stuff. Especially how Lloyd reverse engineered the diode tester. Lloyd=GOAT!
It's always scary to test things after a major repair like that! I'm glad you did and I'm glad you showed us. Thank you!!
The simple fact, that this diode tester exists, tells me: This computer needed al lot of service in the time.
Super cool concept, sorry production was laborious, thanks for seeing it through!
That screw on the front is the Mechanical Zero. When the device is OFF and the meter is reading nothing, you use that to adjust the meter to read exactly Zero. Any other adjustments should be electrical calibrations. What you did may work, but it is not the correct use of the mechanical zero and it may affect the accuracy across the dial.
Yeah I know, I am being pedantic. lol. I do love your videos, and this tester will definitely save you a heap of time.
Lloyd is one smart person! He did an excellent job REing that tester.
You're a LEGEND! Lloyd too! All that hard work and we just get to enjoy the good parts, thank you!
it is a magnificent looking machine. 1950's deco was such a good era for design. It's just beautiful
Just a thought but 4x 9v batteries would be 36v if you wanted to eliminate any active electronics and phantom power draw while not in use. Also if 36v is too high just add a few diodes in series with the batteries as a voltage drop.
In this case, adding diodes in series with batteries would cause an issue if a test was measuring something less then the forward current of the series diodes.
The fact that that test was made, tells us they were well aware that the diodes were a problem. So cool to see that old tester though.
I was the MIL-STAR. MIL STD 1750. FEU (functional equivalent unit) repair man. Many a trip to Trouble W in Redondo Beach and LMSC in San Jose. Ha Ha. A simple Rassberrie Pi project, A DAC & ADC pulse it and read back and check reverse bias test. Looks like a Rube Gold Berg set-up.
16:50 this is what a "debugger" IDE looked like in the 1950s. it was a physical gizmo. this is amazing how far we've come in so little time
The G-15 series is my favourite of the ones you're currently running. I'd never heard of it till I found your channel, but I'm completely invested in this thing. I'm excited to see this one work... soon!
Man, Lloyd is some kind of G15 deity at this point
"You are all a little crazy!", he said, while testing more than 3000 diodes in order to run his vacuum tube computer at home. ;-)
Love the cable lacing in that tester.
As an old fart I always loved the crinkle paint finish on the 50's electronics. And yes I have a early DEC 11/70 in my garage.
Given how rare these machines are now, I am tempted to say that the value (at least to those who collect machines like this) of a working G15 today is even higher. Of course, commercially not so much, as you can get orders of magnitude more computing "oomph" for orders of magnitude less "ka-ching".
Your "see you in the next episode" didn't work for a simple reason. Everyone knows that your videos end with a good portion of your "furry family members". So unless you "roll the fluff", we know the video isn't over. 😄
A working G15, I can't say for sure, but a few have changed hands over the past year. A G15 with MTA2 tape drive, two typewriters, and some other small stuff sold at RR Auctions about a year ago for $60,000. More recently, a G15 with a typewriter, and an MTA-2 tape drive from the LCM sold on Christies Auctions for around $20,000. So, it seems the value of these machines is still pretty impressive, but still well under $100,000 today.
I was wondering how many people I would catch out with the fake ending, but I also know y'all are all really here for the fuzzy family at the end!
@@UsagiElectric The timeline ratted you out. I was curious why you were giving the end while so much time was remaining.
@@UsagiElectric When you first said it I was quite disappointed not going to lie, had me for a few seconds there.
One step closer to a working machine. Thanks for the Videos.
I hope you know germanium transistors and diodes were made well into the 90s in Russia and are still available in the millions as new old stock.
The cable management on this old stuff is a thing of beauty
loved the little gag !
I enjoyed your video. The tube failure you experienced, is the perfect example of why, at Bletchley Park during WWII, when the British turned on the Colossus machines, from the time Tommy Flowers, and his team, turned on the first Colossus machine, the machines stayed on continuously until they were turned off and destroyed at the end of the war. Leaving the machines on, was the key to improving the reliability of the vacuum tubes.
PS: Each mark II machine, had 2500 tubes
Fascinating episode. Look forward to the next one!
You are rapidly becoming one of the world experts in 1950s computer technology.
I have a 70's Elka Concorde 802 organ with many, many diodes also doing a lot of signal control and whatnot. It al runs with similar voltages too: 0V to -27V. So this video is excellent info in debugging some (future) issues!
Phew. The beard is coming back. Success is just around the corner 😅😅
you are the best channel on here, in my view, such persistence, energy and effort
Think about the 1956 factory technician who had to test all the boards needed for a full scale production run. They were looking for the forward / reverse reding for hundreds of boards a week. Fast forward 20 years and you could test each board in 30 seconds, all diodes and all resistors.
I would bet quite a few dollars the factory had automated board testers. I know I had them at Burroughs in the late 1960s. You could build one as a special peripheral to be run by a G15. You could make a simpler version with some stepping switches and limit-indicating voltmeters and ammeters, which were fairly common in those days.
This is a welcome Sunday treat!
This is the beauty of discrete logic:
No special chips that are now obsolete and impossible to find, just standard parts that you can get everywhere.
That test box is nice!
But for testing large numbers of things (parts, modules, whatever) a simple "pass/fail" indicatzor (red light or green light) would have made testing much faster and simpler.
Maybe a comparator that measures the "meter current" with a shunt resistor and compares to a fixed voltage depending on the position of the forward/revers switch?
Probably not worth designing a custom test box, but I probably would have done it.
(Automating something is way more fun then actually doing the work ...)
SUGGESTION: With a fine sharpie, number all the remaining packs with some type of notation such as rack, vertical position, horizontal position- e.g. 3-5-16 May be useful once you get the machine running as you will definitely have failing packs/tubes. Put the info in a spreadsheet for the next guy that is going to service the beast. 😊
The people who gave you this computer must really like you! They even allowed the leg hug. Kudos to you!
Wow that tester is just brilliant! love your show!
18:03 no shame for been crazy. That VC was fun.
It was indeed!
Awesome. The natural follow up question: so how will you be testing the valves?
And the point to point wiring. That's going to be a job!
There are only about 460 vacuum tubes, so it's a much more feasible task to tackle that problem one by one as issues crop up during machine testing. I'm working under the assumption that the majority of them are okay, and we're just going to troubleshoot the machine slowly and surely.
I do have a tube tester, but it's not a very good one, so it won't highlight issues with emissions and balance on the flip flops or things like that. So, if we find a flop isn't behaving and the tube filaments look fine, we'll get into some more in depth testing at that point!
@@UsagiElectric You might look around and see if you can find a Hickok Cardmatic in working order, along with the test cards. They are pretty much idea for this sort of batch testing. The Model 123A had a set of test cards for the 5965 specifically.
Magnificent work, especially on reverse-engineering and documenting the tester.
Also, i noticed you were using Australian Central Time in your scheduling tool, which I can totally understand as a guess at the best average timezone to use for the country, but in actuality, over 80% of the population lives on the east coast, so it would probably make more sense to use AEST unless a disproportionate number of your Australian viewers are tuning in from WA, SA, or NT.
And, yes, technically Queensland is also in a different timezone because they don't do Daylight Savings Time, but we all just make fun of them for that, so I wouldn't factor it into your considerations...
In Usagi's case, I think that might be because Japan uses the same zone as central Australia. So he defaults to that, when choosing time zones for international comparison purposes.
@@PilkScientist I did wonder if better aligning with Japan had something to do with it...
not marking which slot they came out of is craziness to me. that's my first instinct. especially on a legacy system, I'd keep the configuration as close to original as possible just in case the assembly order is meaningful
They are all supposed to be identical, the slots are labeled and the components color coded by function.
It's designed to be worked on like this.
As volvo09 said, the slots are all keyed and all the cards are labeled and color coded by function. Every card should be completely interchangeable with other cards of the same type. I actually specifically chose not retain the same configuration as it's a good way of highlighting failures we may not have caught yet. It can be a little frustrating and time consuming to track down though, haha.
@@volvo09 oh I have no doubt that it was designed to be interchangeable. but with something that old, I would think maybe some board warpage or some corrosion patterns or something might mean that returning something to its original slot could help make better contact or something that wasn't immediately apparent
Even if they are the same card, it is good practice to replace any component/card back in its original location (not just for this project).
I love the G15, excellent progress!
I wonder how many of those testers are left... Can't be many kicking around.
Surprisingly, I think I've seen more testers than actual G15s!
I have one, Bob still has two or three, Lloyd has two, the G15 that sold at RR Auctions a while back had one with it, and so on.
@@UsagiElectric oh wow! That reverse engineering will certainly help anyone else who restores a G15.
I own an airplane from 1956...fewer diodes, still flies great! And yes, it is older than I am...
I think the carbon composition resistors should be taken into account; I have seen one vintage radio / TV restoration site where the person said he checks them and replaces anything that has drifted out of tolerance (gold band = 5%, silver band = 10%, no fourth band = 20%). That might have to be the next step after the diode test if the computer still acts like it hasn't had its morning coffee.
Totally agree with you, Carbon resistors tend to do weird things the older they get.
I'd suggest probably testing all the tubes first, they are more likely to have hard failures that will have an impact. The resistors without digging deeper you won't know if it being out of spec matters. Dud tube and you're going to have a bad day (as they found out(
Are you going to be at System Source on the 18th of October? It would be good to see you again!
Yeah nice jig, I once worked in a repurposed lab and in the back as a similar jig that was used to test disk contoller cards for I think DEC mini comuters, no panel metter just a two banks 32 leds and sbunch os switches to set up the card trst
From the very start of this project when you first gave us an overview of the machine and all those diode cards, I was wondering how in heck anyone was going to be able to check all that. We're getting there. That is, you're getting there. We're looking over you shoulder and trying to keep out of the light.
I had a similar problem once back in the late 1970s or so. I built a test fixture I could plug into my AIM-65 computer, and wrote a Basic program to run it. It could to an almost complete functional test on about 4 different types of cards, and let me test hundreds of them a lot faster than by hand or a manual tester would have let me do.
Wow, nice work and that test tool sure made that quick(ish) work. I bet a good percentage of those carbon composition resistors are bad too. They can go both low and high resistance. Looks like it's full of the little critters. And then, what kinds of capacitors does it use? Electrolytics are suspect but even more so any paper based non-polarized types.
The point contact germanium diode failure is caused by carrier diffusion drift. The dopant elements are attracted to the point of contact causing the diode to have a highly doped region in contact with a normal doped region and a depleted region between. These bad diodes won't switch right in logic ciruits due to shutoff oscillation but they have a secret. They can replace tunnel diodes in most circuits. ❤
About the heat, have there been any mention of how much power usage is? Cant say I’ve watched every video so far, but can’t remember it being mentioned in the ones I’ve watched yet.
I grew up a few miles from the Large Scale Systems Museum and I still live in the area. I considered stopping by for a visit and seeing if my dad would be interested as well since he worked with computer mainframes starting in the 70s. I gotta say though $100 per person to attend is a little much. I genuinely hope they raise a lot of money but feel like they could get more by having a lower cost of entry.
I may be late to the game - but with the Testofon (aka Contitest), you can quickly test any electronic component, especially if you have more than one of them. It works by applying a voltage and transforming the current that flows into a tone for your ears to hear and for your brain to decipher (if I test capacitors, I get a whole chart with the time/frequency axis and a scatter plot on what's good, what's tolerable and what's bad - and it'll fill it in in realtime).
If you got all the diodes next to each other, testing 8 of them both directions will take around 5 seconds.
It can also test transistors. Not great, but I found wonky ones that still worked in circuit most of the time, but either fail when cold or fail when warm.
Like the multimeter, it won't find any that fail the forward voltage test above 4V, but it'll find shorted or reverse-leaky diodes.
You might think of it as a glorified continuity beeper, but to me, the Testofon was a game changer. In a pinch, it could even replace an oscilloscope or logic probe.
Im so happy i stumbled across this video. Great video and delivery. I subscribed!
Dude! I started shouting at the TV when you tried to end the episode the first time! 😅
Hahaha, almost had you!
Being a yinzer, techie, and working at the supercomputing center in college, I have no idea how i have never heard of the large scale systems museum!
The DEC PDP11 is one brilliant machine which I have used at the College of Further and Higher Education at Trostan Ave Northern Ireland.
Booting might have been mostly luck with 2% of all diodes bad.
And, Thanks Lloyd, again!
I remember reading that tube computers (or anything with a sufficiently large number of tubes) couldn't theoretically work because the MTBF is such that there will always be at least one bad tube; the reason they do is because MTBF isn't the whole story, due to production variations some will fail early and some not, if you start with a whole new batch and replace them as they fail eventually you end up with all the better ones. The machine then works reliably until the tubes get into the end of life section of the bathtub curve at which point it is necessary to start again with a new batch. The other thing was never to switch the machine off as it is switching on which puts the most strain on the tubes, obviously not practical for hobby use.
All this used to be observable with incandescent light bulbs in any building with enough of them.
The G15's MTBF was actually quite a bit longer than other tube computers of the era, often going for weeks of 24/7 operation without a failure. There's a few things playing into this. Only having 460 tubes brings the numbers game to a much more reasonable level. Also, the G15 is primarily using indirectly heated Noval tubes, which had much, much better life and reliability than the older octal or directly heated tubes used in machines like ENIAC. Also, filament warm up and start up is the hardest time on a tube, but the G15 has a clever soft start that sequences in, slowly increasing the voltage to the filaments over the course of 60 seconds. That means that all the filaments last much, much longer.
They used a couple of design tricks to extend the tube life. They ran the filament cooler either by a lower voltage or by designing the tube to handle a higher voltage than what was applied to it. Tubes normally ran much higher plate voltages between 45 and 90 volts. It surprised me a bit when I found they were running around 24 volts. Mean time between failure is a worst case calculation and often you will find that parts far exceed the numbers it produces after you get a few early failures out of the way. Running anything below specs will extend its life. I have a chandelier that uses 12 - 40 watt bulbs and was purchased in 1986. I have only replaced on bulb and that was because it broke while cleaning it. The trick is on high it produces so much light and heat that it's always on a dimmer running at perhaps half its full output.
MTBF is a mil-spec calculation based on component type, component count and stress applied to each component. You could have a poor design that never overstresses any parts but fails reliably far sooner than the calculated MTBF. And you could have good designs that run for longer than the MTBF.
@@UsagiElectric I was actually going to ask if you would consider using a soft tube starter. I'm glad to hear they included one, Bendix really thought of EVERYTHING!
@@denawiltsie4412 Incandescent lamp life is inversely proportional to the 13th power of the applied voltage. Running lamps at 50% voltage will extend the life by about 8000 times normal. Since the dimmer is probably a phase control unit, and you probably turn it on and off now and then, you probably won't get this. But if you have 1000 hour bulbs, it is still going to last for a whole lot of years!
BTW, commercial and industrial lamps in the US were designed to run at 130V. Since most power is between 115 and 125, this made them last considerably longer than consumer lamps, but they put out about 10 to 15% less light.
Wow! Amazing you were able to get a tester!
I wonder how common those things are, or if there's an open-source replacement for it.
Look at the lead dress in this thing! Wow. I'm curious though. If you had no documentation, how did you know a 35V battery belonged at that connector?
Glad I could see myself doing X Windows things in the laptop screen ;D
Really happy i could watch this live as well. This was beyond a blast
21:20 - You're still here? It's over. Go home. Go.
Ok Ferris
This was wicked cool.
Always fun.
Thank you
Thanks Lloyd!
You glued the batterie pack onto the factory stamps?
I saw that too and cringed!
He said "Double-sided tape"...
@@Scott-i9v2s
Yeah, it’s still not good for the ink xD
You can make a cheeseball curve tracer with a power transformer and an XY scope... really great for quick go-nogo testing of semiconductors.
Lloyd sure is a treasure to the community and to history. That Bendix tester is sure nifty. I sense a future video will be testing all the vacuum valves. 🤪
Cheers,
Adriel
I see a nice gem there (The multimeter) If you have budget, try to get the UT210E-PRO version. That one has Frequency as an extra, and is more sensetive in the MiliAmpere range through the clamp. Nice habit of them is that they can be reprogrammed to start in DC mode on all ranges, that the backlight can be set to a longer value, or even to continious untill shut off again, and that the Auto shutoff of the meter can be omitted. All that by reflashing the Eeprom with an ESP32 flasher.
Just looking at the construction, this appears to be DTL logic instead of earlier RTL logic (diodes instead resistors) and was much more reliable. I worked for IBM in 1966 and their SLT flip-chips were much much smaller (approx 1/2"x1/2"), and one-chip could effectively replace one of those large PWB boards. The SLT chips had diff-amps available and thus could have myriads of other functions added directly on the hardware, such as flip-flops, monostable delay gates, direct NANDs, and multiple output Interface Drivers. The later IBM70+ systems pretty much killed any of these older PWB based single-card systems. That newer technology was actually driven by the development of solder-flow processes, whereby a huge large board could be wave-soldered in one pass. I lived through this whole cycle, and one of my sons now works on 4-to-7-nanometer scale device technology, and there maybe be yet a chance to get down to about 2-nanometers, before Moore's Law is broken. Now requires below ambient cooling, but the Spintronics people are working in that zone (can be done with CO2 cooling).
While about 400 G-15s were produced, you have to think that the number of diode testers was fewer, and that an arcane piece of diagnostic equipment for a long-obsolete computer would likely be discarded in decades past as its purpose and utility would be meaningless outside the hands of a very small contingent of specialists.
What I'm getting at is that there are likely very Bendix diode testers still in existence, and this one in your workshop/lab/compound (pick your poison) may be the only surviving example for all we know. As such it will be highly sought-after by those working to restore the Bendix lineup of computers, who may shortly be pounding fists against your door to get hold of the unit.
Here I want to recognize the immense value of what Lloyd has accomplished. By reverse-engineering and documenting the tester, he has made it possible to functionally recreate it. Those won't be virtual museum exhibits like yours, but the functionality is the point. A huge round of applause goes to Lloyd for his magnificent work.
The one difficulty may be the slot connectors for the Bendix cards. If nothing commercially available can be made to work, there is the option of 3-d printing, thankfully. It will be a headache but something can be cobbled together. Thanks for another fascinating installment.
Great! 🤔 But how many more elderly germanium diodes will have heart failure in the next few years? Some times the "scattergun" repair method saves you far more future heartache! Especially when you have fairly bomb-proof Schottky's available. (But good job Lloyd! 😃)
Think about it like this:
There were 60 bad diodes out of 3,000. That's just a 2% failure rate.
If you were repairing a more modern machine that had 100 ICs on it, and two of them were bad, would you replace every single IC just to be safe?
A 2% failure rate is incredibly small. I'm sure we'll come across maybe 10 more that fail over time, but we'll tackle those one by one when they happen. The rest of them I expect to soldier on without issue!
@@UsagiElectric Yep! I WAS playing "devil's advocate", but I suspect that the failure rate will increase dramatically as time goes on. Are these diodes a type susceptible to "tin-whisker" failure? This is the main killer of old germanium devices.
Yes, it's a BIG job but as an experienced electronics engineer, I would from now on, probably spread the load by changing ALL the diodes on each board as they subsequently fail.
You may want to install a larger battery or a plug for a wall wort. I suspect that the tester with a meter, inverter board, and others will eat the AA batteries pretty quickly. 6V to 33V is about 6 times the voltage which requires like 6 times the amperage to convert, not counting the no-load current draw of the inverter board. Also I think the meter mechanism takes some power too to move.
Good Sunday Morning!
Blast from the Past is a Great Flick, Funny as Hell
Well Done Sir!
The diode checker meter circuit. When you have a meter with a series resistor it is easier to see it as a voltmeter. That meter is indicating forward and reverse voltage drop across the diode. You can consider the meter resistance and series resistance in series as if it were a resistor in parallel with the dioade with an ideal voltmeter (no current flow) in parallel. The meter is probably 50 microamps at full scale
I believe it said it was 1 MAFS. What you say still holds true, of course.
Such great content, man. Keep it up!
One way you could test those germanium diodes is with an Oscilloscope that has a built-in component tester, my old Dick Smith Electronics DSE Q-1804 20MHz dual-trace analog scope has a built-in component tester that temporarily puts the scope into X/Y mode when you press the component test button, it is great for testing transistors and diodes.
Better way if his scope has no such capabiilty is to build a 5-component Octopus IV tracer
6V transformer, two resistors to bring it down to 1V pkpk, two banana plugs and that is all!
Same functionality, in a pinch
I reckon both of these are probably pointless, given that the normal method of checking the parts made them look om. Did you watch the full video? These diodes are seeing reverse voltages of 24v. Neither of these two suggestions would offer significant enough voltage to show the reverse leakage issues discovered using the official tester. Similarly for the forward voltage drop when they're seeing 30v forward applied.
@@CollinBaillie Of course the simplest modification would be to bump up the test voltage to 30 Volts Peak to peak using a suitable transformer, that's easily done. My own Octopus has adjustable voltage and current limits to check Zeners and diode leakages
UsagiElectric - It's getting exciting that Bendix G15 is getting closer to run the codes through it's ALU circuits... Yay! Still needs a few works, of course, something I am looking forward to when you get back from the trip.
I am glad that Lloyd gave you a hand fixing up that absolutely ancient computer as former Bendix technicians are as rare as hen's teeth nowadays.
(BTW, for weird reasons, UA-cam refuses to let me tag you.)
Great video had me on the edge of my seat, have a safe trip.
I have the same opposite problem trying to fix an old Paia drum machine; it tests as a failure in circuit but working properly outside.
Great job David! One step closer to bringing back the past.....grin