Wow i did not realise this was the first computer Ken Thompson ever used. The change this machine made to the world is quite possibly incalculable, if its what got the guy who wrote Unix hooked
And it gets even crazier when you throw in that the designer, Harry Huskey, worked closely with Alan Turing and the G15 is inspired by Turing's ACE computer. This thing is definitely a historically significant machine, and I'm so excited to see it wake up and stretch its legs again!
I was a freshman at Rose Polytechnic Institute in Terre Haute, Indiana in 1963. (Since then the school name was changed to the Rose Hulman Institute of Technology.) Freshman arrived about two weeks before the rest. As we were on the guided tour of the “main” building, we walked by an enclosed room with a large glass window. We walked up to the window and peered into the school’s computer room. There, in all of its glory sat the school’s blue Bendix G15 computer. I had a class that taught us how to program the G15. Subsequently several other classes required that we create programs to solve problems. This series of videos is a trip down ancient memory lanes. Thank you so much. By the time that I graduated other hardware had been purchased; however the G15 was still being used. Sigh!
I had my first experience with programming on one of these machines back in 1964 when I worked at a weather station in the Marshall Islands. I programmed the computer to process radiosonde readings that had previously been treated manually. Can't wait for you to get that paper tape reader going!
The documentation is beautiful in its own way. What always impresses me about these vintage computers (even 1970s and 1980s home computers) is how rapidly they seemed to be developed from concept to manufacture without (obviously) the benefit of modern computers to assist in designing them and managing their development. A lot of talent and hard work!
Drum bearings try to pull the drum out, and get it on it's side. Then take the bearing facing up, and warm it up a little after cleaning, and apply some EP90 gear oil to the edge of the steel dust shield, to make a thin layer on the seal, and let the bearing cool down, which should allow some to enter past the seal, and mix with the old grease there and revive it. Leave overnight with it up, and next day clean off the excess, and turn so the other side is up, and do the same warm up to around 70C, and apply the oil, and allow to cool and pull some past the seal. Quick fix is to make a sort of dam there if not moving the drum, and fill with the oil, and leave the bearing half covered overnight, so it can seep past the dust shield into the grease there, and give it some lighter oil components again, and reconstitute it a little. Otherwise you need to gently pry the shield off, and then pack with a little moly based grease, and put the cover back on. Look for the same bearings, and order a cheap one, and sacrifice the bearing, using a grinder, to get the shields off them, and thus get new ones if you damage the old ones pulling them off. Have done this to a lot of bearings, especially where a new one was either not available, or needed a lot of specialised tooling not available to set the preloads, and the lube would get you another few years of running out of them.
I appreciate your proper caution in the power-on testing. You are saving yourself much potential grief. Seeing the filament load decrease as each voltage step settles was fun. The heaters increase in resistance as the warm up and that causes the voltage to ramp betweek the steps. Do consider powering the DC supplies externally the first time and ramp up the applied AC voltage slowly to re-form the capacitors and limits any surge current from unforeseen faults. I know you will have already checked the resistance across the DC main busses as the first step. The ghosts of Engineers Past will be watching over you as you proceed 😀
The "bearing noise" might be a dragging co-axial brake on the drum motor. Marvelous to watch it come up again. Wish the yUK Science Museum would find the courage to try the same on their Ferranti Pegasus. Good luck with the Sprague condensors (capacitors) as those are going to be difficult to replace if they are sickly.
It also might be a dragging brush. That style of GE Repulsion Induction motor starts with carbon brushes and switches to induction at speed. There is a centrifugal clutch that pulls the brushes away after starting, and that might be gummed up.
CONGRADULATIONS! A Big first step, glad the drum is spinning and talking! Harry Huskey has got to be smiling also! Wish I still had some of my G15 programs I wrote in the day! Looked for them but no luck! Glad that you are taking slow steps to bring the baby up and working again!
I want to comment on your channel in general. Seeing the various projects and working through them "with" you is like a refresher course in the old technology. Seeing you interact with those such as Curious Marc and building a comaradare with them, priceless!
I would suggest connecting the clock track to one channel and the timing track to the other channel of your DSO and capturing the waveform with the highest memory depth your scope supports. That way you have a backup that could potentially be used to recover it if anything happens.
I second this. That way, if worst comes to worst you have a backup, and you can use it to restore the other drum when you get around to resurfacing it.
It's an excellent idea! However, the clock track is just a simple sine wave with 3,596 pulses per rotation. That can be recreated with any signal generator with ease. The timing track is a little more tricky, but again, we know exactly what it should look like thanks to the documentation. The really tricky part when rewriting those tracks is synchronizing it perfectly to the rotational speed of the drum and motor. At a perfect 60Hz with a new motor and good bearings, it should rotate at exactly 30 Revolutions Per Second or 1,800 RPM. That should result in a clock track frequency of 107.8kHz. But, as we saw on ours, we were rotating a little slow. That means that even if we got an exact copy of the timing track, laying it back down would be not quite possible if the RPM of the drum has changed even slightly. Knowing exactly what the timing track and clock track look like means we can drop a hall-effect sensor on the drum and use a microcontroller to perfectly synchronize everything and drop new tracks down if we need to. So, fret not, we're well equipped to replicate those tracks if anything ever happens!
Magnetic media is really quite reliable, despite claims to the contrary. As long as the material doesn't get heated past its Curie temperature, doesn't get exposed to strong magnetic fields, or has some ridiculously thin/unreliable coating, any stored data should last for a very long time. In the case of the Bendix I think they also went out of their way to ensure long-term reliability, hence the spare clock track.
There exists a number of hard disks there the timming data ie servo is on a separate platter - bad servo data platter - that disk is a boat anchor. AND a number of them (ST256 ? ie 40 years?) STILL WORKS TODAY !!!
My nerves were jangling when you went for the Big Power, ye gods. I never thought I'd see a G15 ever, aside from history books, and suddenly there's 2 and you have one and it's on the way do being powered up! Awesome. All the best to you and Mrs Usagi from across the little pond.
For those Siglent scopes, if you want to see the full sequence, you'll need to zoom out and stop the capture, then you can zoom back in and scroll across until you find the start of the sequence. You can also safely enable the 20Mhz bandwidth limit to help reduce the noise on the trace, and you should also be able to increase the number of points that it captures (although that may decrease the update frequency at the expense of a nicer stopped trace). That said, I'm hugely relieved that you're seeing the expected signals and also filament voltage. This definitely bodes well for the future of this machine. Those definitely look like selenium rectifiers, although I'm not hugely familiar with these early semiconductors. I feel like silicon diodes weren't that common until sometime in the 60s? Maybe be prepared with some extra ventilation just in case. Apparently selenium diodes smell pretty bad when they blow.
Pretty freaking great. Last night you're singing from the moon, today you're resurrecting this behemoth. Thanks for the work preserving history and thanks for bringing us so much joy.
I can totally see him hook up a teletypewriter to it, turn it on, install a short paper tape in the reader, hit start, the machine reads the paper tape and the teletype spits out "HELLORLD"
This is such an exciting project! I love seeing the Bendix G15 come to life. It scares me a little bit to see computers that were built during my lifetime being displayed in computer museums. I was 2 years old when this Bendix was built, but I've worked on some of it's descendants including the PDP-8 and PDP-11 both at the DEC factory and in the field.
I'm really glad you went with solid 2AWG gold and titanium wires with dupont connectors as I would wanted. Better keep the fire extinguisher handy when in use tho
my first job at Control Data (1967) was in the systems refurbishment group (G15's, G20's, 160''s 160A's 8090's 1604's etc). I'd like to tell you that I remember everything about a G15 but I don't, wait I do remember 1 thing... if it's not working, open the doors, turn off the light and look for the blue tube and replace it. in 1968 the US government 31 G15 under contract.
Great video! Your videos encouraged me about a year ago to fire up the PDP-11/23+ and large drives. Had a few bad Sprague caps I had to replace. Hopefully, you will show your process for reformatting the caps in the G-15. Unlike when I powered the PDP-11, I want to test the caps in my Vax 11/30 next before power up which had not been powered up since I purchased it off a BBS swap page in 1994. I am also worried that when I moved, I didn’t lock down the heads. Thanks for the videos.
Here I thought my boss's 1984 Ikegai AX25 was the loudest machine around! Get some ear protection, better safe. Really cool seeing this fired up, thanks for all the effort. Merry Christmas & here's to a healthy 2024!
Wonderful work! Your joy and enthusiasm is infectious! Merry Christmas Eve! I sincerely doubt there are many out there who would put in the level of care and commitment that you put in. It's great to see it!
Even 5 1/4" hard drives from the late 80s and early 90s sounded extremely similar, especially full-height ones. That remindes me I need to spin up my 650 MB external SCSI drive that I used for backing up all kinds of stuff from 68k Macs. I try to run that every once in a while to keep the grease from getting too stiff.
Very cool first start David! Your methodical approach to making sure all elements are actually _ready_ to have power put to them, is vital to getting a working machine in the endgame. I'm amazed that drum has actually retained the clock and timing signals after all these years. All in all, the drum doesn't sound any worse than the old Seagate ST 225 MFM drive in my old IBM AT did when I was running it regularly, and it worked fine! Looks like the caps in that tape drive are a little 'fuzzy'. Probably due for replacement for sure. The electrolytics in the DC bank may or may not come back with reforming.😉 It wouldn't be a crime to replace them with their modern counterparts though, for continued reliability.😊 Looking forward to more on this old tank!😄👍👍
To maintain visual authenticity in a restoration, Mr. Carlson (Mr. Carlson’s Lab) will sometimes replace components by hiding the modern equivalent either inside the original after he hollows out the guts, or else simply leaving the original in place, but taking it out of the circuit and relocating the replacement to an inconspicuous location, usually on the opposite (generally unseen) side of the circuit board or chassis.
For bearings - a lot of bearings are standardized, I doubt that they have used some custom size bearings. There should be a four digit number of the bearing like 6202 or a five digit number. Then there could be some letters like 2RS that indicates a sealed bearing. Try out a replacement on the damaged drum and see if it's feasible. It would be even cooler if you could get someone to resurface the damaged drum.
By coincidence, I visited the UK's National Museum of Computing at Bletchely Park last week and saw the Harwell Dekatron - later known as the Wolverhampton Instrument for Teaching Computing from Harwell (WITCH). It's the world's oldest digital computer, dating back to 1952, and mostly uses telephone exchange technology. Amazing to see it operating in person - if you ever get across to the UK, a visit to Bletchley Park and the NMoC is well worth the time!
Incredible and good work. Both for those designing this machine in the first place in the days, and to you David for the restauration of it! thank you for sharing this!
Wow, what an amazing project! It's baffling to see what engineering marvels were done back then. I had the privilege to see a working Zuse Z23 from 1961(?) executing code in March 2020, in Erlangen, Germany. Also with drum memory, slowly (and noisily) spinning up over a long time... That's the already transitorized variant of the Zuse Z22 (design 1955) that also uses vacuum tubes, like the G15. A Z22 still in working order exists in Karlsruhe, Germany.
congratulations! an exciting step forward for sure. and to all those folks who talk about selenium rectifiers failing and putting out nasty-smelling smoke: have you personally experienced this, or are you repeating something you heard from someone else? genuinely curious to hear from you IF you've personally experienced it.
I guess I remember selenium rectifiers like that from an old movie theater projection booth? Would they have used them? (The projectors originally had carbon arcs, but had long since been upgraded to 1970s era Xenon lamps). My recollection was that they were more or less out in the open where you could touch the plates if your were foolish enough.
I haven't personally seen one catastrophically fail and put out smoke but I have had a weak one which got hot and started to smell. If I let it continue I wouldn't be surprised if it let out all the smoke. I've also heard firsthand accounts of them putting out the nasty smoke (most recently from Bob Anderson @bandersentv) and I'm sure other channels working on old radios/TVs will have similar stories. You can probably find videos showing them failing, just like there are tons of vintage computer channels which have captured failing RIFA capacitors.
@@eDoc2020 Can confirm, selenium rectifiers were used in a lot of old radio equipment. They have a... unique smell when they get hot. Although my understanding is that selenium compounds are similar to sulfur compounds in that they tend to stink to high Heaven. But worse.
This is exciting for me too, because at Skyline High School ( Oakland California ) in 1969 and 1970, I got to play with the Bendix G-15. At that time, I was only interested in having fun, so my skill set never progressed beyond using INTERCOM 550. I had 1 hour each day, and ( somewhat sadly, in retrospect ) my time was spent writing what I called the " G-15 Trash Bulletin. " In the summer of 1969, I was also able to spend many hours , along with a friend, using the G-15 that was at the Chabot Science Center, located ( at that time ) on Mountain Blvd, just a short distance from my home. I am waiting anxiously for the next phase of this project !
Great progress, those bearings definitely need some love. Watch out for those orange selenium rectifiers in the paper tape mechanism, I’ve had many blow up and fill the room with acrid smoke
Wonderful! Very happy to see this working (to the extent it is.) Thanks for your effort on this machine! While I really respect your work here, I don't have much experience with vacuum tubes. Loving to learn from watching you. With mechanical things having bearings, I have worked on them quite a lot. The squeal from ball bearings is almost always due to sliding of the rolling elements due to dried grease. Since you can't get the bearings out of the drum to clean and service them; your idea of using a needle is a good idea. You may want to inject a light oil as opposed to grease, because pretty much any light oil (just a few drops) will help to dilute and soften the grease. Adding grease opens up the worry that it could be incompatible with the existing grease and worsen the problem. I've had good luck adding Lucas Chain lube (aerosol) to sealed bearings to help them out. I would spray it into a syringe and then use the needle to inject it drop at a time. Definitely don't risk spraying the aerosol near the drum and getting oily spots on the coating.
It’s Christmas day down here in New Zealand, I’ve got shingles and my Dad has had chemo so I’m staying away from my family and it’s just me & the cat. This video is the best present ever. Can’t wait to see it up and running, good you cut your teeth on the Centurion, this seems like a very well engineered piece if equipment and quite amazing such a complicated machine still functions after all these years. Merry Christmas to you and yours 🎅🎄💝
About the reforming of the caps, for the domestic devices I usually just use a HV DC bench power supply, limited to about 10mA, and slowly bring it up. I just connect it after the rectifier or anywhere the B+ is located (sometimes I just use a magic eye B+ line, as it is normally exposed). So I doubt you would need to take all the caps out. Also about the selenium rectifiers, they should also be dim-bulbed if possible. They are self-healing, but their healing current can be too large if not limited and they could destroy themselves. Despite what the internet says, those are usually okay. If worn out, their output voltage drops, like a worn rectifier tube... but this happens after 10-15k hours of use
This is like watching a critical scene from a classic science fiction movie, the one where the mad scientist accidentally unleashes hell with their half-finished machine. The only way to set things right is to complete the work. You must run the code to save humanity.
You know you have the good gear when it has oiling points and a sequencer to bring things up in order :) Love the AC adjust knob too. Computers now are so boring!
Awesome. I think I'm going to get to go to System Source when I pick up a car I purchased not far from it. Wish this was there so I could see it. Amazing work. BTW, I'd love a segment on some.of those crazy cars I see in your video like that Honda Beat too!
It's fascinating how the documents are using a bunch of terminology we use or at least understand in pretty much the exact same way today, like "CLOCK" and "SHIFT PULSE," chassis, pulse period, gate, bit, word, et c... Control Data Corporation apparently took over Bendix's computer line in 1963, so maybe they were still building G15s by at least that date.
It's like when people put thermal paste on a CPU ... Yep, we added exactly the amount you thought we ought to have and spread it in exactly the manner you would have preferred!
With sealed bearings that have a rubber shield,you can poke a syringe through the rubber,and squirt a little thin oil inside. Thicker oil can be hard to squirt through the needle of the syringe. If there's enough of a gap with a metal shield,sometimes you can use the syringe to squirt a little oil between the shield and inner race.
Hi! Maybe - once you finish your vacuum tube computer - you can program it yourself to do unit tests on the logic modules from the Bendix G15? Now THAT would be epic! Using one vacuum tube computer to test another vacuum tube computer! Worthy of the next VCF show! Of course, I don’t expect you to lug the entire G15 up there, just a copy of one of the (tube) logic modules. 😂
Merry Christmas! What a gift to watch this today (was traveling this past week). The G-15 series is definitely my favourite one on your channel at the moment, so I was SUPER pumped to hear it start turning on. Wonder how long it's been since anybody heard that on this machine?
you should be able to pop the black seal off the bearing with a pick. if u wanna clean the old grease out use some solvent, otherwise just pack in some new grease.
For the power that's easy! Just use half of a 220 breaker, OR, just a good old fashion screw in fuse. As long as the wire gauge is sufficient either will work fine.
Amazing project and what a milestone. I’ve had to restore a few irreplaceable sealed bearings and have found that with few exceptions you can usually pick the seals of the sides with something similar to dentist picks. The problem is to clean and regressed them you really need to get them out of the assembly (definitely don’t want to contaminate the media!)…. So if disassembly not an option I’m a +1 on a feeding a few drops of gear oil in sideways either through the shields or under a seal lip or with a hypodermic syringe. Though judging by oiler cap at least some are unsealed
It's machines like this that, while revolutionary for their time, show why the transistor was so incredibly important. Such high voltages for relatively low processing power, and so much consumed space basically nullified these machines for personal use beyond a terminal. They had to get way smaller, and the transistor is what allowed that to happen.
Those are selenium rectifiers recommend to change all of them because they will cause you problems in the long run . I’ve been in electronics for 57 years TV broadcast
120v at 50 amps is so weird. Did they not have 2 phase power as commonly back in the 50's? I just can't imagine a good reason to run it at 120v otherwise you can easily divide the 240v 2 phase inside the machine. Maybe when cost isn't a consideration the 120v 50amp was less work for them.
There might have been a few places where only 115v was available in the building but most would have 240 or 208. Relative to the cost of the machine it wouldn't cost much more to have multi-tap dual primary transformers which would enable operation at either of the three voltages with only some simple field wiring changes. It would complicate the power sequencer a bit bit that's not too difficult. Personally I would probably have designed it for 208/240v switchable input and if the customer wanted to run it on 115v only they could get their own step-up transformer. It probably would have been around $100 in 1950s money.
I think on the drum there are just evenly spaced magnetized "spots". The sine comes from the the ramp up and down of the induced field as a given spot moves by the read head.. The timing is needed when setting the track - you need to get the distance between the centers of the spots to evenly divide into the circumference of the disk. So some calculation with the rpm, diameter of the drum and speed of the writes is needed. Ultimately it is timing problem for the write head. Who knows they may have written one mark then used that to set the timing - Check how long it takes to repeat when the drum spins at speed then write every 1/3200 of that or however many they used on the disk.
Merry Christmas and a Happy New Year.. Nice bit of progress. Somebody may have already suggested this but since you have a readable clock and timing signals the first thing I suggest you do is using a decent logic analyser or even a modern scope to capture these signals as a group to serve as a backup incase the timing signal gets damaged. With the clock signal and its backup you should be able to re-write the timing signals onto the drum pretty easily with modern hardware if something happens.
Congratulations! You turned it on and it didn't go bang! Brave man. The only safety device on the supply side of the main switch was ..... ? Anyway, it's good to know you got it this far. For servicing there is a DC safety lockout switch, as I am sure you know, which prevents the DC Reset button doing anything when you don't intend it to. Looking at the G15 Theory of Operation docs, that procedure they promised on page 17 that Customer Engineering could supply for repairing a clock track in the field looks as though it was going to be limited to making a fresh copy of the clock track off the spare track, thus they would automatically be in sync. Making a fresh one from scratch, though ... ? Is the drum motor a synchronous motor relying on the AC mains frequency? I can't see anything else in the schematics. So could the motor be powered independently from a synthesised frequency supply? You quote the motor's specified rotational speed as 1800 rpm. If the clock has precisely 3596 pulses per revolution, and you were getting a frequency at the head output of 106.486kHz, then that drum must have been spinning at 1776.74 rpm. Is your grid AC supply frequency really so low at 59.22 Hz? I thought that the tolerance In the US was +/- 0.05Hz, and that is way off. A figure that low would normally indicate fairly severe grid overloading -- or is this symptomatic of the way you are powering the device? A slower clock speed will mean lengthened pulses in all of the system timing. Are you sure that will be within tolerance, as I see there are considerably precise expectations in the relationship between the various pulses? Might it be a good idea to see if you can drive that drum motor independently of the mains frequency?
Funny... that reset button is exactly the same make as the engine starter button in a Super Constellation airplane from 1955. Only the starter button is red, not green.
when you put the oil in, I suspect some of it was flung away from where it was needed to be, by centrifuge, before the bearings got a proper coating. Probably needed to manually roll the drum a bit before flicking it on. Sounds like it all settled in ok in the end. Never thought I'd be hooked on watching someone waving wrenches at computers. Remarkable channel; where electrickery meets mechanical engineering..
If you can pry out the seal on the outside of the drum you can then pack some new grease and pop the seal back in. Can you film the side of the drum, would like to see it. Thank you and mary Christmas.
And I thought the Centurion series was already fascinating... I never thought seeing such an old piece of tech comming back from the dead would be even more interesting. But it makes me also think... Even it seems a daunting task to restore such marvel, it is still something feasible. It is mainly question of mechanics, replacing dead components and greasing some motors. Of course, as you explain, some issues could be impossible to fix, like a dead read head, or if you lose the clock track. But basically, you might succeed to restore this thing. But what about modern tech? Imagine 50 years from now, a guy trying to make a laptop, a server or a smartphone back to life... That would just be impossible! Between components that you will never be able to reproduce, locked and encrypted software... Even today repairability of our modern tech is already difficult. So in the future? Forget about it. So let us all enjoy your restorations videos, because our kids won't have the chance to witness such thing...
If we're drawing 38 amps then code dictates that we would go for the next size up wire and breaker so 40A that would be the next common size. but then we also have to account for inrush and duty cycle and de-rating so my guess would be a 50A 120V outlet which is uncommon but does exist. however a more standard approach would be a 240V 50A outlet and only use one of the split phases as that would get the ground, neutral and one of the two hots.
Wow i did not realise this was the first computer Ken Thompson ever used. The change this machine made to the world is quite possibly incalculable, if its what got the guy who wrote Unix hooked
And it gets even crazier when you throw in that the designer, Harry Huskey, worked closely with Alan Turing and the G15 is inspired by Turing's ACE computer. This thing is definitely a historically significant machine, and I'm so excited to see it wake up and stretch its legs again!
It's also the first computer that Niklaus Wirth ever reliably used. Mr Wirth contribution to computer science is invaluable.
Bravo! Christmas a few days early! (Bob @ System Source)
I was a freshman at Rose Polytechnic Institute in Terre Haute, Indiana in 1963. (Since then the school name was changed to the Rose Hulman Institute of Technology.) Freshman arrived about two weeks before the rest. As we were on the guided tour of the “main” building, we walked by an enclosed room with a large glass window. We walked up to the window and peered into the school’s computer room. There, in all of its glory sat the school’s blue Bendix G15 computer. I had a class that taught us how to program the G15. Subsequently several other classes required that we create programs to solve problems.
This series of videos is a trip down ancient memory lanes. Thank you so much. By the time that I graduated other hardware had been purchased; however the G15 was still being used. Sigh!
I had my first experience with programming on one of these machines back in 1964 when I worked at a weather station in the Marshall Islands. I programmed the computer to process radiosonde readings that had previously been treated manually. Can't wait for you to get that paper tape reader going!
Love the way you trace schematics. It makes it much easier to understand and follow power and signals through circuits.
Thanks!
It definitely adds a lot of time to the editing process, but it's great to hear that it's worth it!
The documentation is beautiful in its own way. What always impresses me about these vintage computers (even 1970s and 1980s home computers) is how rapidly they seemed to be developed from concept to manufacture without (obviously) the benefit of modern computers to assist in designing them and managing their development. A lot of talent and hard work!
Drum bearings try to pull the drum out, and get it on it's side. Then take the bearing facing up, and warm it up a little after cleaning, and apply some EP90 gear oil to the edge of the steel dust shield, to make a thin layer on the seal, and let the bearing cool down, which should allow some to enter past the seal, and mix with the old grease there and revive it. Leave overnight with it up, and next day clean off the excess, and turn so the other side is up, and do the same warm up to around 70C, and apply the oil, and allow to cool and pull some past the seal. Quick fix is to make a sort of dam there if not moving the drum, and fill with the oil, and leave the bearing half covered overnight, so it can seep past the dust shield into the grease there, and give it some lighter oil components again, and reconstitute it a little.
Otherwise you need to gently pry the shield off, and then pack with a little moly based grease, and put the cover back on. Look for the same bearings, and order a cheap one, and sacrifice the bearing, using a grinder, to get the shields off them, and thus get new ones if you damage the old ones pulling them off. Have done this to a lot of bearings, especially where a new one was either not available, or needed a lot of specialised tooling not available to set the preloads, and the lube would get you another few years of running out of them.
I appreciate your proper caution in the power-on testing. You are saving yourself much potential grief. Seeing the filament load decrease as each voltage step settles was fun. The heaters increase in resistance as the warm up and that causes the voltage to ramp betweek the steps. Do consider powering the DC supplies externally the first time and ramp up the applied AC voltage slowly to re-form the capacitors and limits any surge current from unforeseen faults. I know you will have already checked the resistance across the DC main busses as the first step. The ghosts of Engineers Past will be watching over you as you proceed 😀
The "bearing noise" might be a dragging co-axial brake on the drum motor. Marvelous to watch it come up again. Wish the yUK Science Museum would find the courage to try the same on their Ferranti Pegasus. Good luck with the Sprague condensors (capacitors) as those are going to be difficult to replace if they are sickly.
It also might be a dragging brush. That style of GE Repulsion Induction motor starts with carbon brushes and switches to induction at speed. There is a centrifugal clutch that pulls the brushes away after starting, and that might be gummed up.
That style of capacitors is still being made, by more than one company, so getting replacements is not that big of a problem...
@@PrebleStreetRecordsthat would explaine the noise going aways after a while of running
It's been a long time since "lubricate the memory" has been part of the PM schedule for a computer...
19:26 that's an amazingly clean signal, after over sixty years! 😍
CONGRADULATIONS!
A Big first step, glad the drum is spinning and talking! Harry Huskey has got to be smiling also! Wish I still had some of my G15 programs I wrote in the day! Looked for them but no luck! Glad that you are taking slow steps to bring the baby up and working again!
I want to comment on your channel in general. Seeing the various projects and working through them "with" you is like a refresher course in the old technology. Seeing you interact with those such as Curious Marc and building a comaradare with them, priceless!
I would suggest connecting the clock track to one channel and the timing track to the other channel of your DSO and capturing the waveform with the highest memory depth your scope supports. That way you have a backup that could potentially be used to recover it if anything happens.
I think this is a very, very good idea.
I'm surprised this wasn't done while it was running
@@JoshVennixit may well have been off camera
I second this. That way, if worst comes to worst you have a backup, and you can use it to restore the other drum when you get around to resurfacing it.
It's an excellent idea!
However, the clock track is just a simple sine wave with 3,596 pulses per rotation. That can be recreated with any signal generator with ease. The timing track is a little more tricky, but again, we know exactly what it should look like thanks to the documentation. The really tricky part when rewriting those tracks is synchronizing it perfectly to the rotational speed of the drum and motor. At a perfect 60Hz with a new motor and good bearings, it should rotate at exactly 30 Revolutions Per Second or 1,800 RPM. That should result in a clock track frequency of 107.8kHz. But, as we saw on ours, we were rotating a little slow. That means that even if we got an exact copy of the timing track, laying it back down would be not quite possible if the RPM of the drum has changed even slightly.
Knowing exactly what the timing track and clock track look like means we can drop a hall-effect sensor on the drum and use a microcontroller to perfectly synchronize everything and drop new tracks down if we need to. So, fret not, we're well equipped to replicate those tracks if anything ever happens!
Congratulations and Merry Christmas.
Amazing that the drum still contains the timing tracks after all these years.
Magnetic media is really quite reliable, despite claims to the contrary. As long as the material doesn't get heated past its Curie temperature, doesn't get exposed to strong magnetic fields, or has some ridiculously thin/unreliable coating, any stored data should last for a very long time. In the case of the Bendix I think they also went out of their way to ensure long-term reliability, hence the spare clock track.
There exists a number of hard disks there the timming data ie servo is on a separate platter - bad servo data platter - that disk is a boat anchor.
AND a number of them (ST256 ? ie 40 years?) STILL WORKS TODAY !!!
👏👏👏🤩🤩🤩👍👍👍⭐️⭐️⭐️🌈🌈🌈 Maybe if you sing japanese children’s song to it the bearings will become less angry ;-)
My nerves were jangling when you went for the Big Power, ye gods. I never thought I'd see a G15 ever, aside from history books, and suddenly there's 2 and you have one and it's on the way do being powered up! Awesome. All the best to you and Mrs Usagi from across the little pond.
For those Siglent scopes, if you want to see the full sequence, you'll need to zoom out and stop the capture, then you can zoom back in and scroll across until you find the start of the sequence. You can also safely enable the 20Mhz bandwidth limit to help reduce the noise on the trace, and you should also be able to increase the number of points that it captures (although that may decrease the update frequency at the expense of a nicer stopped trace).
That said, I'm hugely relieved that you're seeing the expected signals and also filament voltage. This definitely bodes well for the future of this machine. Those definitely look like selenium rectifiers, although I'm not hugely familiar with these early semiconductors. I feel like silicon diodes weren't that common until sometime in the 60s? Maybe be prepared with some extra ventilation just in case. Apparently selenium diodes smell pretty bad when they blow.
I know the feeling of excitement when bringing up a piece of equipment that hasn’t been used well over 40 or 50 years
Back in 2001 I ran an old Japanese Punch press machine from the early 60a. The computer absolutely reminds me of that thing. Same color too
Pretty freaking great. Last night you're singing from the moon, today you're resurrecting this behemoth.
Thanks for the work preserving history and thanks for bringing us so much joy.
I can totally see him hook up a teletypewriter to it, turn it on, install a short paper tape in the reader, hit start, the machine reads the paper tape and the teletype spits out "HELLORLD"
I heard some of the cool kids are coding up "HELLORLD!".
This is such an exciting project! I love seeing the Bendix G15 come to life. It scares me a little bit to see computers that were built during my lifetime being displayed in computer museums. I was 2 years old when this Bendix was built, but I've worked on some of it's descendants including the PDP-8 and PDP-11 both at the DEC factory and in the field.
I'm really glad you went with solid 2AWG gold and titanium wires with dupont connectors as I would wanted. Better keep the fire extinguisher handy when in use tho
my first job at Control Data (1967) was in the systems refurbishment group (G15's, G20's, 160''s 160A's 8090's 1604's etc). I'd like to tell you that I remember everything about a G15 but I don't, wait I do remember 1 thing... if it's not working, open the doors, turn off the light and look for the blue tube and replace it. in 1968 the US government 31 G15 under contract.
Great job! Looking forward to the next installment.
> Beware, I live
> I hunger
The bearings should be fairly simple to source. I can make out 105K... on the dust shield, which probably means a metric 25x47x12 (but might not)
4:44 thank you and congrats on your safe and reliable electrical installation
So many selenium rectifiers - so many opportunities for really acrid smoke release!
Great video! Your videos encouraged me about a year ago to fire up the PDP-11/23+ and large drives. Had a few bad Sprague caps I had to replace. Hopefully, you will show your process for reformatting the caps in the G-15. Unlike when I powered the PDP-11, I want to test the caps in my Vax 11/30 next before power up which had not been powered up since I purchased it off a BBS swap page in 1994. I am also worried that when I moved, I didn’t lock down the heads. Thanks for the videos.
Not gonna lie, those are some beautiful looking clock tracks on the scope.
Here I thought my boss's 1984 Ikegai AX25 was the loudest machine around! Get some ear protection, better safe. Really cool seeing this fired up, thanks for all the effort. Merry Christmas & here's to a healthy 2024!
Wonderful work! Your joy and enthusiasm is infectious!
Merry Christmas Eve!
I sincerely doubt there are many out there who would put in the level of care and commitment that you put in. It's great to see it!
That thing spinning up sounds *very* similar to what our old CDC Phoenix on the Wang 2200 MVP used to sound like.
Good times. 🥳
Even 5 1/4" hard drives from the late 80s and early 90s sounded extremely similar, especially full-height ones. That remindes me I need to spin up my 650 MB external SCSI drive that I used for backing up all kinds of stuff from 68k Macs. I try to run that every once in a while to keep the grease from getting too stiff.
That face of pure joy in the thumbnail needs to become an emote on the discord I think
Very cool first start David! Your methodical approach to making sure all elements are actually _ready_ to have power put to them, is vital to getting a working machine in the endgame. I'm amazed that drum has actually retained the clock and timing signals after all these years. All in all, the drum doesn't sound any worse than the old Seagate ST 225 MFM drive in my old IBM AT did when I was running it regularly, and it worked fine! Looks like the caps in that tape drive are a little 'fuzzy'. Probably due for replacement for sure. The electrolytics in the DC bank may or may not come back with reforming.😉 It wouldn't be a crime to replace them with their modern counterparts though, for continued reliability.😊 Looking forward to more on this old tank!😄👍👍
To maintain visual authenticity in a restoration, Mr. Carlson (Mr. Carlson’s Lab) will sometimes replace components by hiding the modern equivalent either inside the original after he hollows out the guts, or else simply leaving the original in place, but taking it out of the circuit and relocating the replacement to an inconspicuous location, usually on the opposite (generally unseen) side of the circuit board or chassis.
Adrenaline rush for that power-on was awesome.
I feel like a wizard's apprentice barely understanding the basics of what my teacher is saying while he shows me barely comprehensible arcane runes.
Be aware that heating and cooling the filaments of the tubes has to be done slowly to prolong their life.
Feliz Navidad a todos. Y enhorabuena por tu excelente trabajo en el canal. Me encantan tus contenidos. 👌🏼
For bearings - a lot of bearings are standardized, I doubt that they have used some custom size bearings. There should be a four digit number of the bearing like 6202 or a five digit number. Then there could be some letters like 2RS that indicates a sealed bearing.
Try out a replacement on the damaged drum and see if it's feasible.
It would be even cooler if you could get someone to resurface the damaged drum.
Great job. I'd recommend a hazmat suit before hitting that reset button.
By coincidence, I visited the UK's National Museum of Computing at Bletchely Park last week and saw the Harwell Dekatron - later known as the Wolverhampton Instrument for Teaching Computing from Harwell (WITCH). It's the world's oldest digital computer, dating back to 1952, and mostly uses telephone exchange technology. Amazing to see it operating in person - if you ever get across to the UK, a visit to Bletchley Park and the NMoC is well worth the time!
Incredible and good work. Both for those designing this machine in the first place in the days, and to you David for the restauration of it! thank you for sharing this!
Re-wiring a room has it's positives and negatives. But, if you stay grounded you should see success.
Wow, what an amazing project! It's baffling to see what engineering marvels were done back then. I had the privilege to see a working Zuse Z23 from 1961(?) executing code in March 2020, in Erlangen, Germany. Also with drum memory, slowly (and noisily) spinning up over a long time... That's the already transitorized variant of the Zuse Z22 (design 1955) that also uses vacuum tubes, like the G15. A Z22 still in working order exists in Karlsruhe, Germany.
congratulations! an exciting step forward for sure. and to all those folks who talk about selenium rectifiers failing and putting out nasty-smelling smoke: have you personally experienced this, or are you repeating something you heard from someone else? genuinely curious to hear from you IF you've personally experienced it.
I guess I remember selenium rectifiers like that from an old movie theater projection booth? Would they have used them? (The projectors originally had carbon arcs, but had long since been upgraded to 1970s era Xenon lamps). My recollection was that they were more or less out in the open where you could touch the plates if your were foolish enough.
I haven't personally seen one catastrophically fail and put out smoke but I have had a weak one which got hot and started to smell. If I let it continue I wouldn't be surprised if it let out all the smoke. I've also heard firsthand accounts of them putting out the nasty smoke (most recently from Bob Anderson @bandersentv) and I'm sure other channels working on old radios/TVs will have similar stories. You can probably find videos showing them failing, just like there are tons of vintage computer channels which have captured failing RIFA capacitors.
@@eDoc2020 thanks. Bob would be a good person to talk to.
@@eDoc2020 Can confirm, selenium rectifiers were used in a lot of old radio equipment. They have a... unique smell when they get hot. Although my understanding is that selenium compounds are similar to sulfur compounds in that they tend to stink to high Heaven. But worse.
As I watch this, I have realized that the heater voltage being ramped up like it is will also extend the life of the tubes greatly.
This is exciting for me too, because at Skyline High School ( Oakland California ) in 1969 and 1970, I got to play with the Bendix G-15. At that time, I was only interested in having fun, so my skill set never progressed beyond using INTERCOM 550. I had 1 hour each day, and ( somewhat sadly, in retrospect ) my time was spent writing what I called the " G-15 Trash Bulletin. "
In the summer of 1969, I was also able to spend many hours , along with a friend, using the G-15 that was at the Chabot Science Center, located ( at that time ) on Mountain Blvd, just a short distance from my home. I am waiting anxiously for the next phase of this project !
To see that sine wave come up! WOW! So impressed with your work- your excitement is CONTAGIOUS! can't wait to see the next steps! 👍😁
Great progress, those bearings definitely need some love. Watch out for those orange selenium rectifiers in the paper tape mechanism, I’ve had many blow up and fill the room with acrid smoke
A super Christmas present indeed! Wishing a very Merry Christmas and Happy New year to you and yours!
Wonderful! Very happy to see this working (to the extent it is.) Thanks for your effort on this machine!
While I really respect your work here, I don't have much experience with vacuum tubes. Loving to learn from watching you. With mechanical things having bearings, I have worked on them quite a lot. The squeal from ball bearings is almost always due to sliding of the rolling elements due to dried grease. Since you can't get the bearings out of the drum to clean and service them; your idea of using a needle is a good idea. You may want to inject a light oil as opposed to grease, because pretty much any light oil (just a few drops) will help to dilute and soften the grease. Adding grease opens up the worry that it could be incompatible with the existing grease and worsen the problem. I've had good luck adding Lucas Chain lube (aerosol) to sealed bearings to help them out. I would spray it into a syringe and then use the needle to inject it drop at a time. Definitely don't risk spraying the aerosol near the drum and getting oily spots on the coating.
Edge of the seat stuff! Brilliant video, thank you!
The fact that anything worked at all is absolutely mind-boggling!
It’s Christmas day down here in New Zealand, I’ve got shingles and my Dad has had chemo so I’m staying away from my family and it’s just me & the cat. This video is the best present ever. Can’t wait to see it up and running, good you cut your teeth on the Centurion, this seems like a very well engineered piece if equipment and quite amazing such a complicated machine still functions after all these years. Merry Christmas to you and yours 🎅🎄💝
About the reforming of the caps, for the domestic devices I usually just use a HV DC bench power supply, limited to about 10mA, and slowly bring it up. I just connect it after the rectifier or anywhere the B+ is located (sometimes I just use a magic eye B+ line, as it is normally exposed). So I doubt you would need to take all the caps out. Also about the selenium rectifiers, they should also be dim-bulbed if possible. They are self-healing, but their healing current can be too large if not limited and they could destroy themselves. Despite what the internet says, those are usually okay. If worn out, their output voltage drops, like a worn rectifier tube... but this happens after 10-15k hours of use
This is like watching a critical scene from a classic science fiction movie, the one where the mad scientist accidentally unleashes hell with their half-finished machine. The only way to set things right is to complete the work. You must run the code to save humanity.
You know you have the good gear when it has oiling points and a sequencer to bring things up in order :) Love the AC adjust knob too. Computers now are so boring!
Awesome. I think I'm going to get to go to System Source when I pick up a car I purchased not far from it. Wish this was there so I could see it. Amazing work. BTW, I'd love a segment on some.of those crazy cars I see in your video like that Honda Beat too!
I passed up orange cat videos to watch this. I was not disappointed.
You can pull those seals and grease no problem, carefully and with the right tool you wont damage them, and can be pushed back in
It's fascinating how the documents are using a bunch of terminology we use or at least understand in pretty much the exact same way today, like "CLOCK" and "SHIFT PULSE," chassis, pulse period, gate, bit, word, et c...
Control Data Corporation apparently took over Bendix's computer line in 1963, so maybe they were still building G15s by at least that date.
Absolute gut laugh @ 4:45
I laughed my butt off at that joke too xD
It's like when people put thermal paste on a CPU ... Yep, we added exactly the amount you thought we ought to have and spread it in exactly the manner you would have preferred!
That’s one of my favorite recent examples of “I’m not getting into this with all of you”.
Merry Christmas from the UK!
With sealed bearings that have a rubber shield,you can poke a syringe through the rubber,and squirt a little thin oil inside. Thicker oil can be hard to squirt through the needle of the syringe. If there's enough of a gap with a metal shield,sometimes you can use the syringe to squirt a little oil between the shield and inner race.
Hi! Maybe - once you finish your vacuum tube computer - you can program it yourself to do unit tests on the logic modules from the Bendix G15? Now THAT would be epic! Using one vacuum tube computer to test another vacuum tube computer! Worthy of the next VCF show! Of course, I don’t expect you to lug the entire G15 up there, just a copy of one of the (tube) logic modules. 😂
Merry Christmas! What a gift to watch this today (was traveling this past week).
The G-15 series is definitely my favourite one on your channel at the moment, so I was SUPER pumped to hear it start turning on. Wonder how long it's been since anybody heard that on this machine?
Adding grease might help, but once the bearing is making noise like that it means the races are damaged.
Woot! That's fantastic! Here's hoping that the rest goes smoothly!
you should be able to pop the black seal off the bearing with a pick. if u wanna clean the old grease out use some solvent, otherwise just pack in some new grease.
For the power that's easy! Just use half of a 220 breaker, OR, just a good old fashion screw in fuse. As long as the wire gauge is sufficient either will work fine.
Amazing project and what a milestone. I’ve had to restore a few irreplaceable sealed bearings and have found that with few exceptions you can usually pick the seals of the sides with something similar to dentist picks. The problem is to clean and regressed them you really need to get them out of the assembly (definitely don’t want to contaminate the media!)…. So if disassembly not an option I’m a +1 on a feeding a few drops of gear oil in sideways either through the shields or under a seal lip or with a hypodermic syringe. Though judging by oiler cap at least some are unsealed
"I am not crazy"
Well, you are. But the good type of crazy :)
Merry Xmas
It's machines like this that, while revolutionary for their time, show why the transistor was so incredibly important. Such high voltages for relatively low processing power, and so much consumed space basically nullified these machines for personal use beyond a terminal. They had to get way smaller, and the transistor is what allowed that to happen.
Those are selenium rectifiers recommend to change all of them because they will cause you problems in the long run . I’ve been in electronics for 57 years TV broadcast
120v at 50 amps is so weird. Did they not have 2 phase power as commonly back in the 50's? I just can't imagine a good reason to run it at 120v otherwise you can easily divide the 240v 2 phase inside the machine. Maybe when cost isn't a consideration the 120v 50amp was less work for them.
There might have been a few places where only 115v was available in the building but most would have 240 or 208. Relative to the cost of the machine it wouldn't cost much more to have multi-tap dual primary transformers which would enable operation at either of the three voltages with only some simple field wiring changes. It would complicate the power sequencer a bit bit that's not too difficult. Personally I would probably have designed it for 208/240v switchable input and if the customer wanted to run it on 115v only they could get their own step-up transformer. It probably would have been around $100 in 1950s money.
Oiling your computer .... who'd ever thought that this was a thing ... :D
I oil my organ. Hammond tone wheel organs had oil cups that fed into the mechanical bearings and had to be kept filled.
Amazing! How did they get the sine wave to not have a gap on the drum? The recording of the clock track must have been extremely precise.
I think on the drum there are just evenly spaced magnetized "spots". The sine comes from the the ramp up and down of the induced field as a given spot moves by the read head.. The timing is needed when setting the track - you need to get the distance between the centers of the spots to evenly divide into the circumference of the disk.
So some calculation with the rpm, diameter of the drum and speed of the writes is needed. Ultimately it is timing problem for the write head. Who knows they may have written one mark then used that to set the timing - Check how long it takes to repeat when the drum spins at speed then write every 1/3200 of that or however many they used on the disk.
if the drum motor is synchronous to wall AC, then any signal derived from wall AC frequency will naturally fit on the drum perfectly without any gap.
Awesome work! I'd love to see a watt meter on that outlet.
IT’S WHISPER QUIET!!
Have a happy holidays
Merry Christmas and a Happy New Year.. Nice bit of progress. Somebody may have already suggested this but since you have a readable clock and timing signals the first thing I suggest you do is using a decent logic analyser or even a modern scope to capture these signals as a group to serve as a backup incase the timing signal gets damaged. With the clock signal and its backup you should be able to re-write the timing signals onto the drum pretty easily with modern hardware if something happens.
Congratulations! You turned it on and it didn't go bang! Brave man. The only safety device on the supply side of the main switch was ..... ? Anyway, it's good to know you got it this far. For servicing there is a DC safety lockout switch, as I am sure you know, which prevents the DC Reset button doing anything when you don't intend it to.
Looking at the G15 Theory of Operation docs, that procedure they promised on page 17 that Customer Engineering could supply for repairing a clock track in the field looks as though it was going to be limited to making a fresh copy of the clock track off the spare track, thus they would automatically be in sync. Making a fresh one from scratch, though ... ? Is the drum motor a synchronous motor relying on the AC mains frequency? I can't see anything else in the schematics. So could the motor be powered independently from a synthesised frequency supply? You quote the motor's specified rotational speed as 1800 rpm. If the clock has precisely 3596 pulses per revolution, and you were getting a frequency at the head output of 106.486kHz, then that drum must have been spinning at 1776.74 rpm. Is your grid AC supply frequency really so low at 59.22 Hz? I thought that the tolerance In the US was +/- 0.05Hz, and that is way off. A figure that low would normally indicate fairly severe grid overloading -- or is this symptomatic of the way you are powering the device?
A slower clock speed will mean lengthened pulses in all of the system timing. Are you sure that will be within tolerance, as I see there are considerably precise expectations in the relationship between the various pulses? Might it be a good idea to see if you can drive that drum motor independently of the mains frequency?
Funny... that reset button is exactly the same make as the engine starter button in a Super Constellation airplane from 1955. Only the starter button is red, not green.
That looks like an Australian AC power outlet!
when you put the oil in, I suspect some of it was flung away from where it was needed to be, by centrifuge, before the bearings got a proper coating. Probably needed to manually roll the drum a bit before flicking it on. Sounds like it all settled in ok in the end.
Never thought I'd be hooked on watching someone waving wrenches at computers. Remarkable channel; where electrickery meets mechanical engineering..
oooooh a new vid on christmas eve, i am glad to see den bendix :) ... i whish you and your family nice holydays!
Merry Christmas!
excellent machine.. doesn't look like it would be hard to emulate that drum.
If you can pry out the seal on the outside of the drum you can then pack some new grease and pop the seal back in. Can you film the side of the drum, would like to see it. Thank you and mary Christmas.
And I thought the Centurion series was already fascinating... I never thought seeing such an old piece of tech comming back from the dead would be even more interesting. But it makes me also think... Even it seems a daunting task to restore such marvel, it is still something feasible. It is mainly question of mechanics, replacing dead components and greasing some motors. Of course, as you explain, some issues could be impossible to fix, like a dead read head, or if you lose the clock track. But basically, you might succeed to restore this thing. But what about modern tech? Imagine 50 years from now, a guy trying to make a laptop, a server or a smartphone back to life... That would just be impossible! Between components that you will never be able to reproduce, locked and encrypted software... Even today repairability of our modern tech is already difficult. So in the future? Forget about it. So let us all enjoy your restorations videos, because our kids won't have the chance to witness such thing...
WESOŁYCH ŚWIĄT 🎄❤
sounds like you have a good year of work ahead in 2024
If we're drawing 38 amps then code dictates that we would go for the next size up wire and breaker so 40A that would be the next common size. but then we also have to account for inrush and duty cycle and de-rating so my guess would be a 50A 120V outlet which is uncommon but does exist. however a more standard approach would be a 240V 50A outlet and only use one of the split phases as that would get the ground, neutral and one of the two hots.
Truly amazing engineering. But that sea of can caps looked truly scary to me. More often than not, problematic after aging for so many decades.
WOW ! Super amazing !
But, when you will bring up the DC voltage, it will run Crisys ?
congrats
Merry Christmas from Finland🇫🇮🎄