Removing the Rotating Drum Memory from the Bendix G15
Вставка
- Опубліковано 30 тра 2024
- In the previous episode, we discovered that our drum had unfortunately suffered multiple head crashes, stopping us dead in our tracks. We’re left with two options: recoat the drum or replace the drum. It just so happens that Bob has a second G15 with a non-crashed drum in it. So, we’ll start with the replacement option, and then once we have one G15 up and running, we’ll get to work on the recoating option. But both options require the drum to be removed, so let’s get to it!
Check out System Source here:
museum.syssrc.com/
If you want to support the channel please hop over to Patreon:
/ usagielectric
Also, we now have some epic shirts for sale!
my-store-11554688.creator-spr...
Come join us on Discord!
Discord: / discord
Intro Music adapted from:
Artist: The Runaway Five
Title: The Shinra Shuffle
ocremix.org/remix/OCR01847
Thanks for watching!
Chapters
0:00 Where did we leave off?
3:41 First things first, a little cleaning
5:23 In-depth look at the Relay and Driver Chassis
8:55 Cool little breakout panel for testing
10:37 Removing the drum
14:22 A better look at the drum on the bench
15:35 A grease trap?
16:38 Still crashed though, so what next?
17:30 What about covering the drum in tape?
20:31 I still think recoating is the best chance for success
22:25 But that’s in the future!
23:14 Bunny! - Наука та технологія
Hi all! Thank you so much for the amazing comments!
I was out of town yesterday, so I'm still trying to catch up, but there are two things I want to address overall right quick.
1. I want to reiterate that as of next month, I will no longer be in possession of this specific drum, it will be in the machine at System Source museum on display. All of the suggestions of companies and people to reach out to are greatly appreciated and I'm making a list, but I won't be reaching out to anyone until I'm sure on a timeline as to when the drum will be back in the shop here. Rest assured, I'm not ignoring your suggestions!
2. I of course would first build a replica of the drum to test things on. Probably a chunk of aluminum with the same diameter, but maybe only 1/4 as wide. This way I can test application methods, machining methods, grinding/polishing methods, etc. without worrying about damaging the original drum. Once we have a workflow hammered out on the test piece, we'll move over to work on the original drum. The goal is to do all of our testing on a sacrificial piece, so that once we get to the original drum, we only have to work on it once.
Maybe. And it’s just a random idea I thought of watching a video about preserving audio tapes in a huge bunker. I saw brands like 3M ,BASF and Scotch tape.
I thought. You might want to contact one of them and ask what they used back in the day.
I spoke with an expert on baked enameling (a process you should check out b/c it uses ionization, there are parallels to tubes) anyway, he said it was most likely electro plated.
I still think the baked enamel process might work better b/c you don't want an electrically conductive film, this guy is an expert on paint, but he's not into electronics, so I couldn't ask him about conductivity)
Your stuff is cool! I sure wish I had a shop like yours :)
Another idea, less destructive.. when you get the chance... the drum surface is 80% ok and at the correct level.
Really what is needed here is a 'filling of the gaps' type operation, perhaps with a flat stainless scraper and some iron oxide paint or paste, you could simply fill in the grooves.
Not saying it'll work - these things are always imagined to be easier than they are, but perhaps a car paint specialist could advise.
The recoating process sounds like something Ben Krasnow over at Applied Science might be into, or at least able to provide guidance for. He's spin coated all sorts of photoresists and so on for photolithography etc and is an outstanding engineer in all sorts of relevant fields. AFAIK he lives in the US, so maybe give him a yell when it comes time to put a plan together?
electroplating might be your best solution. When scientists and engineers experimented with tube computers back in the 50es and 60es in Hungary, they used rotating drum memories just like this computer did. When they had head crashes, they stripped the coating off the drum on a very precise lathe, and used electroplating to deposit a new layer. As i read, they did not use ferrous oxide, they used some king of chromium salt, soluble in water. The finish ended up to be very strong and durable, but needed higher magnetizing currents. The article did not tell the exact composition, i guess it was a trial and error thing. In that era, every computer was an unique design, especially in Eastern Europe.
I'm trying to remember what I knew about magnetic tape. If my 73 year old memory serves me right, the magnetic material in tape was gamma ferris oxide that was laid down on the tape in a solvent and the gamma ferric oxide was oriented using a magnetic field perpendicular to the tape surface. If you have a random orientation of the ferric oxide, then the signal level might not be high enough.
I would bet the magnetic coating was oriented in a strong magnetic field during manufacturing and is not random. Video tape certainly was.
You may want to pick up a paint thickness gauge before you put the recoated drum back in the machine. This will let you test the consistency across the drum surface, so you can rule out high spots causing crashes. These are commonly used by detailers, so they know how much paint they can safely remove when polishing.
An automotive paint thickness gage wouldn't work. The drums surface is conductive. We're talking thousands of an inch not tenths.
edit:
The paint thickness gauges are more accurate than i thought.
It only works if the metal part below the paint is magnetic. The drum is made from aluminium. This won't work
it would be better to simply make a repeatable jig the drum can be placed in and use a tenths (ten-thousanths) indicator. machining solved this problem a long time ago. if it really needs to be repeatable a couple of centres could probably be drilled in the axle of the drum (or may exist already). these are definitely a good idea through the whole recoating process for maintaining excellent concentricity
(:
Why couldn't they just use the lathe to cut the final coat down to the perfect level?
I *love* your idea for recoating the surface of the drum. Just a few thoughts:
1. When you machine off the old oxide you may find the drum is a bit oddly shaped. Unless you know for sure that the drum was machined while mounted to the spindle that it's currently on, there may have been some misalignment between the center of the drum and the center of the shaft when the two pieces were brought together. If that happens then you'll need a plan for dealing with that
2. Please be sure to keep the drum rotating at a slow speed while it dries COMPLETELY. You don't want any of the new layer flowing due to gravity after the coating is complete
3. Practice makes perfect, so if you can find a scrap piece of aluminum about the right size that you can machine first then that would be ideal
I hope for all the best for you. You are a much better and braver man than I. 🙂
Very good comments. One thing that worries me a little about this plan is that the density of magnetic material might be too low. Perhaps the only way to find out is to try it. Remember that what you need is not just magnetic material, but reversible magnetic _domains_. That's what the iron gives you, but not just any old iron. If it's been treated in the wrong way (e.g. heated to very high temperature), it won't have the domains, and it just won't work. Also, I suspect you will need to adjust the heads afterwards, which could be tricky; one twitch and you've crashed the heads again (wince).
@@therealdebaterApply layer by layer? slowly / gently?
@@therealdebater Any old iron? ua-cam.com/video/1m6RlCSeQUc/v-deo.html :)
I think it actually has to be centered on its spindle. If it wasn't the gap under the read/write heads would be constantly changing. It would either crash from the very start or move away from the heads and be too weak of a signal in certain sectors.
Yeah, was going to say, you 100% need to keep the drum spinning while the paint cures.
One option you might want to consider is making a stand-in drum memory from scratch - something you can use to refine the process for cleaning/coating/etc without impacting the original while you work out the details. If you can get THAT working (and clean/re-do the coating multiple times), then you will have a fairly solid process before proceeding to the real one.
Primary concerns on the mechanical side would be dealing with runout from re-mounting things (so turning between centers, and indicating things in REALLY WELL, including having a perfectly "leveled" lathe).
I'm REALLY excited to see this happen, it combines some of the coolest electro-mechanical problems from the era!
You're absolutely right that the most difficult part is going to be the mechanical side of things!
Once we've got a consistent process figured out for that, we can move on to figuring out the correct concoction of magnetic material to spray down. We'll definitely be tackling this first with a replica of the drum that has the same 13" diameter but is maybe only 3" wide, so we can test and iterate ideas and processes really quickly.
It'll be a serious challenge, but I think we can get there!
The right guy to ask would be Ben from Applied Science here on UA-cam. He may not have any experience with coating drums as such, but he did a lot of processes involving coating and chemistry with similar requirements.
There were dozens of comments about that on the last video that he ignored, so it's probably not gonna happen
As for the drum rebuild: I would strongly suggest to get the drum CNC-Ground down to H6 or better tolerances with a Ra of ~0.2µm or better and a cylindricity of 0.002mm or better. The better your starting point, the higher the chances of the rebuild working. The coating itself should be done using the correct binder. perhaps you can find an old IBM patent somewhere describing the binders used in drum or disk coatings. Also make sure to properly measure the coating thickness. If you had the money I would suggest getting the coating sputtered on using a modern magnetron-sputtering system. I would definitely advise against using a lathe. Especially if it hasn't been properly trammed and aligned within itself and to gravity using a machinists level.
I was going to say an ordinary lathe certainly wouldn't be good enough. A drum like this would most certainly have been ground between centers on a cylindrical grinding machine.
I don't think sputtering would work to apply this type of compound. However, your first point is right--this is not a job for a conventional lathe. In fact, I would try to avoid machining the drum at all, as the difficulty of getting the necessary tolerances are not at all trivial.
I would suggest replacing the bearings before doing anything. It would be tragic if after all the work to recover the drum the 70 year old bearings went and you where back to square one.
The coating used is also going to probably have to be very similar to the original one in terms of coercivity etc... or you will need to do things like adjust all the head amp electronics to get reliably recorded data. You can't just slap on any random magnetic coating and expect it to work with the existing read/write bias and level setup..etc.
@@maskddingo1779 In this regard, I think he has going for him the fact that this drum was built in the 1950s, in the early days of magnetic recording, when there wasn't a huge variety of coatings available. My guess is a basic standard-bias audio tape recipe using acicular gamma ferric oxide will work. But, like you say, more "modern" formulations for high-bias tape, videotape, etc., might not work at all.
I am sure spraying is how the drum was originally coated. However, I think high-build primer and automotive clear coat will not be a good choice. If you have only 0.001" to play with, you want a coating that is designed to leave a very thin layer. Both the coatings you mentioned are exactly the opposite--intended to build up thick layers. I think you will find it almost impossible to get a variation in thickness of less than 0.001" with those coatings, no matter how much you polish them.
Maybe a very thin nitrocellulose lacquer, applied with an airbrush? Has the advantage of being easily removable with lacquer thinner if you mess it up.
Also, as long as the drum is already on a precision spindle, I would not take it off and move it to a lathe. A lathe will not, in any case, give you the surface finish you need. Almost certainly that is a ground and polished finish. I would think the best way to restore it would be to spin the drum on its bearings while applying successively finer grits of abrasive papers, films, and polishing compounds.
couldnt it be anodized ?
@@mm-hl7gh I don't see how anodizing would apply a magnetic coating.
Thinking about this a little further, DEFINITELY DO NOT attempt to remachine the drum on a conventional lathe. And I would strongly discourage you from removing the drum from its bearings at all.
If the heads are flying at a height of 0.001, you need to hold roundness and concentricity of 0.0001 or better. That's 4 microns, for the metrically inclined. Maybe Robin Renzetti can hold those tolerances--YOU CAN'T. If you take this drum off its bearings and turn it in the lathe, you will leave it with more runout and a less flat surface than it has now.
Your best option is probably to try to remove the existing coating chemically. Obviously, test any chemicals on scrap aluminum before applying to the actual drum. With luck, you will find that the gouges that have affected the magnetic coating don't penetrate to the aluminum surface.
I would try, in increasing order of seriousness, water, isopropanol, denatured alcohol, lacquer thinner (which I think is mostly toluene), acetone, brake cleaner (the real stuff, which is tetrocholroethylene, not the environmentally friendly stuff, which is typically a blend of acetone and the like), brake fluid (famous for eating paint, should be kind to aluminum), citrus-based paint strip, or if all else fails, methylene chloride-based paint stripper. Check the label of any stripper for acids or bases (e.g., sodium hydroxide, potassium hydroxide), as both acids and bases will etch aluminum.
I agree very much with not taking the drum off its current bearings.
@@danmenes3143 I believe (I could be wrong) that the read/write heads are floated above the drum surface using airflow, so they can maintain proper distance and accommodate some amount of deviation from perfect circularity in the drum's cross section. The patent for this dates to a couple of years before the Bendix G-15, though it's IBM's, so...
After all, these drums were originally manufactured in the 1950s using machinery available at the time, not some kind of magic, so while they were certainly made to very precise tolerances, they're not _perfect_ -- nothing in the real world is! And that's without taking into account the fairly significant changes in outer diameter an aluminium drum of this size would be expected to show just through normal temperature variation. A 300mm (looks about right for this) aluminium ring at 20 celsius is going to expand to about 300.06mm at 30 celsius. That's a radial expansion of 30 microns already, so the idea of a 4 micron tolerance is out the window already.
I used that keypunch when I was an undergrad in the early 70's.
Don’t forget the bearings. A few microns of runout in the bearing could easily result in 1 mil + at the edge of the drum. That’s really a crazy tight tolerance for a machine built in the early to mid 1950s. I am thinking a lacquer with the correct ferrite compound sprayed on is the way to go. I have my doubts about polishing the surface after the fact. I have a feeling how drums and platters were coated is one of those trade secret techniques, details of which are difficult to find in patents. Probably anyone alive who actually worked on coating drums is in the upper 80s to 90 + years old by now.
Yes, the iron oxide could scratch the coating itself. Mixing the iron with clear then sanding could be a nightmare. How would he know the right amount of iron or particle size or resistance of the clear.
@@SoundsLegit71 Thanks for the comment. Years back I had a Western Electric voice announcement unit which could record up to about three minutes of audio on a magnetic drum which moved under a head driven by a lead screw, not too unlike an old Edison victrola. That drum was coated with an orange-red oxide in what appeared to be a fairly soft substrate. I believe the trick will be to find an oxide mix which has the correct magnetostrictive properties to magnetize fast enough and retain enough residual magnetism to provide adequate signal/noise ratio at the read head. That’s probably yet another trade secret. Everything could be right with the coating process and still give zip for output if the formulation is wrong. There is a lot of interesting voodoo in magnetic recording. For example, VHS machines used depth recording which allowed one signal to be laid down on top of another and those signals to be separately recovered by virtue of gap size and head geometry.
1:00 I finally made it to the System Source museum this past summer, and it’s fantastic. I recommend booking a tour, but plan extra time!
I was also finally able to confirm that one machine on display used to belong to me! Bob took a bunch of old Macs (and accessories and magazines) off my hands when I was packing to move abroad from Maryland, and had to get rid of stuff. I couldn’t be more delighted that it all went to a museum rather than the dumpster.
Tim Hunkin has a remastered Version of "The Secret Life of the Videorecorder" on his channel where at the end he talks a bit about the famous stickytape and rust experiment and how they made the magnetic powder. Its not as simple as just using red iron oxide - if I remember right, you start with black iron oxide and heat it for a certain time to a certain temperature such that it partially converts into the red iron oxide and thats the stuff that can hold magnetization while being sensitive enough that the write head magnetic field is sufficient to make a recording.
Right, it isn't simply rust (Fe2O3). I read in a book from 1959, written by a engineer working at Philips, that you need gamma-Fe2O3, not alpha-Fe2O3. And the long and thin crystals are best, especially if aligned along the tape (or drum in this case). But he also says that the process to get this type of oxide is patented by tape manufacturers. Btw.: I got this book (it was decommissioned) from the library of the company BASF, which might be known by some people.😉
Yes, you want the right oxide formula, even if it means you need to get old ferrite tapes, and soak them in thinners to dissolve the oxide from the mylar backing, and then filter the result to make it a consistent product, probably needing a flameproof rod mill to break up the clumps in solvent first. Then filter, and spray onto the drum while it is spinning in the housing, after removing heads and masking off the interior, then leave spinning till the solvent is all flashed off, putting the oxide back on the drum. Ferrite on mylar, probably dating from the 1960's, as that formula is about the right one, and not the high coercivity ones used later, that the old heads likely will not be able to erase or write to.
@@andreasascherl2018Whats very encouraging is that it worked regardless of the very crude way of preparing the magnetic pigment. For use in the drum memory that might already be perfectly adequate, since all that needs to be recorded are bits, not analog signals. For a slow moving tape that is expected to have low noise, linearity etc. it makes sense that the manufacturers aren't so keen on giving away the info on how to make the pigment. Since the heads of a drum memory have large gaps to project the magnetic field over the (compared to a tape) very large distance onto the drum, it also should not be critical that the pigment be a particulalrly fine dust or powder - the size of a bit on the drum is very large.
Love Tim Hunkin! Anyone not familiar should head over and check out his arcade machines and his secret life of x series
Machining the drum on a lathe with required precision will be really REALLY difficult, considering that there's no room for error. You should try a solvent or paint stripper first. Then just polish the surface (if needed)
I noticed that the sides of the drum have the same magnetic coating as the main surface. This suggests that the drum was originally coated by being lowered in a trough filled with the magnetic material in liquid form (just the bottom of the drum) and then rotated to apply material to the entire surface. The centrifugal force would even out the coating over the entire surface of the drum. I suspect this was done in a sealed chamber to contain splattered liquid with some kind of mechanism to raise and lower the trough to start and stop the coating process.
I was thinking the same thing and just posted a similar thought. The shroud that fits over the drum during the coating process would take a little engineering, but would cause quite the mess if it failed in any way. Samples could also be taken from the sides to safely test for enamel or any other binders that would give away how it was done.
Interesting theory!
I find it fascinating that harddrive tech went through a similar format progression as phonographs, from drums to platters
This is for the machine's RAM, not storage
@@pilotsmoe Granted, but consider what it evolved into laid the foundation of hard drives
21:19 In my humble opinion, I believe that an effective method to repair the drum would be to deposit a ferromagnetic material on the surface through a vapor system inside a closed container, something similar to the method used to create optical mirrors for telescopes.
This, vapor deposition is probably going to be the best bet. Spray-coating may have been used more because, with the right tooling, it would be much faster and economical than other methods.
It would be incredibly cool, but I think you would need to give a qualified physicist or chemist about a year to develop the process. There is a lot of complexity, including the exact grain sizes, getting the right mix of oxides (Fe3 or Fe2? Or a mix?) with the right crystal structure, annealing, and so forth. I do seem to remember having cassette tapes, back in the day, that advertised that they were "epitaxial," which suggests some such process. I don't think just sputtering iron, cobalt, or some such onto the surface will give a workable recording medium.
I second the vapor method.
Regarding the re-coating of the drum, I was reminded of an old episode of the Secret Life of Machines. Luckily, _Tim Hunkin_ (UA-cam) remastered these old episodes from the 90's and the Video recorder episode is available to review. Tim attempted to re-create a magnetic film, with some success, and he also talked about how the formulas for the media were a "special sauce" with each company having their own secret formula.
I would also highly recommend Tims other videos, highly entertaining. The original series are all great episodes.
re: re-coating method - my thoughts would be to try some type of electrostatic coating, similar to powder coating...
Cheers,
This is interesting. I've tried looking into the details of how magnetic media was manufactured but there is very little information available. Quite concerning considering how many legacy systems and edge cases there are out there that rely on a dwindling supply of media.
Have you considered PVD (physical vapor deposition) of a ferrous oxide coating? It would give you a far thinner and more even coating. You may be able to find a local company that could handle that for you.
You should consider an electrostatic process to apply the magnetic material to the aluminum drum. In the first stage a mixture of magnetic powder and some kind of PTFE/acrylic dust is sprayed on the drum surface while it is rotating. In the second stage the mixture is "baked" at certain temperature to a solid ring which sticks to the drum. Then you need an optical polishing tool with micrometer adjustment mounted at the read/write head position to get the drum running true before mounting the heads. No easy task.😁
I like your idea, clear coat is soft and sanding clear with iron oxide in it would be a nighmare. I painted two gloss black classic vehicles, sanding with fine grit is key to obtain a mirrored surface with clear. .001 of an inch heck, heat expansion could go past that.
This would certainly be a cool science experiment--although A LOT of testing and tweaking of the process would be needed. I don't think the Harbor Freight powder coating gun will give the needed super-thin coating thickness. I think the binder that laser printers typically use is a finely milled polystyrene powder. Could be wrong, but you need something that melts relatively easily, without breaking down. I think PTFE breaks down chemically before it melts, so that wouldn't work. Not sure about acrylic.
I was thinking electrostatic too - the sides of the drum look evenly coated which probably wouldn't happen with spraying. But the extra level of science needed would be a series in itself.
There's Drum Head Alignment screws. Maybe all of these heads should be pre-set to highest elevated state above the magnetic surface before trying anything with the replacement drum and then probed individually to have them set to a good position. Just as a precaution.
This DEFINITELY needs to be part of the plan not just on a resurfaced drum but ALSO on the exchanged drum. All those heads will need to be recalibrated for the proper height. It is a very manual process which is why I would try to get the machine up and running with the existing crashes if at all possible.
Check the bearings on the drum that might also be some of the cause of the head crashes too.
I see you can buy replacement Bendix drums in the UK. Appears to be for a washing machine though 😂
The method of spraying on a suitably ferromagnetic material should work to be fair.
Just turn the drum between centers on a lathe. (the axel on the drum has visible signs that the axel at least were made in that fashion.)
And beyond that it is as you say a simple spray coating application at a steady RPM. And remember to keep the lathe running while it dries, don't even stop it or change its speed. Since any changes will give the paint time to settle unevenly.
I have a strong feeling that it will work fairly flawlessly. Might though have to play around with it a few times to get to the correct coating thickness. Ie don't be afraid to put a micrometer to the surface for the first couple of runs, and document both RPM, spray time, air pressure, distance and the viscosity of your DIY magnetic paint. Since changing any of those will change the thickness.
Technically to get an even more even coating you would also have to oscillate the spray nozzle back and forth along the axis of rotation. Since te application will have a somewhat bell curve-ish distribution. (However, this effect might be too small to matter, but might be worth keeping in mind.)
Though, another approach is to just coat it thick and cut/"grind"/polish it down to the desired thickness, but I have no clue how that would fair in practice, paint usually isn't the "best" material to machine. (Though, a customer at work grinds rubber parts to +/- 5 µm, and rubber isn't "nice to grind".)
he'll probably want to check how fine the fe2o3 is he gets, i imagine that these companies used pretty small granules (like 10s of nm) size, so having a mortise and pestle for this might be needed which is a pretty reliable way to get to that size
also will absolutely have to dry the powder, iron oxide loves absorbing water so it should be placed in an 100C oven for a bit to dry it out and stored in a desiccator. but otherwise it seems very reasonable, after all its 1950's mechanical computer tech. the magnetic disk/drum was only just recently invented so i can't imagine many specialized processes had been developed yet
@@genderender Or just grab lots of old ferrite tape, and strip the coating off it with a suitable solvent. The grind and filter in the solvent, to get it back to a fine powder suspension, and spray it on.
@@genderender Eh. You are both right and wrong.
Yes, "off the shelf dust" is not always as fine as the seller states it is. And it is true that it risks being too coarse for the application.
But a particle size of 10's of nm is way smaller than anything produced at the time of this machine as far as magnetic media is concerned.
A more reasonable particle size for the time would be noticeably above 1 µm. (and one can technically calculate the needed grain size if one wants, but this comment is long enough as is.)
Secondly. No a mortar and pestle isn't a reliable way to make particles bellow 1 µm as far as most materials are concerned. Let alone 10's of nm.
An actually reliable way to produce particles bellow 1 µm is through chemical means. Simply growing particles around a nucleation sites until said particles reaches one's desired size.
Not that particles of this size is of any relevance to this magnetic drum. (Rough math from my side puts a grain size of 10 µm as overkill, that is if the drum is 1 foot in diameter, but it is likely bigger making the 10 µm grain size even more overkill.)
I love these videos, because it has 2 things I love: Vintage Computers, and learning about Vintage Computers. Keep up the good work, man!
I hate cleaning those resistors. I had them in the Navy on several systems as well as encountering them later on older manufacturing machines. Also, I find it interesting that in more advanced times you find that the clearances and specs of older hardware are actually quite tolerant to the extent that .0001 mil isn't necessarily required. I'm looking forward to seeing your success with recoating the drum as I think (empirically, not based on scientific evidence) that the result might be a bit easier than expected. Either way, thanks for an absolutely great series looking at these older technologies. Some of these I've seen but wasn't given a chance to touch them.
I don't think the 2K clear will level the way you think it will, I would plan on applying it thick and turning it down the way you plan on doing the primer. Then there's the issue of 'are the crashed heads salvageable?'
With the age of the machine they are basically old 1/4in reel to reel heads, pretty large, and pretty robust. Only thing that kills them are open coils, and they are large enough to actually be rewindable, as the wire is large enough to see.
You can really see why discs "beat" drums in the long run... so much easier to spin-coat a disc than a drum!
Ooooh LOOK! BUNNY!
I bet your plan for recoating the drum is exactly how Bendix did it back in the day. Low tech, hand made.
I'm in agreement with others here: consider getting it precision machined and/or built up to exact tolerances as a pristine cylinder, first. Heck, you might find another youtuber has the tools for the job (lots of machinists on this site). As for the coating, the sides look like they're covered in the same stuff. I think that's a clue as to how this was done in the first place and is worth considering. Also, you can take a sample from that side area and test it with solvents to get an idea of what was used in the first place. Lastly, it might have been spin coated anyway despite being a cylinder (and making a colossal mess in the process).
A collab between Usagi and the machining part of youtube to remanufacture a memory drum would be some amazing content.
At first when I saw the sides coated too I thought it was a solid ring of some kind of ferrous oxide compound which could be machined down. Yes I know that's a terrible idea and nobody would do it.
I live only an hour away from System Source. I didn't even know it existed. I will be planning to visit now.
I was going to actually recommend a good spray coat of ferrous paint mix and polish it smooth. So stoked that you are entertaining this idea, have often wondered if it would work to recoat a drum or platter.
I agree with your decision to avoid tape.
Wind aside, that drum's surface will see acceleration in the ballpark of a few hundred g, directly outwards. And the layer needs to withstand both its own weight (at a few hundred g), and the stretching of its entire layer - and not only must it stay intact, but it can only stretch out so much before crashing.
It's probably not very much total force, but it's enough that if "correct" materials and techniques are an option, they are mandatory. And automotive coatings sound (at least to me), much more likely to work than tape. That said, I'd still look around to see if there's anything else within your budget that's actually rated for this kind of load. And definitely would be careful about actually machining the coating. If the bond is weaker than the coating, it'll be pretty hard to machine without flaking off.
This Bendix computer keeps getting more interesting. Since I have to start thinking about simple but modern digital servo processors for a project; AND I tried to build a EM relay/diode computer long ago (didn't have enough relays), AND the Bendix magnetic drum/ punch tape/cards/architecture are too interesting together for too many reasons... you've got me hooked!
I built a weirdly efficient accumulator/shifter/base 10 to binary converter in one memory register and began designing the rest of a processor and realized I'd need way over 500 to make it work so I put it on hold 20 years ago.
4:03 You need some Texwipes and isopropyl alcohol or freon. No residue but it will remove ink and some paint.
12:44 Be careful, products with ammonia can have adverse affects on rotating memory devices. It can delaminate the coating on the drum or platters. Head crashes after housekeeping mopped.
15:08 The scent is acrid with a little ozone.
22:15 That drum will become gyroscopic if it spins up with any speed. A couple different companies make magnetic paint.
A ride on the way back machine was fun. Nice video, good fun.
Just remember to mill the rust into an ultra fine powder, best done in the binding material you'll be using.
Just buy it
Also remember to use the correct type of rust. Fe2O3 (hematite) is paramagnetic but not ferromagnetic so will not work. You need Fe3O4 (magnetite) which is ferromagnetic. Magnetite is actually a black powder, so using red iron oxide is unlikely to work because the latter consists mostly of non-ferromagnetic Fe2O3. Gamma ferric oxide is a ferrimagnetic form of Fe2O3 but requires a special heat treatment of the rust to create in quantity.
I'm going to be the oddball here and suggest making a solid state drum emulator and get the rest of the system up and running before tackling a drum rebuild.
Couple ideas for ya for the refinishing of that drum. 1) Check out Fiber optic Polishing paper for your final polish before coating. 2) Would Electrostatic coating be an option? 3) Spray while spinning in a negitive pressure atmosphere? 4) coat heavy and sand/finish to precision? - interesting to see how this all goes, thanks for sharing this adventure!
That's such a cool old computer, I grew up working on old tube amps with my dad so anything full of vacuum tubes has a special place in my heart.
That thing is a beast, I'd love to see it in person so i could feel the heat coming off all those tubes while its in operation.
The sockets exposed around 4:55 when you pull those relays ARE actually vacuum tube sockets - these were one of four styles used before the introduction of octal / locktal base tubes (the others had 4, 6, or 7 pins, but had two larger pins for keying purposes, and were usually assigned to the filament - the 6E5 magic eye tube is probably the most common example of these early tube bases).
It's strange hearing an American use the word "muntered"......
I love this channel
Take some old vhs tape, wash away the base with acetone/xylene and use the remainder as your magnetic dope for the 2-part paint. Spin the drum as you spray, then polish, repeat. then use a lathe to dimension if better tolerances are needed.
The Applied Science guy probably can figure it out. He has experience in machining and chemistry.
I was one of the commenters who suggested wrapping the drum in a spiral of audio tape. Now that I know the crazy runout specs you need to meet, I think you are correct in assessing the likelihood of success of that method. However, I think the spray method is even less likely to succeed unless you are willing to devote an insane amount of time just to developing the proper techniques, fixtures and materials. Among the things to consider: 1. Magnetic recording coatings aren't just "rust." They are carefully engineered blends of oxides with different chemical formulations (don't forget, Iron can be either +2 or +3), different crystal structures, annealing processes, etc. Do you know what kind of coercivity you need? What other magnetic properties? 2. the particles need to be milled to an exceptionally fine, and consistent size. 3. How can you get enough density of solids in your paint, and still have a coating that levels to a flatness of 0.0001" or so? 4. How do you control the application to get a coating thickness consistent to 0.0001" or so? 5. How to get a coating that will last another 70 years, without flaking off and destroying the heads? 6. All the things I haven't thought of yet.
Still, pulling it off would be a real tour-de-force!
Here's an interesting patent about binders: US3650828A (just put that number into google)
Another interesting read can be found by googling Kreiselmaier_Pigmentation_of_Magnetic_Tapes. Shockingly, this source has actual recipes for magnetic tape coatings, complete with proportions by weight, and supplier catalog numbers for the various ingredients.
Nice. This brings retrocomputing to a whole new level... correction: to an even _older_ level! 😊
Indeed. What's tempting here is that the drum repair should be doable with old metal-working tools. Given the age of the computer in question, the drum was made on a lathe that could have been built as early as WWII. It doesn't get much more vintage than fixing what amounts to CPU registers on 80 year old machine-shop tools.
Man, it's so mind-blowing to see all those circuit components wired together from a time before circuit boards existed (or at least had made it into wide use)!
The squeegee is called a doctor blade. Has been used in the printing industry for many decades.
There are factory notes and patent information with how to spray that drum somewhere. The coating formula/method would hold a patent.
you should experiment re coating the hawk drive platters , and get them to read before trying the drum
19:00 - you're basically talking about Bernoulli Principal, with the B Effect ripping the tape off the drum. It's an early JAZZ drive, without the cartridge.
how about ~20 000 grit paper/stone to bring leveled surface. how about CrO2
Polymer Resins: Various polymer resins are commonly used as the base for the binder. These resins provide adhesion and binding properties to hold the magnetic particles together and adhere them to the tape's base film.
Solvents: Solvents are used to dissolve the polymer resins and make them into a liquid form suitable for coating. Common solvents include toluene, xylene, and various alcohols.
Plasticizers: Plasticizers are added to improve the flexibility and durability of the magnetic coating. They help ensure that the coating remains pliable and resistant to cracking over time.
Surfactants: Surfactants can be added to improve the spreadability and wetting properties of the coating mixture. They help the coating material spread evenly across the base film.
Antistatic Agents: To minimize the buildup of static electricity on the tape's surface, antistatic agents are often added.
can't wait to see the re-coating process!
There may be small chance that there may be enough thickness of material just to true the drum. Refurbishment might have been part of their design. Though, I suspect that after the drum was sprayed, they polished it using a lathe. Though, finger sanding marks, such as those seen in bad bodywork would also be bad for the drum. It's possible that after the drum was coated that they merely, passed a loop of 1,000 to 2,000 grit paper under the wheel and gently lifted both ends up. Machinists sometime use this technique to get that last thou when their machines won't otherwise let them. If you want it shiny, there are telescope grits such as 5-micron aluminum-oxide and 1-micron cerium oxide available. I suppose that it could be charged into a stiff block of felt, held in a lathe toolpost.
As far as spraying it, the challenge is: the proportion of binders to media. It would seem that there needs to be a lot of magnetic material in proportion to the binders. Making sure the spray mixture is well strained is a thing.
Party trivia: The circular nut is a DIN 1804. Amateurs without the proper tool mar with a steel chisel to remove it; but a pro without the proper tool will use a piece soft metal such aluminum or brass and not mar it--or you can just make or buy the tool. You could make one.
Did you figure out what went wrong with the old ‘scope? This is a cool series. Thanks for taking us along!
Thanks for tagging along with us!
Not yet, it popped the main fuse in it. Whether that was because that fuse was the original and it was just tired, or something else was wrong somewhere, I don't know yet. I'll have to sit down and do some proper testing and see what's drifted out of spec!
I have no idea if your plan to recoat the drum will work, but I’m delighted that someone is taking the time to think about this stuff. Such restoration has the potential to keep these machines running for decades to come, if not centuries.
OK, Usagi, you have totally got me down in this rabbit hole--pun certainly intended. It's clear that microscopic examination of the coating is valuable. Don't strip the old coating off the drum until you have developed your process and are ready to coat--you may want it for microscopic comparisons. Particle sizes are sub-micron, so you will be pushing the limits of a good light microscope to resolve individual particles, but clearly there is still a lot you can learn about how they are dispersed and bound. One resource (Kreiselmaier) suggests making tests on glass microscope slides, which had also occurred to me as a good inexpensive source of test blanks.
Also in Kreiselmaier, it seems the way emulsions were applied to magnetic tape are mechanical, not sprayed. "Reverse roll,
knife, and gravure coaters are routinely used." These sound reminiscent of the way ink is handled in a commercial printing press.
Your method for spraying the drum could definitely work, altough keeping your tolerance while cutting down the diameter back to bare metal will be difficult. keeping a 1 thousands of clearance is probably impossible with a normal lathe, maybe ask some machinist youtubers about the issue, they probably know way more than me, and could point you in the right direction. A fun challenge to say the least!
Yes about the only machinist that would have that level is Robin Renzetti, he probably can machine that to better than original runout.
I believe the heads themselves are adjustable for clearance but yeah, if there's 1mil tolerance allowed then that is going to require a large high precision lathe.
@@SeanBZA I bet @robrenz wouldn't try to do it on the lathe, though. Either use a cylindrical grinder, or put the spindle, along with its bearings, on precision V-blocks on a surface grinder.
Tape is definitely a non-starter. Spraying a ferrous coating is the way to go. Good luck!
with a name like "Bendix", it reminds me a bit of futurama.
Usagi: "I'll get this thing to work"
Bendix: "Bite my shiny metal #@^"
(The bot "Bender" on Futurama, that is)
As a mechanical engineer, your solution for manufacturing the coating sounds very good. I would give it a shot!
If the red neck method doesn't work out (im all for seeing that!), maybe give electrolysis a thought? Since the flywheel is a conducting metal, it might be possible.
Very nice construction, almost avionics quality workmanship
I have just watched a article on the Computer History channel called, Origin of the Univac 1103A, at 13:01 there is a picture of a drum being coated. Looks like it was done with the drum mounted in a lathe, with an air brush in the tool post.
Very enjoyable series. I appreciate the enthusiasm.
The lathe idea sounds good, if it doesn’t work you could try mixing a magnetic compound in the form of a paste, apply it thicker than needed and then use the lathe with a scraper set to the exact right running distance, if then run at a slow speed in the lathe, the excess material could be gently scraped off leaving a perfect magnetic surface
My best suggestion for recoating that drum head, is spin it underneath a surface grinder to get it perfectly rounded again, and then electroplate iron onto it for an ultra thin ultra precise ferromagnetic layer
When you come back to Hunt Valley, if you are into BBQ, try out Andy Nelson's BBQ! They are right down the street from Bob's museum, and it is possibly the best restaurant in all of Maryland!
Andy shows up sometimes. If he does, ask him about his time as a Baltimore Colts player!
Seriously... if you don't go, you'll miss out. 🤘
Too bad I live several thousand miles away, I would like to go along. Fascinating project 👍
Despite being less involved than a transitor, I somehow fail to picture how tubes even manage to do any computations. Might be because I was born in the late 90s and these machines don't even look like a computer to me. Not as exciting as the Centureon in terms of seeing an old computer running, but definitely just as exciting seeing a historically significant computer getting to work again. No offense to vacuum tube or mechanical computers, but they just don't interest me in the same way. They're less a computer and more a giant machine lol. It's more akin to somebody repairing old machinery, I'm not even gonna bother picturing HOW it works (like I can for the Centureon and others), but see that it actually works at all, otherwise I'll go crazy lol. Hopefully this project has smooth sailing. My hats off to you for even getting into computing this far removed from today.
I was also thinking of spraypainting the material on before you made mention of spraypainting.
This should be a good way to get it fixed.
Magnetic paint is a commercial product (FerroPaint was the first that came up) - i'd probably take a chance on some of that rather than punting the binder will be compatible
Your enthusiasm is infectious 👌🏻👍. Follows with great interest!
You can clean the drum with alum media at around 60-70psi should clear off old coating . When i have to do mission-critical parts and ceramic coat them, alum oxide is my goto
Debugging this system if one of the cards won't work would be quite a task, I'm glad that scope breakout is there.
Glad to see more on the Bendix!
Dann u guys comment quick
Thank you for the fantastic video. Well done!
Look forward to these every week! I’m thinking you would need an incredibly accurate lathe when you’re talking those tolerances? Can’t wait to continue the journey 🏆
Thanks you for another fantastic video! We saw the sister unit at the Computer Museum @ System Source and Bob told us about what you're doing. I can't wait to see it running when you finish it-he put the best man on the job for it!
When you said about the new coating, my first and only thought was about PVD. I'm sure there are many enthusiasts here who have their own vacuum chamber and can sputter thin films. I don't know much about these particular coatings, but I think it is possible to make em using PVD.
The drum has some important read only tracks though (for timing and such), so those would need to be written correctly on the recoated drum.
I think that's why he wants the machine operational first, with the other drum. With that working, he can probe those signals and recreate them.
Thanks for talking about the bendix ❤
Electroplating is a common metal finishing technique that can achieve each of these objectives. This process involves immersing the aluminum workpiece into an electrolytic solution and introducing an electric current to deposit dissolved ions of another metal onto the surface.
Good luck with the re-coating effort. Really wish it is a success❤❤❤
card reader machine seems like an image scanner, or course you electroplate or vacuum deposit a layer, then lathe it, or lathe it and then coat, of course you use the same method of manufacture as the drum is being written/read, even electro-discharge-machining in liquid.
Your drum looks worse than my G15 drum so I understand replacing it. That being said, I operated my drum for many years with several head crashes. You simply code around (don't address) the damaged channels which is very easy to do. If you're unlucky and a timing track or register track are damaged then (only then) is a crashed track a DOA problem. IMO it would have been worth while to identify exactly which channels are bad before disassembly. Retract the crashed heads of course. At some you're going to have to do it anyway. Also on this video, you mention the tape reader is slow. I don't remember the specs, but in my experience only my MTA was slightly faster. The optical reader on the G15 was ahead of its time and quite snappy. The paper tape punch of course is pretty slow. Keep up the great work.... I am very eager to see more of the G15 project
What is the thickness of the original surface coating? What is the current maximum surface deviation caused by head crashes and other damage? How much adjustment can you get out of the Drum Head Alignment screws?
If the coating is thick enough, you could use an ultrafine cerium paste to polish out the scars on the drum. If the Drum Head Alignment screws give enough adjustability, and the original surface is thick enough, you may not need to recoat!
I will be interested to see if refinishing the drum can be pulled off with a reasonable amount of effort. I am interested in seeing how hard it would be to fabricate a simple hard disk drive from scratch. I want to construct an 8 bit minicomputer from scratch that is a good example of a high end CP/M 2.2 system. I think someone with advanced fine watch making skills could construct a decent magnetic head. My dad had a race engine machine shop, and he built parts accurate to a ten-thousandth of an inch. That's about what you will need to get the drum project to work. Oh, I was just reading some of the other comments. Check the bearings for wear and excess clearance. That could also be an issue.
One thing to consider is the size of the magnetic particles that you use.
Ideally they should be in the range of 10 micron or smaller.
The smaller the particles, the better performance of the media.
Hi Ken. Before resurfacing that drum, you’ll want to get rid of the existing coating. A series of smooth sanding blocks with various grits of sandpaper could be used, starting with 400 grit, and working your way to 1600 grit, followed by lapping compound should work. Machining will leave little grooves in the material; you’ll want it mirror smooth before re-coating. You’ll need to know how thick the modern magnetic material needs to be so the total O.D. of the drum meets the stated design requirements. The “coated” O.D. - the thickness, will give you the O.D. of the drum before coating. This is your target measurement when removing the existing coating.
I think spraying will work, but you should keep the drum spinning at a constant low rpm while it dries (like over night, etc.) to prevent drips or the top from being thicker than the bottom, etc. Also you may want to do it in a mostly clean room environment to keep dust from being hardened into the clear coat.
Oh, and it's so good to see such a boopable snoot at the end your video! That's a cute little bunny-rabbit!
Would be fantastic to get consultancy from one companies that made their living from that, the one that make mechanic drives, or companies like basf that sure could work out the right process and chemestry. But as those are so gigantic, you can expect much. Anyway, this is a work of art, not a computer. It is deligthfull
I think that surface will have to be repaired 'in situ' doe to centering tolerances, assuming the bearings are perfect and lubricated.
Use the bolt holes of that cover to position a tool to smooth the surface, and then a light spray in several goes of iron oxide primer should so the job.
Always interesting how much mechanical precision was around back in the 1950s.
Please check if the iron oxide is magnetic before mixing it in paint. My gut feeling is that it should be more along magnetite powder. Could be wrong too… Yes, I know, far in the future, with many tests in the process. Maybe there is some treatment for making a non-magnetic iron oxide a magnetic iron oxide?
That turquoise looks soooooo nice BTW 😸
He mentioned using red iron oxide which is mostly Fe2O3 (hematite) and not ferromagnetic so wouldn't work. Hope he finds out eventually whatever is the correct stuff he should be using
Kudos to you for digging into the messed up drum and considering how you could repair it. I had a feeling that would be an issue when you first mentioned how everything works off a magnetic drum that had timing signals written on it. That should be fun to rewrite on the drum after a recoat👍. lucky that you have another one to test with. At worst, i'm sure you could emulate it. Very much looking forward to this and I would love to see the recoating attempt.
tim of the secret life of machines had some insight into making their own magnetic tape on the video about the VCR during the little after show segment he did on it. They made their own plastic audio tape by using some tape and sprinkling iron oxide on it. The interesting thing he mentions is that raw iron oxide did not work, and it had to be processed before it would actually work. It's been a year or so since I saw it but I remember him saying something about having to cook/heat it first. Might be worth a watch
Pretty cool video, thanks!
I reckon the spray paint plan is doable, can't wait to see it
I would fabricate a way with a leadscrew that attaches to the drum unit as you currently have it assembled. Then use a very sharp steel cutter mounted on this contraption to turn off all the old oxide coating and aluminum oxide from water damage. Then take the way off, sand the surface of the drum for better adhesion, and spray the new oxide on as you described while spinning the drum and let it completely dry. Then reattach the way and turn the new oxide completely flat. After that disassemble the unit, take the drum out, and clean everything thoroughly to remove all dust and grinding residue from the housing and drum. This last step done with multiple cleaning baths to make sure all traces of dust are gone, while not getting any water into the bearings. Then lubricate the bearings and reassemble.
Good bunny! No poopie on the chair. GOOD Bunny!!!