Advent of Computing - Episode 180 - You Wouldn't Magnetize a Tape!

Episode Date: April 19, 2026

The image of a mainframe is almost always accompanied by it's companion: the magnetic tape drive. For decades magnetic tape served as the medium of choice for computing. It was faster than punch cards, and more available than hard drives. But where did it come from? Is it a borrowed technology like the vacuum tube? Like Advent of Computing? Then check out the after show! Adjunct of Computing is now LIVE: YouTube Spotify Apple Podcasts

Let me disclose something to you. I love old, bad, sci-fi films. You know the ones. They're most often shot in black and white. They're peopled by actors you've never heard of. The plots are paper-thin. And the monsters and props? Well, do I have to say it?

If you get a flying saucer on a visible string, then you're in for a treat. What can I say? I enjoy a certain type of slop. Crucially, I don't really like newer films that are poorly produced, with the exception being the fantastic lost skeleton of cadavera. Seriously, this is my film recommendation for the episode. There's a funny and persistent issue in these films.

They all revolve around some kind of high-tech antics. How do you show that on screen if you don't have any budget? Paper-mache monsters and hubcap flying saucers are one thing. It gets more tricky when you have to show the all-important laboratory, and any good sci-fi film has to have at least one shot of a laboratory. What you get are short hands for technology. If your villain is a doctor with an interest in chemistry, then you get vials and flasks of colored water,

with some dry ice somewhere nearby. If the plot is more biological, you might get a cage of a few wild animals. Or, in the case of X, the man with the x-ray eyes, which is about an evil optometrist, you get a cage with some monkeys and a bunch of bubbling flasks. It gets more difficult when things get into the realm of the digital. Case in point, the 1958 classic The Robot vs. the Aztec mummy. It's bad. It's really, really bad. But it has a robot, so it has to have some kind of technology.

The robot looks to be constructed from cardboard boxes, which, you know, that's fair. You're not going to build a robot for a movie of this caliber. And if you have a robot, you have to have a computer. It's just the logic of the situation. It signifies just how advanced the human human. robot truly is. In this case, the computer is a big old tube radio set

with a bunch of dials and panels glued onto it. Honestly, it's an inspired solution. I think you can track public perception, or at least public consciousness of computers, by using these films as as a bit of an odd crystal ball. There is, of course, some distortion,

but there's a clear trend. Once we get into the 50s and 60s, the computers become an essential prop. We start to see actual mainframes in the back of shots and B-roll of machines at work. But what do you show? What's the digital equivalent of a flask full of dry ice and food die in water? What is visually interesting about a mainframe? In short, blinking lights, sometimes punch cards. flying past, and tapes spinning.

Welcome back to advent of computing. I'm your host, Sean Hasse, and this is episode 180. You wouldn't magnetize a tape. Today, we're asking a simple question. What exactly is the deal with magnetic tapes? But before we get into the episode proper, I'm making my usual plug yet again. The official advent of computing after show, called Adjunct of Computing. is doing very well. Episode 3 came out last week. Episode 4, the companion to this episode,

will come out a week after this episode releases. I'll have links to everything in the description, and if you've ever wanted just a more casual discussion about advent of computing topics, go check it out. I think you're going to really like it. Okay, so to the main episode. I admit I'm being a little playful here. My plan for this. this episode is to track down the origin of the magnetic tape drive. During the mainframe era, open reel-to-reel tape was nearly ubiquitous. The medium was so closely associated with computing that it became part of a visual shorthand. If you ever find yourself watching old sci-fi films, then keep an eye out. Big tape drives are often used as props either in place

of computers or as the actual visually interesting aspect of a computer. This is also one of those old technologies that is nearly died out. No normal user's own magnetic tape storage devices anymore. Tape does still exist, but it's very special purpose. You'll see them used for very large storage setups or for archival storage, but that's about it. The disappearance is noteworthy because tape did adapt to the times. It's not one of those rigid period technologies that get superseded and just dies on the vine. Early microcomputers, very famously, used cassette tapes for data storage.

That's just an update to the more venerable reel-to-reel drives that populated mainframe rooms. It's not like tape died with the first micros. It stuck around for some time. Tape's also just an odd medium. At least, it's always struck me as a little odd. Early computers were, by necessity, cobbled together from non-digital parts. In the 40s and 50s, there weren't really purpose-built components for digital computing. Vacuum tubes came from radio.

Delay lines came from radar and keyboards from telegraph offices. Purpose-built computer components just did not exist. That doesn't come until later, right? To me, it feels like tape should follow this pattern. But is that gut reaction correct? I don't know because I don't know enough about tape. Let's remedy that today. Let's start off with a frustrating question.

When was magnetic tape first used with a computer? Many of the first computers used tape, but it was punched tape. The Harvard Mark 1 used punched paper tape. Zeus's Z3 used punched mylar tape. There's a certain rhyme between punched and magnetic tape, but for this episode, I'm just going to ignore it. I'm pretty sure the story of digital tape starts at EMCC, the Eckert Mouchley Computer Company.

That's the first commercial computer company, founded by the duo behind Eniac. You'll see why I'm only pretty sure as we go on. EMCC is initially founded as ECC, the Electronic Control Company, in 1946. The business plan was simple. Build and sell computers. More specifically, Eckert and Mouchley had,

just been involved in the design of EDVAC. This was the first design for a stored program computer. However, it wasn't exactly a real computer. It was a computer on paper only. The idea was to patent the design for EDVAC, then build a pile of them to sell, or at least build a handful of them. This is still where we're in the period

that people think the market for computing worldwide is like five machines. Eckert and Maltley's plan, in short, didn't work. At least the patent part didn't work out. The result is that EMCC ends up making a number of EDVAC-like computers to sell to the highest bidders. EdVAC is important because the design called for a stored program computer. That is, a computer that has a program stored in some kind of memory. To make that practical, you need a means to store something.

in memory. EdVAC is also important because its design was leaked. In 1945, someone, the jury is still out on exactly who, spread a draft report on the design of EDVAC. That report goes on to be the basis for a pile of early computers. The report is mainly concerned with the theoretical aspects of constructing a computer. It does, however, talk in kind of vague brushstrokes about implementation. This is where we get our first glimpse of the topic at hand. When discussing programming of EDVAC, the report reads, quote, these instructions must be given in some form which the device can sense, punched into a system of punch cards, or on teletype tape, magnetically impressed on steel tape or wire,

photographically impressed on motion picture film, wired into one or more fixed or exchangeable plug boards. End quote. There it is. Steel, tape, or wire. That stands out to me for a simple reason. I can place all the other technologies here. Punch cards are easy.

That's the tabulator's meal of choice. teletype tape is punched paper tape, which ends up becoming a digital mainstate. Motion picture film, believe it or not, has precedence here. In the 30s, Vannever Bush developed a system of encoding digital data on reels of microfilm. That doesn't make the leap to digital computing, but it was a technology that existed in the era that encoded digital information. And plugboards, well, that's also from tabulators and analog days. But steel, tape, or wire.

I've only ever heard those mediums used for audio, not digital. The only thing I could think of would be someone using magnetic tape to record telegraph signals, but after searching around for a while, I can't find any evidence that was ever a practice. Telegraphs tended to use paper tape for storage when needed. magnetic tape or magnetic wire was really just an audio thing. So where is the EDVAC report pulling this tape idea from? Well, there is a related technology that may lead us to tape. That is the magnetic drum.

The shape is different, but the coating is the same. A magnetic drum is, as the name suggests, a drum with a magnetizable surface. It's coated with the same type of magnetized. codeing used on audio tapes. Drums were used for digital storage as early as the 1930s. The first magnetic drum was invented at an IBM-owned company in Austria in 1932. It used basically the same magnetic heads as a reel-to-reel tape deck. But drums tended to be used as more of a short or medium-term storage device. You can't really mount and unmount a device. You can't really mount a drum because the heads have to be very precisely aligned.

And you have to have a lot of heads to make an effective magnetic drum. So when the technology made its way into computers, drum ends up being used for memory, not for long-term or removable storage. I'd like to offer the conjecture that the team that designed EDVAC already knew about magnetic drum. There's already a jump from audio technology to a more digital application when you look at a magnetic drum. I don't think it would have been that much of a leap for the EDVAC crew to think, well, why not take a sidestep here? Why not just use tape or wire for digital storage? After all, drums showed that magnetic media could be used for digital signals.

This is where we get back to EMCC. Their business plan relied on selling computers. In the 1940s, that meant courting contracts with the feds or with huge companies. Their first computer was slated to be Univac. That machine was developed basically as soon as EMCC formed. Funding for it came from the U.S. government, who were slated to be EMCC's first customer. But it turned out, creating a computer from scratch was a lot more expensive than initially thought. To make ends meet EMCC went out looking for some easier money.

They'd sign a contract with Northrop Aircraft for a smaller computer, which would be called Bynac. The plan was that money from the Bynac project would help the Univac project stay afloat, you know, by keeping EMCC from running out of money. An added benefit was that some parts of the Univac project could be thrown into Bynac and vice versa, at least in theory. As a result, there's a bit of a blurry line here. Work done at EMCC in this period could have been put towards ENAQ. either of these two machines. Bynac ends up being delivered before the first Univac. EMCC planned for both of these computers to have magnetic tape drives. Bynac is first, so it ends up being the first machine to use magnetic tape. Developing a tape drive that would work

with a computer, however, was not necessarily an easy process. In fact, EMCC had to start at square one. They weren't able to take existing real-to-reel drives and just plug them into a computer. That didn't work for a few reasons, which, oddly, are all physical. To start with, real-to-reel as a format didn't really exist yet. In the 1940s, tape was still an up-and-coming technology. Some tape formats existed, but I have to stress here, some. The earliest tape decks were invented in the mid-30s. Some of the first units invented in Germany used metal tape. By the 1940s, that had been replaced with acetate tape. But it wasn't like you could go down to the electronic stores and buy a tape deck. That wouldn't really be possible until further into the 1950s.

Even if you could, these early tape drives wouldn't be suitable for digital use. In normal operation, a reel-to-reel tape advances past the tape head at about 30 inches per second. That speed is dictated by very analog constraints. If you go slower, the audio quality will degrade. Recording won't sound nearly as good. If you go faster, well, not only are you wasting tape, you're also positive. possibly damaging the tape. EMCC's new data tape needed to move fast.

It ended up spinning at 100 inches per second. That speed is dictated by digital requirements. The faster the tape can spin, the more bits you can cram onto it. Or, if you keep the bits constant, the faster speed leads to faster reads and write time. however you look at it, you want the tape to spin as fast as possible. You also need the tape to stop on a dime. EMCC planned to format data on tape in blocks. Those blocks would be seekable,

meaning you can instruct the tape drive to read a certain block for you. To facilitate this, the tape would have to be able to spin forward and backward pretty quickly. Now, I just want to be clear, seekable is a little different from addressable. Basically, the setup was such that you could say, read the next block or read the previous block. Still, you had to be able to spin and reverse quickly. Acetate isn't known for its resounding strength. Under tension, it can stretch, deform, and tear.

That made plastic tape a non-starter for EMCC. Metal tape could solve the issue, but early metal tapes were heavy. A heavy tape meant to take a lot of energy to spin up and a lot of force to slow down. Before EMCC could make a tape drive, they had to make a new type of tape, a new formulation. They had to start from square one and work up from there. right from the start, this is a new and high-tech product. That makes tape somewhat special for the period. EMCC was using vacuum tubes that were designed for analog circuits.

Delay lines were borrowed from earlier radar systems. But tape, that was something new and special. It was being built from the ground up for digital computers. The tape formulation project was headed by one Ted Bonn, Now, we don't have a lot of information on Bond himself. What we do know is that he was not a chemist. In fact, he was an electrical engineer. While a neat asset to the project,

that meant he was slightly out of his depth when it came to tape formulation. That turned out to be a very chemistry-heavy project. The tape project is more than just a research lark here. Bond didn't just need to come up with a single tape, it would work in some theoretical digital tape deck. He also needed a process by which those tapes could be mass-produced. That restricted what was possible.

He would take a few passes at the formulation himself. He eventually landed on a chemical method that would coat copper tape in a nickel-cobalt alloy. It didn't work very well. In 47, Bond decided he needed specialized help. That year, he hired Douglas Wendell, Jr., It's actually thanks to Wendell that we have this story. In 2006, he wrote an article about his time at EMCC. When he arrived, the chemical tape formulation was, well, it was an attempt.

Quote, the process required a near-boiling aqueous solution with a short, useful working life, and also required that the plated sample be heat-treated in hydrogen in order to improve the magnetic properties. The resulting test tapes had magnetic properties which varied along the length of the tape, end quote. The trick was to get a sturdy and lightweight tape that had consistent magnetic properties and could be made at scale.

The idea of using a metal base layer would stick around since it met two of those three requirements. It was, however, swapped from copper to a phosphor bronze alloy. A better plating method was the real trick that had to be worked out. Bonn and Wendell worked as close collaborators in this period. Bond made equipment to test out new formulations and helped guide the project. Wendell did the dirty chemical work.

After hundreds of tests, they eventually landed on a formulation. This wasn't an easy road. Bonn had been trying to work off some earlier patents for plating, magnetic media. Patents, however, aren't the same as recipes. They also don't have to actually work. The ultimate solution was a combination of methods. Tape would be electroplated with a nickel-cobalt alloy on a special chemical bath. Then it would be cured under a hydrogen atmosphere. Not the most obvious method, but it was repeatable and predictable. It would produce strong tape. that could actually store data.

From there, the method was scaled up. Bonn and Wendell worked with a larger team to build a machine that would make tape. So they're literally building a magnetic tape factory. This ended up being a huge task in and of itself. The machine took up too much floor space for EMCC's current factory, so they had to rent a new building. One of the more frustrating aspects of this new tape-making machine was quality-controlled. Early on, Bonn had worked out a tool for testing quality of magnetic tape samples.

A similar device was integrated into the new tape machine. As strands of coated metal tape rolled off the line, their properties were tested by passing them over a set of coils. That would tell workers if the tape could reliably store data or if it was bad. bad regions would get marked off and dealt with later on. The issue was those coils tended to overheat. So there was a fun side quest to create a cooling system for these tape measurement coils. John Eckert, then president of EMCC, took personal interest in the part of the project.

Apparently, he helped design the tester used in the marvelous tape machine. That involvement may have been closer to. to micromanagement from Wendell. He became concerned that the Philadelphia water supply pressure might not be enough to cool the coils adequately. I found a pressure gauge and Prez and I went down to the basement to check. I attached a heavy hose to the gauge and held that hose against the water faucet and turned the handle.

We just had time to observe the pressure when the hose got loose and squirted us. It got a laugh from both of us, end quote. Luckily, municipal water pressure was enough for coolant purposes, and perhaps drenching purposes as well. Development of the grand tape machine went on in parallel to development of the actual tape drive. I don't want to talk format quite yet, but let's talk dimensions. The final tape that was actually used was one mill thick. That's not mill as in millimeter.

It's mill as in a thousandths of an inch. What can I say? It's an American company in the 40s. It's an imperial shop. A mill is about 0.025 millimeters, so very, very thin. Wendell points out that this thickness was never directly measured. Rather, it was all done by weight. Tape was a half inch wide.

They could cut a sample to a known length, and they knew the thickness of the raw metal tape. Apply a little bit of math, and you can make a pretty accurate thickness measurement. All this is to say, these tapes were very, very thin. Not only that, they had to have a consistent thickness. Otherwise, tape drives would get gummed up. But they're pretty wide. I mean, one one thousandth of an inch by half an inch?

That's a wild aspect ratio. I said Bynac was the first computer to use magnetic tape. That's true. But it's a bit of a funny and not entirely useful truth. Bynack was a one-off machine. It also wasn't a very successful machine. EMCC built the computer. It was accepted by Northrop, and then it shipped out in 1949.

Now, this is where we get some controversy. It's been widely reported that Northrop never actually. actually used Bynac, that it was somehow damaged in transit or never properly assembled. Jeremy Norman on his website History of Information handily disproves this rumor. He cites a paper from 1950 in which Northrop engineers describe using Bynac, you know, as a computer and not a space heater. I think it's likely that Bynac didn't work at first or was fickle, but it eventually became operational. It's just that once it was running, it wasn't the most sophisticated computer.

This means that in 1949, magnetic tape was being used on a computer. We can argue over how useful the computer may have been, but it did have a magnetic tape drive. So how did that work on Bynac? Allow me to introduce a very, well, I think it's a very funny document. EmCC shipped some documentation along with Bynac. This is the operating and maintenance manual for Bynac, built for Northrop Aircraft. There's just something that tickles me about a manual written for a one-off computer. I give why it exists.

I'd want the same thing. It's just kind of funny. It even has a list of components that include a pile of things that are completely unique. like components that are one of one in the world because there's only one Bynac, and that component list also has two chassis wrenches. It just tickles me, and I can't explain it in any other way. This paper is our main description of how to use Bynac, or at least how EMCC wanted Bynac to be used.

At this point, computers are more experimental lab equipment than Bynac. products. So your mileage is likely to vary. The tape interface, despite these early days, is immediately familiar. Bynac can load data from tape into memory, and it can also write data from memory to tape. That's basically how all storage media ends up working. But I have to stress this here. This is a very new idea. We can also see how this is very very very important. We can also see how this is very similar to the EDVAC report. If you have a stored program computer, you need a way for the device to sense information, and then that information gets put into memory. The mechanism used here is, however, a little unique. So let's talk Bynac. This is an odd machine to say the least. It's

actually two computers in one machine room. The idea is that they'd run the same program and then check each other's results. I think this was mainly done because of concerns with reliability of very early computers. This means we get two computers, two tanks of mercury memory, two power supplies, but only one set of inputs and outputs. That includes, oddly enough, a keyboard and this thing called the converter. The converter is a gateway into Bynax computers. Input from the keyboard goes through the converter. The printer receives data coming out of the converter, and the magnetic tape drive is actually built directly into the converter.

This is actually a clever approach. There's one point of contact with the actual machines. To the computer, this is just a bus that it can receive data from or send data to. The operator controls how data is routed by flipping switch. switches on the converter. That's how you can punch data into memory, record data to tape, or read tape into memory. It really does simplify things a lot. The magnetic tape drive is built into the converter, but it acts as any other input or output. You can flip switches to direct data from computer onto tape or from tape into computer, or even tape to keypad, or I guess

the other way, keypad to tape. It's a little bit it different than later designs, but whatever. This is within the realm of believability. There is one immediately jarring aspect of the converter, however. I want you to imagine what a mainframe's tape drive looks like. If you can't, I'll give you a hand. You should be thinking of a reel-to-reel player built into some large cabinet. It's about five or so feet tall. At the top are the two tape reels vertically mounted and usually behind a pane of glass. That's part of what made these tape drives such cool sci-fi props. During operation, you can watch these tape drives flicking back and forth, spinning,

starting and stopping. It's a very visible technology. The converter's tape drive is mounted horizontally. The reels sit directly on top of the converter's cabinet. And if the diagrams are to be believed, there's no covering. The tape is out to spin in the open, dusty air. The drive is also very simple. It's two reels ahead and some felt pads. It also looks like there's a third wheel that would be used to either tension the tape

or to actually spin the tape forward. Later drives are somewhat famous for their complexity. IBM even develops some things that are close to science fiction sounding tech. So this first pass is very primitive. It's also not nearly as practical. At least it gives me some instant concerns. The dimensions of the converter's cabinet are limited because, well, there's a tape drive on top of it.

A human operator has to be able to reach and manipulate the tape. so the converter can't be all that tall. And the openness of the drive, that makes me feel uncomfortable. It's going to be throwing tape at, well, maybe not 100 inches a second, but it's going to be throwing it pretty fast. It may not be enough to mangle fingers, but I can imagine it would do some damage if you got too close to it. In general, this is just a very primitive tape drive.

You even have to take off the felt pressure pass. had if you want to rewind the tape while it's on the converter. So this thing only really goes in one direction. Not the most useful thing in the world, but, again, it's for a very primitive computer. Another factor that speaks to just how simplistic Bynac was is the fact that it didn't use EMCC's new metal tape formulation. Bynac had been developed while EMCC is making their new metal tape formulation. metal tape from scratch.

That's the top-of-the-line formulation for any digital computer. But Bynac uses plastic tape. That's because the simple drive, which operated pretty similar to an audio drive of the period, wouldn't have worked with metal. Bynac's tape drive, the converter, is just way too primitive. It's also so primitive that it couldn't have benefited from metal tape. Bynac didn't have instructions to control the tape drive at all. Instead, it was simply used to load initial code or data into memory,

or to save the contents of the computer's memory. It's a pure sequential device. Nothing fancy, just something that spins forward and does exactly what's needed to make a computer turn on. This is what I meant when I said, the fact that Bynac was the first computer to use magnetic tape, wasn't exactly a useful fact. It's hard for me to believe that there was no crossover between Bynac and Univac, but Bynac at least didn't get good tape technology.

It got the minimal, useful amount of tech. So what about the top of the line? In 1951, the first Univac computer was finally shipped out. That included EMCC's second tape drive, the Uniservo. That drive is much more complex, and it used the new metal tape formulation, the custom developed and formulated tape specifically designed for digital data. And this is a much, much more sophisticated tape drive than the Bynac converter ever was.

I've been talking in generality so far, but once we get to the Uniservo, we get actual details. Those come courtesy of a 1953 paper from Welsh and Luckoff. It's titled, fittingly, the Uniservo. What's great is this paper even gives us a rationale for the new tape drive. Quote, among the early objectors to computers, the more far-sighted pointed out that even if a machine could be made, its use would be severely limited by inability to converse with it at appropriate speed. it was realized at the time

that a great deal of development work on input and output devices was necessary before a satisfactory commercial computer could be built. End quote. Yes, there were objections to computers. This is actually a super interesting tangent and something that I know will come up in another episode. I did that arguments against programming episode way, way back, so, hey, maybe it's time to do an argument against computing episode.

Anyway, existing I.O. was very slow. Existing I.O. was, well, it was the punch card and punched paper tape. That's severely limited how quickly a computer could operate. If you need to, say, read in 100 records and calculate an average, then you'll spend most of that time reading in those records. That's a pretty obvious spot for optimization. Hence, the pressure to develop a tape that can spin fast and stop fast. And hints, the need for a more sophisticated tape drive.

The Uniservo is nothing like Bynax converter. There are different needs being addressed here. Univac needs an all-purpose and fast data storage medium. This will be used under software control, to load programs, to load data, and to even write programs and data. Bynac just needed something to turn the computer on. So even with all this work going into very custom, very specialized, and very sophisticated storage devices, Bynac gets something way, way more simple and way less useful.

The Uniservo was, as I alluded to, completely computer-controlled. Univac, its matching computer, had instructions to read data and write data to a tape. That, plus all the fixings to make those operations reasonably fast. The name of the game here is sophistication. I keep saying it. It's almost funny going from the converter to an actual cutting-edge device. In general, uniservo just had a different philosophy than the converter. The 1953 paper describes the taped drives as a large and slow memory device.

Note the plural. One Univac could use up to 10 independently operated uniservos. EMCC would service multiple installations that used that maximum number of drives. The idea here was that your bank of uniservos would supplement the Univax 1,000 words of memory. When you use a tape drive, when you issue a read-or-write command, you specified which of the 10 drives you wanted to use. So a running program could, in theory, have access to every single tape that was mounted. That's a large amount of data. But what was this like in practice?

How was data actually formatted and how is it used? Let's start with format and work our way up. First off, Univac was an 8-track system. A tape had 8 data tracks side by side. One track was for error checking, one was for timing, and the remaining 6 tracks were for information that would be accessible to the computer. Those are your core data tracks. These were used by 8 independent tapeheads.

The Uniservo paper describes this as, the tape drive can read and write one character at a time. This is purely sequential access, but it's character sequential, not bit sequential. These characters are ganged up into blocks of data. Each block is 720 characters long. Tape was read and written one block at a time. That block structure is so central to the design of UNIVAC that instructions actually talk about data in terms of blocks.

You issue a command to read a block from tape drive X into memory location Y. You never bash bits straight off the tape. You always bash blocks. Now, to be clear, again, these are not addressable blocks. You can't say, I'd like to go to the 10th block off of tape drive 5. It doesn't work that way. That said, you can actually scan around on the tape. Kind of.

You have instructions to read the next block off of a tape and to read the previous block off the tape. In other words, you can read forwards and backwards. To make that possible, the tape drive supports spinning forwards and backwards. The engineering required for that is actually wild. Again, sophisticated purpose-built device. This isn't borrowed technology at this point. The Uniservo drives tape from a central capstan. Tape is wound from one reel around a spring-loaded tensioner,

up to the capstan in the head, around a second tensioner, and then over to the second reel. The capstan then pushes the tape against the eight heads and spins to move it forward or backwards. The central capstan is the only thing that really moves the tape. But that wouldn't work on its own. The issue comes down to physics. If you drove from the reels, you'd have to use a lot of force to get the tape to spin quickly.

That's because there's a lot of mass on the reel. By the same token, you need a lot of force to get the tape to stop. That would make spinning the tape forwards and backwards very, very difficult. You couldn't go very fast. So EMCC added some slack. The tensioners always make sure there are two loops of tape hanging down, from the reels. The capstan only really has to move the tape in those loops. It's basically throwing this slack tape forward and backwards. Winding a loop forward

a block or back a block doesn't take much force because those loops weigh much less than an actual reel of tape. A block works out to just 5.8 inches of tape. There's a blank space of 2.8 inches between each block. So we could assume that that at any given time, the central capstan only really needs to be able to move about three blocks of tape. That's 22 inches. At one mill of thickness, even though this is metal tape, that's not much mass to move, so it takes not a whole lot of force. The reels are also motorized, but only to support the capstan. The reel motors are there to maintain proper loop length. If a loop gets too long, the correct reel

will take up a little bit of tape on that loop. And if a loop gets too short, then that reel will play out some more tape. These reel motors never need to spin very fast or very accurately. That limits precision and speed to

only one part of the system. The result is the uniservo can wind through tape quickly in both directions. It also leads to this neat characteristic of this drive. A Uniservo's reels flick as it's an operation. You can find videos of this online.

I suggest you go look. As a block is read, the drive flips between forward and backward reads. The reel will flick forward and then flick backwards. It's very distinctive and it looks very different than the continuous motion of an audio tape. The engineering that went into these drives is wild. The loop and flick method is just one piece of the puzzle. Allow me to introduce the buffer tape. Now, you may think this is something to do with data buffering.

That's not the case. So, check this out. During development of the Uniservo, a fatal flaw was discovered. Friction. For the tape to move quickly enough past the tape head, some kind of lubrication was needed. Without lubrication, the tape could catch, slow down or cause undue wear to the components.

This competed against the simple fact that the tape needed to be held pretty tightly against the tape heads. EMCC tried a few different approaches to solve the conflict. One was to simply pre-lubricate the tape. That didn't work. That actually led to another weird issue where the tape was too slippery. To quote, adjacent layers were apt to slip and develop folds in the middle of the reel.

It was necessary to reduce tape to head friction without making the tape itself slippery, end quote. I imagine it also made handling the tape a little dicey, right? I think we can all think of tape springing and slipping off the reels. It just seems like a bad but kind of slapstick time. This was made a little worse because metal tape is more susceptible to folds and wrinkling. than plastic tape. The other funny part of this approach

is that an integrated lubricator was built for the Uniservo. Welsh and Lukov don't say much about this failed experiment, but I like to imagine it was like an oil reservoir that would spread lubrication across the tape as it ran. That would mean an operator, if this had worked out, would have to change their computer's oil

from time to time. That's just delightfully funny to me. The solution that was arrived at is also pretty funny. It's the buffer tape. This is a plastic tape that sits between the Uniservo's tape heads and the real magnetic tape. It's buffering in a physical sense. Welsh and Lukov described this as similar to a typewriter ribbon, but I'm not sure that's the best analogy.

The buffer tape spins past the head pretty slowly. It turns just 10 inches per hour. The plastic tape has much less friction than the metal tape head. So magnetic tape can slide over the buffer tape at full speed. This also solves the issue of wear. Early on it was realized that tapeheads could get worn out pretty quickly. This was due to friction with the magnetic tape. The buffer tape, in this way, protects the tape head.

By making the buffer tape advance, there will always be a clean surface for the magnetic tape to ride on. It's a sacrificial surface. The buffer tape can get scuffed up without the head getting damaged or the tape getting torn. This means the buffer tape is expendable. So you don't have to change Univax oil, but you do need to change out its buffer tapes. I'm not sure if there are any working Univax out there. If there are, I'm sure the buffer tape makes their upkeep. a little bit more difficult.

I want to close this out by going back to the matter of medium. Metal tape was pretty cool technology, but it came with trade-offs. This was even known at the time, from Welsh and Lukoff, quote, metal tape was found to be much superior to plastic at the time of Uniservo development. It still is, although the quality of plastic tape has improved greatly. Plastic tape offers the advantage of being less likely to fold as a result of mechanical handling, order to form as a result of pressure or dirt particles. In quote.

Metal tape is also surprisingly durable, at least if you don't fold or mutilate it. I can't lie, I just kind of want to include this passage because it's wild to me. Quote, a fire test was conducted in which six reels of metal tape were placed in a safe, along with plastic tape, microfilm, and punched cards. The safe was placed inside a furnace. The temperature inside the safe was allowed to rise to 550 degrees Fahrenheit. Film and cards, of course, were destroyed. Mult and plastic tape flowed over the edge of one of the reels of metal tape.

The six reels of metal tape were read on a uniservo without error of any kind, although the oscilloscope or field that the pulses had suffered deterioration of about 10%. In quote. So maybe metals better for disaster recovery. Now, there was a bigger issue with metal tape. EMCC had to produce it themselves. The great part about custom hardware is you can make it exactly how you want it. The bad part about custom hardware is you have to make it exactly how you want it.

You don't get a choice. EMCC could never produce metal tape that was 100% perfect. There were always regions of the tape that didn't work quite right. The solution was to test and mark bad regions of the tape. That's why Eckert was so keen on getting the quality control mechanism to work. If a bad region was detected, it was marked with a hole. Uniservo drives had a photo detector that could see those holes and then skip past the bad region of tape. Plastic, well, acetate, offered some advantages.

Well, really, it offered a different set of trade-offs. Formulations existed so you could use off-the-shelf. tape, or at least more off the shelf than EMCC's custom formulated metal tape. It could hold up to handling a lot better than metal, and perhaps most importantly, it was lightweight. In theory, you could pack more acetate onto a reel than metal, and in theory you could throw acetate much more quickly than metal. That is, if you could avoid ripping your tape in half. The core issue came down to figuring out how to handle tape carefully. It should be possible to use acetate if you can keep it from breaking, and that should give you a pile of

advantages. How would one do such a feat? Well, it comes down to those very special loops. Engineers at EMCC figured out that you don't need to move the entire tape very quickly. By keeping loops of excess tape to either side of the head, you can limit the amount of tape that had to really book it. But the first pass was a little rough. The Uniservo used spring arms that physically tugged at the tape. That worked for metal because metal is very resilient. But that could deform acetate. At least, that's my educated assumption. Around 1949 or 1950, researchers at IBM would take up the tape challenge. Their goal was to create a new tape drive for the upcoming 701 computer, and they wanted specifically to use acetate tapes. That was the pile of trade-offs that they

chose. The tape drive that came out of this project, the 726, bears a superficial resemblance to the Uniservo. It used open reel tapes. It had two large hanging loops, and reels also spun to a distinctive flick. but under the hood, these drives were completely different. The biggest difference, physically speaking, is how the 726 handles loops. You see, there was no mechanism in contact with these tape loops. You won't find any arms tugging them down or any springs threatening to snap the precious tape. IBM went with a wild new device, the vacuum column. Each side of the tape had a rectangular column that the tape drooped into.

Those columns were hooked up to a vacuum pump. Tape fit into the column perfectly, forming a seal. So you end up with a region of negative pressure under the tape. That causes the atmospheric pressure above the tape to push it down. That's a good deal more gentle than a metal arm in a spring because you're spreading the force out on a larger cross-section of tape. The result is that IBM is treating its acetate tapes with more care than Univax metal tape.

The next important feature is tape loop management. That is, how do you make sure each loop has enough tape in it? The Uniservo did this with a feedback system. It's real motors spun when the tension arms detected more tape was needed. IBM didn't use arms, so they needed a different method. The big blue solution is simple, but it sounds like science fiction. Their tape drive used vacuum sensors to provide feedback to a system of magnetic particle clutches. That sounds really cool, right?

But what does it mean? Both vacuum columns had a set of two vacuum sensors. Those sensors could tell the tape drive if they were at. atmospheric or vacuum pressure. If a sensor said it was in vacuum, that meant the tape loop had to be somewhere above it. By using two sensors in each column, the tape drive could tell if the loop was too low or too high. That information was used to spin the appropriate take-up motor on one of the reels. The clutch is the really cool part here. Each reel has its own magnetic particle clutch. Despite the name, these are actually pretty common devices. They're actually used in treadmills and a lot of

exercise equipment of all things. They also show up in some cars and some aircraft. A clutch is just a device that can conditionally transfer power. This is usually done in a very physical way, like pushing two gears together so they line up or tightening a belt against a transfer wheel. Those types of clutches turn on, so to speak, almost instantly. or at least with a pretty good jerk. Magnetic particle clutches work in a completely different way. They are, in very simple terms, boxes full of iron filings. You have two shafts that enter the box, but the shafts don't touch.

Normally, if you spin one shaft, the other would do nothing. It's just in a pile of sand. There's no hope to transfer power through sand. The trick to turning a magnetic particle clutch on is right there in the name. You apply a magnetic field. This can be done with a coil on the outside of the clutch. Pump in a few electrons, and everything starts to stick together. The iron particles clump up and start to transfer force from the input shaft to the output shaft.

There are two big benefits here. The first is that a magnetic particle clutch is fully electrically controlled. That makes it easy to work with in, say, a computer. The second is that these clutches are a little soft. When activated, a particle clutch will start to transfer power right away, but it has a ramp up. It doesn't just turn on all at once. It doesn't jerk. That soft start is nicer to acetate tape.

It prevents the 726 from outright shredding or snapping tapes. It also prevents slippage. There's no hard mechanism that's holding tape in position. And you can only apply so much force to acetate before it breaks. If the tape was jerked, it could break, but if it didn't, it could get pulled past the head. That would interrupt the entire drive. The rest of the drive is just as specialized as its loops. It's also, in general, just a step up from the Uniservo.

Remember how EMCC used one capstan? Well, IBM used two. I know that can sound flippant, but the dual capstan setup is another inspired choice. The 726 had one capstan that's always spinning forward, and one that's always spinning backwards. The capstens are engaged using pinch rollers. These are small, movable rollers that push the tape up against the capstan. To spin the tape forward, you pinch it. against the forward capstan. To go backwards, you pinch against the backwards one.

That makes it so the 726 can switch directions fast. The uniservo's one capstan, its so-called center drive, is a limiting factor here. To advance the tape, the center drive has to, from a dead stop, spin forward. To go backwards, it would have to, from a dead stop, start to spin backwards. switching directions requires spinning down, stopping, and then spinning up. That takes a lot more time than just pinching the tape. The tape can halt in a similar way. The 726 has a pair of stop capstan's. When you want to stop the tape, the moving capstan is unpinched, and the stop capsons are pinched.

Easy, fast, it just works. The other neat detail is how the 726 deals with the problem of friction. You see, IBM would opt for a pressure pad to push the tape against the head. When I first read that, I was surprised. I mean, after all, EMCC had issues using a pressure pad. They had to go to a few lengths to keep the tape sliding well. They had to use a buffer reel, after all. IBM's trick was to use a smooth glass pressure pad.

For IBM, that was enough. Now, I was trying to find some coefficient of friction tables to back up my thinking here. How friction calculations work as you refer to a table that has values for how strong friction is between a pair of materials. But, alas, my tomes fail me. The coefficient for metal on glass friction is pretty high. I'm guessing the coefficient for acetate or plastic on glass is just lower. That's probably why IBM can get away with just having a pressure pad, whereas EMCC couldn't have done that.

Remember, it's not that acetate tape was better than metal tape. The change in materials led to a different set of challenges. That's pretty clear when you compare these two drives. The 726 is, in large part, designed around using fragile tape. It runs slower than the uniservo. it uses these arcane vacuum columns to cradle its precious tape. Take the magnetic particle clutch as another example. Those were not off-the-shelf parts.

IBM designed and manufactured custom clutches. I've been working off a 1953 paper, IBM Magnetic Tape Reader and Recorder by W.S. Bus Lick for this section. He goes into detail on how the internal shapes of the clutches were devised. This is highly specialized stuff. The Uniservo, on the other hand, is much more simple. It can fling tape around by comparison because it's using stronger tape. But the major trade-off, well, that's where things start to look up for acetate. There were already supply chains for acetate magnetic tape from Buslick, quote,

a one-half-inch-wide version of this acetate film tape was obtained from the Minnesota Mining and Manufacturing Corporation for use with the IBM 726. EmCC can't say that. IBM didn't need to make a tape factory to support their drives. They just had to go to market and buy some tape. It wasn't exactly that easy, but it was a lot easier than what EMCC had to put up with. During development of the 726, IBM carried out a pile of tests to determine who they should buy tape from. H. William Nordyke Jr. presents a fascinating paper on the topic of tape testing in 1953. I'm realizing now I didn't say this earlier, but I've been basically living out of one proceedings this episode.

It's a book called Review of Input and Output Equipment Used in Computing Systems, Joint AIEE, E-E-I-R-E-A-M computer conference from 1953. Basically, all the papers I've been using this episode come from this review. Anyway, the Nordic paper. IBM initially had issues getting a hold of long runs of tape that would actually work. Nordic puts numbers to this, but it basically comes down to signal-to-noise ratio. All tapes have a certain background noise, a hiss. If you've ever listened to a cassette tape, you know,

what I'm talking about. For a tape to be useful for storing data, it needs to have a pretty good signal-to-noise ratio, that hiss needs to be at a certain level of quietness. The hiss is a property of the tape itself, and it will vary on a tape's length. So you can get regions of bad tape. EmCC faced the same issue, hence the factory test and punch setup. IBM wanted to ship reels, with 1,200 feet of tape. To do that, well, that was a hit or miss process. Nordic recounts the development of a test rig that sounds similar to the setup at EMCC. Tape would flow through the rig, and it would tell you when it encountered a bad region of hiss.

The IBM crew's first idea was you just splice out bad tape. But that was, well, that would not work, but it was so. soon discovered that the usual splices would not withstand machine operating conditions and storage conditions. Spices using the conventional splicing tape, no matter how carefully made, showed a tendency to pull apart along the diagonal cut and to ooze adhesive at the ends of the splicing tape and in the gap between the ends of the magnetic tape. End quote. Even with all the engineering and worked that into treating tape well, the 726 tended to rip apart splices. And oozing, well, forget about it. Imagine the havoc that could cause. You could wear down

your tape head that way. This is another one of those relatable experiences, right? Nordyke and his co-workers just spent, I'm sure, a lot of time getting their first run of 1,200 feet of tape all spliced up. They even had to build their bad spot detector to do it. They run the tape through the detector one last time, and everything comes up green. They've done it. They made their first working tape. The only thing left to do is actually use the tape in a computer. They mount it in a brand new 726.

And the futuristic machine rips it apart at the literal seams. Even though IBM could just go to market, that didn't mean the tape they bought was suitable. Nordike made it sound like these tape tests went on for quite a while. Early on, mass-produced tapes had different types of defects, but with time, things changed. During the project, tape got more reliable, and Nordike started seeing different types of defects. Oxide nodules were a major issue at the start of the project, but those disappeared and were replaced with contaminations from dust and stray particles. So, yes, acetate offered the advantage of mass production,

but this was still so early that mass production methods were still in flux. Magnetic tape wasn't some solved problem. It wasn't a commodity that could be adapted to computers in the same way that vacuum tubes were. The fact is that tape in the 1950s was the cutting edge of storage technology. All right. That does it for a dive into early magnetic tape. This one did surprise me.

Tape is unique among early digital technology. So many components of first-generation computers are borrowed. But tape? Not really. It had to be adapted to work with digital machines. That adaptation took a few forms. Bynac offers a very direct approach. That machine took existing magnetic tape,

a plastic kind, and use it for data. But the implementation was primitive. It couldn't do a whole lot with existing tape and its very simple drive. Bionac is a bit of a blur due to the incomplete information we have. It used acetate tape, but it was being developed side by side with a more sophisticated system. EMCC's larger machine, Univac, used custom-formulated tape. That was a medium made from scratch to be used with computers. It ran in drives designed for the peculiarities of metallic tape. IBM would work out how to use mass-produced acetate tapes, but it took some doing. Their first drive was a very specialized and sophisticated machine. At every turn, except maybe Bynac, we have very new technology in play. Tape didn't start out as an easier cost-reduced

solution. It was top of the line. Thanks for listening to Admin of Computing. I'll be back in two weeks' time with another episode. If you like the show, share it with a friend. Support me on Patreon. Go to Admin of Computing to get links to everything, and you can even listen to the official aftershow adjunct of computing. As always, thanks for listening and have a great rest of your day.