If you think of a 1960s mainframe computer, it’s likely that your mental image includes alongside the cabinets with the blinkenlights, a row of reel-to-reel tape drives. These refrigerator-sized units had a superficial resemblance to an audio tape deck, but with the tape hanging down in a loop either side of the head assembly. This loop was held by a vacuum to allow faster random access speeds at the head, and this fascinates [Thorbjörn Jemander]. He’s trying to create a cassette tape drive that can load 64 kilobytes in ten seconds, so he’s starting by replicating the vacuum columns of old.
The video below is the first of a series on this project, and aside from explaining the tape drive’s operation, it’s really an in-depth exploration of centrifugal fan design. He discovers that it’s speed rather than special impeller design that matters, and in particular a closed impeller delivers the required vacuum. We like his home-made manometer in particular.
What he comes up with is a 3D printed contraption with a big 12 volt motor on the back, and a slot for a cassette on the front. It achieves the right pressure, and pulls the tape neatly down into a pair of loops. We’d be curious to know whether a faster motor such as you might find in a drone would deliver more for less drama, but we can see the genesis of a fascinating project here. Definitely a series to watch.
Meanwhile, if your interest extends to those early machine rooms, have a wallow in the past.
Youtube has its downs, but it’s great to see a silly project like this one come along every once in a while.
One thing he seems to be missing, those tape drives of old controlled both spools.
However, there were backup tape mechanisms which controlled both spools and used a “standard” audio cassette (with a notch in the middle of the top to identify it to a couple of micro switches), I’m pretty sure they write and read lot faster than a standard audio cassette recorder by using four or more tracks and possibly multi tones as well as running the tape faster.
I guess that would be cheating though?
Usagi has one of those Techtran units. He got it working and is using it to store and retrieve programs for his DG Nova 1200.
We used IBM 3420 tape drives in the data center I worked in as an operator in the mid 80s. We had 16 of them.
Give that a casette tape has around 50 dB SNR and 15 KHz bandwidth (according to random internet searches) the Shannon capacity is around 250 kbps, so it should not be that hard to reach the required 50 kbps speed he is targeting. That is on a single channel, multiple channels of course gives more capacity.
… as well described not even a quarter way into the video.
Way to say you didn’t watch the video without telling us you didn’t watch the video.
Do you also comment authoritatively on books you haven’t read?
He didn’t comment on the video though, but on the article?
The point of an article like this is to summarize the essentials of the matter for discussion. If the article itself can be summarized as “here, watch this video”, or, that you demand the reader to go to the source to get the information that should have been provided already, diminishes the article – deserved or not.
Who cares, it’s an interesting detail for those of us who aren’t about watch the full video.
I worked with these tape drive 40 years ago on a Univac mainframe. It was very funny!
My recollection is that the reels were behind glass as well, for better pressure control . Cool
Project though ! De-case some HD platters and make your own scale removable drives! The look of 80s Big Iron in the Glass Box with literally sucky tape drives and the Roulette Wheel of Removable Storage!! Clean those heads or have a job crap out! And when the reel hub lock failed — “ticker-tape parade!” We had one operator who was strictly forbidden from touching removable hard drives (0 for 4 @$2200 1982 dollars) ! #Good times
The vacuum columns controlled the actual movement of the tape past the heads. The motors would provide coarse motion as the vacuum columns controlled the fine movement of the tape. As I remember, there was a lot of service time involved in keeping the columns clean and well sealed. The motors were VERY large analog servo motors.
vacuum columns controlled the actual movement of the tape past the heads
Uh, no. Capstans did the actual transport. The vacuum columns were just the buffer so the reel motors could keep up with the fast starts and stops.
Another part the operator had to clean! 4,8,12-trichlorotriflourobenzene (I may have the isomer #s wrong. ) A SuperFund toxin IIRC! Good times!
Nice Warren Zevon reference
There were three engineering challenges for high-speed tape drives back in the day:
1) Getting the tape from zero to full speed in half the length of the inter-record gap on the tape, and then back down to zero in the same length. This was accomplished mainly with low-inertia coreless motors directly driving the capstans. In many drives, the capstans had the tape wrap around most of the capstan to provide enough static friction (along with the tension provided by the vacuum columns) to operate without pinch rollers, further reducing inertia.
2) Supplying the tape under tension to the capstans, which was the job of the vacuum columns. This eased the work on the capstan motors, because they only had to accelerate and decelerate a few feet to tape rather than a whole reel.
3) Supplying tape to the vacuum columns as required. In the IBM drives this was done using constantly running motors with bidirectional clutches, since the tape had to be moved in both directions.
Seems like the part of the job this project addresses is the least demanding of the three, and no mention was made of producing a specific amount of tension to the tape to get the tape pressure against the head right.
Why the vacuum columns? If his goal is as stated he doesn’t need to stop and start the tape record by record. He could make a streaming tape drive.
You don’t get a running start. That would be too easy for the point – the original device he’s copying did it to improve seek time on tape, so you could quickly roll backwards and forwards to act more like a random access memory (within a short section of the tape).
What point is that? The function of the big tape drives that had vacuum columns was to be able to have random access to any block on the tape. This is because magnetic tape was used to augment the main memory of the computer, allowing processing more data than the main memory could hold. As memory became cheaper and larger amounts became available, the role of tape drives shifted to mass storage, and streaming drives were developed that did this with much less cost, space, and power. You always made sure there was a long enough leader on the tape (after the Beginning Of Tape mark) so that you COULD get a running start. DECtape was an early example of a tape drive being directly driven without vacuum columns. These were the “floppy disks” of their day, and were used for personal data storage and backup. When I started working on computers, part of my work was maintaining streaming drives that used the same 3600 ft. reels of half-inch 9-track tape that the big drives did. These could back up a 35 MB hard drive in just a few minutes. IIRC, I used a 600′ tape as the system backup for my bench Unix machine, which I used often because Unix 4.2 was infamous for corrupting its filesystem (and thus the operating system) on any improper shutdown, meaning any shutdown that resulted from pretty much any hardware failure.
Anyway, unless the “point” was to try to make something that looks like ancient computers seen in ancient movies, vacuum columns would not be the best way to do this.
Also note that the stated objective was to LOAD 64 kB in 10 seconds. This does not sound like a random-access application. A streaming drive is definitely indicated. Also, there shouldn’t be any concern about how much tape you had to get up to speed, because NOBODY storing programs on cassettes would start anywhere in the first couple minutes of tape, where you had all kinds of dropouts. Ask anybody who actually did this in the 80s.
Moreover, the tape buffer allows you to run the tape faster than the cassette reels actually turn, because you’ve got extra tape in the buffer.
At normal cassette speed, if you have about 20 cm of tape already spooled out, you can run the tape 40% faster than the reels for ten seconds before you run out of buffer and have to slow down.
Maybe true for a cassette tape implementation, but ultimate speed in the reel-reel drives were certainly not limited by the reel rotation rate. At least not the motors: The tape itself would have quite a bit of centripetal force acting on it though, which the vacuum would need to act against.
For the mechanical point of it, the regular tape mechanism has the pickup reel pulling on the tape with tension against the capstan roller that controls the speed of the tape past the head, and also pulls the other reel behind it. There’s variable friction with both reels, which results in varying tape speed and flutter, which is actually more limiting to the data rate than the theoretical S/N ratio of the tape material itself.
By de-coupling the reel winding mechanism from the capstan with a slack tape buffer, you get much better speed control of the tape going past the read head, which lets you get closer to the theoretical data rate limits of the tape.
Given the robustness of the coding scheme against jitter, I suspect the more pragmatic reason for the vacuum tape buffer is so the tape doesn’t snap.
Isn’t jitter just like adding noise to the signal? If a regular good cassette deck has 0.1% flutter, that would be equivalent to raising the noise floor to -30 dB. That would put a hard limit around, dunno, 7-8 kbit/s?
Except for us humans of course, because we don’t care about being exactly on-key when it come to music. The computer could deal with it just the same, but the early computer systems weren’t fast enough to do the calculations.
kiloBytes, actually.
“.. and also pulls the other reel behind it.” Remember there are vacuum columns on both sides, so each reel motor is only concerned with keeping the height of its vacuum column within its operating limits. The tension on the tape is set by the area of the cross section of the column multiplied by the difference of air pressure on either side of the tape in the column. THAT is what the reel motor has to pull against, and it is independent of what is going on on the other side of the capstan.
Also, while the motors and the reels could probably turn quite fast, they didn’t have infinite acceleration. With the tape in buffer, you can start reading the tape while the big motors are still spinning up, so you don’t have to rewind for a bit of runway to pick up speed.
For the task of “read 64 kB off a tape”, this can happen at any time, so the device needs to be in a state to read the data now. If it first has to reverse, pick up speed, and then read the data, that’s going to add more seconds.
Way before Snowden, in the days of 9 track tapes, there was a persistent rumor among database professionals that amounted to:
‘The NSA (and it’s precursor) has records or every phone number/person you have ever called more than once (now known as metadata). They have a nationwide double linked list of acquaintances/friends. It is the worlds largest relational database.’
I expect this rumor, now known fact, was rooted in all the head up their ass local phone companies that were sending a tape or three of call records to Virginia every week.
I know, off topic.
Fun project, looking forward to the followup videos.
As for “the fastest” tape: the sinclair ZX microdrive could do 15KiB/s.
I quite liked the Idea of replicating the way those old tape spools worked with a standard compact audio cassette, but 15 minutes of video just for the vacuum was a bit disappointing. I’m not going to watch part 2.
That bit was interesting because it detailed how you actually get the vacuum – and that was the point of the video. If you start by assuming you simply have it, where’s the beef?
Just the comparison between an axial fan vs. centrifugal fan, and then detailing how the centrifugal fan actually performs in different configurations, has much wider appeal than just this particular application. It gives you a concrete sense of how different fans perform, so even if you’re not interested in building a tape drive yourself, this knowledge has other applications such as for building a fume extractor for soldering or a liquid cooling loop for a PC – and that’s a hallmark for a good informative presentation. It gives you something useful. Viewing this video, you actually understand something.
I’m definitely looking forward to part 2 and will not be surprised if there will be a part 3 4 or 5. This fellow has a nice way of approaching a certain problem and finding interesting solutions to solve them. The video about video on the Commodore PET is an excellent example.
To be honest I do not care about the problem that he’s trying to solve, nobody wants/needs such a device in the time and age of digital tape replacements and 16 MB REU cartridges. It will be noisy, incompatible with existing games and programs and one of a kind. This project is not about the destination, it’s about the trip to get there. And learning how and why things were done in the past. But, you have to admit (if you are a retro computing enthusiast) that this project is pretty cool it it works. and it would steal the show on any 80’s home computing exhibit.
Can’t wait for the follow up videos.
Cool project, cool approach.
Now they have to look into static build up and X-Ray emission perhaps, as per Sellotape in a vacuum.
Copal made some of these crazy fast DLT tape drive motors #PN12-60119-01 Bet that would spin out fast enough.
I used those Philips Super Ferro as a kid!!
I hate tapes. I was like a busy waiter at a honky tonk running with tapes of electronic CAD designs for updates or changes to feed a VAX serving workstations or another mini in search of old data entry batches.
I had a Coleco Adam decades ago, the tape drive in it would zoom back and forth, it always reminded me of the older reel units you’d see on the mainframes.
The coleco adam tape drive is a nice concept and the automatic tape play/stop/fwd/rev is awesome to look at. As a kid it must have felt like having a real spy-movie-like computer center in your own home. I had a C64… with a datasette… I had to press all the buttons myself, it was pretty low-tech but it loaded the games I wanted to play.