Inside A DEC Hard Drive

A lot of technology from the not-so-distant past doesn’t resemble modern versions very much. For a case in point, look at the DEC RS08 disk drive meant to pair with a circa 1970 PDP-8. Paired with an RF08 controller, this was state of the art, holding 262K 12-bit words with a blistering access speed of almost 63K/second unless you were plugged into 50Hz AC when it was closer to 50K/second. [Uniservo] had the disk unit, but not the controller. Someone else had a controller, but no disk drive. So [Uniservo] is shipping the disk to its new owner in a move worthy of a Reeses’ Peanutbutter Cup. The problem? The disk is super fragile and shipping is risky, so he decided to remove the platter for separate packing. Good thing for us, because we get a peek inside.

The nickel-cobalt platter looks like a thick LP record with heads underneath. As you might guess from the data transfer specification, the motor was just a common AC motor that rotated the platter against the head.

It wasn’t unusual to wear out the platter and flip it over to use the other side. There’s no telling what a replacement platter would cost. The device was the first drive DEC developed themselves. Unlike a modern disk, the heads actually made contact with the platter.

Because the heads are on the surface, disassembly was tricky. There’s a chance that crud on the platter could scrape the surface if the platter were to move. Once the platter is out of the way, you can see the many heads. There were 128 data tracks and six timing tracks (three were spares). These mount on springs, and at first glance they look like Molex connectors with tiny coils within.

In a day when gigabytes can fit on your keychain and are virtually free, it is amazing to contemplate the engineering and cost that went into a device like this. You can only imagine what people in 50 years will think of our disk drives.

We enjoy the PDP-8, although our version doesn’t need a physical disk drive. If you ever wonder what all the lights and switches are for, we can help.

40 thoughts on “Inside A DEC Hard Drive

    1. It was a Fixeded Head Disc FHD.
      It means that the head didn’t move and every track had it’s own head.
      I worked on a similar huge drive (128/256/512 heads) which had an compressor to LOWER the heads toward the plate when rotational speed was OK.

        1. That was just to make sure the track sectors were still in roughly the same spot, assuming the drive tracked the 0 sector from the spindle or motor position below and not by some bit encoding scheme… My concern is if it doesn’t get remounted is the exact same way in other dimensions it will be off balance

          1. The platter had already been flipped once since it was balanced at the factory, Marking the screws so they go back in the same holes they were found in is just being cautious. Maybe it makes no difference. This RS08 is not exactly a precision machined device.

  1. Take your complaints to the makers and hackers posting everything on YouTube. Maybe you can give them an alternative source of income to help them take you seriously.

    1. Since when is it about income? (Sarcasm, I know the “community” will praise sparking and adafruit all day long for selling a 25 cent shift register for 4.99 but YouTube’s are evil)

  2. > You can only imagine what people in 50 years will think of our disk drives.
    If it is helium filled spinning rust not much, because the helium would have leaked away and the head would be welded to the platter with the high vacuum (~0.0038 torr) inside.

    If it is data on a typical consumer SSD, they won’t retain your data for more than a year if unpluged at 30 C (~86 F). Enterprise SSD if powered off would loose data after 10 weeks at 30 C (~86 F).
    (ref: )

    The hardware will survive, but in the case of the helium drives the data will be nearly inaccessible and in the case of SSD’s the data will be long gone, unless it was stored in very cold storage.

      1. “Charge trap flash”, is the technology at their heart, same as EEPROM and even EPROM.

        The problem is that they are shoving fewer electrons into smaller charge traps, so it is due to defects and leakage. Smaller traps, are faster to charge, smaller traps are cheaper to manufacture because less silicon is used. And instead of just “on” and “off” being stored more information is being crammed into each cell by varying the number of electrons trapped in each cell ( ).

        At the current data densities on ALL storage devices, EVERY single read has FAULTS that are automatically corrected by the Error detection and correction layers buried deep within the devices.

        At the end of the day all things will eventually break, higher temperature equals faster leakage. It is probably related in some way to the Arrhenius equation, which when applied to electronics roughly translates to “Every 10°C (18°F) increase in temperature reduces the life of electronics by half”.

          1. Since if you look at the numbers from Intel, there is a direct related between temperature and data loss, I’d say that there is probably no direct relation to cosmic radiation (nor butterflies flapping their wings.

      2. If you look at page 27 of the linked PDF, the data from Intel, appears to half for every 5°C (9°F) increase in storage temperature before data loss. And the other interesting thing from that page is that the hotter the temperature that the device was when the data was written, the longer it can be stored!

  3. You erroneously confused the term “access speed” (more properly called “access time”) and “transfer rate”.

    Transfer rate is how fast you can read or write data from/to the drive. Typically measured in bytes per second.

    Access time or speed is the time to find the data to read or write. On a hard disk with a moving head, it’s the time for the head to move to the right track plus the time it takes for the disk to rotate the start of the track under the head. Since this is a fixed head drive, it would only have the rotational latency.

  4. You reminded me of a program (for MFM/RLL disks on PC’s) called spinrite that allowed you to modify the interleave to maximize the throughput, for your computers specific ISA controller card. That if you were reading data from the disk to the buffer on the disk controller, that there was enough time (in terms of distance or gap between tracks) between tracks to allow for the buffer to transfer the data to the computers RAM, before the next track was directly underneath the read head. So instead of 1,2,3,4,5,6,7,8,9,10…. (1:1 interleave) where you could read track 1 no problem, but would totally miss track 2 and have to wait for the disk to spin/revolve fully around again to attempt to retrieve the data from track 2. So you may end up with something like (7:1 interleave) 1,.,.,.,.,.,.,.,2,.,.,.,.,.,.,.,3,.,.,.,.,.,.,.,4,.,.,.,.,.,.,.,…

    And the same with writing data, in fact you could end up with a different optimal interleave for write (7:1 interleave) to reads (5:1 interleave). And the interleave with the longest duration would be chosen to make the best of a bad situation.

    1. Gosh, talking about interleaves brings up bad memories of trying to teach this concept at the community college level a decade ago (it was still testable on several major certification exams back then).

    2. if you were lucky enough to own an interface-card from something called a “hard-card” (or a card with simillar software), then you already have many software routines built-in to the ROM of the interface-card that DO NOT NEED DOS OR DEBUG, including an auto-interleave feature in a GUI (text-mode).

      if im not mistaken, regular cards had ROM but no way of executing without a direct CPU call (meaning you need a boot-disk with the debug util and know where in the mem-map your interface-card’s ROM lay, and where to execute within that chip) and often lacked features, such as text-mode GUI, and auto-interleave, which can take HOURS as it formats in EVRY interleave then chooses the fastest

      modern hard-drives are ONLY LOW-LEVEL FORMATTED ONCE AT THE FACTORY, the format you do using the DOS format command is actually a high-level format, this is not possible if the LOW-LEVEL format is not done or incompatible with the connected card, of course because modern hard-drive electronics are combined onto the same circuit-board as the interface-card, you should never have this problem. (IDE is an expansion BUS, not a HDD port).

      as far as i know, most modern hard-drives do not even include low-level formatting software on the drives ROM, and the ones that do require you to either use DOS DEBUG or solder a serial-port connector and level converter to access the command-line of the CPU or uC inside your hard-drive.

      back in the day, you used to have to back-up and re-LOW-LEVEL format (and re-interleave, and re-partition, and re-high-level format, and re-store data) every 5-10 years to ensure reliability.

      because modern HDDs are so reliable, even high-level re-formatting is technically optional, thats why there is a “quick-format” option included in “modern” OS’s. although i usually prefer full-high-level-format (to detect bad-sectors), ive actually deleted (deltree) and sys’d FAT32 drives before instead of formatting and those drives still booted and held data from before OS re-install for over 10 years after the re-install.

      just dont rely on a cavalier, if you have one, your either very lucky or your data is never to be accessed again, most get “stick-tion”, you need a security-torx bit and a clean room to fix it, assuming the heads parked in the parking spot, and you have spare heads… (lukily modern drives auto-align every time they spin-up unless the alignment track is damaged… by sticktion? darn)

      1. oh, and dont forget the possible bump on the spot where thge sticktion happened, if the heads hit that at speed, you can forget about your data, maybe the parking stuff can be forcefully disabled, but then what about the auto-align (unparking) procedure that happens every start-up? you gonna need a whole new circuit-board andor ROM to get that thing recovered, im not sure if drives back then had serial ports.

  5. I at least partially agree. I’m disappointed that so many HAD posts are just links to a video somewhere that I just don’t want to take the time to watch. It’s almost exciting when a HAD piece is an actual article that someone has written.

    1. I often listen to music while browsing, and I’ve become more and more annoyed by having to pause the music to listen to the video so some dude can cheat their 15 minutes out of Google to game the visibility algorithm that decides which videos go viral.

      I think it was Veritasium who had a good video about the subject. There’s basically so much content being shoved onto Youtube that most of it is never even seen by anyone unless the algorithm decides to push it to you.

  6. As someone who worked on DEC’s Westfield RK06 production line as a temp test technician summers during grad school, I’m getting a kick out of this.

    We didn’t do engineering internships at school, so I sort of did my own. I learned a whole lot during those summer jobs and I thank DEC and the Westfield management for the opportunity to see how real engineering results in a shipping product. I don’t think they sold many RK06s…it was almost immediately replaced by the double-density RK07.

  7. “A lot of technology from the not-so-distant past doesn’t resemble modern versions very much.”

    Well, I guess “resemble very much” is subjective. I see a lot of resemblance there to the inside of a fairly recent non-solid state drive.

    It reminds me of a washing machine. I remember as a kid going to the Henry Ford museum and seeing antique washing machines with the big, fully visible spinning drum. I thought it looked almost like some sort of alien technology and nothing like the plain white boxes in my house. Then the one at home broke and I saw the repair guy remove the shell. And there it is! Just like at the museum but with a few more wires, sensors and electronic devices.

  8. When I was in High School in the 1980’s, a classmates dad worked for Exxon who donated an “old computer” to our school. I think is was a IBM system 7, or such. When the school saw that it needed 60A 3 phase power, it was decided to pass it on to someone else. What I do remember about it was the disk drives had hydraulic stepper motors for the heads, with an open pool of hydraulic fluid in the base of the drive. I never did get the chance to see them spin up.

  9. Around 1980, I removed two working PDP-8i machines from a London Hotel. Each had one of these drives and the two computers were connected together. They ran the Front of House Billing system. The operating system was custom, written by the Hotel Chain’s programming team. It was located at the top of a spiral staircase, 1 rack for the PDP, another for the RS08 and third for the Intercommunication. Each unit drove 3 terminals and a line printer.

    My job was to disassemble it so it could be passed out to truck we had waiting outside. (Central London, just off Piccadilly Circus) Having got all 6 units onto the truck, we were informed there were another two pairs in the warehouse, so we returned to the College where I worked with 18 Racks.

    As I had seen the one I disassembled working, that was the set I reassembled over 6 months. The disc drive still worked, despite the rough treatment. A few years later, it was replaced with a Prime 2250 (Rabbit). The Dec had 10 years in hours on the counter.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.