Hackaday 68k: So You Want A Kit?


It’s yet another update to the Hackaday 68k, the wire-wrapped backplane computer that will eventually be serving up our retro site.

This is also a demo of Hackaday Projects, our new, fancy online documentation tool for all your adventures in making and tinkering. Did you know we’re having a contest on Hackaday Projects? Make something sci-fi, and you’re in the running for some really good prizes. There’s soldering stations, o-scopes, and a lot of other prizes being thrown at the winners. It’s awesome. First one to build a working Mr. Fusion wins.

In this update, I’m going to go over the beginnings of the video board, why Hammond enclosures are awesome and terrible at the same time, and some thoughts on turning this into a kit or product of some type. Click that, ‘Read more…’ link.


The Video Board

Like I’ve said before, I’m using the Yamaha V9938 video display processor as the graphics chip on this computer. It’s the easiest way I can get an 80×24 text mode – perfect for that *NIX goodness – and should be able to pull off some cool demoscene stuff. It’s pin compatible with the V9958, so I have that option, and it’s also fairly simple to interface to the rest of the computer:


That’s from the V9938 Technical Data Book. Big PDF warning there. On the right side of that graphic is the DRAM interface for the video memory and the pins for video output. There are a few different configurations ranging from 16k of VRAM to 128k of VRAM. Of course I’m going with the 128k option, using a quartet of TMS4464 DRAM chips I picked up from Jameco. Here’s the schematic of the VRAM interface:


V99X8I don’t want to wire wrap that. It’s a lot of fiddly, short bits of wire. It’s also extremely simple and won’t be seeing any changes in its design.  The solution to my laziness is, of course, to make a PCB.

Because the connections to the V9938 are just an 8-bit data bus and a few control signals, and the output is dead simple for composite output, this greatly minimizes the amount of wirewrapping I’ll need to do. Even if I tried wrapping a V9938, I’d run into a problem: the pin pitch isn’t 0.1″, rendering all my wirewrap adapters useless.

If you’re wondering about the physical size of the board, it’s just a wee bit larger than an Arduino. Yes, I know what you’re thinking, but a 128kB V99X8 will not fit on a standard Arduino shield. 16kB, maybe. In any event, when I get these boards back, I’ll have a go at driving it with an Arduino. Just because.

Mechanical Considerations

Yeah! Fancy Video!

Import DXFI’ve never had any luck with Hammond enclosures. The die-cast aluminum guitar pedals? I’ve ruined dozens of them drilling holes for pots, switches, and jacks. The enclosure for the 68k is no different. It’s a beautiful case, no doubt about that, but I am cursed with a mystical ability to always mess up the drilling, painting, or some random thing when it comes to Hammond enclosures.

The original plan for this backplane + enclosure combo was to have a small extension board on the front (hence, “frontplane”) that broke out the power and reset lines so this computer would at least look the part of an early 80s homebrew computer. Also, having a power and reset switch on the outside of the case is a good idea anyway.

PhysicalBecause of the complete failure of my ‘frontplane’ plan with the stock front panel, I’m going for something much, much cooler: a custom CNC’d aluminum panel. Right now I have holes for a power switch, a reset button, and a 5mm LED for power indication.

The basic circuit for this frontplane is very simple: The lines on the backplane are broken out on a huge 2×32 pin header. There’s also a small three-pin header for the PS_ON and POWER_OK lines for the ATX power supply. Ground the PS_ON line with the switch, and the power… uh, turns on. The PS_ON line provides +5V when the power is on. Attach the LED to that.

The reset circuit is the same from the CPU board: a Maxim DS1812 supervisory and reset circuit in a single TO-92 package.


Creating the milled front panel and new, improved frontplane was an interesting exercise in mechanical design. First, I created the front panel as a 3D model, exported the top view as a DXF, and imported that into Eagle. Then, I took the board file for the backplane and overlaid the holes. Then it’s just a simple matter of removing the parts and traces from the backplane I don’t need – everything except the pin headers – and making a board.

Apparently the Dimension layer in Eagle has a keepout. This board is far too simple for me to care about doing it properly.

So there you go. Fun adventures in mixing mechanical design with circuit board creation. This, like just about everything else relating to the 68k project, is up on the github.

Oh yeah, a kit

For some reason I can’t comprehend, a lot of people have asked if I’m going to make the 68k into a product, or at the very least a kit. I don’t quite understand the demand; the fun of homebrew computers is designing and building them.

That doesn’t mean I won’t entertain the idea. In its current form, though, a 68k kit would be absurdly expensive, take hours and hours to assemble – the RAM card alone would be three or four hours – and would have an extremely high number of unsatisfied buyers. It only takes one misplaced wire to screw the entire thing up, you know.

So, an improved,  single-PCB kit is the only option. This is months and months in the future, but here’s what I’m thinking:

  • Uses the currently-in-production 68SEC000
  • Already assembled.
  • A MiniITX or MicroATX motherboard form factor.
  • Uses 30 or 72-pin SIMMs for the RAM.
  • Some sort of expansion port.
  • User-updatable ROM.

That last bullet is the sticking point. I’ve been turning this around in my head for a while, and I can’t come up with a good way of doing it. The problem is I need a small amount (~64kB) of EEPROM or Flash that can be accessed on a parallel bus. That means 15 address lines, 16 data lines, and control signals. I need a way of reprogramming this, in system, with few additional parts, cheaply.

The obvious solution is to throw a big FPGA in the system for address decoding, an SPI bus, and in-system reprogramming of the ROM. That may end up being the eventual solution to this problem, but I’m thinking there’s an even more clever and cheaper way of doing things.

I’ve toyed around with doing the whole ‘in system ROM reprogramming’ thing in a 6502-based retrocomputer, and it is possible by using a microcontroller and a bunch of shift registers to program the ROM. This takes up a lot of board space and is extremely kludgy.

Another solution would be something like this amazing retrocomputer that actually should be a product. It uses a 40-pin PIC microcontroller as the RAM, ROM, and ACIA. It is, without question, the most innovative project in the retrocomputing world for the past few years and presents an interesting solution to the problem of in-system ROM programming: just put the ROM on a big microcontroller.

Are any of these ideas the right solution? I don’t know, because I’m not designing this computer as a product right now. This problem has been bothering me for a while, and I’d love to hear some more ideas. In any event, there’s plenty of space on my ROM board to prototype some in-system reprogramming. Come up with a good idea and I might put it in.

That’s it for now. You can continue to follow the progress of the Hackaday 68k over on Hackaday Projects. Be sure to comment and give a skull to the project. Seriously, give the project a skull. I’m losing to [Mathieu]’s Mooltipass project in the skull department. I need more skulls.




43 thoughts on “Hackaday 68k: So You Want A Kit?

      1. Is it possible? Of course there is few limits like speed and that it is so new versus 68k but there could be something like microcontroller in between.

        1. Modern SD is plenty fast enough for the needs of a 68K… if you could run it in 68K mode, In SPI mode (what most folks run when hacking around with SD cards), you’re limited to around 2Mb/s or so. It’s simple enough that this could be done with a simple microcontroller or even a CPLD. Keep in mind, back in the Mac days, Apple used the SCSI chip essentially as memory — the SCSI chip (they weren’t even using one of the better ones) controlled /DTACK directly.

          Another option would be to use a bootstrap micro. Rather than having the micro proxy for the SD card, it could read that SD card and dump some number of sectors into RAM, then hand the reigns over to the 68K and shut itself down.

          Of course, you don’t need much ROM, so why not just use NOR flash? You’ll need a flash-write routine for the 68K, and JTAG or something for the first programming, but that could be updated from files on an SD card via an all-RAM loader routine. Usually you want some boot-block code that doesn’t ever get overwritten, capable only of loading something from SD. So I’m still using SD, but it’s only as mass storage, not run-time ROM.

  1. Cool! You are making good progress. :)
    I was very suprised when people asked me for a Kit as well. Thus there will be a Kiwi kit in summer. :) Good luck!

  2. >The obvious solution is to throw a big FPGA in the system for address decoding,

    The obvious solution is to use a part with the SEC core and all of that stuff built in.. like I don’t know.. the MC68VZ328.

    1. MC68VZ328 is obsolete and not produced any more, most part vendors are not carrying it anymore. Not to mention that it comes only in a 144pin TQFP or a rather big BGA package :(

      Actually the older 68ks are easier to find on eBay and such than these Dragonballs, unless you want to desolder one from a Palm Pilot.

      The address decoding can be trivially handled by a bit of glue logic or a small CPLD, e.g. XC95xxXL series from Xillinx, or if you really want to go old school, by a GAL device.

      1. >MC68VZ328 is obsolete and not produced any more

        They are the same around the same price as the SEC from parts brokers.. and there are trays of them on ebay.

  3. A big FPGA? A tiny FPGA is enough to implement the entire 68K, V9938 and support logic. You can pick up an Altera Cyclone II based dev board (with 8K LEs), complete with VGA port, PS/2 keyboard port, 512KB of SRAM, 32MB of SDRAM, 8MB of flash ROM, an SD card socket, ADC and DAC chips, DS1302 clock chip and everything else you need for less than $80 on eBay.

  4. Oh I so would love a kit version f this. That sad – why ot just mount the backplane on standoffs? That way you could chuck say an inch or two of space between it and the back wall of the case.

  5. This might be a stupid question, so apologies in advance if this is the case, but what kind of video signal does it output? Or, in other words, what kind of monitor (TV?) can I connect it to? This might be a very good part for my experiments too, but I’d like to connect it to a VGA monitor ideally (even if this requires a couple more chips).

  6. Just use a pair of 39SF040 chips for the upgradable ROM. These are 8-bit parallel flash memory. You erase and write them in 4K blocks, all in software, by writing some magic numbers at the right address. When not trying to write them, they behave just like a standard ROM chip. No need for any additional logic.

    There are other flash chips like this, for example the AT29C040A, available in DIP packages for more wire wrapping fun. You can find them with different erase block sizes (256 bytes up to 16K) and also in smaller capacities if you don’t want 1MB of flash memory.

  7. Instant gratification (well, sooner than later):
    – minimig
    – kiwi (if you don’t mind the 8-bit data bus)

    Willing to wait and spend $$$:
    – s-100-based 68000 board

    I like the idea of using a micro to act as ROM, memory interface, I/O controller, etc…
    An atmega2650 has 86 I/O pins available. That sounds like enough to do the job on a 68000.

  8. I would be interested in a kit but only if it uses modern day hardware and not a propeller chip. An Arm chip can create VGA.

  9. I do like the idea of a kit. Single board… I don’t mind assembling unless there’s wire wrapping involved!

    I would do so man enthrawling OS projects with this. Brian’s right that designing and building is a huge part of the fun. But I think seeing first hand what the original programmers were dealing with on a system like this is its own set of engaging projects.

  10. >but I am cursed with a mystical ability to always mess up the drilling, painting, or some random thing when it comes to Hammond enclosures.

    They have CAD files for the enclosure. Plan it out in CAD where you want to put hole. Print it out on paper, use a glue stick to attach the paper. Use a punch to mark the hole. If you are worry about wandering bit, drill a tiny pilot hole first and use masking tape to protect painted surfaces.
    I use Black & Decker Bullet® Drill Bits for drilling. They have a finer pilot bit so they don’t wandering as much and they tend to give a cleaner cut that require less deburring.

  11. FYI: There are already FPGA based 68xxx FOSS projects out there that have better specs with (closer to) modern day part and properly layout PCB.

    MIST – Atari ST / Amiga fpga CLONE:
    FPGA for 68xxxx, SDRAM, VGA, audio; uses ARM7 chip for peripheral and USB support. 199.99 EUR comes with preassembled PCB + a case!

    Minimig – Amiga clone
    FPGA for Amiga chipset + external 68xxx CPU/Microcontroller. VGA output. $163.51 US

  12. How about using a type of flash memory which was designed for this explicit purpose: reprogramming and in-place execution? That is, NOR flash. For example, Micron M29F200. 48-pin TSOP, 20 dedicated address lines, 16 dedicated data lines, and control signals.

    1. There is the issue of preprogramming those TSOP parts defore soldering them down for initial development as neither the 68xxx nor the parallel FLASH memory comes with easy in system programming interface.

      There are sockets for them BTW.

      1. Okay, it also comes in SOIC. There are widely available “clothes-peg” style SOIC adapters.

        Depending on bus architecture (not familiar with 68k), it may also be possible to program them in-system, e.g. by adding an extension slot and making the programmer act as a bus master.

        (By the way, I’m really disturbed how the cheapest way to access a flash memory through a parallel interface is, apparently, “throw in a big FPGA”. That’s literally the most expensive possible way except, perhaps, using an ASIC. ಠ_ಠ)

      2. The problem for the parallel FLASH is the high I/O count required for programming, hence CPLD/FPGA/microcontroller solutions.

        You could use 74XX595 for driving the address/data line and read back with a 74XX597 if you want to. Those chip have tristate output and you can put the 68xxx in reset and/or use bus request to force it to tristate its memory interface. You can Bitbanging the programmer on a parallel port. You can make a programmer for it in a slot with discrete, but by the time it is done, you are probably better off putting it in a FPGA/CPLD as you can reprogram them and add a lot more functionality for the system. Why stop at programming when you can have VGA or even the CPU core inside the FPGA?

        I built a programmer with ATMega8 with V-USB in about 3 hours on a bread board with most of the time spent on stripping wires as I have already the PC host software, but I was building a quick & dirty programmer not a 68xxx computer in the first place! So easily to get side tracked and add complexity because the parts are being in system programming friendly.

        1. What? To quote the article:
          > The problem is I need a small amount (~64kB) of EEPROM or Flash that can be accessed on a parallel bus. That means 15 address lines, 16 data lines, and control signals.
          I offered a chip which suits these requirements almost exactly. (It’s a bit too big, but I couldn’t find a small flash with a 16-bit interface. AT29C512-70JU is a 64k flash with a 8-bit one. Also, costs less than $1@1.)

        2. What about using… the 68K CPU to do this?
          The NOR-flash chips are designed to interface with a CPU bus. There are address pins, data pins, and the usual read/write signals.

          Just preload the NOR-Flash (well, ok, to do this you need some other device that can program it) with some boot code that sets up the CPU, copy itself to RAM, then runs from there, configures a serial port, downloads a binary over xmodem and write it to the NOR-flash. And there you have your in-system programming.

          Once you have a more complete OS loaded this way, you can use other ways to reprogram the memory as well (loading from whatever mass storage device you use).

          1. >Just preload the NOR-Flash (well, ok, to do this you need some other device
            >that can program it) with some boot code that sets up the CPU,

            Indirect programming of a small serial bootloader to the flash using boundary scan on a CPLD or FPGA connected to the address and data bus would also work.

        3. The way this project is done tells me that he is a person that rushes ahead – prefer to wire wrap a 68xxx machine instead of properly designing the whole system before implementing it. Having a fully debugged monitor/bootloader (e.g developed with a emulator) isn’t likely MO. There is going to be a dozen of crash/head scratching/burn cycles depending on the skill levels.

  13. I would want a math coprocessor to be included with this kit. I like doing ridiculous floating point operators and vector drawings. I would be interested if it did come with the coprocessor. It looks nice, and I hope it’s successful.

  14. Seeing the “moved everything .03 inches” on github reminded me of my little 3D print for a PIR sensor so much lol, it was actually around that size (.025 but rounded to .03) which was the radius of screw holes. I measured every aspect of the board for at least 24 hours. If I didn’t self-impose size constraints to save money it’d be a lot easier, but I (stupidly) thought it’d be easy to get any size screw and the box was just slightly off (so irritating). Having a 2nd go w/ it I could make it perfect. Just needed to drill it though.

    I can’t get in your head as far as what you want for the enclosure, just take a deep breath and make cheap proto’s; because that’s the easy part. :p

  15. You could have two separate sockets on the board, one that responds on 0xFF0000/0x0000 and the other at another address and just physically swap them when you want to use the other EEPROM for boot.
    Or modify the reset vector circuit and add an AND gate with one input the BOOT signal and the other from a switch. The Switch would also control one or more bits on the addresses decoders for the ROMs.

    This way to reprogram the BIOS, you’d send data to a specific address (Whichever one the secondary chip is listening on) then power-down, flip the switch, and power-on to use the other ROM.

    Some modifications:
    Add a couple flip-flops at a specific address and use the to toggle the WE pin on the secondary ROM and maybe one to control which ROM is the boot ROM (This Flip-flop should be battery-backed so that this value is preserved between power cycles)

    1. I used 2 D type F/F to latch in the mouse buttons the /reset get deasserted. One of them drives the Bus Request for the 68000 and the other one on the 68020 accelerator to select the CPU. The other F/F was to select the ROM. You don’t need to power cycle the computer as reset (aka cold start) is a safe point to swap CPU or ROM as the system is initialized to a known state. There are non-volatile I/O chips if you don’t want dip switches. e.g. PCA8550 – “4-bit multiplexed/1-bit latched 5-bit I2C EEPROM DIP switch”

      I would imagine that something like that can also be used to drive the upper address pin if you want a single set of ROM for fail safe/dual firmware. As [whitequark] pointed out FLASH do not come in small sizes these days.

    2. In my first paying gig, I added a quad OR gate on my upper 4 address line to modify the ROM page of the boot rom in bootloader mode. The Fxxxx page of the ROM is used in bootloader mode regardless of what the actual ROM address was driven (aka aliasing), so all the vectors are mapped there. Bootloader is small, so 1/16 of the chip is not a limitation.

      Copy the code from FLASH to RAM, you can then turn off the mapping signal and reprogram the ROM (as FLASH is not accessible in programming mode). You can write protect the bootstrap FLASH page by locking it if you want to.

      The beauty is of this is that only 1 set of ROM chip is required and the whole ROM space is also accessible. 68000 requires 16-bit ROM, so you don;t want to add a 2nd set of chips.

  16. I would look at NVRAM. It might be more parts, but should be simple. Have a battery and bootstrap the contents in somehow. Switch panel bootloader if you’re hc. Or just a separate microcontroller (raspi, ****duino…) board for the “flashing”.

  17. As someone who’s currently attempted to build a 68k machine for myself, something I’ve wanted to do for many years, I’m a little disturbed by the suggestion that it should be a pre-made kit, or should contain microcontrollers or FPGAs or whatever.

    I know we all do things for different reasons, but to me the fun of such a machine is that it’s built using relatively simple and reasonably period appropriate parts. If you put an FPGA or a microcontroller in it to emulate half the machine, you may as well go the whole hog and do away with all the other parts, it’s already become a completely different project by that point. People have built many different complete computers using almost only AVRs or PICs or Propellers or FPGAs, and I don’t want to put them down for it, but I think it’d be a shame to turn this 68K project into something else.

    There are a million ways I could have already completed my project, had I given in and used microcontrollers as interfaces/processors or FPGAs as glue logic/a bus arbitrator/etc, but that feels far too much like cheating to me. I’d be screwing myself out of the potential for understanding which comes with using those original parts, and I wouldn’t get to experience some of what computer designers in the 80s went through in trying to build their machines. Building your own “retro” computer is a brilliant way around the difficulty in understanding the massive complex, super high speed and highly integrated PCs of today, despite the parts being old or difficult/expensive to come by, it would be a shame to throw that opportunity away by “cheating” in my opinion.

    Still, each to their own, I guess if you’re a software guy then I can see the value in having a new and exciting platform to explore, or… I dunno… if you just like the idea of owning a mostly useless trinket to brag about? :D

    1. *attempting, *68K, *the understanding. *massively…

      Hey, HAD, how about that edit function for the tired, caffeine-deprived and rabidly-foaming-at-the-mouth comments? :P

    2. One of the problems of building your own computer without it being a clone of existing ones is that you have to write every single darn piece of software. It gets a bit tiring after the initial excitement worn off and now you have a piece of expensive piece of “trophy” sitting on the shelf doing nothing. Unless you are a software guy that has the NIH symptoms. Been there with old non-popular 8-bit computers.

      If you instead of going the FPGA path, you can explore the different configurations. At some point you can also emulate other machines and run their software. You can even make it into something useful like a logic analyzer etc.

      As for the old chips… You can learn even more if you implement them on your own with just from the datasheet timing diagrams and register definition as specs instead of just buying them and wiring them up.

      1. Fair points well made. I think my point, if I had one, was probably that it seems a shame to muddy the historical context of the thing, and that I really l like the visual context, being able to see “this bit connects to that bit, which sends data via that bus to these bits” and so on. As I jokingly mentioned above, I really do believe in “each to their own”, I can see value in FPGAing it or throwing in microcontrollers as peripherals, but I don’t want to see the old methods get lost because of that. I dp appreciate all efforts to compute in the style of ye olden computer though, so good luck to anybody partaking in such projects. Just, maybe leave this one as it is? :}

  18. I was thinking about doing basically the same thing and now that he’s gone that way I may too. Those Yamaha MSX graphics chips are nice and easy to use almost like a few programmable registers I had one of them running on a sinclair QL in 198?. Which dates the tech in the whole project.I actually cheated and used this project’s framestore board for the Graphics


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s