An easy to build programmer for 2704 / 2708 / MCM68764 / MCM68766 EPROMs

Introduction

Some time ago I needed to program an MCM68766 EPROM to upgrade the BIOS in an IBM PC 5150, and in the process discovered that hardware which can program these is rather difficult to come by. Despite having a collection EPROM programmers, including one fairly expensive model (which apparently could, but wouldn’t do it in practice) – in the end none were up to it.

So I set out to build my own. Knowing that the pin-out for 2708 EPROMs was quite similar to MCM68766 – I thought I’d tick that off while I was at it, as this is another type that is very difficult to find hardware to program.

The original prototype – designed late 2016 and mentioned on vcfed.org forums early 2017

Why is MCM68766 difficult to program?

There is nothing algorithmically complicated about programming one of these – the main difficulty centres on the high Vpp voltage (26V). Most cheap EPROM programms have a max Vpp of 13V. That having been said, there are some cheap programmers which support the 2716 EPROM, also needing a Vpp of 26V, so quite why MCM68766 is such a difficulty, may also depend on other things, such as pin arrangements, or perhaps just limited demand for it.

Why is 2708/2704 difficult to program?

This type of EPROM is genuinely more difficult to deal with. It also needs the same high Vpp voltage, as well as an additional cocktail of voltages. +26V, +12V, +5V and -5V being the full list. All of these are required during programming – quite a headache for the designers of universal programmers. Models that do support it, require a special adapter which doesn’t come cheap. This adapter typically supplies the extra voltages which aren’t practical to provide through the universal socket. Like the MCM68766 – they are algorithmically simple to program, so no complicated software stuff to worry about.

Design

The physical design is in the form of an Arduino Mega shield. I chose this because it was large enough to accommodate all of the components, and there would be enough I/O pins without needing I/O expanders or other complicated stuff.

The only supported Arduino is the Mega 2560 R3 or compatible.

When designing it I was aware that I was likely the only person that’d want something that did both, so have allowed for it to be constructed as a dual 270x/MCM6876x programmer, or 270x only / MCM6876x only.

When built in its dual type form, selection between devices is made by a 6PDT switch, which software can query before powering on the device, checking against the selected device in the user interface, limiting the chance of accidentally blowing an EPROM (which will certainly happen if the switch is in the wrong position).

When programming 270x EPROMs, the Arduino must be supplied with a regulated +12V supply as the shield does not generate its own +12V rail. There is a software check to ensure this voltage is correct before powering on the EPROM.

There is also a footprint for an RS-232 connection + MAX232 but this is not required for most applications as the Arduino can provide a serial interface through its USB port.

Constructing it as a dual type programmer

All components must be fitted except for the RS-232 section.

Constructing it as a MCM68764/MCM68766 programmer

I figured it would be easier to just show a picture of this. All fitted components are as specified on the schematic, unless omitted or replaced with a wire link.

SW1 is hard wired in the left hand position.

Constructing it as a 2704/2708 programmer

When constructing as a 2708 programmer, only the following components can be omitted: SW1, J1, U4, D2, C1, C8, C9, C10, C11, R9, R16.

SW1 must be hard wired in the right hand position.

Buying one pre-built

I don’t sell these things myself. Search for “2708 programmer” on eBay. I’ve seen a number of examples of people selling my design pre-built. There is now at least one “compatible” design out there – different hardware, but using my software.

Host software / firmware

Please see the main project page for more information.

Schematic

Can be downloaded here.

Gerbers

There are two versions of the gerbers. A “Long” version which has the exact dimensions of an Arduino Mega. Choose this if your PCB house doesn’t charge extra for exceeding 100x100mm dimensions. Total size is 101.5mm x 53.5mm.

The “Short” version has the tab at the end clipped to keep it under 100mm. Total size is 99.95mm x 53.5mm.

BOM

Can be downloaded from here (or here – CSV).

2708 Voltage range warning

When developing this shield I found that some 2708 devices are fussy about the voltage on the +12V rail – they won’t be damaged but they’ll not program successfully. The designers of the Arduino Mega have helpfully fitted a reverse protection diode (D1).

While this does prevent hapless newbies from blowing their Arduinos, unfortunately in our case it reduces the 12V supply down to 11.2V. For reliable 2708 programming either this diode must be replaced with a wire link, or the Arduino should be powered with a 13V supply.

82 thoughts on “An easy to build programmer for 2704 / 2708 / MCM68764 / MCM68766 EPROMs”

  1. Hello,

    I hope you don’t mind, but I am going to have some of your shields made up by Oshpark for myself and a couple of friends who are involved in the repair of Vintage computers.
    I have sourced some MCM68764/66 EPROMs and this will be great for replacing or upgrading old 24-pin DIP ROMs.
    If it’s OK, I’m going to share the ability to have the boards ordered from Oshpark as needed.
    No one, other than them, will make money off this and I will have the board design credited to you.

    Thank you for your work in designing this shield.

    Regards,
    Richard

    Reply

  3. I built your programmer and everything worked on the first try!
    I’m really excited about having programmed an EPROM 2708 for my old computer.
    Thank you so much, you were incredible, very good … and sorry for my bad English 🙂
    Greetings from Italy

    Reply

  8. Built today your 2708 programmer, worked from the first power on, great project, thank you for sharing. Was curious, because I used a 12V Zener instead of a 11V and 220uH inductors instead of a 180uH, but the voltage values were ok, and I was able to program a 2708 EPROM. Are I on a “dangerous” way with this componets?

    Reply

    1. The changes you’ve made wouldn’t make a lick of difference to operation. The zener is to keep Vgs below 20V but still allow fast gate charge time, so 12V will still be OK.

      As for the inductor, also not going to be an issue. Going higher can result in higher losses (+overheating) and slower regulation response time, but at 220uH it’s still going to be well within operating margins.

      Reply

  10. Thanks for the effort.
    I wanted to display the checksum display automatically when loading the binary file or reading out the EPROM.
    Couldn’t you get the source code after all?

    Reply

    1. Perhaps you could look at writing a specific programming tool which does exactly what you want? I’ll never be able to please everyone! I’d certainly be happy to share it here. Happy to provide any protocol information needed.

      Reply

  11. Would be nice if you shared the sources, so that you can customize the software according to your needs.
    Kind regards

    Reply

  12. I wanted to make the software a little more comfortable.
    For example, immediate CRC display in the user interface, use as a pure 2708 programmer and possibly an editor.
    I did not want to expect all of this from you.
    Kind regards

    Reply

    1. If you are looking to create a highly customised interface – I’d ask that you build it from scratch rather than customising mine. If you are in a position to integrate a binary editor – this should be no difficulty for you.

      As I’ve previously said, I do not intend to release the source for the Windows UI as I don’t want bastardised versions of it all over the place, which I then have to support.

      I have added the source for the command line version which should be a very good starting point:

      https://github.com/inaxeon/hvepromcmd

      Reply

  13. I last programmed a 2708 with a borrowed burner for my Commodore 64 club. 1990 or so. Always wanted to play with them since.
    I’ve been mostly wanting to burn spare chips for older test equipment and other stuff.

    Ideally I would like to just read older chips and burn to some pin-compatible new stuff. I assume eeproms for 2708 and newer are available.

    I have unused burners somewhere, but was going to order a modern burner like the TL866.

    Also for my Heathkit HERO 1 robots and et-3400a microprocessor trainer and peripheral adapter roms.

    Reply

  14. Hello Matt,

    I am maintaining old school arcade game pcb. As you know, it is difficult to use programmers who read and write old eproms now.

    I’d like to use your 2708 eprom programmer. I know you don’t sell it yourself, but it’s very difficult because there’s no place to buy it. I know one seller on eBay, but I can’t make a purchase because of the delivery problem. He said that shipping from Germany to Korea is difficult due to costs and customs clearance.

    I would appreciate it if you could sell the finished product or kit for me. Prefer finished products to minimize errors in production. I will pay for the cost and delivery charge by Paypal.

    I live in South Korea. I look forward to your positive answer.

    Thank you.

    Reply

    2. Hi,

      is anybody ordering/making these pcb’s?
      i am interested to have one…

      struggling with my 2708 here

      thanks,

      Wim

      Reply

  17. Hi Matt,
    congratulations to your outstanding projekt.
    I built a programmer for my Arcade repairs and it worked from the first try!

    Thank you very much for sharing your project!!!

    Reply

  20. Hi Matt,

    I built the 2708 version of your programmer. I’m getting “Chip is not blank at 0x0000, Data 0x00” on what I believe is a blank chip. If I save to a .bin file and open in a hex editor, I’m seeing 00 in every position. All the voltages looked correct when I ran the self test.

    Any guesses what could be wrong here?

    thanks
    Charlie

    Reply

      1. Hmm, the test looked good last night, but now I’m seeing wrong voltages in the tests. I do have 12v coming from my power supply.

        5v is reading 4.2v, -5v reading .3v, 12v reading 1v, 4.2v reading 1.8v. Sounds like I managed to burn something out!

        Also, no more buzz sound on the 25v step. This looks to be pulsing between 0v and ~10v, but it is difficult to tell the exact high number

        I had not tried the read test yet, (I’m assuming that is the final one with the resistors).

        I’m going to go back to my bench to inspect my soldering job for any mistakes and also double check that I didn’t put something in the wrong spot.

        Anything jump out at you as possible failed components that could do this?

        Reply

  21. My minipro programmer has a max VPP of 21V, this is not enough for some of the 2716 eprom I have. I have built this programmer (thanks for sharing you design by the way) and am thinking it should be able to support these eproms too. Is it possible to add support for the 2716?

    Reply

    1. I was looking at this recently. Unfortunately it’s not just a software tweak, significant change to the hardware would be required.

      Reply

      1. Yes, I was just looking at the pinouts and realised that VPP /CE and A10 would need to be switched around. Maybe it’s easier to design an adapter board that switches these pins to the right place? Still would need some adjustments to the software, or have a jumper tie A10 to gnd or vcc and program both halves separately.

        Reply

        1. Software changes not an issue, having an adapter would be a bit of a pain. It may make more sense to revise the board so that the switch changes between 2708/2716(32) and support MCM6876X with an adapter, given that few build it for this. But, considering that 2716 capable hardware can be had for $30 – I don’t think many would bother.

          Reply

        2. Interestingly the 2732 can be supported with a software change. Maybe it might be worth knocking up a small adapter for 2716’s…

          Reply

          1. If the 2732 can be done with only software, that would be great. Hoping you can add this feature and maybe even the 2717 with adapter board.

          2. I may be wrong about that. The sticking point is that all 2732’s that I could see are 21v Vpp, so there’s be a small hardware change to allow software selection of the programming voltage.

            Probably one for a future revision of the shield I think…

          3. Actually I have many 2732’s here that are 25v Vpp so I am hoping for support for these. The 21V Vpp 2732’s I can program with my TL866.

          5. There are many 2732’s with 25vpp I found a list here with different brands.

            2732 2732 EPROM* normal algorithm, 50msVpp=25.0V, Vcc*=5.0V
            AMD, Fujitsu, Intel,
            Mitsubishi, NEC, NS, Oki,
            UMC
            2732 2732 EPROM* normal algorithm, 5ms Vpp*=25.0V, Vcc*=5.0V TI
            2732 2732 4k*8 EPROM* 25VPP many
            2732 2732A EPROM* normal algorithm, 10msVpp=21.0V, Vcc*=5.0V TI
            2732 2732A EPROM* normal algorithm, 50msVpp=21.0V, Vcc*=5.0V
            AMD, Fujitsu, Intel,
            Mitsubishi, NEC, Oki, ST,
            Toshiba
            file:///C|/홈페이지/new_021010site/public_html/body/PARTS/EPROM.htm (4 / 12) [2003-04-01 ｿﾀﾈﾄ 5:14:20]
            EPROM
            2732 2732B EPROM* intelligent algorithm Vpp*=12.5V, Vcc*=6.0V Intel
            2732 2732D EPROM* normal algorithm, 50msVpp=25.0V, Vcc*=5.0V Toshiba
            2732 27C32 EPROM* normal algorithm, 50msVpp=25.0V, Vcc*=5.0V NS
            2732 27C32A EPROM* normal algorithm, 50msVpp=21.0V, Vcc*=5.0V Fujitsu
            2732 27C32B EPROM* interactive algorithm Vpp*=13.0V, Vcc*=6.0V NS
            2732 27C32H EPROM* normal algorithm, 10msVpp=25.0V, Vcc*=5.0V NS
            2732 27P32A EPROM* normal algorithm, 10msVpp=21.0V, Vcc*=5.0V TI
            2732 P2732A EPROM* normal algorithm, 10msVpp=21.0V, Vcc*=5.0V Intel

  22. Update on my issues above. It turned out the MC34063 was failed. I replaced it and now my voltages are correct.

    Unfortunately I am still getting “Chip is not blank at 0x0000, Data 0x00”. The resulting bin file is all 0s which I think means blank. The hardware tests all pass with the correct voltages and the read test with resistors works correctly for all the inputs.

    Any suggestions on what else to check?

    PS – Here’s a link to the bin file if that is of interest

    https://1drv.ms/u/s!AqiyUGOT8PbNw-tEeGj1u-EIkMvWZQ?e=ODDmBn

    Reply

  23. I just have one 2708. I will trying ordering some others in case this one is bad.

    I figured I have nothing to lose with trying to program this 2708, so I just turned off the blank check option and tried programming it anyway. It made a few passes and U1 exploded in a puff of smoke and sparks! Below is a pic of the U1 aftermath…

    I have U1 socketed and was able to replace it quickly. The next time I tried it was able to make all 100 passes but it doesn’t look like anything was written. I will try with a different 2708 and see if that helps

    U1 explosion!
    https://1drv.ms/u/s!AqiyUGOT8PbNw-tJrwbP2DYdOMMLjQ?e=XrR7YL

    Reply

  24. When burning a TMS2708JL Eprom, an error message occurs.
    The comparison immediately after the burn process reports “Verify error”.
    But if Verify is called individually after burning, it is correct.
    Can you insert a pause before the automatic verify in the software?

    Reply

  25. Hi Matt – You were right. I had a bad 2708. Replacement 2708s seem to work fine.

    One issue I noticed is blank / unused memory areas of the programmed chip have “FF” in them. I see this if I write a .bin file that is smaller than the total size of the 2708. This leads to a different checksum when comparing the chip contents to the original bin.

    For example, if I write this 2 line source file, (rom #9 from the arcade game, Magical Spot)

    https://1drv.ms/u/s!AqiyUGOT8PbNw-tzHjdzqXG2hdPQHA?e=cNS9C9

    it seems to write fine, but I then get this slightly different file when reading the 2708 and saving as a bin.

    https://1drv.ms/u/s!AqiyUGOT8PbNw-t0PJZXLhlLBvZZfg?e=RmrjwU

    If I view the original file in a hex editor, there are just 2 lines of code. The saved bin has those 2 lines, followed by many lines of “FF”.

    Is there any way to eliminate all the “FF”s in this situation so the files are identical?

    Reply

        1. Looks correct to me. 2708 hold 1024 bytes, so that’s what it should read back.

          The write checksum now includes the 0xFF padding, so will now match the read checksum.

          Reply

  26. I’ve been struggling to modify a TL866II+ to program some of my vintage eeproms. Namely 2716 and 2532. I ended up making a circuit that looks a lot like your VPP drivers. Alas, not much repeatable success with the 2532. I’d be curious to hear a write up on your CE, OE and VPP circuit theory here. And a mod of your programmer for 2532. Vpp=25 btw (and drops to 5v for reads)

    Reply

    1. There isn’t much theory to it. There are quite a few different approaches out there. Most common is NPN+PNP combo, works for most. My design will work with PNP transistors in place of the FETs provided a base resistor is added (with the exception of Q6). In my case I’m using P-Ch FETs because Rds(on) is lower than comparable circuit using a small PNP transistor, which gives a faster and cleaner pulse, but it probably doesn’t matter. It may well be possible to achieve comparable performance with a NPN+PNP transistor combo. I’ve not looked into it in detail.

      The 11v Zener diode isn’t needed, and I haven’t included this on subsequent designs. The idea was a circuit which charges the gate as quickly as possible, without exceeding Vgs(max). A simple divider circuit does the job just fine.

      I’ve also seen people use opto-isolators to drive Vpp pins too. Not an approach I’d use, but it definitely does work.

      Back in the heyday of these chips, engineers spent a lot of time fretting about this stuff because FETs were rare exotica and the performance of available PNP transistors was quite poor. These days just about anything will work.

      Reply

  27. Please help programmer 2708 what upload on windows fimware to arduino for programer 2708 file for upload and software for upload? Thanks usb version
    Get send to email? Thanks

    Reply

  29. Ooh, I really had trouble getting past “Operation timed out”.

    I could not succeed to burn the bootloader on my Win10 either. Had to pull put an old Mac which worked without problem. Still had doubts that it would work with the Win-GUI, and guess what? It didn’t. I still got “Operation timed out”.

    Hmm, tried to run the GUI as Administrator, and everything was butter smooth!

    So, if you get the “Operation timed out”, try running the GUI as Administrator!

    Cheers,
    Ola

    Reply

      1. Well, my Win10 is probably locked down in some “smart” way by my IT department, and I suppose that may be the case for others users as well.
        Just wanted you to know that this could happen.
        Thank you for an excellent project!
        Cheers,
        Ola

        Reply

  30. Hi
    your programmer works great for MHB8608 MHB8708 is the same witk 2708 and question get update win software for MHB2716 read get on your programmer only problem programming file is 2k is too large thanks a lot

    Reply

  31. Dear Matt,
    I have an Eprom Burner for your software built. To start burning I need to switch – but I cannot find a switch. For a MC Eprom???
    Could you help me, the seller did not answer.
    Regards
    Manfred

    Reply

  32. Which settings do I have to consider when programming an MCM68764.
    When starting the burn process, the error message appears:
    “The maximum number of per- byte exceeded….
    mfg

    Reply

  34. Hello
    Would it be possible to switch off branch D3-R6 via an Arduino port when burning an MCM68764 so that the full range of the programming voltage (0-25V) is applied to pin 20?
    Maybe an additional transistor.
    In the current circuit, the programming pulse only goes back to a minimum of 2V. Should go down to 0V.
    This prevents my Eprom from burning.
    I compared it with a data I / O programmer.
    Kind regards

    Reply

    1. That would require a revision of the design. There are quite a few builds of this shield out there, and this is the first report of this I’ve had, also I have 5 different samples of this chip which program successfully. Also, in the datasheet the pin only varies between 5V and 25V during programming.

      Is it possible to fix it by just removing D3?

      Reply

  35. Wouldn’t it be possible to set a signal to an unused port before programming to L and after programming to H, but before verifying?
    It ought to be feasible.
    Kind regards

    Reply

    1. Let me know how you get on. There is a small list of improvements I’d make for a V2 of the shield as it stands, perhaps this can join them.

      Reply

  37. The problem is probably elsewhere.
    I have spikes on programming voltage.
    Let’s see where they come from.
    The improvements in V2 would be interesting.

    Reply

    1. There is nothing interesting on the V2. Just minor tweaks.

      The problem you are seeing, I actually spent quite a bit of time studying it. It is only seen on one sample (the purple brazed ceramic one pictured above).

      After the programming pulse is applied the algorithm tries to read back the data, however it reads 0xFF regardless. After about 250ms the bits “settle” back to the correct state, far too slow to use a program-then-verify approach.

      I did try forcing the bus to 0x00 after each byte but this did nothing. The incorrect read-back data is held internally within the chip and driven onto the bus. The question is how to make the data bits settle faster after programming on these earliest revisions of the chip.

      Reply

  38. The programming of the MCM68764 only works for me after I have inserted a 1uF capacitor from collector Q3 to VCC. Programming was not possible before.
    Got it by accident.

    Reply

  39. Working on fixing programer would a broken d1 diode cause the 25vpp not reach pin 20 as it is present at d4

    Reply

Leave a Reply

Your email address will not be published. Required fields are marked *