Eurorack modular 8085 computer

[ChristofferB]

Finally have some progress to show off!
Got my Schroff 19'' subrack assembled, and backplane figured out, so now it's just a question of wiring the boards.

I think one of my first boards should have an IDC connector at the very edge, to connect directly to my logic analyzer. Thus I can check function without having a complete I/O system. The LA (TTI LA4800) has 74HCT input buffers, so I should just be able to run some ribbon cable directly from my bus to the LA.

Here's some photos of the system so far (with cameo of a ZX spectrum that I'm gonna use as serial terminal)
The empty 3U enclosure beneath the card cage is unoccupied at the moment. Might become floppy station at some point. The flat 4U top box is destined to become a front panel (altair/pdp-11 style)
having its own driver card in the subrack.

[MK14]

I will be VERY jealous, if you make that large top panel, look (similar to) like this:



With tons of flashing computer lights and switches.

I did not even know you could use the ZX Spectrum as a terminal. I did not think it had RS232 or anything like that. I was never really involved much with the spectrums. Maybe it is a bitbanged port, or add on interface, or something ?

[ChristofferB]

Wow, that does look incredible. Though having a control/indicator panel that large, one gotta make very sure every light has a function - else it's just like carrying around a slide rule without knowing how to use it.

I'd really like using hex encoder wheels or knobs instead of binary switches though. They're just pricey if you don't find them used.

You got me there. The ZX spectrum doesn't have a serial port, but the expansion interface does!
Still less of a sin than hooking up a modern pc with a usb->rs232 converter to a system that's supposed to be retro.

[MK14]

Apparently, (from memory), the Cray1 series of super computers (in their day, 1970's), had just such an OPTIONAL flashing lights panel(s). But it was mainly for show, since (if I remember the rumor correctly), management (who PAY for the computer, but DON'T have a clue about it!), need to see something like that (flashing lights panel), to feel they have just wisely spent, $25,000,000 (or whatever the price paid, was).

You can get some very nice hex/dec/oct switches. I could not quickly find a good image of them. But the image I have found, is somewhat close/similar.

[ChristofferB]

Hah! That's amazing. Maybe I should just have a 4U 19'' panel above my machine with blinkenlights and no connection to the machine - just being run by some sequential logic and a clock...

Yes that was the encoder wheels I was talking about. But those are strictly in the "scavenge only" section of what is affordable, usually.

Maybe just 16 way rotary switches, but they're hard to find too. Cheapest solution would be to go octal with those. But that hurts my brain too much.

Cheap sliding switches would be the most economical option, but they're just so nasty to use (and mount).

[MK14]

Yes, you could make a big string of randomly flashing neons, by using a few passive components. No semiconductors would be needed. Fun to look at, but completely non-programmable!

I've had a look around for the switches.

These are (claimed to be) brand new, hex, thumbswitches. From the USA, at about $5 each, but you may be able to make a bulk price offer. Offer function is available on page.

I don't know what you consider "expensive". Switches like that have *ALWAYS* been expensive (if I remember, correctly), even when they were commonly used/available. Presumably there is a lot of work to manufacture quality ones, and the potential sales are relatively small. Since only test equipment and similar, commonly used them.
Most/all mass produced stuff (for the consumer market), DOESN'T use them.

http://www.ebay.co.uk/itm/PHGA-3114-CHERRY-ELECTRICAL-SWITCH-THUMBWHEEL-HEX-BLK-MATTE-/111229210264?hash=item19e5c70e98:g:zO4AAMXQWzNSmk5C



On the other hand, the following type is what I was originally looking for. As it seems to allow, faster/easier setting. (Keypads are probably quickest, though).

[ChristofferB]

Those are pretty nice, even though it looks as if you'd need a stick to hit the up/down buttons.

I'm sure they've always been expensive, mechanical components like that often are, and it's difficult to lower that cost without making them really crappy. Those indicators/reduction gear/friction brake knobs for multiturn potentiometers are very pricey too.

Actually, if one didn't mind putting some design time/circuitry into it, it wouldn't be too hard to replicate their function with logic. A 4 bit hex up/down counter is only a bit of TTL logic, and a HEX display for that can be accomplished with an EPROM programmed as a look up table.
That's actually a really neat solution! and I get the 7segment blinkenlights too!

The 74169 would work great for that, actually, having a positive and negative pulse button for the U/D line, and then maybe some ordinary switches to load a value, if one wanted.

[MK14]

It wouldn't necessarily be so easy to replace the mechanical switches, with mass produced (general purpose), electronics.
Because the up/down buttons, would need debouncing circuitry, be able to cope with a huge range of logic voltages (from really low, to high, e.g. 3.3V, 5V, ECL, very low V, etc).
If the item was battery operated then mechanical digits, would consume zero power. But bright led displays (to show the switch positions), would probably consume a fair bit.
Quality push switches, get high prices as well, which it would still need.
Also if there are already other led displays in the item. Having even more for the input switches (hex), may be too over-powering.

The modern solution, seems to be LCD screens. But the cheap, dot matrix (such as 2 lines by 16 character) ones, seem to be NOT very clear (in MY opinion). The contrast is usually NOT very good, and they tend to be too small, to easily read. E.g. Compare simple calculators, with relatively good/readable LCDs, with programmable, 2 x 16 char, calculators. The LCD screen (sometimes graphical, as well), is usually of somewhat poor (unreadable), quality, in my experience.

The final button set, I showed above, are VERY usable, WITHOUT a stick. But then, I assume, you will be some distance from your rack computer. So it could be different, for you.

The keyboards (in my experience), from the 1960's .. 1980's, were of much better quality, than the modern (membrane/cheap) ones. They had a substantial, quality/solid feel to them. (But I guess poor quality ones, were available, then, maybe).
Probably the reed relay ones, made nice/quality keyboards, when they use to commonly do keyboards, that way.

BUT you still can get quality, mechanical keyboards, such as Cherry (PC). So things are not too bad.

[ChristofferB]

You're right, it'd consume a fair bit of juice. It'd need 3 EPROMS alone, just as lookup tables.
But the 74193 (better than the 169) just has an "up" and a "down" line - add a 74121 to each button gets the necessary debouncing - sure, it'd be TTL or CMOS only, but if other logic levels were needed, adding in a level converter wouldn't be terrible.

They're likely usable, but if one has a front panel where you'd need to punch in an entire bootstrap routine each time you turn on the system, ease of access becomes vital.

-I agree that older keyboards (with actual separate switches) are much nicer. Though, one of the lousiest keyboards I've ever tried is on an old 1974-vintage dumb terminal. Mechanical plastic plungers with conducting foam pads pressing pcb switch pads. shrieky, creaky and terrible.

[MK14]

Is there any way you could make an extractable, hex keyboard + led displays/lights ?

I.e. You could leave it clipped to the racks, front panel, and it would look cool/nice.

OR you could pull it off, and put it near, where you are sitting. For convenient/quick, boot program entry. Until your paper tape, floppy or whatever boot mechanism, you want to use, later. Or Eproms etc. Is ready and in use.

Alternatively, you could have two. One for the front panel, and another for placing on your desk. Via a plug in ribbon cable, or something.

Just a set of seven segment displays and a pre-made hex keypad, are both low cost and fairly quick to build. If you end up wanting two of them.

Part of the real pain in the neck, and TERRIBLE ergonomics, was a feature, of real-life, early computers. (The ones with front panel switches, which MUST have them used, to get the initial boot loader, inputed). Presumably (I think), every time the computer is re-started, you have to put the boot loader in from SCRATCH. Roms/Eproms, were NOT always around, I guess.

[ChristofferB]

A hex keypad is of cause the most handy, but that requires some interface and a scanning routine. Thus having a hex keypad for entering a bootloader would be silly - you'd need a bootloader for the keypad to work! The original switch front panels accessed the memory directly, halting the processor and tri-stating its busses. That's probably too involved, and will require a lot of circuitry -even worse: a lot of thinking!

I have a nice hex keypad with a ribbon cable and DIP header I'll use, I think. Then I'll have just the 7segment displays in rack. I've found an old intel 8279 IC. That's a very interesting keyboard/display controller, the one used on the intel MCS-85 and -88 microcomputer trainer, I might use that. Only issue is I only have the one, so if it's bad, or dies in the process, finding a spare might be difficult.
Actually I just found some on Ebay. They're not so rare afterall. I think I'll do that. They have a lot of different modes, from all 64-key keyboard to all 7-seg displays and most combinations.

[MK14]

There are "tricks" to using raw (unencoded) hex keypads, WITHOUT boot loaders. Typing straight into memory, with few or no latches.

There are various solutions, possible.

E.g.
Diode (steering) from the keyboard switches to create the hex digits. (Simple diode ROM, if you like).

Allowing the MCU to create the address pattern, by artificially giving it an I/O map, which is all zeros (or all ones), via pull ups (or downs). While typing in the boot loader program. It relies on the "NOP" or nop like instructions, via the continuous $00, or $FF's, making the MCU reliably increment to the next address location (in a kind of single step mode). How exactly you do this, and its viability, depends on the specific microprocessor.
The 4 bits of hex information from the keyboard, can then be written into ram (assuming it has separate 4 bit organisation, otherwise at least one 4 bit latch would be needed). Via some glue logic and/or clever circuitry.

E.g.
If I remember correctly, some early microprocessor stuff, did just that. To allow simple switches, to enter programs, WITHOUT needing to worry about the addresses. Such as the "SCAMP" computer (from memory, sorry if I'm wrong).

BUT you are quite right. It is NOT easy to use the hex keypad without a boot loader program, already being present.
Some of these technical challenges are NOT so obvious, as we take so much for granted these days, computing equipment wise.

[ChristofferB]

Wow, that's actually amazing. Never knew you could do that. Sounds logical enough, but yeah, there must be considerable design obstacles.
Here's the chip, in an example circuit, I'll use:
http://8085projects.info/KEYBOARD-AND-DISPLAY-INTERFACE-USING-8279.html
It seem to have only very few registers, and it handles all 7segment refresh, keyboard scanning and is interrupt driven! no need to waste cycles with reading keypads!

The ergonomics might be a hassle, though. I might go with a sloped desktop box instead of a rack unit. Dunno yet.

[MK14]

The 8279 is an interesting chip. With the right software, it will be useful, for what you want to do.

This is part of the reason, why modern day FPGAs and microcontrollers are so useful/versatile. If you were doing this as a modern design. You could just use a simple microcontroller or FPGA/CPLD to "create" the hex keypad and the displays.

The next $64,000,000 question is going to be, how small/tiny can you get the boot loader program to be, in order to just load in (or allow to be typed in), a "proper", BIG boot loader program ?

[ChristofferB]

Yeah, that's right, but programmable logic kinda goes that step too far in taking all the fun out of it.

I plan on using a programmable interrupt controller for function keys. That must be the easiest solution. So a minimal bootloader would be:
 - Initialize 8279 (couple of bytes)
 - Initialize interrupt controller (8259 I think) (couple of bytes)
 - on interrupt from keypad, save data to addr, increment addr, repeat
 - run from addr when "run" interrupt key is depressed

I think that could be accomplished in less than 32 bytes.
So two 2x 4-16 decoders and a diode matrix memory could run this without an actual ROM chip.

That would be almost a cool enough waste of time to try!

[MK14]

Yeah, that's right, but programmable logic kinda goes that step too far in taking all the fun out of it.

I plan on using a programmable interrupt controller for function keys. That must be the easiest solution. So a minimal bootloader would be:
 - Initialize 8279 (couple of bytes)
 - Initialize interrupt controller (8259 I think) (couple of bytes)
 - on interrupt from keypad, save data to addr, increment addr, repeat
 - run from addr when "run" interrupt key is depressed

I think that could be accomplished in less than 32 bytes.
So two 2x 4-16 decoders and a diode matrix memory could run this without an actual ROM chip.

That would be almost a cool enough waste of time to try!

Sorry, I was NOT clear enough in my original post. I meant "*IF*" you were doing it in the modern style. Of course, since you are NOT doing it that way, this section of this thread, is working out ways of doing it. I.e. FPGAs/CPLDs/modern MCUs are rules OUT! (Sorry for any confusion).

Yes I agree, diode ROMS, are real fun. A long time ago, that is one of the methods, they use to use, to do stuff like that. Also because masks are VERY expensive (so only suitable for high volumes or clients with very deep pockets) and originally EPROMS and PROMS did not exist. Diode matrix/steering ROMS, was the way they had to do it. But there were other solutions, I guess.

Another solution. Which would be authentic (time wise, I think). Would be to use a matrix plug board, of the type they use to use in higher end, analogue synthesizers. Hence you would only need to create the boot loader, ONCE, via the plug board. But could experimentally/playfully mess with it, to your hearts content.



But don't worry. Being realistic, I'm sure you are much better off, using the steps, you have recently posted about, to solve the problem. I am really just making conversation here.

If it had been a computer made out of RAW transistors (i.e. 1950s, and NO integrated circuits, whatsoever!), ONLY. Then maybe the plug board would have made more sense.

[SL4P]

...
The next $64,000,000 question is going to be, how small/tiny can you get the boot loader program to be, in order to just load in (or allow to be typed in), a "proper", BIG boot loader program ?

Ahhh. I remember back in my days with DEC, working on PDP-11s, we used the front panel toggle switches to thump in about 20 words - then reset the address, and release the RUN switch - to boot the disk controller, and everything from there on was rosy!
Slightly later machines had an Intel 4004 based 'front panel emulator' that loaded the boot code with just a couple of switch actions... but no blinkenlights!

[MK14]

To @ChristofferB

We were talking about your Proms (Intel), and a programmer for them.

The following vintage magazine, seems to include a programmer project for them. I'm not 100% sure if it applies to the ones you have, and/or includes enough details to use. But probably interesting, anyway. It goes into a HUGE amount of detail, into these "new" at the time, Proms. (Maybe you have Eproms rather than Proms, and I don't know if that is similar enough ?).

http://www.americanradiohistory.com/Archive-Byte/70s/Byte-1976-05.pdf

Example post, where we were talking about it:

Well most EPROM's you find in gear has these metalized, it looks, labels.
I have never had any trouble with accidental erasure, but I'm a bit cautious about photographing a board with flash that has unprotected EPROM's. Might be nonsense, dunno.

I've by the way recently aquired nearly 200 intel 1702 EPROMS (the first commercial eprom, if I recall correctly), but they're 256 bytes and impossible to program, requiring 0V, +5V, -9V,+47V and +56V for programming.

[ChristofferB]

Great! that might be useful!
Well they say prom's but they clearly have windows - thus must be EPROMS. Think it's similar enough. I'll keep that in the documentation archive!

Using a smaller microcomputer system to bootstrap could be interesting- maybe an intel 8051?

-The plug matrix is a cool idea, I actually thought about making one such using 3,5mm mono jacks, with diodes in the jack plugs connecting rows and columns.

[ChristofferB]

Diode rom matrix boards are difficult to make on strip/hole/veroboard. The layout is awkward.
Back in the days, there were dedicated cards, with traces lain out for diodes to be fitted easily.

[MK14]

Diode rom matrix boards are difficult to make on strip/hole/veroboard. The layout is awkward.
Back in the days, there were dedicated cards, with traces lain out for diodes to be fitted easily.

Possible way of making a cheap matrix/switch plug board, diode array.

Could you do it, via headers and 2 pin shorting links, to make a super cheap/quick/easy matrix board.

I would imagine the diodes can be done. Maybe you can get SIL/SIP diodes, example shown below:
Possibly soldering them on the other (reverse) side of the board, if you are happy about that.

[MK14]

But a microcontroller, such as the 8051 or something, would also be a good solution. Potentially doable in the 1980's, so your retro/vintage qualifications/requirements, are still met!

[ChristofferB]

You're right, that would be a completely viable way of doing it! I think I'll go for actual rom booting for now. Last time I tried diode matrix roms I lost all interest half way through:

[ChristofferB]

I just noticed a rather large design fault with the system. It has very poor handling capabilities for multiple interrupts. I've added four interrupt lines in my backplane bus, deviating from the standard ECB bus pinout, simply by removing unused pins (like D8-15, I'll never go 16bit in this system), called GPINT0-3 that can either be inputs for the 8085's many interrupt lines (all the RSTx.x's) or a 8259 interrupt controller. With that, I could have only one interrupt line going to the cpu.

Not having multiple interrupts would be a nightmare. Also, one could just have a 8259 per board, but it'd be hard to mask between them - two boards could still interrupt at the same time. Now I can have a master 8259 on the CPU card, branching to 4 out on other cards, giving up to 32 interrupting peripherals.

The other option being having the 8259 on the CPU card, and then have a secondary (ribbon cable with DIP headers?) bus running all the interrupts. It still may be necessary, but it's not very elegant.