stardot.org.uk

Among the Acorn System range, I find the original System 1 holds appeal for its sheer simplicity (compared to it's successors): no operating system, no complex video or DRAM refresh circuitry, just a 6502, 1 KB or so of SRAM, some buttons and a calculator display (plus tape interface). Sadly, as they sell on eBay for well over £1000, I'm never going to buy an original.

Some time ago, while reading about the history of the Acorn System range, I came across this photograph of Sophie Wilson's original prototype ('Hawk') through the mos6502 Google Plus group:
That's pretty much the whole System 1 on one board: it's missing the tape interface, and doesn't have the pair of 2114 SRAM chips fitted (though the sockets are seen in red), just the 128kB provided by the INS8154 I/O chip (similar to the 6522). It probably has a EPROM chip (or a second INS8154) hiding behind the calculator display.

That photograph, in turn, reminded me of a disused circuit board I saw lying on a shelf at work:
Both are prototyping boards sold by Vero, using the Verowire (or Road Runner) system. Varnished copper wire dispensed from a plastic wiring pencil (essentially a hollow biro with a spool of wire on top) is wound along plastic guides and around component legs, where they are soldered together. The varnish burns off under the heat of the soldering iron, making an electrical connection. See this view of the board's underside:
Best of all, I'd also found a wiring pencil and spool of wire! At this point I started doing some serious plotting...

Most of the components of the System 1 were either already to hand or readily available. I had some 6502s and most of the logic ICs, and an old faulty Rockwell 8R of mine happens to use a display with the exact same pinout. I also had some spectra-strip and loads of DIN 41612 connectors. The major issue would be obtaining some INS8154s, as at least one other person has found. There were some places off in the far East that claimed to have some NOS parts lying about, but at a price.

You may have noted that my Verowire board is smaller than Sophie's. Mine is Eurocard-sized (160x100mm) with five rows for ICs, Sophie's is slightly taller with six. The original System 1 used two 160x100mm Eurocards. I wouldn't be able to fit a whole System 1 on my single Eurocard, but I did want to keep open the possibility of expanding my System 1 into a full System 2/3/4/5 at some point in the future.

One approach was to leave the cassette interface off (saving 7 ICs) and put it on another card if I wanted it. This is something that was done with the Acorn System: the System 2 used the System 1's front board for the cassette interface only, accessed through the euro-connector. Unfortunately, try as I might, I couldn't find a layout to fit the rest of the ICs (along with the keyboard and display). Fitting the keyboard buttons along my card's five rows (while keeping a sane layout) was particularly troublesome.

Dropping the 74S571 chips that hold the monitor program (which could be put on the EPROM instead) didn't give me enough space to make things work. Equally, I wanted to keep the 2716 EPROM in order to have somewhere to store finished programs that didn't require loading from cassette every time.

In the meantime, I started to clean up my old board. I wanted to reuse the plastic guide strips, and cutting/desoldering all the wire was a particularly fiddly job:
I also came up with a layout that would fit all but the keyboard and display on my Eurocard:
Components labelled in Yellow are from the 6502 board, those in green are from the Cassette/Display board. The empty space near the 6502 is for the clock crystal and associated components. There's likewise space around the cassette ICs for the various resistors and capacitors needed for that part or the circuit. The holes labelled in pink are for pin headers which a separate keyboard & display board will mount. The extra board will be smaller than the normal Eurocard, and probably made from stripboard. The address selection socket will be accessible with the keyboard on, but it will have to be removed to change the EPROM. If the board gets placed in a Eurocard rack, I can simply remove the keyboard. A notable omission is the LM340 voltage regulator: regulated 5V power supplies are ten a penny these days, so it was an early sacrifice in making this layout work.

I've made a slow start on wiring the card, and progressed far enough that I felt ready to write up all the above for you to read. I'll make another post later this week to show you the progress to date.

Nice built.
I am looking of maybe building an Atom on 100x160mm Eurocards.
This one is nice as well though.
Only thing is the "bubbly displays" seem to become rare.

Kind regards, Bram

Great project!

So, a week on, here's another update:

Having stripped the old verowire board of sockets and wires, I needed to get three decades or so's grime off the copper tracks, if I was to have any hope os soldering anything onto it. Brasso to the rescue!
I'd made a diagrammatic wiring sheet, which had a few minor revisions from the photo-based layout I posted last week. I'm filling it out as I wire it (more or less). Aslo pictured is my wiring pencil, and a few of the more obscure chips I'd sourced for this project, a pair of INS8154s and 74S571s.
The INS8154 is roughly a counterpart to the 6522 VIA, in that it provides 16 I/O lines for the CPU, but it lacks a timer, and has 128 bytes of SRAM instead. It was used in Science of Cambridge's MK14 computer kit, alongside National Semiconductor's INS8060 CPU. The on-board RAM was probably very popular in the late 1970s (The MK14 only shipped with 256 bytes) but the insurgence of large DRAM chips by the 1980s, and the INS8154 fell to the wayside in obscurity. The two INS8154s pictured were NOS shipped from the USA. (They're unlikely to be relabelled chips, as the plastic has dust and scuff marks that indicate many years of sitting in a component drawer.) Only one is needed for the System 1, IC2 for interfacing with the keyboard, display and cassette circuitry. The second is only used for interfacing with other devices through the eurocard bus. As I was having to pay a king's ransom in postage for one chip, I decided to buy two: even if I only need one I'll have a spare one to hand.

The 74S571s are fusible PROM chips, each with 512 x 4 bits of storage. These are programmed by literally blowing fuses inside the chip to program zeroes into the devices memory (if the fuse still conducts, it's a one). These held the monitor program in the System 1 and the prototype. I have no means of programming these PROM chips (a run-of-the-mill EPROM programmer won't know what to do with them) but I'm not concerned: I can put the monitor program on an EPROM at first, and then worry about the PROM chips once the System's working.

Time for a few more gratuitous progress shots:
I've wired up all of the Data bus (D0-7) and roughly half of the address bus (A0-6), including connecting them to the bus interface. (The DIN 41612 connector isn't soldered on yet: I'm leaving it loose for clearance until I solder all the wires for side A and B. Here's the front of the board:
This picture illustrates an unforeseen problem. I'd left the area in the lower left bounded by the EPROM and 6502 sockets for mounting a crystal. My mental impression of a crystal resonator is of a small through-hole silver package roughly the size of a fingernail (HC-49 package) or even smaller (HC-49/US). Unfortunately, it seems that 1MHz crystals are rapidly going the way of the dodo. The only 1MHz crystals I could find were large, bulky ones (HC-51), such as the blue one to the left of the board. While these are the same size as used on the System 1, I don't have the space to fit one on my board. (Remember, there'll be a 7445 below the crystal, driving the keyboard and display. Moving that onto the keyboard card would present another set of problems.)

Alternative solutions include using a capacitor instead (an option shown on the original circuit diagram), using a ceramic resonator, or using a full can crystal oscillator. These DIP-shaped cans require a power supply, but generate the whole clock signal on-board and output a square wave on their output pin. The silver square to the left of the board is an example, and the one I'm currently minded to use. It'll fit in the space available (just about), and will then leave me with a much simplified clock circuit.
With the resistor and crystal/capacitor removed, and the crystal can connected directly to Phi0in, Phi2out will just go through a pair of inverters on its way to the rest of the circuit. Is it necessary to double-invert the clock? Is the pair of gate delays required? Is it desirable for the 74LS04 to drive the signal, as opposed to the 6502? If It isn't necessary, I could choose to wire up the spare NOT gates to extra pins on the address select socket, allowing me to invert some signals there to rearrange the memory map. Or perhaps I'll leave it as-is, even if it isn't strictly required, to match the original System 1 design. I'll have to decide soon, as A7-9 need to be wired up to whichever address select socket I choose to use.

There's quite a lot gone on since the last update...

First off, the issue of double inverters on the clock. I took my thoughts to the wise heads at forum.6502.org (which included a few familiar names!) and thrashed things out a bit with them.

Looking at the System 1's original clock circuit, there's a question of why a second inverter was used to generate the clock signal for the board, instead of using Phi2out directly (on pin 39 of the 6502). The three main possibilities were:

• timing: ensure the 6502 clock is two gate delays ahead of the rest of the board
• buffering: let the 74LS04 feed the current to all the board (and expansion bus) instead
• "brain fart": misunderstanding, caution or some other reason that made sense to the designer at the time

The System 1's clock circuit bears a striking resemblance to the example circuit given in MOS' original datasheet for the 6502, so it's likely that the example was replicated as-is. (Clock circuitry using logic gates was apparently a bit of a black art back then, and many designers left well enough alone.

It was very interesting to note that while the Atom uses a different clock circuit (a 4MHz clock divided with flip-flops, fed into Phi0in, not connected to Phi2out) it keeps the two inverters on the Phi2out line! I guess they were kept out of caution: if the timing etc worked for the System line, it should work for the Atom too. And if it worked for the System and Atom, it'll do for me too.
Having come to a decision over those two inverters, I've wired up the rest of the address bus, and also the glue logic and clock circuitry. Last night saw the power supply and decoupling capacitors wired up to the left half of the board, and a quick NOP test run on the CPU. (The data bus was hardwired to 0xEA, the opcode for "no operation", and a very cheap scope was used to probe the address lines as the CPU ran through its address space finding NOP, NOP, NOP...) The oscillator was generating a clock signal, and most of the 16 address lines were oscillating as expected. A4 was oddly quiet at 0v, and the 6502 was running noticeably warm to the touch, so I'll need to go back and check my wiring (and/or try another 6502) when I revisit it.
I've also designed and (almost) assembled the keypad and display board. This board (that will sit above the 6502 and 8154s) is made using strip board, mimcking the layout of the System 1 as much as possible. I decided to use bare metal wire to connect tracks in order to keep with the old-fashioned look. The buttons are labelled with some "wire-labelling stickers" I picked up from my local Maplin as they were closing. The PCB of the display will need a bit of trimming to fit next to the reset button, then it'll be ready to be soldered on. There are two rows of header pins (20 and 12 pins each) soldered on to connect to the main board. I didn't anticipate when coming up with the design that attaching header pins to the non-solder side of a stripboard would be troublesome. My solution was to push some bare stranded wire down through each hole, solder it to the track, then push the header pins up through the hole. I soldered the header pins to the loose stranded wire on the upper side, and then reflowed the solder again from the lower side. It gives electrical continuity, and seems to hold the pins on securely. We'll see if they stay on when the it's plugged into the main board...

New update: I finished construction last week, and powered up the System to find that, surprise surprise, it didn't work. On powering up I got an inverted "A" on one segment of the display, and when pressing the reset line it gave no response. I've since had a few evenings of debugging my work, and found a comedy of errors:

• The pad for Address line A4 under the 6502 was shorted to ground (probably old solder from the previous project)
• Address lines A5 and A6 were tied together by a whisker of verowire at the RAM socket
• Having misread my diagram, the keyboard's reset button was connected to pin 6 (NWDS) instead of pin 8 (NRST) on the edge connector - thus no response from the reset button

At this point the machine would respond to the rest button: the inverted "A" was replaced by an overscore, and the address and data buses showed some life. I checked my ERPOM's output by hand on a breadboard, and all seemed fine, but there was no response from the keyboard/display when running. Examining the INS8154, I found that after resetting it was not setting the scan drive lines (A,B,C) as output lines. At the 7445, these pins floated up to 1~2V, whereas input D was tied to ground. Based on these inputs, it was driving output 7 low, thus enabling the final digit. The INS8154 was not being initialised correctly, but why? I swapped out the 6502 and INS8154, with no change. I began to suspect that ringing on the address and data lines was causing the monitor program to go awry, but setting up the output definition register (for port B: the segment drives) is the third instruction after reset!

The breakthrough came when I realised that I'd connected my System's R/W line to pin 35 of the 6502 instead of pin 34. Pin 35 is not connected, according to the datasheet, so my R/W line was floating. The net result of this was that the NRDS signal triggered correctly for read operations, but the NWDS signal never triggered, even when the 6502 wanted to write data to the INS8154 to set up its output definition registers. Moving this wire to the correct pin gave me a series of dots on the display after reset, just as expected.

Hooray!

It soon became apparent that I'd mis-designed the keyboard, and keys 1-E gave the wrong characters (but the display was operating correctly). After rewiring either side of the keyboard (easier than rewiring the keyboard itself) everything was in full working order. Here's a photo of the System displaying the contents of memory location FE00 (beginning of the monitor program):

I've had a few minutes to work through a few of the example instructions from chapter 2 of the User's manual. Typing in a program by tapping out opcodes punctuated by the ^ (or +) key feels really satisfying, it's far swifter and natural than I'd have expected. (The placement of the keyboard on the left side of the board was a conscious decision: as a left-handed user I prefer to have it over there.)

I intend to have some fun playing around with my new System before I give any thought to completing the build. While I have all the ICs needed for the cassette interface, I've almost used all the Verowire that was on my salvaged pen, and I'll need to buy some more before I build it. Programming the PROMs will also have to wait until I come up with a method of programming them (unless there's anybody here that has a programmer suited to 74s571 PROMs). I may make a portable power supply out of a 9V battery and a 7805 in the meantime, however.

A side-goal in this project was to try using an old prototyping method that I'd never used before: Verowire. The ability to wire point-to-point makes for denser layouts than regular stripboard, but laying numerous address and data lines in parallel along the plastic trunking makes for a lot of ringing and crosstalk. I can see why this method fell out of favour in the past: wrapping and cutting the wire can be fiddly, you need plastic strips as well as a suitable board, and I doubt this method would work well with higher frequency circuits. Also, the plastic strips are absurdly expensive, ~£1 each, and I used at least twenty for this project! Still, it's been a fun experience trying a different method, and I'll keep hold of the wiring pen for possible point-to-point repair work in the future.

(If I expand this System 1 into a rack-based System 2 in the future, I'll probably have a go at wire wrapping the backplane: I have several wire-wrap DIN41612 sockets, and a wire wrap tool sat in a toolbox.)

Great result!

Great to hear that it's working

Mark

Congratulations! I love reading about this kind of project.

Thanks for the messages. I've had a bit of simple fun with using assembly opcodes to pump hex values into zero page and calling the monitor's DISPLAY subroutine: The screen's not big enough for "hello world", at least not without scrolling the display and introducing timing loops. Was " hello world" even a thing for programmers 40 years ago? I've no idea.

(Incidentally, according to the Acorn User Guide, the suggested character for "i" is segments a, b, c and d lit for a sort of reverse capital C. Use of segment c alone, as I did, is suggested for a comma.)

Another thought: have we many/any other System 1 users on the forum?

(Hello world dates from at least 1972!)

BigEd wrote: ↑

Wed Aug 08, 2018 8:07 pm

(Hello world dates from at least 1972!)

There we go, then. Of course it was an original part of the "C" culture - I should have remembered that, having taught people C, with reference to its history. (That's what happens when I type faster than I think.) I was idly wondering if it might have been a part of computer culture in North America in the 1970s, and not yet over here, but that Wikipedia page has an uncited claim that it was used in BCPL in Cambridge back in the 1960s, so it seems that it was known across the world.

My Atom 2k14 lacks the two inverters at phi2 and it works perfectly without them. But maybe that is because of the faster, modern components used.

Fantastic result - love your work.



Ade

Kazzie wrote: ↑

Wed Aug 08, 2018 11:36 am

Programming the PROMs will also have to wait until I come up with a method of programming them (unless there's anybody here that has a programmer suited to 74s571 PROMs).

I'm a little nervous in saying this, since they are OTP and to my knowledge I don't have any spares, but according to the device list for my programmer, it can do those PROMs. I've no reason to doubt it, I've done some earlier 74S series PROMs like those found in arcade PCB colour circuits.

(If I expand this System 1 into a rack-based System 2 in the future, I'll probably have a go at wire wrapping the backplane: I have several wire-wrap DIN41612 sockets, and a wire wrap tool sat in a toolbox.)

For these kind of projects I've designed and ordered a bus board as well as an extender card some months ago :

guddler wrote: ↑

Mon Sep 03, 2018 10:44 pm

Kazzie wrote: ↑

Wed Aug 08, 2018 11:36 am

Programming the PROMs will also have to wait until I come up with a method of programming them (unless there's anybody here that has a programmer suited to 74s571 PROMs).

I'm a little nervous in saying this, since they are OTP and to my knowledge I don't have any spares, but according to the device list for my programmer, it can do those PROMs. I've no reason to doubt it, I've done some earlier 74S series PROMs like those found in arcade PCB colour circuits.

That's interesting. Does you programmer support National Semiconductor's DM74s571 or the Tesla MH74s571? The original System used the DM, but they're getting rather rare outside the USA these days, and are at least ₤10 each. Tesla's Czech-made chips are commonly available online for a few pounds, but though they're fully compatible when reading, they use a very different programming method.

I'll make a note to go and look in the actual programmer software. Always better to look in there rather than rely on device lists from the internet.

Will let you know.

@Kazzie,

Do you want to stick to the original using the proms ?
As you could consider throwing in some pals/gals.

Kind regards, Bram

bprosman wrote: ↑

Tue Sep 04, 2018 1:03 pm

@Kazzie,

Do you want to stick to the original using the proms ?
As you could consider throwing in some pals/gals.

Kind regards, Bram

I've got a pair of MH74s571s already, so if I'll be programming something I may as well have a go at those in the first instance.

There are some other people looking at creating their own programmers for the MH variety, see this link: https://www.vintage-radio.net/forum/sho ... ?p=1057809

As the PROMs are fairly small, I'm playing with the idea of programming them by hand with a simple breadboarded circuit, following Tesla's instructions. The only critical bits are the timings for applying and removing the programming voltages, which I could do with a microcontroller switching a few transistors.

In an ideal world I'd program my System 1 to do it, but I need to solder up the second INS8154 to get the extra I/O lines first. It would also improve my sanity to get the cassette save/load circuit built before I start working on any significant assembly coding...

If you get stuck, I can program these for you. I have a Logical Devices Chipmaster (look-up the list) which programs most things. I’ve programmed lots of that part before using a Phillips part number. Note: The Czech parts do have ‘drop-outs’ !! Nothing to do with the programming algorithm, though.

Dave H

daveejhitchins wrote: ↑

Sun Sep 09, 2018 4:27 pm

If you get stuck, I can program these for you. I have a Logical Devices Chipmaster (look-up the list) which programs most things. I’ve programmed lots of that part before using a Phillips part number. Note: The Czech parts do have ‘drop-outs’ !! Nothing to do with the programming algorithm, though.

Dave H

That's a kind offer, thank you.

I may take you up on it at some point, but I'd like to take a shot at a DIY attempt in the first instance. Any such attempt will have to wait for spare time, as I've also got an EDSI hard drive swap/backup in my Archimedes, two faulty AND01 floppy drives and a hard-driveless Mac Classic II on the project table. So it may not be this side of Christmas.

Edit: Doy you happen to recall the Phillips part number you used? The posters on the other forum I linked to may be interested.

Kazzie wrote: ↑

Mon Sep 10, 2018 12:04 pm

Edit: Do you happen to recall the Phillips part number you used? The posters on the other forum I linked to may be interested.

The Phillips part is: 82S131.

Here's a link to some alternatives.

My programmer will do all the alternatives . . .

Dave H

daveejhitchins wrote: ↑

Wed Sep 12, 2018 12:09 pm

Kazzie wrote: ↑

Mon Sep 10, 2018 12:04 pm

Edit: Do you happen to recall the Phillips part number you used? The posters on the other forum I linked to may be interested.

The Phillips part is: 82S131.

Here's a link to some alternatives.

My programmer will do all the alternatives . . .

Dave H

Thanks for the details.
Looking at the programming procedure for the Philips 82S131 here, it is apparently programmed with VCC at 8.75V and the output bit at 17V, pulsing /CE to program. The MH74S571 (as descibed on the forum I linked to earlier) expects the output bit to be grounded, and VCC at 10.5V, pulsing /CE to program. Somewhat different techniques, but I can see how the techniques could be compatible, as in both instances the non-programmed bits will be at 5V, thus within ~5V of VCC (or floating, in which case there'll be no current flowing). The significant difference appears to be the polarity of the effective voltage. I might not have guessed it'd work, but you've got the badge to prove that it does.

Dear All,

Inspired by your great efforts I've also decided to try and build a System 1!


1/ Before I look at getting some eurocard form factor strip-board, is there anyone on the forum who has looked into laying out a System 1 CPU and Keyboard + Tape PCB?


2/ Also, I need to ask someone on this forum if they can very kindly program the 74S571 PROMS for me?

Alternative is to save up for:

https://www.rapidonline.com/Catalogue/S ... y=beeprog2

...although I won't be able to consider this until at least April, both for budget reasons and for a few other projects that might benefit


Will aim to document progress once parts list complete.

Best wishes

anightin

anightin wrote: ↑

Sun Jan 06, 2019 12:04 pm

2/ Also, I need to ask someone on this forum if they can very kindly program the 74S571 PROMS for me?

PM sent . . .

Dave H

The DM74S571 are fairly commonly supported and easy to program (my output-compatible MH74s571's are a bit more obscure).

I wasn't sure from your post whether you were planning on using one Eurocard or two? My experience in trying to fit everything on a single Eurocard is that the keypad eats up a lot of space, and the DIL-oriented spacing on my veroboard made things awkward. If not for the horizontal ground/power planes on my board I calculated I could have fitted all the CPU board's components and the keypad/display on a single board, putting the cassette interface on a separate board. If you're willing to do point-to-point wiring on a perfboard rather than a regular stripboard, you could definitely fit them on.

Depending on your intended use and how accurately you want to recreate the exact board, you may be able to reduce your chip count to help things fit better. You don't necessarily need both the PROM and EPROM sockets, for example. The second INS8154 isn't used by any of Acorn's System expansion cards, it's purely for you to interface to your own hardware (and you get to write the software to do it too). Also, you could choose to use a single 8-bit-wide SRAM chip instead of a pair of 4-bit-wide 2114 chips (giving 1KB): you could opt for a single HM6116 (2K) or even a HY6264 (8K), with some minor modifications to the address-decoding logic. (At a glance, the modifications for 2K look fairly straight forward. Let me know if you'd like me to put my thoughts to paper.)

I look forward to hearing your thoughts.

Hi Kazzie,

Thanks for your thoughts so far, I'm sure there will be more questions as the build progresses

I'm looking to use 2 Eurocards as planning to create a CPU card and a separate keyboard/display and tape card.


Progress so far is that I managed find most of the components, about 50% of them including the second euro-card are on order.

I haven't managed to find the correct spectra strip using google, so will have to resort to rainbow ribbon cable unless the 8 way stuff you found is still sourceable?

Thanks Dave, PROMS will shortly be in the post, much appreciated!