"Arcflash" A3000/A3xx/A4xx/A5xx/A5000 in-system programmable ROM -- any interest?

[myelin]

http://myelin.nz/arcflash

Status: testing first v3 boards

I've finally figured out a design I'm happy with for this -- posting here to see if anyone else would be interested in one.

This will be a board that plugs in to the four ROM sockets on an A3000, and act like a DualOS/MultiOS board, plus let you flash new versions of RISC OS (or old versions... so you could have a machine that'll boot up with Arthur, RO2, RO3, the test ROMs, your own RO build...) over USB without having to open the case.

Features:
- 16MB of flash memory, so you can store 4-8 different OS ROMs at once (4 x 4MB, 8 x 2MB, 16 x 1MB, or any combination).
- In-system programmability -- you'll be able to connect your Arc up to a laptop via USB, and update the OS images while it's running (or powered down)
- SYS "OS_UpdateMEMC", 64, 64 support, even under RISC OS 3, for some extra speed
- Support for all 4-ROM machines (A3000, A3xx, A4xx, A5xx, A5000 with adapter).
- The ability to make 4MB of flash visible to the machine by wiring in the A21 line, so you can fit more stuff in your RISC OS build
- Not enough room for this one: MAYBE a fast USB serial port for off-machine communications (HostFS maybe?) using the extra address space, if I can fit this into the CPLD

Cost will be about $100 (£78) assembled including shipping worldwide; a bit less if you're in the USA. Originally I was hoping to offer a kitset version, except the board has several hard-to-solder BGA chips, and the work of packing up kitsets is about the same as assembling the board, so this isn't going to happen unless you live in the SF Bay Area and want to come over and help me assemble your board.

Caveat of course here is that I might not actually finish the project, but if I do, it'll be more or less as above

UPDATE: Keeping track of who's interested (from replies to this thread) - I'll confirm all this before actually ordering anything; this is just to give me an idea of who to talk to when I do have something working.

Who has boards:

• me: 1 x v1 (A3000), 4 x v2 (A3000, A310, spare x 2)
• IanS - #3, v1
• steve3000 - #4, v2

Who wants boards:

• IanS - one or more
• hubersn - 1 of the A310 type (has carrier board), 3 x A3000 (most important), 2 x A5000 type, 2-5 x Risc PC.
• paulv - one
• trixter - one (assembled)
• lcww1 - one
• JonC - 1 x A3000, 1 x A3020, 2 x A5000 (1's an Alpha), 1 x A440, 2 x A310, 1 x RISCPC
• 1024MAK - one or more for A5000, A30x0, A7000/Risc PC.
• Cutter - 1 x A3000
• markrwj - 1 x A3000
• Nutter - 1 x A3000
• sirbod - 1 x A305 (later: A440/1, A4000 and A5000)
• steve3000 - 1 x A540
• Sniffler - 1 x A3000, 1 x A5000
• AndyMc1280
• stevebubs - 1 x 4-ROM, 1 x RISC PC
• Zarchos
• Kazzie - 1 x A420/1
• anightin - 1 x RPC600, 1 x RPC700
• zfpsystems - 1 x A310, 2 x A30x0 (1 A3020, 1 A3010), 1 x A5000, 2 x Risc PC (both are RPC 700, MK3)
• IanJeffray - 1 x A400

[IanS]

Sounds like a great idea. I'd be interested in one (or more).

It would be nice to make the connections a bit configurable, so it could be fitted to an A3xx without ROM carrier board (the carrier board picks up a few address lines that aren't available as standard on the rom sockets).

I'd be happy to help test it on other 4 roms machines (other than rom carrier board issues, I can't see why it wouldn't work).

[myelin]

It would be nice to make the connections a bit configurable, so it could be fitted to an A3xx without ROM carrier board (the carrier board picks up a few address lines that aren't available as standard on the rom sockets).

Ah, excellent... glad I asked for feedback, because I'd completely overlooked this! Looks like pins 30 and 31 are tied to +5V on the A3xx sockets, so I'd want to put some jumpers in there that could be hooked up to pins on the buffer chip (IC28 on A3xx) for LA19 and LA20. Looks like the A4xx has the same jumper setup as the A3000, so this would only be needed on the A3xx.

[IanS]

The fitting instructions for the carrier board show you need to pick up LA18, LA19, and LA20 from IC28.
http://www.simtec.co.uk/products/AURCB/ ... itting.pdf

carrier.PNG (15.19 KiB) Viewed 4810 times

I don't know if the spacing between rom sockets is the same on the A3000 and A3xx, I'll try and check over the weekend.

[hubersn]

I've finally figured out a design I'm happy with for this -- posting here to see if anyone else would be interested in one.

I'm definitely interested. Might even be interesting for the A5000 to switch between custom, 3.11 and 3.19.

However, I would be even more interested in a Risc PC variant! 3.7, 4.0x, Adjust, 5.xx...

Now I would only need an additional switchable CMOS that automatically detects the RO version active...

hubersn

[myelin]

The fitting instructions for the carrier board show you need to pick up LA18, LA19, and LA20 from IC28.

Odd... Looks like the A310 has LA18 connected; LK12 lets you connect it to pin 2 or pin 24 of the ROMs, like LK17+LK18 on the A3000: http://www.drobe.co.uk/archives/ftp.aco ... /PART2.GIF

Anyway that's nice and easy to add... I'll put it on the to-do list.

I'm definitely interested. Might even be interesting for the A5000 to switch between custom, 3.11 and 3.19.

However, I would be even more interested in a Risc PC variant! 3.7, 4.0x, Adjust, 5.xx...

Now I would only need an additional switchable CMOS that automatically detects the RO version active...

Looks like the A5000 has a similar set of ROM sockets, although running at 12MHz so the single-cycle sequential access time will be a bit tighter... 83ns clock instead of 125ns, so a 42ns shorter cycle, squishing the usual 60ns down to 18ns, which is a bit unlikely. Should work fine with 2 clocks per read (158ns rom access time) though, which will be a nice bump from the standard 3 clock timing at 12MHz (although maybe less obvious on an ARM3 with cache).

I'd be very much interested in doing one for the Risc PC, if the Archimedes version works out -- I have zero access to Risc PC hardware at this point though, but let me know if you have any leads on some

Good point on the CMOS... that's a relatively low speed I2C device so I'm sure it would be possible to emulate with a microcontroller. A whole separate project though!

[IanS]

Odd... Looks like the A310 has LA18 connected; LK12 lets you connect it to pin 2 or pin 24 of the ROMs, like LK17+LK18 on the A3000: http://www.drobe.co.uk/archives/ftp.aco ... /PART2.GIF

I wonder if they did it this way so all the jumpers related to the ROM sockets were on the carrier board and not dependant on the main board jumper settings. So just to reduce complication for the end user.

I don't think the main board jumpers are obscured by the carrier board, but that could be another possible reason. The carrier board is not something you want to add/remove a lot, so to avoid the problem of the on-board jumper being wrong after the carrier board is installed.

[IanS]

I don't know if the spacing between rom sockets is the same on the A3000 and A3xx, I'll try and check over the weekend.

The spacing of the roms on the A5000 is definately different to the A3000, from online pics, it looks like the A3xx spacing is similar to the A3000.

[myelin]

The spacing of the roms on the A5000 is definately different to the A3000, from online pics, it looks like the A3xx spacing is similar to the A3000.

I can't find any good close up top-down shots of A5000 motherboards, but it looks like they're a bit closer together than those on the A3000 - is that correct? The gap between adjacent rows of pins on the A3000 motherboard appears to 0.2", and I'm guessing it might be 0.1" on the A5000. Would love a good pic with measurements if you have an A5000 handy (or if someone reading this does!)

[paulv]

I'd definitely be interested in this. Bring able to switch between the various OSes more easily is definitely something I'd like to be able to do.

Paul

[trixster]

I think I’d be interested in an assembled board for an A3000, this does look very intriguing

And also an assembled board for my A420/1 if that machines is compatible, please.

[lcww1]

Yes, please!

[JonC]

Very interested in 1 or more of these please!

[IanS]

The gap between adjacent rows of pins on the A3000 motherboard appears to 0.2", and I'm guessing it might be 0.1" on the A5000. Would love a good pic with measurements if you have an A5000 handy (or if someone reading this does!)

It's not a great pic, but it does show the pins in an A5000 are 0.1" apart. So probably quite difficult (if not impossible) to make one board fit in A3000 and A5000.

Edit to add more pics.
It looks like LK12 was not fitted as standard on the A3xx PCB, so that's probably why they picked up LA18.

[myelin]

Those pics are perfect; thank you Ian!

Looks like I’ll need to consider component clearance for the A3xx and A4xx when working out the board shape here. I need to look more carefully, but it looks like the podule backplane might block off the side of the board that was where I was planning on putting most of the chips — which would extend harmlessly out to the left on an A3000.

(Also wow! Didn’t expect this much interest!)

[sirbod]

Also wow! Didn’t expect this much interest!

This was one of the pre-reqs for RISC OS 3.20. Although that's some way off due to other issues, this would allow RISC OS 3.11 / 3.19 to be patched and tested on physical.

[steve3000]

This sounds like a great project! Could the bare PCB be designed with holes positioned for either A5000 or other Archies? So you just need to design one PCB and keep option for A5000 users to fit the header pins to the correct holes...

[IanS]

This sounds like a great project! Could the bare PCB be designed with holes positioned for either A5000 or other Archies? So you just need to design one PCB and keep option for A5000 users to fit the header pins to the correct holes...

Tricky to do, you waste a lot of PCB space with holes you aren't using, and with spacing as it is, different rows of pins occupy the same space.
On a 0.1" grid, the pins from one chip occupy the same space as another row of pins from another rom.
e.g.
Coloured blocks are rom chips, "X" denotes a row of pins, the "?" shows where there is a clash.

[steve3000]

On a 0.1" grid, the pins from one chip occupy the same space as another row of pins from another rom.

True... but is there a way round this by careful design, because are not all pins are required from each ROM? E.g. the data lines are needed for all ROMs but address lines could just be taken from a single ROM...

...so could an overlapping row be managed somehow.

[IanS]

Looks like I’ll need to consider component clearance for the A3xx and A4xx when working out the board shape here.

For completeness, here is the A540 pic, though there is so little space around the sockets it may be tricky to accomodate.

A540

[hubersn]

UPDATE: Keeping track of who's interested (from replies to this thread) -

• hubersn - one (also interested in models for A5000 or Risc PC)

Talking about numbers...I would like 1 of the A310 type (I have a carrier board), 3 of the A3000 type and 2 of the A5000 type. A3000 would be the most important variant for me.

Risc PC, I would need to think about a little longer. A maximum of 5, but probably only 2.

Have fun
hubersn

[1024MAK]

I don’t have any A3xx or any A3000 machines. But if a version that works in/with any of the following is produced:

• A3010
• A3020
• A4000 (once I have recovered it from the garage and discovered if it is working or can be repaired)
• A5000
• A5000 33mhz Alpha
• A7000+
• Risc PC

Then I would be interested in one or more...

[JonC]

I'll update my numbers seeing as everyone's getting their orders in!

1 x A3000
1 x A3020
2 x A5000 (1's an Alpha)
1 x A440
2 x A310
1 x RISCPC

So when's the factory opening?

[myelin]

I could possibly handle the A5000/A3000 pin overlap by using a few more pins on the CPLD and making it redefine them as appropriate... I’ll figure that out when I do the PCB layout. That said, it looks like there might be enough interest in an A5000 version that I could just make another PCB with pins moved.

The tricky thing is that I don’t have an A5000 or any of the two-ROM machines to test with. I guess any profits from the run of A3000/A3xx/A4xx boards (which I believe will be a single model) can go towards acquiring a RISC PC to test the next version on It looks like the RISC PC and A30x0/A4000 may all have the same spacing, so with any luck I’d only need three board versions (two socket, four socket, and A5000).

[Cutter]

Also interested in one for an A3000

[markrwj]

Please put me down as interested in an A3000 board. Thanks!

[Nutter]

Interested in one for A3000.

[myelin]

Making some progress here!

This is the most complicated board I've designed, by quite a long way. I managed to use all 117 signals on the CPLD, and daisy chaining 25 address/control lines through two BGA flash chips was an interesting process! Hopefully this will solder OK

Here are some renders of the board so far, for anyone who's interested. Passes DRC and all signals are connected, although I need to shuffle things around a little (keeping crosstalky signals apart, getting rid of burrs and stuff produced by the interactive router, adding info on the silkscreen) before sending it to the fab.

Top:

Bottom:

[IanS]

Here are some renders of the board so far, for anyone who's interested. Passes DRC and all signals are connected, although I need to shuffle things around a little (keeping crosstalky signals apart, getting rid of burrs and stuff produced by the interactive router, adding info on the silkscreen) before sending it to the fab.

Could you add some pins to the top end of ROM 4 (e.g. Pins 1-4), just for mechanical support, as the usb connector will be near there?

[myelin]

Could you add some pins to the top end of ROM 4 (e.g. Pins 1-4), just for mechanical support, as the usb connector will be near there?

Good idea! I'll do that.