Colecovision on the M5? Why yes!

So far so good!

And yes, they do all run on actual M5 hardware.

todo: give frogger a title screen :D

M5DEV

I was asked if I made a cart that would enable development on the M5. I didn't but then some time passed and I did.

It's a fairly simple re-work of the Multi-II cart with data and some control lines broken out. And it's red. I've been paste-curious for a while now and I finally caved and ordered a stencil along with the PCBs. I've had a stencil before from the OSH people, hence the board retainers, but it didn't work out for me. This is from JLC and I have to say I'm very impressed with it. I've loved JLC for a long time now and despite trying other houses I keep going back.

I don't have an oven but I do have a very basic hot air rework station. Even at this introductory level  the process is simple and quick. I use a low temperature paste from the bay of E. The board can be pasted up, populated and cooked to perfection with a few waves of the hot-air-wand in a smattering of minutes.


But does it work?

Oh yes.

Here it is connected to an SD-X. A useful little board if I do say so myself. I should build some more really. It's very crudely integrated at the moment but the plan is to refine the interface.

This is the very first time my game BiggOil downloaded to the M5 using the board.

Right now the M5DEV decodes I/O requests into a simple read strobe and the SD-X listens out for it. When it sees the read request it puts a byte on the data bus. Repeat 8192 times. The SD-X is easily fast enough to service the IO requests and the result is an 8KB ROM image loading in under a second. 600mS to be exact-ish. It can go faster, I'm being a little conservative right now.

The plan is to either make a dedicated board to interface with M5DEV or keep it SD-X based. The former idea could be cheaper but more work, and the latter would give more options. I intend to add WiFi via serial using an ESP series chip. This could provide a basic user interface. The work cycle would be compile, boot the m5 into a helper ROM, upload code to sd-x, download code to m5. A round-trip of a few tens of seconds I reckon.

Oh yes and the M5DEV is also a fully-functioning first-class M5Multi-II!

M5-Multi-II ... II

A couple of tweaks have reduced the size of the board by nearly 2 square inches, meaning that the medium run service from OSHPark is now feasible.

I just need to adjust the size of the printing on the front and move a capacitor so that the EEPROM socket doesn't foul on it but other than that I believe it's good to go!

Sord M5 - Multicart V2

Finally the ennui has passed and some time was devoted to seeing what I could rescue from the ashes of the original M5 multi-cart.

Long-time readers may remember this board from a few years ago now. It's a complicated little thing, with lots of chips and passives and switches and chips and passives and things. It was a pain to make and to top it all the failure rate was in the high 70%.

There were a number of factors that contributed to this situation and I dare say that I could probably get it working now that I've had some support and debugging help from a friend and fellow enthusiast - Bas at BetaGamma.

The main issue with the board was that the ROM images needed to be stepped through one-at-a-time by pressing the button on the front. It was a chore and the reset method I devised was at best unreliable.

What I really wanted was a menu-driven design which was simpler. So I came up with this.

Simple, eh!

Gone was the PIC microcontroller and its supporting hardware. Added was my CPLD of choice, a Xilinx XC9500XL series chip. 5V tolerant, and in my experience utterly bomb-proof.

The M5's cartridge slot contains a few signals which are very useful - IOWRITE and EXTIOB. The former is what you'd probably guess - asserted when an IO write is in progress by the Z80. The latter is a signal originating in the M5's memory controller custom chip, and is asserted when an IO access is made to a port in the range $70-$7f. Under normal circumstances nothing in the system writes to this IO port.

My plan was to watch for writes to one of the $7x ports and capture the data bus content. This would be used as a base page number for selecting any one of the 64 8K pages of rom in the multicart's EEPROM. Most of the M5's carts are 8k, with the exception of BASICs F & G, and FALC - the M5's spreadsheet.

The M5 has 3 ROM select lines on the cartridge slot. ROM1, ROM2 and EXT-ROM.

ROM1 selects 8k in the region $2000-$4000

ROM2 selects 8k in the region $4000-$6000
EXT-ROM selects 4K in the region $6000-$7000

The CPLD outputs 6 address lines used as bank selection on the EEPROM. The bank number output is a sum of the base bank number set by the Z80, and a 2 bit number formed by ROM2 and EXT-ROM being bit 0 and bit 1 respectively.

With this logic in place and a suitable menu program written and debugged using MESS, I finally had the multicart working that I'd originally imagined.

A fairly simple job of re-working the original schematic and board in EAGLE, another simple job of uploading the design to my favourite short-run fab-house and a final simple job of waiting for 2 weeks and here is the result:

Perfect. Purple.

I just need to sort out a manufacturer for a mid size run of boards then I'll be selling these.