And now....with USB!

Been too busy recently to do much with this but had a couple of mammoth wirewrapping and code hacking sessions this last week.

What I've done is add a Vinculum VDIP2 module. This makes interfacing with USB storage devices (and other USB devices) a doddle, it's got its own firmware that handles all the FAT32 misery and just presents you with a few simple APIs to open, seek, read, write bytes to and from files etc.

I was given the inspiration by RetroClinic's DataCentre interface for the BBC Micro. This lets you fill up a USB stick with all your favourite games as disk images and use them from the BBC just like normal disks.



So far I've kept the interface dead simple, and use a PIO and bit-banging to do all the bus signals, though I'll probably have a go at marrying this up closer to the z80 bus directly to speed up larger transfers. (Though it only takes a couple of seconds to load up a full 32k)

The biggest challenge has been to port CP/M. There's not a lot of room in the BIOS area to fit in all the stuff I'd like so I think I may have to cheat and move some of the code for accessing the Vinculum into ROM to free up a bit more RAM and allow for a little image file selection menu.

Anyway here's a couple screen shots of CP/M 2.2 running (via a serial link). It's read-only so far, though to save me constanly corrupting my usb sticks when I make a mistake!

fungal fun

When I got back from Retro Reunited today (wish I could have stayed longer), I decided to pick out a beeb from my spares pile to fit a new DataCentre card to and take down to London as my one allowed "toy" in the flat!

But I was confronted when I pulled the machines from the shelf with something that looked like a scene from the Donald Sutherland version of the Invasion of the Body Snatchers!

I've got 4 beebs all covered in this stuff! I've started by dismantling one and giving it a very good clean with IPA but so far all I've been rewarded with is a folorn beeeeeeeeeeeeeep!

Looks like I need to get the damp sorted out!

Dom

And now in living colour....

All 64 of em!

Lo-Res Graphics!

The machine has been running reliably for some time now - running a "Dewdney's game of life" program continuously for 5 days at 5Mhz - it should be good for six but I haven't got a crystal handy! Anyway now its going reliably I decided to add a PIO chip which was very easy - those lads at Zilog made some very nice and easy chips! But then I needed something for the PIO to do so I decided to build something to interface with one of my other hobbies - NBTV (narrow bandwidth television). I've already built a televisor (a mechanical television) and concocted my own colour standard so why not build a "graphics card" that can output NBTV signals! So far the graphics card comprises ten resistors and a transistor to combine 7 bits output by the PIO into a single analogue voltage (2 bits for R, G and B and another bit for sync). An interrupt routine runs off the CTC which outputs a pixel every 50uS. This works ok but jumps about a fair bit if there are any other interrupts for the serial port or the other CTC timers. A program can run (slowly) in the foreground and update the contents of the screen. The resolution is 48x32 pixels with 64 grey tones. Next up I'm going to add colour and make a proper graphics card that generates NMIs instead of interrupts so that the other interrupts don't mess with the timing!

Improved memory logic

After a bit of a redesign I've now got both CMOS and NMOS working at 4MHz reliably. It turns out that the memory board logic had too long a propagation to the ROM and RAM0 lines - caused by having two cascade '138s and lots of '02s and '04s in the way.

After a bit of head scratching I've now got the control logic down to the attached diagram and this seems to be a lot more reliable at speed and leaves a nice WINDOW signal which will be used as the window for the graphics memory when I get round to trying out the VGA card.

Dom

Working again

I've now got everything working at 4MHz with either the NMOS or CMOS part. This after much messing around!

I've now got a "proper" clock module giving 8MHz as the main clock and drive all the other chips from the XTAL/2 line (to get proper 50/50 duty cycle). Also the clock is now buffered seperately before going to the bus.

I still am getting the odd glitches though that disturb interrupt operation but I'm not sure why!

Anyway I've now added a CTC chip and am exercising this with some code. More schematics and code to follow soon....

Once step forwards...one step back

The step forward:

I've now got IM2 mode interrupts working - and a decent improvement in throughput!

The mistake I'd made was in the buffering of the Z80, not the software. When an IM0 or IM2 interrupt happens the z80 asserts /IORQ and /M1 together and the device puts the vector on the bus which the z80 reads, however the z80 doesn't assert /RD so my old buffering logic was not letting it through! I've now changed the buffers to work on /WR and WR instead of /WR and /RD. (new bits in blue boxes on the schema).

The step back:

I bought a CMOS z80 (8MHz) from ebay in the hope of improving on my clock speed but this hasn't worked out, the new part is only capable of 2MHz! No idea why yet though I think I'll try adding some pull-ups to the data lines between the buffers and the z80....can't hurt can it?

Dom

IM2 interrupts not working!

I tried getting vectored interrupts working last night but after several bleary eyed hours in the early morning realised it was not my dodgy software but my even dodgier hardware!

I'd not realised that the Z80 doesn't assert /RD during an interrupt acknowledge cycle, meaning that the buffers are not allowing through D[0..7] to the processor. This was confirmed by rewriting the software to work from one vector at I*256+0HFF which worked (all lines are pulled high when /RD and /WR are not asserted)

Tonight I'll try changing the buffering arrangement to work solely on /WR....

Full speed ahead....

I've finally got everything working at 4Mhz and almost working at 6Mhz.

I improved the routing of supplies and decoupling on the memory board (that actually made things worse!)

I added pull up resistors to the memory data lines (buffered) on the memory board and all is now well.

The thing seems to clock up to 6Mhz but then crashes at exactly the same point each time I reboot so it could be a memory/speed issue or it could just be a bug in my transmit code that is only showing up at higher speeds....

Source for z80 monitor

The sources for the monitor for my z80 machine are here:

http://www.authorityfile.co.uk/personal/z80/z80-monitor-20090719.tar.gz

This gives enough CP/M like functionality to run MBasic and BBCBasic

It gets better....

Now I've got a proper Basic working! R.T.Russell's BBC Basic. This allows proper programming and outperforms MBasic in many areas and it uses a lot less memory!

Now for some real programming...

Yay! I've now written a monitor program to test things out it consists of:

• A basic CP/M BIOS - no disk stuff yet
  
• A "monitor" program - a hacked down version of CP/M 2.2 with no disk routines (to fit in 4k rom window)
• An intel hex loader

With this I can now run a few CP/M programs, the screen shot above shows MBasic running, with this I can write some real programs ;)

Now to see if I can get any shareware games working!....

Full Duplex - With interrupts

I've now got Serial I/O working fine on one channel. I've made two ring buffers for receive and transmit and the interrupt service routine services the SIO and places stuff in the buffers. I can transfer huge files looped back through a simple program without losing a byte so it should do for now - though it is slow. I think I need to look at how it prioritises TX and RX as with the processor running at 1MHz and a baud rate of 19200 it doesn't quite give the throughput I'd hoped for more like 1000 cps. Still it will do for now if it is reliable.

For handshaking I've now tied the DCDA line low, ignore the RTS line and instead use DTRA on the SIO as a RTS/RTE this is controlled by the interrupt routine which turns DTRA high when there are only a few bytes left in the ring buffer and reenables it when the buffer is emptied.

Now to write a simple monitor so I can upload code without having to pop the Flash ROM out every five minutes - I'm bound to drop it in a cup of coffee or bend the pins sooner or later....

Serial I/O I think I've worked it out

The SIO uses the old sense for the RTS/CTS lines (asymmetric) so I will have to use different lines. So I think I'll use DTR on the SIO to control data in and tie DCD to ground to auto-enable the receiver and see if that works.

Here's the I/O schemas so far:

Serial IO Working...well sort of!

I've made up a bit of ground on my Serial I/O card today and managed to print a few messages from the Z80 to a computer running Hyperterm in both 3-wire (no handshaking) and 5-wire (RTS/CTS) handshaking - however I've can only get 3-wire working in the opposite direction. I think I need to bridge some signals at the PC end to use just RTS/CTS or possibly use different signals from the SIO instead of its own RTS.

On the memory glitches I've now got it working 100% fine for all but the first bank which goes through a lot of different gates to get the boot rom select to work, I think I may have too many gates in the path and it is not settling quickly enough so I may have to rethink that bit!

Memory problems

There are still a few problems with memory, running a memory test routine at 4MHz throws up a few spurious errors every now and again, fewer at 2MHz and non so far at 1MHz....more capacitor swapping might be in order. I'll get the soldering iron warmed up!

I've got a simple PS/2 PIC microcontroller program working - well it will display the scan code of the currently pressed key on a row of LEDs and send / receive keyboard commands. I'll hook this up to the PIO when I get it going.

First though I think I'll have a crack at getting the SIO going. I've got an SIO/2 chip here and a MAX232, should be enough to get a terminal interface going!

Back working again

Today I decided to try replacing all the metal foil caps with ceramics, after about half way through (I'd only done the memory board) I decided to check to see if the thing still worked at low speed and to my surprise it now works at full speed. So two conclusions:

a) It wasn't the metal foil caps causing the problem - the processor, IO and monitor board are covered in them still
b) It must have been some dodgy wirewrapping / soldering near the decoupling caps. I probably melted through a wire sufficiently to allow leakage but not a dead short which was causing problems!

Two feelings after this - elation at getting things working, despair at all the time lost messing about!

Anyway I've now got the LCD interface working reasonably well and written a little scroller - sorry no video, my crappy camera will not take videos of the LCD. Also written a memory soak tester - so far all my 32K RAM tests good!

Next up:

PS/2 keybaord interface
RS-232 interface
CGA/VGA card
SD/MMC Card interface
CP/M
Graphical LCD interface

Not sure which order I'll attack these in but I'd like more info on my intended VGA card (pictured below). Specifically I'd like to know more about the dip-switch settings at the rear. I can see these causing headaches if I get them wrong!

The card is marked MAGIC VGA, CT-8490 and uses the Cirrus Logic GD510A Chipset.

Progress of sorts...

After much messing I've:

• added extra decoupling capacitors - this seemed to make things worse! [I used metal film caps as I've got loads but these may be making matters worse - I'm getting some ceramics today]
• tried removing the clock signal from the bus to make sure it was getting to the z80 cleanly
• tried different z80 chips
• replaced all the 74LS244 buffers on the processor board
• tried adding pull-up resistors to the data bus (tried 10k and 2k2)

This all made no real difference with the processor often randomly jumping to FFFF or 03FF and register load/saves containing random data

Finally, in depsperation, I added a set of 74LS244 buffers to the memory card databus and can now get reliable operation but only up to 1 MHz whereas before all this started going wrong I could acheive full 4MHz if not 100% reliably.

Next up I'll try swapping all the decoupling caps for 100nF ceramics (should have lower inductance and better a mopping up spikes?) and check that I haven't got any decouplers connected to the wrong lines!

Anyway after all that messing I got my LCD I/O Port working at about 1am, enough to display a hello world anyway!

Grrr...

A very frustrating day, first I blew my last LCD panel by getting the power connections mixed up...

Then about 6 hours trying to figure out an intermittent problem that has now become permanent:

• Stepping through program slowly / by hand register contents are random
• Sometimes works at full speed
  
• Sometimes randomly jumps out of program

Tomorrow will have a few checks:

• Is interupt /NMI getting generated? I've not noticed any IORQ + M1s - these lines should be pulled high
  
• Is ROM corrupted - had trouble then it worked then trouble so doubt its the ROM
• More decoupling!
• Buffer memory data bus?

Very frustrating!

It lives!

Yay after a few false starts I've got all the 7-segment displays wired up and the system now functions.

I had to do a fair bit of debugging and re-wiring to get all of the displays to work properly - I shouldn't sing along to the radio when wirewrapping.

When all was well I retried the test program from yesterday and this still didn't work. After a lot of debugging and getting nowhere. The thing was following the program flow but writing out random data on all MREQ+WR's and IORQ+WRs from the processor which I put down to my processor buffers being wired wrong - then I spotted that it was wrong *differently* each time....after much messing about I tried a different processor chip - hey presto it all works! Only a couple of hours wasted!

Here's a picture of the board halted (A=E12F D=ED, during an M1 cycle read)
And below is a video showing me single-stepping through a test program, then in the second half the test program running at full speed (4MHz). The program outputs repeatedly to the top 7 IOnn lines to make a simple fading routine to make each LED light up bright then gradually dim - well that's the idea anyway.


Next up I'll tackle a simple LCD board so I can do a "hello world"

PS: Sorry I made the videos "upside down"!

First Run...

Yesterday I finished wiring the processor board and I wired up a new board with a 24 LEDs and buiffers to help diagnose any problems (they cover the control and IO[0..7] bus signals) I'm going to also wire up 6 x 7 segment displays to display the current A[0..15] and D[0..7] lines as hex (well as the odd hexadecimal that the 74LS47 makes)

However I couldn't resist (it was taking so long to wire up the seven segment displays) so had to plug in and give the thing a test.

First there was a problem with the thing just hanging which I managed to trace back to the WAIT signal always being asserted, this due to me not wiring the pull up resistors to Vcc but instead to GND. Wiring up all those LEDs has already proved its worth! I'd never have spotted this so quickly without them!

Next I tried running at full speed, the ROM is blown with a program that should reset the memory controller then access each of the IO1..7 lines (not 0) over and over. When I ran it these all lit up dimly! Success...maybe

Next I tried single stepping, this was less successful firstly it gets stuck after a reset.....then I found that reset only works by holding down reset and doing a number of clock cycles. Then it kind of seems to work but is not running the program I intended there's a few M1 cycles showing up but sometimes it will do a lot of cycles without and M1. I'm not sure if this is due to the single step switch bouncing - I'll try a bigger debounce cap or if there's something bad happening. Also the IO1..7 lights don' light up as expected so I reckon something odd is going on!

Next I tried wiring up the most significant digit of the address lines to it's 7 segment display this should be showing E all the time but actually jumps about all over the place which suggests to me that there's a problem....anyway I'll try and connect up the other digits and see if I can work out what's going on.

Back later with some pictures...

Schemas

After a few days of construction - processor board part built (z80, buffers and clock generator) I've made a few changes.

I spotted a few mistakes on the memory chip select logic - still not sure I've got it quite right, when it boots the ROM bank at 3F000 (in the ROM) should be mapped at both 0000 and E000 in the Z80 address space. Any write to IO0-32 should then remap the ROM if bit 6 is set then RAM will be mapped at 0000....we'll see if this works.

Another change was to add 40106 to detect the reset and clock switches as normal TTL logic gave noise at the switch over (the 40106 is slow and has schmidtt inputs).

Lastly I moved the clock signal to be away from other lines and between two ground lines - it was giving a lot of cross-talk to other lines at 4MHz and as this is the highest frequency line I supposed it would be better away from the other lines. If the other lines prove to cross-talk too much I may have to completely re-think the bus!

Anyway here's the schematics so far....

Backplane mk.I built

Here's a rather blurry picture of my first cut backplane. The system bus is fed along the IDC cables each card mates using a 64 way eurocard connector.

The board plugged in is the memory board which I built in 2007. I've still not tested the board yet other than to check that the supplies go to the correct pins on each chip!

The memory board has only 32k RAM fitted but the design allows for 48k RAM. The empty IC holder is for the 256k Flash ROM which can be paged into the top 16k of memory space and into the bottom 8k after a hard reset.

Z80 Homebrew Computer

Hello,

In this blog I will be keeping brief notes of my progress in trying to design and build a Z80 based computer.

I'm in no way holding this up as an example of how things should be done I just hope it will be of interest!

I first started this project back in 2007 and managed to build a RAM/ROM board with boot controller and ROM segment selector (I'll explain in a later post) but then I moved house and have never managed to get round to doing more.

Before I start I'd like to give some thanks to people who have donated many of the components I'll be using:

From uk.comp.vintage back in 2007:
Little Paul - wirewrap sockets
Barry Ruck - ICs and dip switches

From vintage-radio.net forums (2009):
GMB (George) - Wirewrap sockets
AC/HL (Bill) - Wirewrap gun and bit!
Brianc - Loads of ICs and sockets!

Also many thanks to the people on the vintage-radio forums and the comp.sys.cpm, uk.comp.vintage and comp.sys.acorn.* and bbc-micro mailing list who have given help and advice on programming and electronics over the years.

Dom