Saturday, 30 April 2016

PET Diagnostics - Upcoming Project Preview

This is a preview of an upcoming project. Not sure if the name will stick, but at the moment it's called PET Exerciser (UPDATE: it didn't, it's now called 6502 Diagnostics). This started when I was repairing a few PETs and was using a NOP generator. This idea of these is the hard wire the NOP instruction onto the processor databus, the end result of that is it just steps through all the addresses sequentially, continually. That is useful to check address decoding, and see if chip enable pins are pulsing in the right way. The trouble is the 1MHz 6502 based NOP generator is a bit slow to check high address bits on a scope (at least one without digital storage), but too fast to check by eye.
I got to thinking, could I adjust the clock rate fed to the 6502 to get the pulses into a more usable range? Before I started looking at that, I thought that it would be easier to just bypass the 6502 all together and use a couple of 8 bit counters to cycle through all the address bus at whatever speed I chose. I did look at a few options for that, but it was going to be a bit messy as it would have ended up with 4 x 4 bit counters and a pulse generator, which was maybe going to be an 8 pin microcontroller. I then thought I could just use a microcontroller. And lo the PET exerciser diagnostics was born.
I then got to thinking, if I had a few more pins on the microcontroller, I could access the databus and some of the control signals as well. Hmm. interesting. The prototype doesn't look as neat as that final PCB will.
Under that mass of wires is a Commodore PET, and on the breadboard is a micrcontroller, the DIP version of the one I will use on the final board. The microcontroller is connected to most of the pins in the 6502 socket. There isn't an actual 6502 involved any more. Before I got around to writing the code for the NOP generator. I was interested to see if I could actually drive the PET hardware to do anything interesting.
This is the first test was a screen of incrementing characters, but as I got the hang of writing the PET video RAM, that quickly grew into a character table. If nothing else than to help me work out what characters to write the the video RAM to show on the screen - it's PETSCII, not ASCII. Since I can now write to RAM, I can also poke things, and one useful poke on the PET is to change character sets from business mode (with uppercase and lowercase letters) to graphics mode (without lowercase but more symbols. This sets CA2 on the VIA which controls the high address line on the character ROM.
Here you can see the PET board I was testing has a faulty character ROM, some of the symbols are not displaying correctly in graphics mode. OK, this is proving rather interesting and should be very useful. What next? Well, let's test the RAM. I'm currently working on a 2001 board, so I've only set it to test the first 8K. I've split it into 1K blocks and tested by bit to help narrow down the faults.
The test algorithm I'm using is fairly simple, I'm just poking various values and reading back. I'm recording any bits that are different from the values written, and here it has picked up 2 errors on one RAM chip. This matches what BASIC was showing when the PET was running, it only detected the first 1K of RAM, so was showing around 800 bytes free. I've removed the chip I think it is, and retested.
Good, it's showing 400 errors (that's 400 hex, 1024 decimal), all in the lower nibble, which is exactly the area served by the 2114 chip removed. I tried to lay the screen out to match the orientation of the chips, so the second row of 2114 chips, right hand side is the culprit.
I replaced the chip and retested and it's now fine, and when I return this to 6502 control, it should detect all 8K of RAM again. The TMS2114L-20NL I have used there isn't an exact replacement for the TMS4045-45NL, but is enough to prove the point whilst I try to locate something better. I'll probably use one of the pair that are the video RAM, and replace both of those with faster 2114 chips.
You can see where I was also starting to test the ROMs. I've gone for a simple 16 bit checksum of the ROMs, and that seems to be working consistently. I've added a lookup table of CRC codes with descriptions for known good ROMs. I've got it plugged into one of my 6502 ROM/RAM boards for testing, so I have been able to get CRCs for most of the standard ROM sets already. This is a good way to identify faulty ROMs.
Here some of the 6540 ROM chips are working (I've given the chip numbers of the good ones), but the others are coming back with different CRCs. The CRCs have also changed on those as I have repeated the test, so I guess they are faulty, or since it is only the x000 ones that are working, there may be a problem with the BA11 line that controls the x800 ones. More testing required.
I've added in a test for video RAM before the screen is drawn, and inserted the results into the memory map. I'm testing the full 2K region, but the second 1K is a mirror of the first, apart from 80 column systems. I've also expanded the address ranges to test the 32K and 4K ROM sets on a 4032 system. These two modes may be a different build, jumper selectable, or maybe by a menu?
I'm getting a bit carried away, there are so many possibilities here, and I think this is going to turn into a very useful piece of kit. There any many more things I plan to add, such as scanning the keyboard to test the keyboard, and also to control a menu system to cycle between various tests. I can also add testing for datasettes, IEEE-488 drives and even printers. I need to look into initialising the CRTC on later PETs so they can be used as well, and also reformat the screen for 80 column use. I could even include a simple BASIC interpreter....maybe not. It would also suit many other 6502 based machines, including the VIC20, BBC micro, and maybe some of the Atari 400/800 series. Anywhere there is a 6502 processor and a display that I can work out how to drive.
So just to recap, this is being generated on a real PET, with the PET exerciser in the 6502 socket. It's a microcontroller with it's own ROM and RAM. No 6502 code is executed. No actual 6502 is used. No PETs were harmed in the testing of this product.

UPDATE: I've decided to go with PET diagnostics. I have also added a reset button and disconnected that from the onboard reset, instead monitoring that to show faults with onboard reset circuits.

UPDATE 2: I've changed the board to just 6502 diagnostics, as I will be looking at other systems to use (KIM-1, VIC20, BBC etc.), I'll keep PET diagnostics as the on screen title for the PET version of the software. I've removed the reset switch, but kept the reset monitoring. I've added some DIP switches to allow screen mode to be set and maybe some other options. The PCBs have now been ordered.

UPDATE 3: The boards have arrived, but the PCB manufacturer has managed to mess up my design and rotated the ground pad of the main IC to short out several of the pins, including the +5V rail. This is what I had designed:
This is what arrived:
Now waiting for a second batch.

UPDATE 4: Still waiting for the boards, in the mean time, I've redesigned it a bit, with a bigger microcontroller. This now appears as a USB device and shows status in a serial terminal. This also has some analogue inputs so I can check the rise time of the power and reset pins (to check problems like I had with the Oric 1 repair recently).

UPDATE 5: I'll write this up properly once I have had time to do further testing, but the V1.1 boards have now arrived, with no pads rotated as far as I can see. The first one is now built up and being tested.
This is how it would be used, plugged into the CPU socket of a Commodore PET.

More on the testing of these new board.

The new boards are available now.

If you want to support this blog, you can donate via Patreon or Paypal, or buy something from my store.

Sunday, 24 April 2016

Acorn A3000 USB keyboard

This is a USB keyboard conversion for an Acorn Archimedes A3000 computer, and it has been a long time in the works. I initially planned to build one of these when I picked up an A3000 with battery damage, Unfortunately with that one I was able to fix the damage, so that one is still running as a real A3000.
I have now picked up another one, the damage here is worse, the battery leak has eaten through too many track this time.
This one will getting converted. The keyboard looks reasonably straightforward, two 20 way membrane connectors.
On the A3000, these are connected to an 8051 microcontroller which generates serial data. This is presumably left over from the older Archimedes machines which had a separate keyboard. I did consider making a new board which used that 8051 to run the keyboard and a small microcontroller that processed the serial output. That arrangements works well on the Amiga 500 and Atari ST where the keyboard controller is actually part of the keyboard module.
Instead, I've got for just reading the keyboard as a matrix as normal. I have a standard USB keyboard controller with enough pins (just) to cope with this. The Acorn documentation shows the keyboard controller and connectors. It should be easy to work out which way around they go as the top membrane is missing the rightmost four contacts. All the rows are along the top, and columns and LEDs on the bottom row.
The five LEDs (power, disk access, num lock, caps lock and scroll lock) are connected directly to the membrane The resistors to drive these were on the board, so I needed to add a resistor pack to limit the current to these, driving them from the unused pins on the top row to avoid cutting tracks.
When I built this up and tried it out, it sort of worked, but not all of the keys were working, and none of the LEDs. My wiring wasn't the neatest, and I was a bit short of time, so I presumed I had messed something up with the wiring. I thought it would be easier on a proper PCB, so I designed one to that spec and added it to the list to be ordered with my next batch of PCBs.
Roll on several months, and the new board came back. I built one up, much neater now, with a surface mount resistor pack. The trouble is, that didn't work either. Same problem. After some further investigation, it turns out I had made the assumption that since both connectors were drawn the same way around, and one above the other, that this was the way they were laid out. Well, I was wrong. The top one is drawn with pin 1 and the gaps on the right, and the lower one is drawn pin 1 on the right and the LEDs on the left. It turns out, on the board the lower one is flipped. It is fitted to the board with pin 1 on the left and the LEDs on the right. [Expletive deleted].
So much for the new board. I went back to the original multi-purpose board and removed the resistor pack and wired the resistors to the other end, in a much neater way it turns out.The code had to be shuffled around now as the row connections were all different. It's much easier to change the code to read the pins they happen to connect to, rather than rewiring the connections. They have used three different value resistors in the schematic, so I followed suit. Odd since they are all green LEDs?
Guess what, that worked perfectly! The keyboard mapping was fairly straightforward, most of the keys match a modern layout (well, when I say modern, I mean 'the same as the IBM model M', which came out five years before A3000).
Now that they keyboard is working, I can get on with the plan to install a Raspberry Pi running RISC OS Pi. So it's not really a casemod as such, I've just upgraded the RISC processor on the A3000 to the one on the Pi, and also added more RAM and disk space etc. Oh and HDMI video output and LAN. It's actually quite an old Pi, just happened to be the one I had RISC OS Pi installed on.
More on that once I've sorted out how it's all going to fit together. The A3000 power supply appears to be working, a few caps will need to be replaced, but that could supply 5V for the Pi. The speakers in the case could be wired up.
Not sure about the floppy drive. I've been considering making a USB to floppy adapter for a while, but haven't got there yet.

The revised boards are available now.

If you want to support this blog, you can donate via Patreon or Paypal, or buy something from my store.

Tuesday, 12 April 2016

Sony HB75-P MSX USB keyboard with USB joystick

I was sent this Sony HB75-P computer to convert into a USB keyboard. This is an MSX computer, not one I've converted before. MSX was a nice idea, but didn't really take off. Rather than a single manufacturer like Commodore or Sinclair, MSX machines were made to a standard specification designed by Microsoft (whatever happened to them?). MSX machines were produced by several companies not usually known for computers, such as Sony, Phillips and Toshiba
I think this would have been the MSX-1 specification, Z80 based, 64K of RAM and a cartridge ROM system. But what I have here is an empty case.
The keyboard has nice, individual mechanical keyswitches, and a reasonably standard arrangement of keys. These form an 8x9 matrix with separate reset switch and caps lock LED. There are two cables with JST XH connectors on, 9 way and 13 way.
I've used one of my standard USB keyboard controller boards here, one with extra rows of holes to deal with connectors of various shapes and sizes. The pads are using the standard 2.54mm (0.1") pitch, but the XH connectors are 2.50mm pitch, they just fit although the 13 way is a bit tight due to the incremental error.
The board fits into the lid of the case with some self adhesive standoffs. The power LED is wired to the board, with an extended cable.
The MSX would have had two 9 way D joystick ports, so I've installed one of my USB joystick boards where one of those would have been. I used one of the boards I designed for Spectrum conversions as it fitted nicely in the space. The power connector side of the board is unused.
To avoid having two USB leads, I fitted a USB hub into the case. Two internal ports for the USB keyboard and joystick, and two external ports available.
The MSX had an internal mains power supply with a nice clunky power switch. I've wired that so it controls the power to the USB hub. When turned off, it disconnects the USB devices.
USB input is via another board designed for Spectrum conversions, again ignoring the power input connections.
That's all working nicely, testing with a USB power monitor.
It's all low voltage, but since there was quite a bit of metalwork around, I tied it all together with earth straps.
With all that wired together, plug in a joystick, plug the unit into a PC and switch on and the USB devices appear, one USB keyboard and one USB joystick.
There are more things like this available in my Etsy store, or contact me if you would like something custom built or modified.

Sunday, 3 April 2016

Nissan Leaf - Three Years On

For the last three years, I've been driving around in a Mark I Nissan Leaf electric car, and it's been great. Three years ago, there weren't that many of them about, but they have certainly picked up. When I got it, things were a bit uncertain, there were lots of neighsayers warning the battery pack would go bad and cost half the price of the car to replace, and also that it would never be practical etc. I was pretty confident it would be fine, but to minimise the risk, I took a three year lease. At the end of that the car goes back to the leasing company and if it does need a new battery, that would be their problem.
Well, here we are three years and 9,000 miles later, and none of the doom and gloom has happened. Far from it. It's been a great car, very easy to drive, very cheap to run and maintain, and no massive battery replacement bills. Running costs have been minimal, servicing seems to be mainly replacing the pollen filter and the screen wash. Charging in many places is free (many car parks, Ikea, Metro Centre, my parents house), and checking back on my chargeyourcar charging card, I have paid in total £3.00 in the last three years. Yes, three quid. The most I have paid in one go was in a business park where a 75% charge on the Leaf's 24KWh battery pack cost £1.05, between 1p and 2p per mile. My old faithful diesel Golf was about 10p per mile (+ servicing, road tax etc.). I don't have figures for how much I spent charging it at home, but at 10p per kWh, it would cost £1.80 for a 75% charge, working out between 2p and 3p per mile. Given that, and all the free charging, I would guess the 9000 miles has cost me in total less than £100. Read that again, nine thousand miles for under one hundred pounds!
I'm quoting these as 75% charge figures as I generally try to recharge when it goes down to 25% charge. That is usually showing 25-30 miles left at that point. I have learned that the range estimates on the Leaf are frankly rubbish. Often double when you would actually get. In practice, driving with no consideration for economy, and enjoying the car and it's quite impressive acceleration, I could get around 40-60 miles before it got down to 25%. This was fine for me, I do lots of short local journeys, so this is great. From Tynemouth, I've driven down as far as Billingham, and up as far as Alnwick. These both pushed the limits a bit. However it was comfortable driving to Durham or Newcastle Airport etc. The Mark I Leaf starts in 'D' (drive) mode, which is full performance. There is an 'E' (economy) mode, that reduces the acceleration and the heater to conserve energy. It's still nice to drive like that, but that is more effort to engage (press the gear knob down twice), so I didn't use it much unless I had to. I think the newer Leafs reverse this to encourage greater use of Eco mode.
The online 'carwings' app shows charge and estimated range has got a bit better, but is still flaky, and suffers from the same poor estimates and still shows percentages that can only be one of 13 values, one for each of the 12 bars on the display, so no additional granularity. There was a bit of publicity recently about the lack of security on this. I was interested to see whether there was any more detail available in the raw data responses, but didn't want to get my IP address blocked by trying it. Credit to whoever created that animation, I wonder if I am the only Leaf owner with a collection of Commodore PETs?
Towards the end, the range has seemed to have dropped a bit. Not sure why. There are 12 bars on the side of the 'fuel' gauge which show battery condition. Mine is still showing all 12. I have seen some other Mark I Leafs for sale with 30,000 miles on the clock and the battery condition dropped to 11 bars. As an example, I was trying to work out the numbers. One day, I was down to 2 bars, so 17% charge, 83% to fill. I charged it on a 3.3kW charger (a 16A supply), and in 4 hours 15 minutes, it was full. That is 4.25h x 3.3kW = 14kHw, it should have been almost 20kWh. I suspect this is an instrumentation problem, so it may have been reading empty when there was quite a bit of charge left, but I was never confident enough of that to run it flat. As I say, the instrumentation of the fuel gauge on the Mark I Leaf is a bit rubbish.
In all that time, I never actually ran out of charge. That was always the first question, 'what happens when you run out of charge on a motorway'. Of course the answer is the same as a petrol or diesel car, people point and laugh. The trick is also the same as with a conventional fuel vehicle, plan to not run out of fuel. I won't name any names, but in the three years I've had the Leaf, another member of my family has run out of diesel twice.
So all in all a great car. I would certainly recommend a Leaf to anyone who drives less than 50 miles a day. That's about 90% of car owners apparently. Cheap to run, very easy to drive, low maintenance costs. I've got a loan of a petrol car at the moment and I'd forgotten how much more effort it was to change gear all the time, go out of your way to stand for five minutes filling the tank with foul smelling flammable chemicals and then paying £60 for the privilege. (N.B. that's £60 for a tank of petrol to maybe do about 400-500 miles, I got 9,000 miles for less and two tanks full).
Like any used cars, the prices on used Leafs are quite low these days. There aren't many of these Mark I Leafs around, but it is well worth considering a used Mark II, which has a better heater, better range etc. It is now the end of the lease, and the lease company (Arval) has messed me around quite a bit offering to sell me the car at the end, then saying they couldn't, then saying ask again 30 days before the end of the contract and so on. The final conclusion is no, they won't sell me the car. Bit of a shame as I would have been happy to buy and keep it. The car has now been collected and I am soon to take delivery of my new Electric car. No, not a Tesla model 3 (I'm waiting for the model X), it will be another Leaf, one with more batteries, bigger range, seemingly better range estimates and a better stereo. More on that when it arrives.