-

Retro Challenge 2024/10 – Schrödinger’s PCB

The ZX81 Module I designed in 2016 has been an example of Schrödinger’s PCB. It has sat around for 8 years in both a state of working and not working. Only when it is observed will the actual state be determined. In this wrap up post we will finally plug it in and find out.

The first step was to finish assembling the board. If you remember back to the start of the month, or indeed 2016, you will have noticed that the PCB was partially assembled with 4 resistors, 2 transistors, crystal, chip socket and header pins. All the values are now on the schematic, so building up the rest of it was straightforward. I don’t know why 2016-Me only fitted 2 of the transistors. They are all the same value and I have a strip of 50 of them. I did find an error with the footprint I used for the composite jack, but this was easily fixed by turning it 90 degrees.

I burned a ROM with the first 8k duplicated in to the second 8k bank. Sadly my pricing gun that I use for ROM labels doesn’t have a Z, so it’ll have to be a 2X81 now :-)

The next big dilemma was about the current limiting resistors – the green blocks on this diagram.

My understanding of what they are for is that they protect the components if there are conflicting values put on the bus. For example, if the Z80 was doing a write with 0xFF on the data bus, and at the same time the ROM was doing a write and put 0xFE on the bus. This situation should never happen, however, if it did, the D0 pin on the Z80 would be 5v and the D0 pin on the ROM would be 0v, so current would flow from the CPU to the ROM faster than the CPU could source it or the ROM could sink it. Having resistors in there will limit the current, and potentially save the parts.

The logic in a stock RC2014 means that there can only be one device writing to the address or data bus at any one time, so this kind of situation couldn’t crop up. However, with something like the original SD Memory Dump Module, the AVR chip was under software control. So any screwup on the software side could mean that both the AVR and another device could write to the bus. So current limiting resistors were important for that.

The ZX81 schematic confused me though. The CPU and RAM sat on one side of the address bus resistors with the ULA and ROM on the other side. But the CPU and ULA were on one side of the data bus resistors with the ROM and RAM on the other side. I do not fully understand exactly what the ULA does, but the implication is that it could mess with either bus in some way. The RC2014 backplane certainly doesn’t allow for that combination of modules with resistors between them. Rightly or wrongly, I decided to separate the ULA from the ROM, RAM and CUP via resistors.

Rather than solder a load of resistors to the backplane though, I put header pins on there and then used sockets with badly soldered surface mount resistors on them. That way I could swap them out or replace them with jumpers if necessary.

So, finally I had a completed module, a programmed ROM, backplane with resistors on it, and no more excuses for not powering it up.

So I did….

And…

Nothing.

Well, nothing with a combination of heat from the ULA. That is probably not good. I only powered it up for a few seconds, but it clearly got hotter than it should. So rather than risk doing any damage, I decided to go back to the schematics and look closely at everything on my module compared to the original ZX81 schematic.

It didn’t take long before I found a discrepancy. The transistor at the top has the emitter connected to the NMI signal on the ZX81, but my module had it connected to the WR signal. I don’t know exactly what the effect of this would be, but I know it isn’t going to give the desired outcome!

Luckily the fix isn’t too difficult. One track needs to be cut, and a wire needs to be added.

So, with that fixed, everything should power up just fine, right? Right? Sadly, no.

The ULA didn’t get hot, though. So that was definitely better. There was also a regular pulse on the composite output that looks a lot like a line sync signal. But there is nothing else.

Time to look closer at what I did in 2016. Whilst there were very few components soldered on to the PCB back then, 2024-Me assumed that those were at least correct. Which turned out to be a mistake! The ZX81 uses a 6.5MHz crystal. However, I had soldered a 7.3MHz crystal on to the board. Why? I really have no idea. I also have no 6.5MHz crystals either. The closest I have is 6MHz.

As video signal is derived from the crystal, the value has to be pretty accurate. Without the right crystal this is never going to generate a white screen with a happy little K cursor in the bottom corner. However, a poke around the data and address lines makes it look to me like the rest of the computer isn’t running either. So even with a picture being generated, I doubt it would include a K cursor on it.

Schrödinger’s PCB has finally been observed, and it is currently not working :-(

Why? I do not know. There are several aspects that might be playing a part. So for Retro Challenge 2025 I will certainly need to address these

  • 6.5MHz crystal needs to be obtained
  • ULA needs to be checked. I don’t think the few seconds where it got hot killed it, but I cannot assume that it is totally undamaged.
  • Limiting resistors need to be understood better
  • Shadow copies of the ROM might need to be in the upper memory. I don’t think this is necessary, but it is a deviation from the original ZX81 design.

My decision in 2016 to base my module on a ZX81 because of the simplicity of a ULA might be flawed. In hindsight, the ZX80 might have been a much better place to start. There are a lot more chips in that design, but they are all regular 74 series logic, and each part of the circuit can be broken down and understood. That will demystify the Black Box of the ULA at least. I will leave it down to 2025-Me to decide if he wants to carry on with this module, or start afresh with the ZX80 schematic.

-

Retro Challenge 2024/10 – Defeating The Riddler

The schematic for the ZX81 Module I designed in2016 had been designed in a very old version of KiCad, and hadn’t survived all the upgrades too well. This resulted in every component being replaced by a question mark in a box. Luckily I had some photos of the schematic from previous Retro Challenge posts, as well as an expectation as to what should be resistors or capacitors or whatever.

Things now look much better, and actually make some kind of sense.

2016 Me didn’t consider the importance of putting component values on the schematic. Or the PCB. My module was based on the original ZX81 schematic though, so it was pretty much a case of matching up components and updating their values.

The observant amongst you will notice that the ZX81 schematic has a lot more stuff on it than my module. These changes can be categorised in three ways;

  • Components which the RC2014 already have, such as the Z80 CPU, ROM or RAM
  • Components which need to be added, but are not part of the module, such as the keyboard or data/address bus resistors
  • Components which will not need to be included, such as the TV modulator including the USA/French options, or power regulation.

In addition, I noticed my module design had some extra components that were not part of the ZX81 schematic.

More specifically, this is a transistor and resistor. This actually matches the composite mod that the donor machine had. So, well done 2016 me for thinking ahead and including this!

Something which is a bit more concerning though is the way that the keyboard lines all seem to be shorted against the TV Out netlist. I think this might be an artifact of the schematic rescue, as I am sure that 2016 Me would have noticed that. However, I still need to test the PCB in case this fault has made its way through to the board.

So, with a donor ULA, and all the component values, I have everything I need to finish the module. There are other important things that I need to do to the RC2014 though. Firstly, I need to burn a ROM.

You may remember from a recent post that I did a partial ROM dump from the donor machine. This was a simple BASIC program

10 FOR F=1 TO 200
20 PRINT PEEK F;
30 NEXT F

I remember reading that the ROM chips used on old machines were almost, but not identical pin out to the 27C EPROM chips, and that it was a couple of data pins which were swapped over. Or something like that. So my worry was that, for example, D3 and D7 on every byte was swapped. So if I could compare the first 200 bytes or so then I would know if everything matched, or if the difference was consistent.

Luckily, everything matched with the ZX81.ROM file I have.

Although converting decimal to hex becomes less and less fun as you go. But at least I know that I should be able to just burn that image to a ROM and it *should* work.

The memory map on a ZX81 is a ZX81 is interesting. The ROM shows up in 4 locations, and the RAM is mirrored too. A13 is the only line used to decode if ROM or RAM is accessed. Therefor the memory map looks like this;

FFFF
      RAM (Shaddow)
E000
DFFF
      RAM (Shaddow)
C000
BFFF
      ROM (Shaddow)
A000
9FFF
      ROM (Shaddow)
8000
7FFF
      RAM
6000
5FFF
      RAM
4000
3FFF
      ROM (Shaddow)
2000
1FFF
      ROM
0000

As I understand it, Only the 8 ROM from 0x000 to 0x1FFF and the 16k RAM from 0x4000 to 0x7FFF actually matter. However, my plan is to burn two consecutive copies of the ROM and set the Pageable ROM Module to a 16k bank size. I will also use a 64k RAM Module with RAM starting at 0x4000. This will mean that the shaddow ROMs from 0x8000-0xBFFF will actually be RAM. I very much doubt that anything makes calls to that address range expecting it to be ROM, but as long as it isn’t critical to bootup, I can write a routing to fill that RAM with the ROM contents.

So, with all of that sorted out… I just need to solder some components, burn a ROM, and put it all together.

-

Retro Challenge 2024/10 – Finding a donor

If you remember from my initial post at the start of the month, the first thing that I needed to do in order to make progress on the ZX81 Module was to find a donor for the ULA. Some things can’t be rushed though, so this has only just happened. Luckily I have several candidates to choose from.

I had some constraints though. Obviously the ULA had to come from a machine that worked. Ideally one that had undergone a composite mod. And preferably not the original one that my dads school friend bought back in the day and passed on to me a few years ago. (The ULA will only be borrowed to check that things work, but I would rather not risk damaging one that has sentimental value)

This one seemed to fit the bill. Apart from the fact that it is housed in a full size case with mechanical keyboard, there is nothing special about it. I think this may well have had various expansions or add-ons back in the day, but those are long gone now.

Luckily, it also has also been composite modded. In particular, this has been done via a transistor, rather than just sending the video signal out of the socket. This means that I can test that this works with my monitor before taking the ULA. I know that some modern monitors are fussy with ZX81 signals as there is no back-porch generated. Adding the circuitry to do that isn’t overly complex, but if I can keep things simple then all the better.

The TV image is actually really sharp, and certainly works with my monitor. So it looks like I have a donor chip!

Before removing it, however, there were a couple of checks that I wanted to make. Firstly, I have a file on my computer called ZX81.ROM which I think might be exactly what it sounds like. When opening it up in a hex editor, though, there is nothing readable in there at all. I would have expected to see something like the copyright message or the keywords listed. So I will take this opportunity to dump the first couple hundred bytes of the ROM to the screen so that I can compare it to the file I have.

I also want to take a quick look at the output signal on a scope.

Everything looks good and just what I would expect from a video signal (without a back porch)

So, with no further ado, out comes the ULA, ready for its new mission in an RC2014

Next up, I need to finish the schematic so that I know what value components I need to populate the rest of this board…

-

Retro Challenge 2024/10 Update 1 – The Riddler

When I failed to complete the Retro Challenge in 2016 I finished by saying that the dog ate my homework. That sounded like a reasonable enough excuse.

This year any failure to complete is because The Riddler ate my schematic!

One of the first things I wanted to do was familiarise myself with what 2016-me did regarding the ZX81 module I designed for the RC2014. I have 3 PCBs here, and whilst they look great, they lack some info, such as the component values. So I opened up the schematic I created and discovered that all of the components have been replaced by ?? boxes!

To be fair, this isn’t the first time I have seen this situation. This was probably created in Kicad v4 or possibly earlier, and one thing which Kicad was terrible with back then was keeping track of component libraries whenever there was an upgrade. I am now using Kicad v8, so there have been a lot of upgrades.

I might be able to install a fresh copy of an old version and hope that has the right libraries and I can map them in to work with the schematic. I think that most of the components are resistors and capacitors though, so even if that worked, I wouldn’t have the values anyway.

I think it will actually be easier to convert the schematic to the v8 standard and manually change each of the parts to the resistor, capacitor or diode as approprite. I still won’t have the values, but the thing I do have is the original ZX81 schematic that this module was based on, so it should be a case of matching things up.

Mid-blog update

I looked back at the intro post that I made and noticed that the photo I reused from 2016 was a picture of the partially assembled module that was sitting on a schematic printout. The values can be seen too!

So that makes life easier!

Not only that, but another blog post from 2016 also has the full schematic (less component values) so that helps too!

The RC2014 has come along way since this module was conceived 8 years ago. I think some of the stuff that I had learned about the ZX81 back then had guided some of the decisions that I made later, which actually helps this a lot.

Firstly, the Backplane 8 and Backplane Pro both have the ability to add resistors on the data and address bus, as the ZX81 did. There is now an RC2014 keyboard which mimics the ZX81 or Spectrum layout and has the diodes onboard, so that makes things easier too. The Pageable ROM module supports 16k blocks, which will help if anything on the ZX81 needs to access the shadow ROM in 0x2000 – 0x3FFF. And, lastly, the 64k RAM module can be set to start at any address. With the ZX81 having RAM mapped from 0x4000 – 0x8000 this is perfect.

-

Retro Challenge 2024/10 – Intro Post

It has been a while since I entered Retro Challenge, but I think I remember what to do. It goes something along the lines of stating at the start of the month by saying that you are going to design a ZX81 module for the RC2014, and then at the end of the month report back in with minimal progress. Looking back, I seem pretty good at that. Retro Challenge 2016/1 started off like this and ended like this. Then later for Retro Challenge 2016/10 it looked like this.

I hope you will excuse me for using the same image I finished on 8 years ago, but, honestly nothing has changed. The ZX81 module is still in this state.

So, before we get to far ahead of ourselves, lets take a look at how we got to this point. My original goal was to recreate either a ZX Spectrum, ZX81, ZX80 or Jupiter Ace module for the RC2014. They are all fairly similar architecture, although the Jupiter Ace has weird dual port memory RAM. They all have a Z80 CPU, some ROM, some RAM and a clock, the same(ish) as theRC2014, so those parts of the schematic can all be ticked off. The ZX80 does everything else with conventional 74 series logic. The ZX81 is almost identical, except that 74 series logic is all smooshed in to a ULA. The Spectrum is similar except the ULA does more stuff (which can also be done with 74 series logic).

I ended up choosing the ZX81 as a good place to start because it is pretty much just a ULA chip with a little bit of support circuitry. The ULA is no longer manufactured, which will mean taking one from a working ZX81 to get this up and running. There are, however, modern FPGA replacements available, so I might well end up getting one of those to swap in at some point. So in January 2016 I translated the relevant parts of the ZX81 schematic in to KiCad and designed a module around that. I got some boards made, but that was as far as things got. Then later that year I soldered in 9 components.

To get started this year I need to do two things. Firstly, familiarise myself with the ZX81 circuit and the one I designed 8 years ago. I seem to remember being fairly confident it would work, but at this stage I have Schrodingers PCB. It both works and does not work simultaneously, needing a test to clarify the actual state. The second thing is that I need to check my stock of ZX81 and select a donor one that the ULA can be borrowed from. I probably want to make note of some of the pin voltages or signal traces before taking it out so I can check those when it is in the module.

There is a chance that I will end up waiting for parts, so the secondary part of my Retro Challenge challenge is to talk to my RC2014 with as many of my retro computers as possible. This may include but not be limited to;

  • Atari Portfolio
  • Cambridge Z88
  • Psion Organiser II
  • Psion Organiser 3a
  • Sinclair Spectrum (with Interface I)
  • Sinclair Spectrum 128
  • Sinclair Spectrum +2 or +3
  • Sinclair QL
  • Amstrad PPC640
  • Toshiba T1200

I think this will be limited to serial communication, and luckily I have a few MAX232 chips and boards here. I expect it will mainly be using the device as a serial terminal for the RC2014, but it would be nice to maybe send a Spectrum loading screen from the RC2014 to the Spectrum. It might sound trivial, but understanding how the serial ports work on those devices, and finding/writing software to use them could be the challenging part.

-

Retro Challenge 10/2016 – Still a work in progress

So, who remembers my entry for Retro Challenge in January?  It was quite devoid of effort and results I think you’ll agree.

Well, this time around, I’ve taken that theme and pushed it even further!  Even less effort and much less result.

Since January, I have had a few opportunities to carry on with the ZX81 Module.  Because I wanted this to be a Retro Challenge project, I deliberately avoided doing any work on it, so I could save it all for October.  Well, October started and I was overwhelmed with non-Retro Challenge stuff.  But after a couple of weeks, I had a spare afternoon, and decided to dedicate this to the ZX81 Module!

So, I dug out the PCBs I had manufactured in January and fired up the soldering iron;

retrochallenge10-2016_1

Time to wrack my brains and work out what my plans had been 9 months earlier.  I had failed to put any component values on the PCB, or even my Kicad schematics, so had to rifle through original ZX81 schematics to work out what these should be.

Now I knew what components I needed, it quickly dawned on me that I didn’t have very many of these at all.  This didn’t deter me though.  I fitted everything I had; IC socket, pin headers, some resistors and some transistors that may or may not be compatible;

retrochallenge10-2016_2

And that’s about it.  I have ordered the rest of the parts now, but not had an opportunity to fit them.  yet.

One thing I have learned from this experience though, is that if I get any opportunity or urge to work on this, then I will seize it.  Not wait until the next Retro Challenge.  If I’d done work on this months ago when I had the chance, then this would more likely be a trouble shooting / tweaking / developing challenge.

One other good thing that has come out of this though, is that when I started this in January, I had a few design changes or additions required for the RC2014 for it to work.  The Backplane 8 has the ability to add resistors between the CPU and the RAM.  The Universal Micro Keyboard has the right connections to work with a real ZX81 or this ZX81 Module (including diodes).  The CPU Module also now has BUSRQ, WAIT, NMI and WAIT pins broken out which is required for this.  These changes should all make things much easier.

Despite my poor efforts this time around, I have thoroughly enjoyed seeing what everybody else has achieved with their Retro Challenge.  Good work everybody else!

-

Retro Challenge 2016 – My Dog Ate My Homework

So, all the way back in deepest darkest December, I announced I would enter the Retro Challenge 2016 competition that ran throughout January.  Those of you that followed by blog or Twitter account when I did this in 2014 will know that I blogged and Tweeted relentlessly for the whole month, but, this time around, almost nothing.  Obviously, I’m keeping some secret about an amazing breakthrough or something, right?  Well, truth is, I’ve done almost nothing.

Things started well, and on 1st January, I designed a new backplane for the RC2014.  Although I hadn’t studied the circuit diagrams for the ZX81, Jupiter Ace or ZX Spectrum yet, I knew that there were resistiors between the Z80 CPU and other devices.  The stripboard backplane I’d been using had served me well, but it was time to progress to a better solution, and one that could be adapted better to my needs.  Knowing that PCB delivery times could be against me, I thought it best to crack on and get this  ordered.

Screenshot from 2016-01-31 16:01:00

The basic circuit is very very simple – however, I wanted to get this just right, not only for Retro Challenge 2016, but for other possible RC2014 uses.  Essentially, there are 8 40 way connectors that are linked straight through – however, the data lines and address lines for the leftmost 2 connectors and rightmost 2 connectors are separated by a pair of pads.  These can either be shorted together for up to 8 commoned connectors, or have resistors soldered across them.  I also added a power connector and the option of either running 5v directly in to the board, or regulating a higher voltage down via a LM7805.

(more…)

page 1 of 1