How to go on vacation with peace of mind!

I would simply say: with as few nagging concerns as possible. 

And the best way to achieve that is to finalize certain projects that are at a conclusive stage. 

Since December 2025, I've been working on two main subjects: an FPGA-based Drumulator emulator, and porting an MSX cartridge design to FPGA. Both have been important because, first, I really had to level up my VHDL skills. Implementing the ZILOG CTC controller in VHDL — complete with the full vectorized interrupt system — was a genuine challenge. But without it, I couldn't have implemented even the Drumulator's Data potentiometer. 

That's the software side of things. 

On the hardware side, I also had to design a PCB for the MSX cartridge, this time for the larger FPGA — an Efinix TRION T20. And that meant tackling my very first four-layer PCB design.

What's more, TRION FPGAs are somewhat trickier to implement than Gowin FPGAs, for instance. That said, it really depends on the specific chips and other factors that can certainly nuance that view — but for this MSX cartridge, it definitely held true. With the physical space constraints of a cartridge, plus the fpga size and the number of passive components needed, sticking to two layers was simply out of the question.

I actually reached out to a PCB specialist who quoted me over €1,500 just for the design work. No doubt he was a pro, but that was way out of my budget — even when considering more affordable options from Asia.

So I had no choice but to tackle it myself!

It took a fair amount of time to lay out the board in KiCad, assemble it, and start testing. But at least I had the satisfaction of seeing the FPGA communicate with the JTAG probe almost right away.

After that came the implementation of the internal Sapphire processor, and finally the interfacing with the MSX bus.

A professional would say none of this is particularly remarkable. And I'd agree. But when you're a weekend hobbyist with no formal background in the field, the road is obviously a bit less 'smooth sailing.'

But there it is! I finally got my first MSX machine to boot on my new cartridge. And I have to say — it makes me absolutely thrilled!

It took a few hours of work just to be able to take this photo — and then I had to adapt the VHDL code I'd written during my tests with the Efinix development board to this new hardware. Unsurprisingly, it didn't work straight away, but thanks to Efinix's debugging tools, I finally managed to get the cartridge up and running:

In step '1', the MSX computer places address addr_IN = 0x4000 on the bus. This is the first address where a valid cartridge signature is expected. In step '2', the MSX reads from the cartridge with SLTSL_IN = 0 AND RD_IN = 0, and finally in step '3', the cartridge returns 0x4241 — which is the signature, i.e. 'A' and 'B' in ASCII.

You may notice that in step '2', the address used to access the internal FLASH on the cartridge is not 0x4000 but 0x2000. That's simply because the FLASH has a 16-bit data bus. Bit '0' of the MSX address is ultimately used to select which of the two bytes should be presented on its data bus. This is also why the FLASH returns 0x4241 — both bytes are read out as a single 16-bit word.

So, a bit of internal address translation going on.

I should also mention that during my initial tests, I used small delays to handle FLASH writes. These aren't standard components — writing to them requires a specific procedure and wait times. To keep it short, this time I'm using the FLASH's #READY/BUSY signal directly for all operations, so I'm getting as close as possible to the FLASH's maximum speed. And I have to say, writing — and thus transferring data from the PC — is now significantly faster.

From time to time, I also work on other little projects — like the CODY computer, for instance.  

I just find this tiny machine really interesting and very affordable, and since all the sources are available, it's a great source of creative fun and distraction that brings me back to my 80s computing pleasures. And that's precisely what this little computer is all about. So I adapted a USB keyboard adapter for it.

I used a solution that could definitely be improved, especially by using a Pi PICO. And since I've just managed to get the PICO development environment up and running, I think I'll do a little study on replacing it with that chip.

Finally, I was able to type things on the keyboard and run the first little program that works.
 

I also sent out a USB expansion board for an AKAI sampler to be fabricated. I honestly have no idea whether it'll work, because Akai used some pretty odd tricks on their original board. But at least this time, the backplane connector should be correctly wired — because despite double-checking everything, I still managed to reverse the ROWs. So the whole thing was upside down. You don't get to be a "Sunday tinkerer" for nothing!

I should get the board back when I return from vacation. I'll still have a few days to test it. And if it doesn't work — well, so be it. I'll have done what I could with this board and Akai's rather baroque design choices.

December 2025 was a big month for me. That's when I really started picking at artificial intelligence about VHDL, and it helped me make huge strides on the subject.

But beyond the technical side and the knowledge I gained, what really mattered was the freedom to learn. 
I don't know how things work in other countries, but Western societies tend to be pretty rigidly class-based. France is a perfect example — it still runs on a kind of "Ancien Régime" model where, if you're a nobody (which I am), you're effectively locked out of higher education. I'm oversimplifying, but the core truth is there: the education system is basically a class-filtering machine that only lets the upper classes access the kind of studies that lead to a "normal" professional life.

"The big shots land the fancy seats
Right on
The rest get Saint-Maur, Châteauroux Palace
Nowhere to go"

(Saint-Maur and Châteauroux are prisons, for real.)

InFrench original : 

"Les cadors on les retrouve aux belles places
Nickel
Les autres, c'est Saint-Maur, Châteauroux Palace
Plus d'ciel"

(Song by Alain Souchon ‧ 1988)

Artificial intelligence currently lets you break out of that situation. Which is a huge pleasure for me, because it makes me more creative. But for how long? That's a whole other debate.

Msx Cartridge, first run.

The MSX cartridge saga continues...

As a reminder, here is what the cartridge should look like once finished. I had the printed circuit board manufactured with a request to have all passive components in 402 packages placed, because doing that by hand: why not, but actually, no!

On the other hand, I had to solder the FPGA myself. Even though it's a 144-pin package, this type of package is very easy to solder. I've already had the opportunity to do it before; a little practice is enough to carry out the operation without difficulty. I also asked for the flash to be placed on the board.

The FPGA functionality tests did not reveal any particular design issues. I just had to remove a resistor on one of the JTAG signals — a useless resistor that I had added without much thought on my original schematic.

The first programming of the FPGA therefore went without a hitch — see previous articles for more details. The famous 'Hello World' worked on the very first try.

So the next step was to program the Sapphire RISC-V processor core provided by Efinix into the FPGA, in order to implement the whole file-reception program from the PC, as well as the routine to write that file to the Flash.

When I do feasibility tests, the code I write is never very clean. Rewriting it for the final implementation allows me to follow a cleaner structure.

To check that the application code for the processor was starting to work properly, I simply programmed the UART for communication with the PC, the processor's internal timer to generate a one-second interrupt, and the timer interrupt routine to send a CRC-16 character (actually the character 'C') to the PC. Indeed, it's upon receiving this character that the Ymodem transfer protocol can start transmitting.

Because, no, I'm not using the Windows copy-paste file system over USB media. That system isn't designed to be controlled — it either works or it doesn't, with no visibility. So I preferred a more controllable transfer method.

And that's when the real trouble began. Although the code running on the processor started correctly and displayed the program prompt along with a few occurrences of the CRC-16 character, the program would stop without any intervention from me.

For a main loop consisting of while(1); that's a problem.

So I spent an evening eliminating possible causes for this malfunction. Then I remembered a small electrical issue I'd encountered during my initial tests a few months ago — a minor problem with ground loops and/or noise. This issue was obviously causing disturbances on the JTAG link, which in turn would crash the processor.

I inserted a USB isolator dongle on the USB hub and plugged the JTAG dongle into it.

And this time, no more problems. The processor runs its program without any issue.

So, my very first 4-layer PCB I designed, even if it's not the pinnacle of the art in this field, isn't causing any issues so far. The code I developed on the Efinix evaluation board works as well. Up until now, the most complex systems I had produced were circuits around Gowin FPGAs, which are simpler to implement. For this MSX cartridge, the whole circuit is nevertheless a bit more complex.

I must say I'm quite pleased with the result :

The Cody Computer : With an USB keyboard?

From time to time, I've mentioned the little computer designed by Frederick John Milens. The main reason for my interest in this small machine is that all the source code is provided, it's very simple to build, and for learning what microcomputing is, or what it was, or what it might well become again, it's an absolutely brilliant tool.

I repeat myself, but for me, the only "small issue" with this machine is its keyboard. In my opinion, it's really not very practical, and it adds an unnecessary extra cost to its construction.

So much so that I didn't actually build the keyboard. As a result, in the few months I've owned this little computer, I've never been able to type anything on the keyboard to test its functionality. Until now, I only had its startup screen. That's already a good sign that it's working, but still, a bit frustrating.

So, a few months ago, I developed a small electronic board to adapt a standard USB keyboard to the Coddy. For the interface, two approaches are possible. The first is to connect directly to the keyboard connector, thus simulating the keyboard hardware with I/O ports on a microcontroller.

This principle works, but obviously, since the computer's keyboard scanning and the microcontroller's I/O matrix scanning are asynchronous, there are rare occasions where a key press is missed.

The other solution is to present to the computer, one way or another, the key code directly returned by the USB keyboard. But this solution, although more reliable than the first, requires having full access to the computer's resources, because it involves not only some hardware modifications but also system software changes.

Since the designer of this computer provides absolutely all the source code, I opted for this second solution.

The hardware principle used is simple: a USB-to-serial converter translates the codes from a standard USB keyboard into codes available on a serial port. A small microcontroller retrieves these serial codes and presents them in parallel on the computer's data bus.

A tri-state buffer allows placing this code on the data bus using an additional selection signal generated from the Coddy's VIA selection signal, discriminated from the VIA by taking into account an extra address signal. Basically, I'm using one of the VIA's memory maps to place the I/O registers that provide the keyboard code.

And that's it.

From a software perspective, it's "simply" a matter of hooking into the existing system code by adapting certain routines according to the new hardware.

Said like that, it sounds simple, but in fact it's not really, because even though all the source code is available, you still need to understand how it works and its subtleties. And adapt to the 6502 assembly language, which I had never used since until now I've always worked on Z80 or 68000. This caused me some initial confusion. But then again, when you know assembly language, switching from one processor to another isn't difficult.

And so, after a few hours of studying and testing, I finally have a keyboard connected to the Coddy. 

And, compared to my previous article where I showed the very first signs of character reception but where the whole processing of commands entered via the keyboard was not yet functional, now it is fully working.

And now?

Well, not everything is actually functional. The "break" that allows stopping a running program hasn't been implemented yet. This system uses certain hardware lines from the original matrix keyboard to generate the Break. I haven't implemented it for now. After the few hours I spent on the BASIC assembly source, I don't think it will be very complicated to implement.

My interface is a bit complicated. I use a small USB-to-serial conversion module, then my decoding and data presentation board that interfaces with the Coddy's data bus.
I think there's a way to do it much more efficiently by using a Pi Pico with the TinyUSB library.

Since I managed to set up (without difficulty) the Pi Pico development environment with Visual Studio Code, I should be able to greatly simplify all this into something better!

And yes... Unlike my first attempts at installing Visual Studio Code with the Pi Pico package for VSC, I was able to complete the installation without any problem whatsoever. It just goes to show that some Microsoft solutions work almost intuitively!

And I almost forgot to thank Frederik for a very handy little detail in his code: PUTHEX.
This routine allows displaying in hexadecimal at the cursor position any data previously placed in the A accumulator.
A very handy little utility for real-time debugging.

 

Some headway despite the heatwave.

Despite this intense heat, I managed to make some progress on my developments. And to kick things off, a magnificent failure. Well, it's still progress, I suppose. This concerns the USB expansion board for the Akai S5000/S6000 sampler. My first prototype doesn't work at all. Mind you, that does make it easier to find the cause. I did indeed make a beginner's mistake on the PCB. It's so obvious, even though I checked my schematic several times, that I'm simply not going to reveal it. So I've corrected my circuit. Now I just need to get it re-fabricated.

There's no visible difference on this new board, except that I've moved the USB-C connector out a bit more, because once the current board is installed in the sampler, I realized it was missing just a tiny millimeter to properly plug in an external cable.

Luckily, I do have some genuine progress to report, particularly with the MSX cartridge. I won't recount the whole development story of this project here, but suffice it to say that I recently moved over to FPGA. For that purpose, I switched to an Efinix TRION FPGA. These FPGAs are new to me and are a little trickier to implement than the GoWin FPGAs I used before, but they're also less expensive.

Given the tight space constraints of an MSX cartridge, and since I'm using a 144-pin device, I had to opt for a four-layer PCB. The first spin of this board isn't perfect, but it's still well enough laid out to allow programming the FPGA through its JTAG port.

The image below shows the result of programming the FPGA via JTAG. In reality, it's a little more involved than that. Using the JTAG interface alone, you can program the FPGA's internal RAM 'on the fly'—which is great for development. However, it also means that whenever the FPGA is reset or powered off for any reason, that RAM is wiped, and the FPGA then expects to find an external EEPROM holding the configuration bitstream to reload its internal RAM and boot up. This is called SPI ACTIVE mode, because the FPGA itself initiates and manages the copying process from the external EEPROM.

So how do you program that EEPROM when you haven't added a dedicated connector for an external SPI programmer? The answer is simply to use 'SPI over JTAG' mode. And that's where the magic happens. During programming, the bitstream is written directly into the SPI EEPROM through the JTAG port. Once that's done, a simple reset of the FPGA is all it takes for it to fetch its configuration from the EEPROM—and that's exactly what this image illustrates.

Here's what my development setup looks like:

I should also mention that to program this FPGA, I'm using a commercial interface I bought myself on Aliexpress JTAG probe. Although it comes with the FTDI chip, it doesn't quite meet the specific requirements for programming the Efinix FPGA. One signal is missing from the output connector, but luckily that signal is available on the internal connector on the PCB. You just have to route that signal to the external connector, and the interface works flawlessly in JTAG mode with the Efinix programming software.

View of the programmer interface:

I'm quite pleased to have gotten this far from the initial study I did on an Efinix test board. It feels like years have passed since my successful tests back in December 2025, even though it's only been just over six months. Since then, I've made a lot of progress with VHDL and have other developments underway. Needless to say, I'm once again having to go back to the drawing board when it comes to implementing the Drumulator in FPGA. Plus, with the constraint I've imposed on myself to use as few American components as possible—would that force me to find other solutions? No, it encourages me to do so. And this constraint actually opens up potentially more interesting avenues. I have a few tests to run regarding the digital-to-analog conversion of the Drumulator, and I hope to be able to decide on the new direction to take based on the results of those tests.

My MSX development system in December 2025:

Despite a few hiccups that come with electronics development—especially since I'm more of a weekend hobbyist than a fully certified engineer—I'm pretty happy with the result. But still, I feel it's not going as fast as I'd like. I hope to ramp up my efficiency over the coming months.

Today's thought: US tariffs applied to France. Thank you, Donald Trump.

Today's reflection: US customs duties.

I really enjoy retro-computing. The question I hadn't really settled until now was: with retro components or with modern components? After the Z80 disappeared, the only 8-bit family left was the 6502, still manufactured by Western Design Center (WDC).

I'm currently spending some of my free time on the little Cody computer and the Memo-1, both based on the W65C02S from WDC. On the other hand, I also took an interest in the Commander X16, also based on the W65C02S, because I found that machine interesting.

To keep going with the subject and especially to build the Memo-1 board, I just placed an order for a few W65C02S processors and peripheral circuits from Mouser.

And this morning, the big surprise. Between the shipping costs and the taxes already paid, which already came to €28 for a total component price of €67.35—i.e., fees equivalent to 41.5% of the product amount—UPS has just billed me €49.14 in customs duties. The total fees for this order therefore amount to €77.14, or 114.53% of the value of the goods.

To be perfectly clear about that last sentence: I paid €77.14 in shipping, taxes, and customs fees for a component total of €67.35.

This is quite simply armed robbery. That reminds me exactly of Apple's business practices, which is why I have never bought any Apple products.

- My decision is simple: components of this kind from the United States—It's over.
- Retro-computing with retro components—It's over.
Period.

Americans are definitely no longer my friends!

Thanks to whom?
Thanks, Donald!