lundi 25 septembre 2023

Finally, my idea of ​​publishing all the progress of a project in a single post is not a very good idea. This makes the blog less dynamic and ultimately does not reflect the development activity as it really is. 

So I'm coming back to the form of publishing a post for major progress on a project.

I haven't posted anything for a few months now. I actually have several ongoing projects that are at various stages. The holidays forced me to take a break for a few weeks. In this generally gloomy atmosphere, you have to stay motivated and continue to 'think'!

dimanche 9 juillet 2023


It's been a few months now since I subscribed to the Facebook group 'minimalist computing'. Personally, I believe more in this theme than in free software.

Why that? Because in absolute terms, free labor cannot exist. Whoever spends time studying a solution rather than cultivating his garden can only meet his basic needs through the intermediary of a a currency of exchange that he will have acquired thanks to his study and with which he will be able to obtain what he needs with people who specialize in this type of service :food; clothing; health etc. .

For me, there are three types of services:

  • Paid services at very high cost.
  • So-called 'free' services.
  • The 'intermediate' services at an 'affordable' displayed cost.

What do i mean?

As everyone knows, 'when the service is free, then you are the product'.

This is understandable when it comes to the sale of finished goods, but what about when it comes to services?
In fact, it is very simple. This involves considering the product in terms of a service (the IT solution), but also the resources needed to set up this service. 

From here, the three possibilities mentioned at the beginning of the post emerge.
  • The services or solutions that you will pay a high price for, will allow you to really improve your productivity. Quite simply by interfacing with a complex system by offering you a simple method of accessing it. You can then imagine creating your 'own' business because this interface will take care of all the aspects that are not the responsibility of the business application you want to develop. So you can focus on YOUR real work.
  • The so-called free services. Well they don't exist. In this type of business, you will often find the resources in open access. But the other part will be a mountain of documentation, very often disparate, erroneous or even contradictory. As your project evolves, you will be forced to find human resources who can handle different aspects that you did not originally foresee, as well as many other problems.
  • And finally the "intermediate" services which bring together aspects of the two previous offers. Namely that you will be led to finance applications at a certain price, supposed to provide you with simplified access services to what you wish to develop, but which will be voluntarily so disparate, so poorly documented, so unmanageable that in the end, you will still have to create a team to study your project.
To image everything, the first type of service will provide you with CAD software, for example, which will allow you, after having studied it, to easily develop mechanical parts, for example. The entry parameters in the business are simple: the price of the solution and the time you need to understand it. If you are an experienced person, you will be able to assess the situation.

The second type of service, the free one, is also very simple: You are in front of the Himalayas, you are not a mountaineer at all and you want to go to the top. Basically you want to create something around Linux, because it's free: good joke! The Sherpa will know how to be paid dearly for his help. In this case, it's very simple, you are good for the asylum. But there are people who are motivated by the impossible, even in full consciousness, so...

And the third solution: the most cynical. Yes, because it will make you believe that by paying for an environment that is not very expensive, you will get the expected service. Of course, for me, excellence in this area is Microsoft. You will pay for a solution, which will quickly lead you to see that your project cannot be built and will require you to form a team of experts. 

Whereas in the two previous solutions you can 'play' in full consciousness, i.e. you pay dearly and possibly you even take credit for it, but there is a strong probability that you will reach the end, or else you will not pay nothing, but you're young, have a lot of time to waste and are ready to live 10 years with pizzas... much in the third solution, the worst, you pay, with your savings, you are forced to surround yourself as your project evolves, and you realize that in fact you will never deliver the customer on time because everything is done so that you cannot assess the workload required to develop your solution. In fact, it's the best business tactic to nip any possibility of competition in the bud. microsoft is not the only one to work in this way. In general, all the companies at the top of the Nasdaq operate like that: smoking and 'easy' access to the offer, then, when the fish has bitten, we strike and there, it's dead. Either you pay if you can afford it, or you die. Usually you die. The American way actually...

So... Arrived here, what solution is offered to you if you want to develop something? I would say, a key word: simplicity.

Back to basics, ie using simple but effective technologies that you can easily master on your own and whose cost, in terms of learning time or money invested, remains within the realm of reasonableness. Reason why I appreciate the notion of minimalist computing. This is also the reason why working this way is for many, working the way of the 80s.

I wrote well to work "the way of"... Because with the systems accessible today, the performances have nothing to do with those of the 80s. For example I am currently working with a small RISC-V microcontroller 32 bit at 144MHz. How does that compare to a microcontroller from the 80s, like the 8031 ​​for example, delivering a painful 1Mips at 11MHz?

And I come to the 'profound' reason for this post. Namely the implementation of the basic 8052 on a fast microcontroller. In fact I already made this type of implementation in the early 90s on a Dallas DS80C320 processor. This processor is about three times faster than an 8031 ​​at the same frequency, and is able to run at 33MHz, which results in nearly 10 times faster than an original 8031 i.e. the equivalent of a basic 8031 ​​but at 100MHz

'What the fuck'? Today, an 80C320 at Mouser is $20.62 : oh my god!!!

STC, a Chinese company, manufactures 8031-type microcontrollers with 1T architecture, ie the majority of instructions are executed in 1 or 2 clock cycles. That is twice as fast as the DS80C320. It is capable of running at 33MHz, the equivalent of a basic 8031 ​​but here at 400MHz, and can be found for around $1.30. 

Okay, 400MHz is an impressive number, but that's just to give a clock comparison to the original 8031. If you divide 400 by 12, you get about 33MIPS. Inevitably, almost 1 MIPS per MHz, at 33MHz, that gives 33MIPS. To be realistic, it's probably less than that since some of the instructions are executed in 2 clock cycles.

It should be checked at runtime, but let's expect 20MIPS at 33MHz, this gives a basic interpreter 20 times faster. This is not nothing, knowing that the BASIC-52 interpreter already works well on a standard 8031. Hence the temptation to implement BASIC-52 on this STC processor, in this case an STC12C5A60S2.

Yes, because it's much more fun to go on this kind of adventure than to spend hours looking at uninteresting Facebook pages!

So why this particular STC12C5A60S2. 

  • Because it contains onboard flash.
  • Because it contains on-board RAM.
  • Because you can find cheap development boards.
  • Because it is programmed very easily by its serial port n°1, thanks to the software available from STC.

The dev board ordered.

The dev board received & the Dallas DS80C320

The programing software

All this to get to this:

The previous image would try to think it works. In a sense yes. In fact, I have a problem accessing the RAM memory. I cannot record Basic lines. The input goes well but the LIST statement does not list anything. I can no longer assign a value to a variable. I clearly have a memory access problem.

However the interpreter starts well, tests the available memory well and moreover displays it correctly. The interpreter requires 1KB to run, which it is. I haven't researched the problem yet.

Because the objective of this Sunday afternoon was simply to see if I could establish a connection with the interpreter on an 8031-type microcontroller, which therefore does not have all the resources of an 8052-type component.

But STC had the good idea to propose a baud rate generator instead of the timer n°2 only present in the 8052 type components. It was therefore enough to code the initialization of this generator in the source code of the version 1.3 of the Basic interpreter to successfully establish the connection and see that it works.

What else needs to be done to make it work well. And well find the reason why the RAM is inaccessible. I haven't looked in detail at the changes that were made to version 1.3 of the Basic-52. I know that there is a detection of the type of processor that I removed to have the place to put my initialization code of the Baud Rate generator. I suspect 'stuff' must be happening in this piece of code.

If I can determine the problem, I will also have to review the EPROM programming instructions. Because in fact I intend to make a small card equipped with a FRAM type memory. In this case, only the EPROM 'type' memory transfer instructions would have to be kept.

In short, it's a fun little subject for Sunday afternoons. Despite the 'fun' aspect of what I have just described, this system, which I could operate at 33MHz, could provide very good performance while offering disconcerting ease of access both in terms of financial cost and knowledge term. That's what I call the fourth way!

The fourth way : financially affordable, easy to use, very good performance, very good durability.

lundi 26 juin 2023

A downloadable MSX cartridge.

After having studied a universal timer based on the operation of the TMS1122, and while waiting to receive the printed circuits board of the definitive version, it is time to continue the study of the MSX cartridge.

The reasons for this type of cartridge: I knew absolutely nothing about the MSX stadard. Out of curiosity, I mounted an MSX board developed by Sergey Kiselev :

This MSX computer works very well. However, the problems started when I wanted to test an application, typically a game. Which cartridge to insert since I don't own any original game cartridge?...

I tested several 'universal' cartridges like the Carnivore 2 witch was defective, the Rookie Drive NX which is supposed to emulate a floppy drive and the MegaFlashROM which is the one I managed to use. 

The common point of these cartridges is that they are not at all easy to use. Most of the time, it's a matter of copying the desired program to an SD card and then inserting this SD card into the cartridge. When starting the computer, it is then necessary to press keys at certain times to access the menu of the cartridge, allowing the program of the SD card to be recognized by the MSX computer, most often by copying the contents of this SD card in a memory space addressed by the MSX computer. It works, but it's anything but practical. And to consider development, it's downright impractical, hence the presence of MSX emulators for Windows for example.

So, I decided to try creating a cartridge that would work like a memory emulator. That is to say that the memory of the cartridge could be loaded from a PC and directly accessible by the MSX computer after RESET. Simple in fact!

Note that with this project, I have no idea where I'm going. I have never programmed the USB interface of the processor used (always RISC V). My experience with USB dates back a few years when I created a MIDI multiport using an STM32 processor. For now, I have created an MSX cartridge and have just implemented the few components needed to make the microcontroller work.

This is the 3D representation of my cartridge and...

This is the hardware realization with just the necessary components to go on the adventure. The first step will consist in programming the microcontroller to behave like a mass memory then to use part of the internal SRAM as memory accessible by the PC.

This is already a nice challenge I think...

vendredi 31 mars 2023

Retro-computing : TMS1122NL

Until today, I used to post topics in no particular order. As a result, a project was sometimes scattered over long periods of time on this blog, except for my repair of the Korg T1.

I published on Facebook a few days ago a photo taken of a printed circuit board that I intend to emulate one of the first microcontrollers, namely the TMS1122, a circuit from the 4-bit TMS1000 family from Texas Instruments.

I was asked if I had a blog on the development of this circuit in order to follow the construction. Hence my decision to make dedicated articles. I will see in time if it suits the form of this blog.

So :

- 03/31-2023 'A small project around the TMS1122'.
- 05/05-2023 'What... since the last publication'.
- 06/07-2023 'What... since the last publication'. 
- 06/18-2023 'The prototype is approuved'. 

A small project around the TMS1122

Why such a project?
In fact, a very long time ago, I made a device around this circuit. It was in fact my first 'processor' based achievement, in 1983. I was so impressed by this system that I kept very good memories of it and the desire to use again the same concept for years.

But what does this TMS1122 do? 
It is quite simply a 4-output weekly programmable clock.
And that's all?
Yes, but he does it very well. Its simplicity of use is disconcerting.

Let's go, what can such a clock look like?

The image is taken from a scan of a French magazine: Electronique Pratique. It is this clock that I also made and used for years: not very sexy!

Presented like this, we realize that the device is quite bulky. We also note that the front panel has many keys plus LEDs and 7-digit displays. The 4 outputs are available on the rear face to supply power (mains) to 4 devices.

Aspects that I find a little annoying: the size; the obligation to connect the 4 devices directly to the clock case without the possibility of a remote control; the display that does not manage the seconds. This last point is important because it can be interesting to put a device in operation for only a few seconds.

The interesting aspects: a clearly readable, simple display which indicates absolutely all the necessary information. A sufficient number of buttons to manage the device without having a headache, even without instructions.

The goal of my realization is therefore to reproduce this clock but by removing the disadvantages and adding the functionalities that I think are absolutely necessary.

Of course, the TMS1122NL is no longer available today. And anyway I could not use it since it does not have the management of seconds, which I absolutely want to implement. Incredible, according to WikipediA information, about 26 million of these circuits, TMS 1000, were sold by T.I. per year, from 1974 to 1979, or 130 million! The TMS1122 being a TMS1000 version programmed to be this famous 4-output clock.

TMS1122NL, the brave!

This raises the question of processor emulation, or application emulation. In fact the question does not really arise. Even if I emulated the processor, since I don't own the application binary because it is factory-written in the circuit, I couldn't run this binary anyway. (Or else, help from a veteran of T.I. who would have the binary in question...)

It is therefore a question of recreating the operation of this TMS1122 programmed to be a clock, on a totally different system. Consequently, the question of the choice of micro-controller arises. There are a multitude of possibilities, in particular the circuits of the ARM type, which I use regularly, or others... I've used many different types all these years. This time, I decided to try the adventure on a relatively new type of processor, namely a RISC-V, In this case, a CH32V203C8T6 from WCH. It's trendy, I know...

Without further ado, here is a three-dimensional view of what I will try to achieve and make it work:

As you can see, the size of this circuit is much smaller than the bulky machine presented at the beginning of the article. Also, there only seem to be two ICs on this design. And you are not wrong. No transistors or other logic circuits either. In fact, the small integrated circuit is an integrated real-time clock. So apart from the processor and the clock circuit, there is no other active circuit on this circuit board. Compared to the original TMS1122 design, will it work?

The original design, provided by T.I.

My design:

My schematic is very different but it contains all the elements of the original application as proposed by Texas Instruments. All switches, LEDs and displays are present. And here, not only 4 7-segment displays, but 6 since it is planned to display the seconds.

At this point, I don't know if this schema is valid. I don't even know if I wired the right signals for programming the processor. I did some tests with a study board that I previously purchased, but despite the fact that I managed to program the processor of this study board the same as what I plan to do here, a doubt remains, which I have not been able to dispel with the documentation. I'll see...

A few days later!

First I mounted the whole board. I started with the processor to verify that I could access it using the programming dongle and the development environment. It has been. So I placed the rest of the components, and this is what it looks like:

On the other hand, what I had anticipated about the programming pins of the processor turned out. I use some inputs / outputs of the processor to control elements, LEDs, buttons. However, some outputs used are also dedicated to the debug port. It is therefore necessary to deactivate this function within the processor to be able to use these I/O as such. In doing so, the possibility of placing the processor in debug mode using the external dongle is deactivated. So I had to connect two switches to the board to put the processor in the right mode using a hardware RESET.
This is the reason for the 4 wires soldered to the board.

Programming the system is not very easy. To simplify the hardware to the extreme, I used very few components, so no buffers, no transistors etc... I also routed the board so that the tracks are the easiest to route, with a minimum of vias.

All this implies that everything must be managed by software. So it took me some time and some headaches. Because in fact I was also discovering this processor. The development environment does not have a graphical interface like that of STmicro to configure the hardware resources of the processor. Everything has to be done 'by hand'. On the other hand, and thanks to the many examples provided, this way of doing things allows you to know very precisely how the processor works, which is not necessarily the case when everything is 'hidden by a nice graphical interface.

Where am I today?

Well, all inputs/outputs are functional. Communication with the real-time clock works well, but it was no fun hand-programming a half-duplex SPI port on a single I/O. Basically, the SPI bus of the processor always uses two distinct I/O. The management of the keyboard works very well, which was not necessarily the case with the TMS1122 or the bounce of the switch contacts was not correctly treated: many false bounces and therefore a somewhat delicate entry of information. I also finished the behavior of the system when entering information.

It now remains for me to create the command interpreter which will have to execute either the orders directly, or place them in the database. Then to realize the scheduler of the database, which will be responsible for the realization of the programmed actions.

On to the next post...

> 05/05-2023 < What has been achieved since the last publication.

Well, database management has been completed. All the functions necessary for the final construction of the program are now established. There will probably some modifications to be made because during development, I realized that I could make some operational improvements.

I also made a front face for this application. I just received it. I must say that for a test shot, the rendering is quite correct:

I still have to find a plastic film to improve the contrast of the LED displays.

In order to keep the project small, I split the construction into two parts. The first, the one presented to you, manages the user interface as well as all the functions of the clock. However, neither the power supply nor the inputs/outputs are present. This is the subject of the second part of this clock, here:

This part will mount directly to the back of the main board. It contains everything you need for a backed up power supply as well as different types of I/O including the 'famous' relays present on the original TI concept. I added some specific inputs/outputs to make this part of the construction as flexible as possible.

I also think I can use this part in a different implementation of the TMS1122 concept, I think using a graphic display.

What remains to do to 'practically' finalize this project: the re-writing of the firmware in a stable version, the actual realization of the I/O board and its test. It still represents work.

> 06/07-2023 < What has been achieved since the last publication.

Well, I rewrote the entire system program. Indeed, the first draft consisted in checking the software feasibility corresponding to the hardware options that I had implemented.

Once this verification was validated, I first had to improve the sub-parts corresponding to the various hardware implementations. On the other hand, I also modified the general functioning, namely that absolutely everything that happens is managed by interrupts and no longer in a main loop reacting to various semaphores coming from multiple interrupts. Finally, I also added some features that are not original in the TMS1121/1122 in order to improve the general workflow.

I just have to check the general operation of the system and correct any bugs. But it will now be much simpler with the new software architecture.

In addition, I received the power supply/relay board as well as the screws necessary for the general assembly. I have not yet tested the power supply board but have already assembled the assembly to check if all the dimensions were respected.

It fits very fine :

I had planned to mount it in a Euro Rack type box. This type of rack is elegant and highlights the general appearance of the TIMER, here next to a Behringer module. Of course, in the case of mounting near a PLC box, this Timer can be mounted in a specific box.

I placed a few too many screws for my taste. I still have to check if the keyboard works as well with fewer fixing points. I also have the screen printing to be improved. All these adjustments will be made on the final version which I hope will be available within a few weeks.

I am also testing the power/driver board. It attaches to the back of the Timer board. Here is what the final assembly looks like:

The Timer is designed to work for a certain time, even in the event of a power cut. This board will also eventually receive some serial links to allow the Timer to communicate with different peripherals. I plan this for a future version. I only assembled some of the components because it is only a prototype for the moment. Everything seems to be working fine...

Even if it is a relatively simple project, I admit that it still takes time when you want to do well and you have to manage all the aspects of such a project. But hey, that's also what's interesting...

> 06/18-2023 < The prototype is approuved!

The two-way communication between the clock card and the input/output card works perfectly well. I can now check the battery status directly on the front panel.

About battery, the whole system is able to run at least three whole days on the 16550 battery, relays not activated in this case. Which is more than enough for 'normal' use of this type of equipment. From memory, the original version, equipped with the famous TMS1122NL, was just able to run about a day with two 4.5V flat batteries in series, and even then, on condition of turning off the clock displays. Whereas in the case of this new version, the displays constantly show the time.

It only remains for me to make the few adjustments that I think should be made on the printed circuit boards as well as on the front face to propose the final version of this clock.

As a reminder, this clock is extremely easy to use and in the current version, offers 4 programmable outputs over the week. 40 program steps are available, which doubles the capacity of the original TMS1122NL based clock. Once this device is really finished, I plan to study various extensions that seem interesting to me to integrate, such as temperature control and other kind of measures ...

vendredi 24 mars 2023

For TEST : a downloadable MSX cartridge

In the 1980s, there were several families of computers. Among these was the family of MSX machines. The interest of the MSX was that it was an open standard allowing the manufacture of compatible machines by a certain number of manufacturers.
This type of computer has not been used for a long time, but remains a very good gaming platform. Msx is now trying to evolve, especially with the MSX-3 initiative, but I'm not a fan of this evolution.

So, and in order to facilitate the development of applications under MSX/MSX2/MSX2+, I am trying to create a cartridge that can be downloaded from 'PC' via the USB bus.

here's what it might look like:

mardi 28 février 2023

MSX Omega board : first game!

Difficult start to the year. I haven't published much in 2023. But here's a little new that delights me: I was finally able to launch and play a game with my Omega board.

The Omega board :

I didn't have a keyboard at the time, and given the cost of building this board, I decided to use a standard USB keyboard. So I made an adapter to connect a cheap commercial keyboard to the Omega:

The results of using this adapter were satisfying, but I didn't know why I couldn't use the Omega board correctly. And then someone gave me a new BIOS that was visibly better suited. From that moment, I was able to use the MEGAFLASHROM SCC+ SD cartridge to install the Galaga game on it.

Why Galaga? because a user of my USB keyboard adapter project ( told me some problems with Galaga. So I solved these problems and tried this game with my corrections.

The system being fixed and tested for 'small' USB converter bugs:

There you go, the game works perfectly. Maybe I'll make a little video, but before that, I'd have to improve at this game where I'm particularly bad!

Champagne: first time I run a 'real' game on my 'own' computer!
Little video of the game. It's obvious, I'm not an inter galactic champion ;-)

Small note anyway. You have to know a lot of little things to successfully operate such a system. Nothing complicated but the documentation is fragmented. You often have to try to put the pieces together to understand how it all works. This is what I 'reproach' the most to this type of system. Too bad, because it's really fun to work with.

UPDATE 03-07-2023

As I have to perform various tests on several games in order to fine-tune the operation of the USB keyboard converter, I realized that this requires somewhat tedious cartridge loading manipulations. I am therefore in the process of preparing a somewhat particular type of cartridge which will be very easy and quick to use. I say no more for the moment. If you are interested in a simple and easily reloadable cartridge, let me know in the comments, indicating a price that seems 'reasonable' to you.

mercredi 4 janvier 2023

Kitchenaid coffee grinder repair : thank you China

Unusual article on my blog, it's true, but as I enjoy troubleshooting well-built devices that seem to be repairable, I'm sharing it here.

In fact, because the media space on the subject of sustainable development is inversely proportional to the possibilities of real reparations. The main objective being display rather than reality, particularly in France which excels in the art of lying! So, a kind of denunciation here...

I have had this excellent coffee grinder for more than ten years:

Over the past few days, a pungent-smelling puff of smoke has escaped from the engine compartment.

Dismantling the device is very easy. I removed the electric motor and could see that the smell was indeed coming from this motor. I checked the 'coals' to see if they were not worn. They were, but there was enough left to keep the engine running smoothly.

I therefore concluded that there must be a problem with the winding. I have not tried to start this engine again. When it has smoked, it is better to replace it, it is much less risky.

Here is the motor :

Nothing special, it is a motor that operates under 230V DC. I thought it was not going to be very difficult to find its new equivalent. In fact, I was wrong. It was not easy to find a new model!

KitchenAid's after-sales service for Europe is located in Belgium. So I called this service, and on the phone, I had a lady who told me that the coffee grinder model was so old, that she couldn't find the motor reference.

The only thing she could do was give me the phone number of a French after-sales service company affiliated with KitchenAid.

I suspected that she was getting rid of the problem with diplomacy. I still called this company, which was of course, unable to help me.

Important information. I have concluded from calling this service company that if your KitcheAid appliance is not under warranty, the effort to find a repair solution will be non-existent.

So: the repairability of KitcheAid devices is only guaranteed for a period of two years! After this time, if the part is no longer in stock, it's dead! This is absolutely unacceptable. A period of ten years should be possible, given the quality of construction of this type of device. Mine must be approaching fifteen years, I understand, but still: two years!!!

So what to do? Thanks China!

It is not easy at all to find this type of engine reference. In fact, I ended up finding a site that presented an engine whose images corresponded exactly to the one I had to replace. Needless to say that I greatly prefer to rely on real references, but impossible on this engine. So I had to rely on the images shown and ordered the motor while mentally preparing myself for this not being the right reference.

But hey, the price of the engine plus shipping was €39.2: not very risky.

And this morning, a very good surprise was in my mailbox :

On the left, the old engine, on the right, the new one.

From this point, I removed the drive gear from the old motor, as well as the power connector, which I placed on the new motor :

Once that was done, all that was left to do was reassemble the coffee grinder with the new motor, which is very easy to do:

If I count the time it took me to disassemble and reassemble the device, it must have taken me about an hour and a half. I took my time cleaning the entire grinder when reassembling. If I had to invoice my time, it would have cost around €30 or €60 (€30*2) if it had been done by a company. By adding the €30 of the engine, it would have normally cost me less than €100. 

The company would obviously have charged the repair as close as possible to €200. Which, for a person who knows nothing about it, is profitable compared to the €300 that this new machine costs today. (knowing that in any case, even this type of grinder in new condition is no longer repairable since the engine is no longer available in the field of standard maintenance, but that, nobody knows...).

Conclusion: the repair cost me less than €40, which represents less than 7.5% of the price of a new machine. It's a very good deal. It didn't take me long for the repair. However, in the future, I will avoid buying this type of brand which makes a good product but unfortunately at the operating cost which is potentially far too risky.

About China: obviously, my approach is not at all in line with the objective of finance, namely to produce at the lowest possible price and impose the highest possible selling price while requiring replacement as often as possible. Yet this is the image that the standard citizen has of China. Personally, I don't have this image. For me, it is a country that can also produce very good products, I use GoWin FPGAs for example, and which also has such a huge stock of parts that it is often possible to find parts there spares unavailable in the West. 

China has become an important country on the international scene. This displeases in the United States, but hey, it's the game! China is a bit too much of a dictatorship, but hey (again), are western countries really democracies? Hum... I have some doubts here...