Chimera Interest

[mos6507]

I have to decide whether or not to try to continue reviving the Chimera project. Chris Wilkson has agreed to look over the IP but it will take some time for me to compile the fragments that I have available for him. Before I do this, I want to run this poll to see how much interest there currently is.

 

For those who don't know what Chimera is, please visit my blog.

 

http://www.atariage.com/forums/index.php?a...103&cat=176

 

Thanks.

 

BTW, this is relevant to the poll. We were hoping to keep the pricepoint of this cartridge between $30-50. Last time I did the math, the raw cost was inching close to $30 and that's not counting packaging. The move to the bigger ARM and using a 4-layer board really impacted things. If it can't be sold for under $50 then I won't even consider doing it. I'd actually like it to be considerably lower than $30 but it doesn't seem possible to do without cripping it.

Edited July 5, 2007 by mos6507

[Shawn]

I would buy these boards and\or games that are big enough to utilize them.

[Quadhorn]

I've looked through the blog, but it's a little confusing and fragmented; could you give us a brief summary, please?

 

Edit: Never mind; finally found it on the third page.

Edited July 4, 2007 by Quadhorn

[A.J. Franzman]

I've looked through the blog, but it's a little confusing and fragmented

Probably because it's in reverse chronological order.

[mos6507]

I've looked through the blog, but it's a little confusing and fragmented

Probably because it's in reverse chronological order.

 

Yes, you're going to want to start from page three and work back. The first two posts go over the main stuff and the newer posts add some additional features which we discovered we should be able to do on the hardware.

 

Another useful thread:

 

http://www.atariage.com/forums/index.php?showtopic=96885

[Quadhorn]

@AJF: Aye, I was having a 'stupid day'.

 

Chimera looks cute, but a little over-kill if it's main intent is to provide a new cheap Supercharger alternative.

[mos6507]

Chimera looks cute, but a little over-kill if it's main intent is to provide a new cheap Supercharger alternative.

 

 

We ran the numbers up down and sideways and kept coming back to a combination of parts that just so happened to enable a lot of extra stuff which I was determined to exploit for the sake of maximizing the utility of the cartridge. Most of these new features would be "firsts" for the VCS and were really exciting. (We had fast bidirectional serial, read-write queues, and magic writes running successfully.) Once we started discussing the project here on AtariAge, it seemed like people became much more interested in these new features than they were with the strict Supercharger implementation. In fact they started asking for even more than we were planning to give (namely more queue memory). So after much handwringing we bumped the ARM up a notch and went to the 4-layer board, hence the current design. We were so close yet so far, that it's kind of hard to think about scaling things back. However, if someone were to magically come forward with a design that could be built for $15 or less that lets you store a multiload Supercharger game on internal flash or load through RS232, I'd love to know. This is a challenge of creative engineering and I am just relaying what Delicon's intuition was on this. Others may have innovative ways of working around these constraints that he couldn't see.

Edited July 5, 2007 by mos6507

[Quadhorn]

We ran the numbers up down and sideways and kept coming back to a combination of parts that just so happened to enable a lot of extra stuff [...]

Fair do's. Aye, if extra stuff's there anyway, use it.

 

Hmmm, $15 is kind of tight - any idea of intended volume? Single quantities are always rediculously expensive. Is there room on the CPLD for a simple bankswitch method, IP processor, RS-232 controller and FLASH programmer? You'd obviously still need an RS-232 transceiver, parallel Flash RAM, power supply, sockets, passives and PCB to complete it. 'Might be able to get away with a 2-layer board given the virtually point-to-point layout, with everything else inside the CPLD. The "processor" would just need to be a simple state machine to handle getting data from the RS-232 to the Flash programmer. I reckon a Spartan-3 3S50 could easily handle it, but then you're looking at $10 for that alone, plus all the extra support components and hassle it brings in. CPLD's I don't know enough about, straight off. Have you looked at something similar already or is it worth me taking a look at?

[mos6507]

Hmmm, $15 is kind of tight - any idea of intended volume?

 

I wanted to make a batch of at least 200-250 with no preorder requirement. These board would be used for both standalone and embedded varieties. Standalone means sold as a blank (or with the Starpath games on it). Embedded means it would be sold with a single homebrew on it and branded as such. The most likely first game to put on Chimera would be Eric Ball's Leprechaun. All the levels would be contained on internal flash so it would look and act like a regular cartridge but would be equally programmable (depending on how the game wants to handle it). If I wound up having to subsidize this (i.e. take a loss on each unit) to keep the price down then I'd have to limit things to about 500 or so units before dropping out. Someone else could take over at that point if they wanted. Odds are by the time 200 boards run out the parts used would be cheap enough to make a second run much cheaper.

 

Single quantities are always rediculously expensive. Is there room on the CPLD for a simple bankswitch method, IP processor, RS-232 controller and FLASH programmer?

 

You can look into it, but it seemed like it would result in spending nearly the same amount of money and losing a lot of functionality in the process.

 

I think it's too late to make wholescale changes in approach. This project has already been in on-again-off-again R&D for a couple years. I really wanted to see whether or not it could be completed as planned. And I'm just here to make them and sell them. Someone still needs to volunteer to do the finishing work on it. So it's really hanging by a thread.

 

BTW, the cart doesn't require external power unless it's needed to supply power to external peripherals. Most external peripherals would not need this.

Edited July 6, 2007 by mos6507

[Quadhorn]

You can look into it, but it seemed like it would result in spending nearly the same amount of money and losing a lot of functionality in the process.

 

I think it's too late to make wholescale changes in approach.

True enough ... but I was considering having a stab at that "magical $15 design"; I've got something else in mind I want to do eventually, but this might be a simpler starting point.

 

Someone still needs to volunteer to do the finishing work on it.

How much work are we talking? And how much documentation is available? You might get offers if people had more of an idea what they'd be taking on.

 

BTW, the cart doesn't require external power unless it's needed to supply power to external peripherals.

Out of interest, how much current can you get out of the A2600? I've asked elsewhere before and had no response; you sound like you've had hands-on experience with this.

[mos6507]

True enough ... but I was considering having a stab at that "magical $15 design"; I've got something else in mind I want to do eventually, but this might be a simpler starting point.

 

I am not qualified enough to say what is or isn't feasible without some degree of uncertainty. Just bear in mind that a tremendous amount of brainstorming went into what we specced out. It's just that it was limited to Delicon's ideas (and a few of my) ideas. Coming into this project fresh, you might find an innovative path, but you also risk going down dead-ends and winding up with a similar design as this.

 

Someone still needs to volunteer to do the finishing work on it.

How much work are we talking? And how much documentation is available? You might get offers if people had more of an idea what they'd be taking on.

 

The biggie is I have no VHDL or ARM source. No schematic datafiles. I just have compiled files, notes, and screenshots of schematics. So some reverse-engineering and extrapolation is required.

 

What I can do is start compiling some materials and put it online somewhere with the understanding that it will be a constant WIP. I'll just let open it up so everyone can see it and see what happens.

 

If you want a headstart, I would read up on the NXP LPC2368 which is an ARM7 microcontroller running at 72mhz:

 

http://www.keil.com/dd/chipinfo.asp?did=4152&bhcp=1

 

Most of the R&D up to this point involved picking the most appropriate LPC chip, figuring out the best way to wire it to the CPLD and to peripherals, and coming up with programming strategies to allow it to run in parallel with the VCS with minimal latency involved for things like DPC queue functions.

 

I have faith that all of the wiring decisions were ironed out with version 10, but the software still needed a lot of work.

 

Out of interest, how much current can you get out of the A2600? I've asked elsewhere before and had no response; you sound like you've had hands-on experience with this.

 

The 2600 power supplies vary from 400ma to 500ma. Even if you put a more powerful brick on the VCS, there is a voltage regulator inside the VCS which prevents the cart from pulling too much power.

Edited July 7, 2007 by mos6507

[supercat]

Is there room on the CPLD for a simple bankswitch method, IP processor, RS-232 controller and FLASH programmer?

 

My 4A50 design includes a very powerful banking system, plus the ability to in-circuit program flash using a special header, in a Xilinx XC9536XL. The Chimera did not include a chip with a "bus keeper" circuit (present in Xilinx 3.3V-supply parts but not 5V-supply parts) If the Chimera design is reworked, a bus keeper should be included.

[mos6507]

My 4A50 design includes a very powerful banking system, plus the ability to in-circuit program flash using a special header, in a Xilinx XC9536XL. The Chimera did not include a chip with a "bus keeper" circuit (present in Xilinx 3.3V-supply parts but not 5V-supply parts) If the Chimera design is reworked, a bus keeper should be included.

 

He added a resistor pack between the CPLD and the cartridge with version 9.2. We used this to verify Magic Writes.

[A.J. Franzman]

Out of interest, how much current can you get out of the A2600? I've asked elsewhere before and had no response; you sound like you've had hands-on experience with this.

The 2600 power supplies vary from 400ma to 500ma. Even if you put a more powerful brick on the VCS, there is a voltage regulator inside the VCS which prevents the cart from pulling too much power.

I seem to remember reading that the 400 mA or 500 mA was only a labeling difference and the supplies are actually physically the same. I believe the 500 mA version is far more common anyway; I can't remember ever seeing a 400 mA version (perhaps one of the three Heavy Sixer versions?)

 

Anyway, the console with a 4k cartridge typically uses around 300 mA or less, so you should expect to have at least 200 mA at 5 volts avilable to the Chimera.

 

The regulator in the console is a 7805, which I believe can supply up to an amp if properly heat-sunk (but only Heavy Sixers and some Light Sixers have true heatsinks; others use a metal plate or copper pour on the board, and I've come across one or two that did not even have the regulator screwed down to the copper pour).

[mos6507]

I posted the PCB layer images and a parts list on my blog.

 

http://www.atariage.com/forums/index.php?a...103&cat=176

 

I'm starting to think reviving this may be a real longshot, though, because none of the parts are labelled on these images. I can figure out what some of them are but not the small discrete components. Maybe it would be clear to someone who knows what they are doing from context?

[Quadhorn]

Thanks to all that have answered.

 

The 2600 power supplies vary from 400ma to 500ma. Even if you put a more powerful brick on the VCS, there is a voltage regulator inside the VCS which prevents the cart from pulling too much power.

I'm still not sure I've got the answer I'm after; is that the _adaptor_ that provides 400-500 mA in _total_ to the A2600, or is 400-500 mA how much you can actually pull through the cartridge slot's Vcc line without killing / brown-outing the A2600? I think AJF might have answered that, to be fair.

 

@AJF: 300 mA drawn with a 4kB cartridge ... hmmm ... any idea how much of that's the console and how much is the cartridge? I know modern S/NV RAM'll only use 10's of mA, but I have no idea how power-hungry the older chips were.

 

Btw, it looks like you can fit a UART on that $2.50 XC9572XL ... with a little space to spare. Cutting down the UART a little (don't quite need everything for this - fixed baudrate, remove some buffering and error reporting and have only one or two fixed responses to the host instead of a full TX line) and with a little tweaking, there might _just_ be enough space on there for a NOR RAM programmer (simple state machine, off-load sequencing to the RS232 host machine) and some simple bank-switching ... maybe the smaller RAM-free methods. I'd prefer to avoid using two CPLDs if one can do the job. If only the next CPLD up were only $5, but where I can find it it seems to jump to $10...