In the process of designing and editing the schematics for my ForthBox Kestrel, I realized that other folks here might want to also use a 65816 in their projects, but might not want to bother with the bus demultiplexing. Moreover, people might want to interface slower logic to an otherwise faster running CPU, so that the system can run fast when accessing fast memories, but slow down when appropriate to access older or slower hardware.

I am thinking of releasing the CPU component of the Kestrel as an independent product for those who are interested. The cost of the parts alone totals close to $16. The cost of the printed circuit board will obviously add to this, and then there is a small amount of profit that I will be adding in as well. I'm not looking to make 2000% profit on this. But I do need some kind of profit to justify continued development and technical support.

The features are as follows:

* W65C816S-14 microprocessor running at 12.6MHz.
* Kestrel asynchronous bus (3-phase) supporting arbitrary throughputs up to and including 12.6MBps.
* No on-board RAM, ROM, or I/O capabilities; no address decoding either.
* Available as a kit, or pre-assembled (for an additional fee)

Basically, as far as your project is concerned, the CPU module is a raw 65816, but with an asynchronous bus interface, and with a fully demultiplexed 24-bit address bus at your immediate disposal.

I envision this product is great for the following applications:

* Prototyping and development.
* When you need the ability to switch CPU modules for other kinds of CPU modules (e.g., switching from 65816 to Rabbit 2000 for example)
* If you need a good backplane expansion standard for your project, and don't want to manufacture a CPU card yourself.

What do you folks think? What do you folks think is a fair price for such an item? I welcome your input in this matter. Thanks!

I like the idea, but don't know if it's something worth pursuing.

The big thing is, you want to give more benefit than just the circuit layout. It all depends on what your circuit ends up being. If it's hard-to-find parts or a CPLD/FPGA/GAL/PAL that folks wouldn't have the programmer for or requires the use of SMT, it becomes a gain. If it's just a collection of through-hole parts, you aren't going to give the person who is purchasing the circuit enough value to really justify buying it.

This was actually my line of thinking too, but I figured I'd pass it by the folks here to see whether they thought similarly. If it's one thing I've learned in trying to start my own businesses before, it's that the weirdest ideas are sometimes amazingly marketable.

You mention CPLD/FPGA circuits. I'm thinking of investing in the programming software and hardware required for programming such circuits. Not necessarily right away, but probably by year's end. What kind of devices would you or others be interested in seeing made available?

I too have been looking at CPLD as a way of reducing the amount of discrete logic that I have to work with. I came across a simple experimenter's kit at http://www.al-williams.com/pldhome.htm (usual disclaimer) which has a header card that includes a JTAG interface. The kit I've got can cope with two different Xilinx devices XC9572 and XC95108. The latter has a reasonable number of logic elements on it and is reasonably priced. They're also available in PLCC format which while not really suitable for wirewrapping seem easy enough to work with on a PCB.

I only received the kit last week and am currently away from home on business so it's going to be a couple of weeks before I can even assemble it to have a play!

Incidentally, the microcontroller demo boards from WDC use this device and they supply the logic equations. Note, I'm not suggesting that I will copy this design but I'm thinking of using something similar at the heart of a new board.

Simon J

Well, I think, to avoid me from bankrupting you with bad advice, that you shouldn't spend much on CPLD/FPGA software. Thankfully, there's some inexpensive stuff to lure you into using the various lines, at least. I know that Atmel's CPLDs are reasonably inexpensive.

The best idea is to solve in a CPLD/FPGA the same sort of obnoxious problems that people keep solving and replace parts that have all gone obselete in favor of others. I mean, if you put too much on a chip, it starts to ruin the fun of building your own system from scratch, you know?

1) Simple framebuffer VGA/LCD interface. Should act just like memory that happens to display video. If done well, it would most likely be useful for PIC/AVR projects, too. I mean, there's a 640x480 monochrome display at http://www.earthlcd.com/ for $35. There's lots that can be done there and one of the big things that everybody keeps spending time trying to do, especially given that the remaining available chips really suck.
2) ISA bus interface good enough so that we aren't stuck with 6551s for UARTs.
3) A "done the right way" 65c816 upper-8-bit address decoder. (Which almost gets back to what you were selling, except a little more minimal and done as a single chip)
4) Floppy drive interface. It looks pretty easy to bit-bang an IDE or PCMCIA interface, but people haven't done much with floppy interfaces for a while. Although that might just call for some simple glue and an existing SMT super IO chip on a module.

I currently am learning how to do behavioral modeling with Icarus Verilog. It's quite the learning experience. I'm currently working on the Kestrel's first piece of custom hardware: a simple UART.

Some features include the following:

* Frame sizes from 1 to 14 bits. Most other UARTs restrict you to only 5 to 8 bits at best. A 14 bit frame is obviously incompatible with most devices; but with compatible hardware, you can use it to include ECC redundancy, for example.

* Data rates freely programmable from 27 bps to 787kbps (yes, that's 787,000 bits per second).



I was thinking of using Cypress or Atmel for CPLDs, and Atmel or Xilinx for FPGAs.



This warrents two responses.

First, I agree; and I will be developing replacement chips based on customer requirements.

Second, the Kestrel is a beginner's kit; I can't have too many surface mount components in the design, because the hand of a beginner isn't going to be as steady as the hand of an experienced electronics kit builder. And surface mount components all require precision when soldering. It also reduces the cost of the kit, letting me get more profit for the value (which is important, considering I have only $15 in my bank account at the moment).



The Kestrel has been re-spec'ed to have DMA-fed video with the following (maximum) configurations: 2, 4, 8, and 16-colors at 320x240/320x480, and 2 or 4-colors at 640x200/640x480. Colors are out of a palette of 256 colors. VGA interface. NTSC and/or PAL interfaces are possible, I suppose, but they'd take a lot longer to develop and debug. I doubt I could pull it off without help from others.

The custom logic that makes this happen will be available as discrete components for the independent project as well.



Actually, the WISHBONE interface is openly documented on the opencores.org site. And upon reading it, I realized that I had basically and completely re-invented it with my asynchronous logic for the 65816. Like my interface, WISHBONE implements a three-phase asynchronous protocol that can support single-clock data transfers. And unlike my interface, it's fully specified. While intended for use as an interconnect inside FPGAs, if you tie the DI and DO (data IN and data OUT) buses together, you can easily get a slot-capable expansion back-plane.



This is also on the design board.

The idea I'm following is that every "core" I develop for the Kestrel will also be made available as a discrete component in chip form.

All 65x-support chips will belong in the FTS10xx family.



A floppy interface would be an interesting challenge. However, the cost of floppy storage is prohibitively expensive compared to a harddrive. Floppies get maybe $0.50 to $1.00/MB, while harddrives are pennies per MB, or less. However, I acknowledge that floppies are convenient for sneaker-net applications.

I will definitely consider these. Each of these are things I've always wanted to hack on in the past. And if I can earn some small profit to help support myself in the process, all the more better!

Another thing: I didn't get my usual e-mail saying someone had posted to this topic. Why is that? I always used to!