Any interest in custom 6502-like CPU chips?

[enso]

OK, I am getting carried away from the original idea - a pre-programmed FPGA. Apologies.

[Dajgoro]

Can't just start a new topic for that?

[GARTHWILSON]

If the point is to get it really fast (like 100MHz), it won't work with any existing hardware it could plug into anyway, so there's no need to stay with 5V. The 40-pin-DIP 6502 pinout might not work at all at 100MHz anyway though.

[BigEd]

Hi Enso
certainly you could build something like your sketches. History tells us that one person taking an idea and running with it works more often than two people trying to agree on something! As Dajgoro says, a new thread, or a continuation of the previous board-idea thread, would be better. One difference between your idea and this thread's idea is that we are sticking with non-BGA parts, because hobby-accessible soldering is part of the spec. I do realise that you have already shown that BGA is not completely inaccessible to the hobbyist.

Garth, Chuck:
It sounds positive to me, that Garth (for example) could design a board around an FPGA, publish the design files, and then Chuck (for example) could take that design, and build it either modified or unmodified. In the course of Garth's design process, we would converge on an FPGA netlist and pinout which suits that design: we are hopefully wise enough to make it general enough that modified designs would be possible and attractive (*). If anyone at any stage wanted to build and sell a handful or a truckfull of boards, that would be a good result.

I agree that for confident use, the FPGA design files would all need to be published, with an open enough license that anyone undertaking a design could take a safe copy. All the cores discussed already have an open enough license to allow us to work that way: all the HDL I mention as available is available under redistributable terms. Ideally, as mentioned above, at least one board design would be made similarly available. That doesn't preclude someone from making a proprietary board design and even a proprietary ROM, if they are so inclined, but of course an open sourced ROM would always be my preference.

Garth: in the case of the 8-bit cores, it would make sense to put at least zero page and stack on-chip. It would be easy to provide a configuration pin so that on-chip memory could be enabled or disabled. It would, for example, also be easy to provide a pin which decodes some slice of address space, such as $FExx, to make it slightly simpler to decode addresses for I/O devices, while at the same time providing some on-chip memory at high addresses for boot. In my view, we should be able to converge on how much (minimal) extra hardware(**) to put on-chip, without getting carried away and making a system-on-chip with a spec that doesn't converge. Note that Michael's core presently uses 4 cycles for an off-chip memory access.

Cheers
Ed

(*) pin count permitting, I would much prefer to provide 32bits of address and data at the outset, with the understanding that an 8-bit core uses only a subset of them.

(**) a 6509-style off-CPU but on-chip bank register connected to bank pins or high address pins would be an idea for the case of 8-bit cores. We'd be constrained by the core's ability or inability to signal the type of memory access: code or data.

[enso]

Understood. I don't have good non-BGA footprints in my current setup... Interestingly, 1mm-pitch BGAs should be easier to solder down than non-BGA <1mm. I've soldered .5mm QFPs and it's definitely harder to position them (not that it's that hard with decent magnification).

It's a good point that 3.3V is good enough for new designs; we can't really be pin-compatible anyhow for old designs.

I also would like to point you to ths $27.99 XC3S250E board:

Core3S250E-intro.jpg (60.91 KiB) Viewed 1190 times

and its larger XC3S500E brother $32.99:

Both are on EBay and directly available from Waveshare (a little hard to deal with...)

[BigEd]

Thanks enso, I've posted that in our FPGA dev board thread.

[Arlet]

Suppose if someone got one of those boards with a JTAG connector, do we have any good suggestions for a cheap programming cable ? The original Xilinx USB JTAG adapter is quite expensive. My first programming I did with a parallel port cable that was really cheap, but nowadays parallel printer ports are rare.

[enso]

Arlet - search EBay for 'xilinx usb' - there are many under $50 cables there.

[GARTHWILSON]

How big is the file, and what type? Can the cable and programmer be made easily enough? I program microcontrollers with my workbench computer and it is not dependent on the PC's ports or OS etc..

[ChuckT]

BigEd wrote:

Garth, Chuck:
It sounds positive to me, that Garth (for example) could design a board around an FPGA, publish the design files, and then Chuck (for example) could take that design, and build it either modified or unmodified. In the course of Garth's design process, we would converge on an FPGA netlist and pinout which suits that design: we are hopefully wise enough to make it general enough that modified designs would be possible and attractive (*). If anyone at any stage wanted to build and sell a handful or a truckfull of boards, that would be a good result.

I agree that for confident use, the FPGA design files would all need to be published, with an open enough license that anyone undertaking a design could take a safe copy. All the cores discussed already have an open enough license to allow us to work that way: all the HDL I mention as available is available under redistributable terms. Ideally, as mentioned above, at least one board design would be made similarly available. That doesn't preclude someone from making a proprietary board design and even a proprietary ROM, if they are so inclined, but of course an open sourced ROM would always be my preference.

I would like you to please, please, please consider some of these audio chips for a Single Board Computer using the Spartan or Xilinx FPGA chip with the 6502 Core:

Wolfson ADC WM8782 and a Wolfson DAC WM8523

See:
http://dangerousprototypes.com/2012/11/ ... -breakout/

LA2610 analog surround sound processor
http://www.datasheetcatalog.org/datashe ... LA2610.pdf

I also saw this Gameduino used with the MSP430 Launchpad as a gaming platform and I wouldn't mind interfacing it with an ARM or 6502 board:

http://dangerousprototypes.com/2012/07/ ... amingpack/

I already have several ARM boards, several ARM chips, a Jtag connector and I'm going to order a ST Link after the holidays because it is my goal to build a single board computer. I've already ordered a PS/2 to breadboard connector, an SD card to breadboard component and I have several LCDs to work with. I have been planning and wish to make this a reality. I'm also looking at that VGA board for the 6502 and other chips.

Microchip also has some serial SRAMS:

http://www.microchip.com/wwwproducts/De ... e=en559066

I'm willing to promote the 6502 but these are some of the components I wish to work with so I am constantly working at choosing components and Microchip will also give you information on how to use them if you give them a project idea because they've already given one user information for using it with video.

If you or anyone know of any better or comparible components, please let me hear from you.

[ElEctric_EyE]

ChuckT wrote:

... please, please, please consider some of these audio chips for a Single Board Computer using the Spartan or Xilinx FPGA chip with the 6502 Core:

Wolfson ADC WM8782 and a Wolfson DAC WM8523

See:
http://dangerousprototypes.com/2012/11/ ... -breakout/

LA2610 analog surround sound processor
http://www.datasheetcatalog.org/datashe ... LA2610.pdf

I also saw this Gameduino used with the MSP430 Launchpad as a gaming platform and I wouldn't mind interfacing it with an ARM or 6502 board:

http://dangerousprototypes.com/2012/07/ ... amingpack/

I already have several ARM boards, several ARM chips, a Jtag connector and I'm going to order a ST Link after the holidays because it is my goal to build a single board computer. I've already ordered a PS/2 to breadboard connector, an SD card to breadboard component and I have several LCDs to work with. I have been planning and wish to make this a reality. I'm also looking at that VGA board for the 6502 and other chips.

Microchip also has some serial SRAMS:

http://www.microchip.com/wwwproducts/De ... e=en559066

I'm willing to promote the 6502 but these are some of the components I wish to work with so I am constantly working at choosing components and Microchip will also give you information on how to use them if you give them a project idea because they've already given one user information for using it with video.

If you or anyone know of any better or comparible components, please let me hear from you.

I'm putting this in my DSS thread for my future reference. Thank you Chuck.

[GARTHWILSON]

ChuckT wrote:

I would like you to please, please, please consider some of these audio chips for a Single Board Computer using the Spartan or Xilinx FPGA chip with the 6502 Core:

Wolfson ADC WM8782 and a Wolfson DAC WM8523

See:
http://dangerousprototypes.com/2012/11/ ... -breakout/

These are high-end audio parts, so the number of users who would be interested in having them on an SBC would probably be very limited. They have SPI though, so they could be put on a separate board connected via 65SIB, an interface which is easy to put on the SBC. Seven devices at once can be put on the same 65SIB without adding hubs. It's very hobbyist-friendly too, which was one of the design goals.

You definitely don't need 16-24 bits to get good audio for gaming though. Even 8 bits can give 50dB SNR which is all the high-end cassettes gave on good music tapes not many years back, and 8-bit sampling can give much, much better frequency response and distortion figures than the cassette at high record levels. My separate computer speakers on my desk which are about 7" high are hardly good enough for 8-bit.

[BigEd]

I don't think we're talking about an SBC in this thread: we're trying to find a way to get variant 6502-like CPUs into people's hands. We may need a reference design for something minimal, if only to test that our designs and pinouts are OK, but I don't think it makes sense to pile features onto such a reference design. Enough to boot, show that external RAM works, and (optionally) some minimal peripheral. Better than a peripheral would be an expansion connector. The reason the peripheral is optional is that we could integrate a VIA or ACIA onto the FPGA, if we wished. Even if we don't, we can load a ROM which exercises the CPU without any peripheral: we need nothing more than an 'OK' output pin.

Back to the subject: I'll probably edit the head post, but note that it initially put forward the idea that someone with FPGA expertise could ship a programmed FPGA to someone without, for money. The second idea was to use conventional (volatile) FPGAs and to ship only a programmed EEPROM containing the bitstream. The third idea, by Arlet, was to suppose that programming a serial EEPROM over SPI is easy enough that we only need to produce instructions.

If using SPI is significantly easier than JTAG, we should use that route. We don't want the end-user to have to use any Xilinx tools or any commercial JTAG adaptor: we want the simplest setup assuming only that they have a PC running Windows or Linux, and the ability to breadboard electronics.

There's a USB-SPI adaptor from robotelectronics for about £20 http://www.robot-electronics.co.uk/acat ... B_I2C.html, and hobbytronics do the Bus Pirate for about the same price: http://www.hobbytronics.co.uk/bus-pirate - do have cheaper or simpler ways to program and verify a serial EEPROM? Almost certainly we can bitbang from a parallel port - maybe even use the control signals of a serial port - I think SPI can be run as slow as necessary. We don't expect our end-users to program very many times.

Garth: you asked how big a bitstream file is. It's about 2.6Mbit, apparently. That's 330kByte.

Cheers
Ed