Novice looking for pointers - design simple cpu

[123jack]

I'm interested in designing a microcontroler to my own specification but I have no idea
what technologies I should be researching to do this. Do I want asic/FPGA or something else?
What design software could I start with etc.

Could someone point me at the correct areas to look at and maybe suggest tools
and/or info that will lead me to it.

Obviously I'm competent with programming/low level work etc and electronics.
I'm looking for info on which technologies and design tools I could look at for this
type of design work. I just don't know the best place to start.

If it helps I'm funding this personally so low cost initially is beneficial - I will be
looking to develop a very simple CPU model with additional on board modules.
(Something like a microchip PIC but to my own specs)

Any suggestions gratefully recieved.

Thanks
jack

 

[ads-ee]

Low cost? Then you've posted to the wrong section. Creating even a simple CPU design in an ASIC flow is NOT going to be cheap.

You should probably do this design in an FPGA either Altera or Xilinx. If Altera still has hardcopy you could have your design converted to a full metal mask (pay $10's of Ks to do it).

I would probably go with an Altera Cyclone V or Xilinx Artix and/or Spartan 6 device and use the free tool suites both vendors have. Also there are evaluation boards for all three and they vary between $100-$300 depending on features and size of the FPGA.

 

[c_mitra]

It depends. The good old intel 8040 is a good place to start to look for specifications. I hope it is still available.

 

[123jack]

Thanks to both for the suggestions - I'll look into them.

ads-ee
Reading between your lines would it be correct to think FPGA would be a good prototype development
which would be made more efficient by then converting to ASIC - or are we talking something
totally different? Would the design be transferable at all.

Would one technology be faster (runtime operation) than the other?
High speed for simple logic operations will be important in the final product.

The eval boards sound good -

If anyone has any more to say please don't hold back.

 

[ads-ee]

Yes use an FPGA for prototype development, but if you plan on making an ASIC I hope you have a few $100,0000 laying around to pay the ASIC vendor to make masks for you and even more if you use a more recent technology node like anything under 1um. You'll probably need millions to run on Intel's 14nm fab ;-)

Just code everything in VHDL/Verilog using behavioral code and you will be fine. You might have to port things like RAM over to the ASIC vendors RAM and the I/O cells too.

FPGAs are fast but an ASIC will be much faster, but due to the NRE costs unless you plan on making 10's of thousands of parts and selling upwards of 100's of thousands of the ASIC it might not be cost effective. Also you ask for high speed but don't clarify what you consider high speed...what 300 MHz?, 400 MHz? (still within FPGA capabilities), 500 MHz, 800 MHz, 1 GHz (way outside FPGA capabilities).

 

[123jack]

Thanks ads-ee

Your speed examples give me something to work work with - by speed I'm thinking
high speed chip to chip comms. I'm planning a 3 wire system - something akin to the
old transputer designs if you're familiar with those.
As far as costs go - I think I have a handle on that now - thanks. Ball park is all I
need to start. (I really had no idea)

I've just been reading the altera website you suggested and have just got the links for their
e-books. I'll give them a read and maybe book myself on a course or two.
I'll check out the other links to compare later.

Looking at their fpga chips (very quickly) they all seem massively over-spec for what I need.
(Thinking I need a 6502 cpu when all I can get is an Intel quad core to give an idea of what I mean.)
That being said I see no reason I can't start like this - the pricing is much lower than I had imagined.
I'd better check out VHDL/verilog then - just 2 more languages to confuse my already addled brain

Thanks again.

 

[ads-ee]

As you still haven't defined how fast is fast...

You could use a slow Microsemi/Actel part, (they really are slow). Don't expect to run at even 100 MHz unless you pipeline like crazy.

Yes even an Artix-7 15 part is still 20,000+ FFs.
A Spartan-6 XC6SLX4 has only 4,800 FFs, which is still huge for a 6502 like CPU (which by the way can't be implemented directly in an FPGA without lots of modification as it used latches. Here is an interesting paper on someone implementing a **broken link removed** on a Cyclone II.

Most of the parts (from Xilinx/Altera) are huge now days as they are on things like 22 nm (or below) process nodes. You would have to go to a much older part (which cost even more than the current parts in some cases) to find a small Xilinx/Altera FPGA.