Hi, I don't know much about this sort of thing, but I have been looking for a system that runs on a die that is of a larger process like the 65c816 (0.6um) which is for long long long term stability purposes.
I wonder, I have heard that using it's full 16mb potential one would have to use bank switching and stuff, would all of this be handeld by the c compiler?
Also, would this mean that more chips than just the 816 and ram chips are required to run the full 16mb?
I have tried to do research but honestly being just introduced to this most of it went over my head.
Thanks

The '816 can be a bit confusing when first approached. Here's my current understanding of its characteristics in native mode (which is strongly recommended):

1: From a hardware perspective, the extra 8 bits of address space (relative to the 6502 family proper) are handled, without increasing the pin count of the device, by multiplexing them onto the data bus lines. They are commonly referred to as the "bank address". A small amount of external hardware (a transparent 8-bit latch and a small number of 74-series gates) is needed to capture and forward them to address decoders and memory devices during each clock cycle. A distinction between program and data accesses can also be detected (VPA pin), and optionally treated as a 25th address bit - but this is most useful if you implement virtual memory paging, as otherwise it would be hard to load a program into memory.

2: There are essentially four address spaces on the '816: Program, Direct, Absolute, and Long. The latter gives you a completely flat 24-bit address space, whilst the others can be treated as shorthand and/or performance optimisations within that space.

3: Program space prepends a Program Bank number to the nominally 16-bit PC for all instruction fetches. The Program Bank number is changed by JML, JSL, RTL, or by taking or returning from an interrupt; it is not incremented when the PC wraps around from $FFFF to $0000. So you must arrange your code in 64KB modules, but it's easy to switch between them. Interrupt handlers (or at least their entry points) must always be in Bank 0.

4: Direct addresses, and their close cousin Stack-Relative accesses, are 8-bit offsets from a 16-bit pointer register, and are always in Bank 0 (even if the offset and/or any indexing causes the 16-bit sum to overflow). This area is primarily used as a sort of extended register bank to compensate for the small number of registers internal to the CPU, and is relatively fast. The Direct Page Register can easily be pushed on the stack or swapped for another copy, so different routines can be given independent Direct address spaces to work with.

5: Absolute addresses are 16-bit and have the Data Bank Register prepended to them. This can be used as a shorthand (relative to Long addressing) for accessing larger (up to 64KB) chunks of data than will comfortably fit in Direct Page.

6: It must be understood that, in general, address 0 may refer to four completely different locations in memory, depending on whether you use Direct, Absolute, Long, or Program addressing modes - and even a fifth if you count Stack-Relative. Unfortunately, the established assembler syntax doesn't make it easy to unambiguously distinguish between them (as the distinction doesn't exist on the 6502 itself). But this is something a good compiler will take care of for you - though I make no representations as to whether any particular compiler is a good one.

Welcome, nodelink. I have struggled to formulate a reply to your query: there may be a lot you don't know and your decision about your project may be a difficult one especially if your priorities are not clear.

If reliability is paramount, you should know that ICs are extremely reliable once they have been burnt-in, which is usually done in production. You should then be concerned about the reliability of your system construction, of your power supply, of other components such as capacitors, and of course the reliability of your software. So, choosing an '816 primarily because it's made on an older process may not be a great justification.

By all means choose the '816 if you like it, but as noted by chromatix, it's not a simple machine with a simple programming model. A PIC32 might be a better choice, or an FPGA. Everything depends on your real priorities: time, money, performance, support, tooling.

The Space Shuttle had multiple computers and a means of comparing their outputs. This can be very reliable.

The Clock of the Long Now uses a mechanical calculator with great attention to materials and construction. They hope to run for ten thousand years.

The solution space is very large!

Welcome.

The data sheet shows the basics of decoding the bank byte and multiplexing the data bus. We've had additional good discussion about it in the topic "Managing the 65816 multiplexed bus"

The 6502 is definitely one of the best-documented processors in history. It was not very friendly to C compilers though, and the common cc65 C compiler turns out extremely inefficient code. Word on the street seems to be that WDC's C compiler for the 65816 is somewhat better, although the '816 was still not designed from the ground up to be particularly compiler-friendly either. Still, the very things that seem to make the 6502 a poorly suited target for a C compiler also make it easy to envision solutions in assembly; and my experience is that the '816 is even easier to program than the '02 not just because of the extra instructions and addressing modes that give you more freedom, but also because when you're constantly dealing with 16-bit quantities, you don't have to keep taking them apart and handling them 8 bits at a time to "get them through the door" so to speak.

I cannot pass up the opportunity to recommend the "Programming the 65816—Including the 6502, 65C02 and 65802" 6502/65816 programmer's manual by David Eyes and Ron Lichty. This is definitely the best 65xx programming manual available, and a must-have for every 65xx programmer! It starts with the basics, followed by architecture, the CMOS 65c02's many improvements over the original NMOS 6502 including added instructions and addressing modes and fixing the NMOS's bugs and quirks, and then the natural progression to the 65816; a thorough tutorial, writing applications, then very detailed and diagrammed information on all 34 addressing modes, at least a page of very detailed description for each instruction, with info on every addressing mode available for that instruction, then instruction lists, tables, and groups, of all 255 active op codes, plus more. 469 pages. From Western Design Center. (.pdf) Note: There were many problems with the earlier .pdf version that were not in the original paper manual; but in late March 2015, WDC scanned and OCR'ed the paper manual and posted the new, repaired .pdf.

Some of the other things you allude to like address decoding are discussed in the 6502 primer which is about building your own 6502 computer. Although it's geared to the 6502, many of the sections can be applied much more broadly. It was written to address questions and problems that kept coming up on the forum. It has 22 sections arranged in a logical order.

Regarding stability, Bill Mensch, the owner of Western Design Center (WDC) which owns the 65c02 and '816 intellectual property, intends to make these processors available indefinitely.

_________________
http://WilsonMinesCo.com/ lots of 6502 resources
The "second front page" is http://wilsonminesco.com/links.html .
What's an additional VIA among friends, anyhow?

Thanks to all of you for your incredibly nice and in depth responses . I apologize for my unclear initial post, I shouldn't have mentioned any specifics (Like banking) as that just shows how confused I was and might have confused some readers.

Whilst your explanation did clarify this, seeing the scope of this, I believe I must find a different platform.
The 65 series is something I could sink a lot of time into because they are beautiful pieces of hardware and technology, but I think progressing my project is paramount.

As for the reason for creating this post in the first place, if any of you knows a similar component, that has easier memory interfacing I would love to hear about it!!!
I have looked into the Zilog eZ80, but that thing consumes a ton of power according to the data sheet.
I also looked at the 68000 but the versions that do still exist of it aren't really going to be produced much longer and it also seems to consume much more power than for example the 816.
Modern Microcontrollers often totally fit my bill, but they have no memory interface so that makes them worthless to this project

Some PIC32's are perfect, but I have found that the PIC32mx which seems like the best choice (Because of it's memory interface, I need at least 16MB of RAM, not just flash) however it's process is smaller than other PIC's and it's reliability and fit data as supplied by Microchip is correspondingly worse than the rest.

See I know that most components on a board will last thousands of times shorter than an integrated circuit, but I don't have any of those components. This project aims to create extreme long time reliability far beyond that of most reliability goals.

I love the 6502 especially, that thing is amazing. I don't know too much about the design specifics of course but it just has this quality I don't know how to describe it. Perhaps it arises from my perceptions of reliability, low power consumption, history etc/
Too bad it can only address 64K
Thanks to all!

I don't know all the details, but the F4 ARM-based microcontrollers from ST have loads of I/Os with many modes (as is normal), and one of those modes does dedicate a number of pins to an external memory interface ("FSMC external memory controller"). So, it can be done, and I imagine their later offerings have similar facilities. I like the ARM's simplicity, and of course it's really well supported. At assembly level it's not too different from a 6502.

Therein lies my dilemma, those chips are great but they are processed on tiny nodes, often 90nm or smaller.
I know that my obsession with feature size may seem irrational and it probably is but it is just one of those quirks I am having to work around.
Luckily we are not talking about a commercial product here, however we are talking about something of scale so a few hundred chips will be needed or more.
Thanks!

_________________
http://WilsonMinesCo.com/ lots of 6502 resources
The "second front page" is http://wilsonminesco.com/links.html .
What's an additional VIA among friends, anyhow?

This is a really neat option if you decide to go this route. I got a development board a few years ago for and it was a great investment. You get a big LCD, 168MHz M4 ARM with 2MB built in flash and USB, and 8MB DRAM.

Do you mean there is data that shows PIC32s are lower reliability? Can you post a link?

I don't really understand why you are focusing on process size but if you really want to do this you might consider some obsolete models apart from the 68000 that were produced in the 80s and 90s. Some of those might have been made on a larger process. I don't know which if any of these meet your requirements but I believe they all have external buses and 24 or 32 bit addressing. It would be a place to start looking at least:

- 32016 (including low power microcontroller versions with integrated clock like NS32CG160)
- Intel i960
- 80251
- XA-S3
- H8/500
- H8S/300
- TMS320 (some models it seems)

I should also mention that your life might become pretty unpleasant if you decide to do a big project with one of these since there is probably little to no community support and software is probably antiquated. They all come in PLCC or DIP, which is what I was looking for when I found these. You might find more options if you expand your search to SMD parts.

If you want to buy one of these chips, you can try eBay or the CPU Shack Trade List, which I have bought from (very nice guy and shipped right away.)