Call for final Elf99 design input

From: Ethan Dicks <erd_at_infinet.com>
Date: Fri Dec 4 12:05:08 1998

 
> The idea is to map a part of the EPROM permanently to $0000-$001F
> as replacemet for the original prom, or to suply a 32x8 RAM (like the
> 1824, since we want to use as much 'original' Harris stuff as possible

Using as much Harris stuff as possible is not one of my design goals.
Simplicity is. At the moment, the only Harris part may be the 1802
mostly due to cost. The Harris stuff is between $5 and $10 per chip.

> Yep, but AFAIR I never taked about the starting address - my fault.
> I assumed a varable starting adress.

That was the gist of my questions. It appears that we were using terms that
the other was not fully understanding. The term "memory space" that I used,
refers to a region of memory that is designed to contain something, but may
or may not be full. For example, a 4K PET has 32K of RAM address space, but
4K of RAM. It's trivial to add another 28K of RAM, not so trivial to add
RAM outside of that 32K address space. A 68000 smart serial card I used to
make had the 16Mb divided into 4 address spaces: RAM, I/O, shared memory (it
was a DMA card on a PDP-11) and ROM. Even though no card ever had more than
2Mb of RAM (the oldest card had 32K of 2114!), all of the later cards had 4Mb
of the memory map blocked out for RAM.

It is easy to divide the memory space into 1/2 RAM and 1/2 ROM. It's harder
to divide it into 3/4 RAM and 1/4 ROM. It's even harder to divide it into
3/4 RAM, 1/8 ROM and 1/8 memory-mapped I/O, and it's hardest of all to make
the division user configurable. A 48K Elf might be cool, but none of the
ancient software knows much more than about 4K, leaving plenty of the
intended 32K left for new stuff. Besides, I want to stick with one RAM chip
and one ROM chip, for simplicity.


> Ahh ok. Hmm so if we want to use the 'switch-boot' a RAM based function
> can never be used ? or do you only map the first 256 (or less) bytes of the
> EPROM to $0000 ?

No... the entire ROM maps at $0000 and $8000 until the first time that A15
gets set, then the flip-flip clears and RAM banks in at $0000 and ROM remains
at $8000. This is easy to implement and fairly clean. No I/O address needs
to be hung up as in the VIP design (62).

> Adding two sockets wouldnt be a big thing I guess, but one could build
> a system with only the 256 Bytes of storage :)

Technolgically speaking, it's trivial. The issue is real estate. If I have
to make the board bigger, the cost goes up. I'm thinking about designing
and releasing a design for a piggyback board to plug into the 6116/62256
socket to adapt a pair of 1822/2101's. I wouldn't *make* that board, but
anyone who wanted to use 256x4 RAMs should be able to make their own.

> > The space taken up by two 40 pin chips is a bit much when I can implement
> > two cheap ports (one in and one out) with octal latches.
>
> You're right.

Now... OTOH, Allison Parent's mentioning of the COSMAC eval kit has pushed me
back to considering dual 1852's. It's still the most expensive option, but
software compatibility for any sort of existing software base is nothing to
be lightly dismissed.

> What about adding a latched solution for all parallel ports like mentioned
> in one of the App notes (I think it was with one of the latches).

If I understand what you are suggestion, Allison suggested that, too. You mean
decoding additional secondary ports from a primary port, yes? The reasons
not to do it are a) increased complexity of design and b) increased complexity
of programming. I have not utterly dismissed it, but neither am I
overwhelmingly convinced that it's necessary.
 
> > > > o A 1854 UART
>
> > > Data transfer via Q is way more fun (Hi Alison :).
>
> Serious, this was inteded as a joke - Even with software controll
> 2400 Bd should be no problem at all - and for the LCD, I would realy
> like to use one of these ASCII input LCDs

They cost more. I can get 4/8-bit PIO displays 16x1 for $3 each. I have
an ASCII LCD - PIC-an-LCD from B.G. Micro. The board was $12, the PIC
controller was $12. I like it a lot. It sits on my Linux Internet router.
Sticking with an easy to implement LCD connector means that the user can
hang a *cheap* display right off the Elf.

> Anyway, if we use your 32K RAM plus 32K ROM design, I thougt
> about adding a kind of CS unit to provide several CS signals for adding
> various memory or memory maped areas. For example if we supply
> 8 CS lines for 8 8K areas, one could bild a system like:
> 2K EPROM at $8000 (needed to perform the 'boot') and
> 256 Bytes RAM at $2000 and a memory maped
> 1851 at $4000 ... or some other configurations.

Ack! Too complicated. I'm after *simple* here. While your concept
here is very powerful,

> My view of the Elf99 is more fun and hobby than real applications.
> Playing with unusual components and using 'real' 18xx stuff is
> part of this flair. A 4xxx based I/O is just less 'sexy' than a 185x.

It is and it isn't, depending on the nature of the playing. After all,
you can't stick a logic probe inside a 185x chip. You *can* watch a 4308
latch the data and observe how a parallel port works, if you care to.
Besides, the 1852 is nearly 10 times the cost of the 4308.

> Its like with some of our vintage equipment - connecting a Terminal
> to an Altair gives a diferent feeling than to any 12 MHz 64180 system,
> even when running the same CP/M and the same Wordstar...

Yes, but even a real RCA product, the VIP, used 4308 latches for PIO.
 
> > I still have yet to hear why the 1855 Multiply/Divide Unit is worth
> > the real estate. Sure, it's a neat chip, but unless it has a purpose,
> > I can't see including it.
>
> Gee the purpose is to multiply and divide - what else :-)

Well.. Duh. I meant, why go the trouble of a bank of hardware integer
multiply units with no practical use in mind beforehand. For example,
if I were building a PIC project board, I'd contemplate ways to attach
a thermistor to play with computerized thermometer projects. I am at a
loss to suggest to the user what they could do with a bank of 1855's to
justify 1/6th of the space on the board.

> If you just think about the real world usability of an Elf99,
> we shold stop it and go for a 80188 design of an equal
> level (I like the 8018x CPUs)-

But the Intel architecture isn't sexy like the 1802. You can't build
a ROMless 8018x system with 256 bytes of RAM. As complex as the Elf99
is turning out, I know it can be partially populated to produce a very
good approximation of the original Quest design (leave off the address
latch and all I/O except the Q LED). Someone wishing to build the
whole design can stuff every crevice with parts and have an Elf beyond
our wildest dreams of 1977.
 
> And for both chips (MDU and RTC), I don't want them to
> be esential - just optional. And I think they offer a lot
> of possible usages (what about a Elf Basic using the MDU
> for speed ?).

Optional isn't the issue. The cost of the board space is. I've already
planned on lots of optional stuff (PIO, UART, more than 256 bytes of RAM).
I'm not opposed to optional things. I am just trying to give as much
expansion as $99 will allow. If the board costs more, it loses its
audience.

> (BTW: Just look at the 'front panel' thread - a MDU would offer
> a lot of new and excitieing lights :)

Too complicated for a $99 design.

> USD 2 for a 16x2 (or USD 7 for a 20x4) including controller ?

If by controller you mean a Hitachi 48440 or similar chip that sits between
the raw LCD and provides the user with a font and parallel (4 or 8 bit)
interface, then yes. $2 to $9 is a typical range, depending on new or
used or backlight or not.

> Thats pretty cheap. If this is true, a 20x4 should be included
> for shure. A LCD without controller is not usefull, since a permanent
> scann to support output counters the basic static Design.

I've got a 20x4 on my Linux box (as I said). I chose to spend $24 extra
(I scrounged the power supply and box) to keep the programming from a UNIX
machine simple. There is no such requirement on a toy board.

> The apeal lies in having an Elf like computer, but with all the fancy
> stuff to draw fun from programming and playing. Thats one main
> reason why I suggestet the library and the library boot

No argument about the library boot. I spent lots of time entering in
games from the VIP manual (I still have the tape!)

> So my idea of the Elf99 is an easy to use experimentation / training
> kit with a lot of interesting features - NOT any kind of very usefull high
> powered controller. Something to learn about a microcomputer for
> children...

So far we are in agreement. I don't expect anyone to buy this and build
it into an Internet Coke machine or anything.

> ..and the MDU is just a part of this idea

In the abstract, I agree, but I still see no practical use for the MDU except
as an expensive, flashy toy that would only appeal to a few people (less than
5%). I'm shooting for 80% or 90% audience acceptance for a feature. So
far, that's 8 bits in, 8 bits out and some sort of serial port (plus the
required switches, displays and Q).

> In fact, if the design is a in this direction I plan to use two of them
> for children in the age of 11 and 14.

They would be a good audience. I got my hands on a Quest Elf when I was 11.
It nand the PET profoundly affected my life and career path.

-ethan
Received on Fri Dec 04 1998 - 12:05:08 GMT