Hello,
Last year, thanks to Garth Wilson's site and its great amount of information, I successfully built my first 65C02 single-board computer which is to become the heart of a scientific and programmable desktop RPN calculator.
It's a very basic SBC, inspired from the one found on top of the "circuit potpourri" page : 65C02 + 65C22 + 65C51 + 27C256 (only one switch-selectable 16K half used) and a 32Kx8 CMOS SRAM.
You can see it on the picture, with its LCD and keyboard interfaces (I put some piece of foam on the EPROM to avoid the running program to crash when taking the photo).
The decoupling capacitors are on the solder side, directly wired to each IC. I used W-W wire and soldering -not the best method, I admit. I succesfully ran it up to 2.6 MHz, but kept a conservative 1-MHz clock rate. Lee Davison's EhBASIC ran without trouble for more than 3 months 24/24, computing decimals of e. Then, I started developing my calculator software around C.R. Bond's floating-point package. Software is finished by now, all I need to do is a custom keyboard + cabinet. (by now I'm using 2 4x5 Grayhill keypads)

Now to my question : as soon as I power it off, all data in RAM is lost. I perfectly know I could use a "zeropower" SRAM to get it continuous, but I wonder if there's an elegant way to use a 3V lithium cell and a bit of electronics ? At first I thought about using Schottky diodes in the RAM VDD line, but that would result in incoming signals getting higher than the power supply -bad !
So, any suggestion would be much welcome ; I add that the MAX 1232 you can see below the 1-MHz can oscillator next to the CPU, has voltage-monitoring features which I still don't use.
(the logic behind all that : it's much easier and cheaper to replace a dead coin-cell than a whole zeropower RAM)
In advance thank you very much for reading and for your help !
Regards,
Marc

You might want to have a look at this: http://www.digikey.com/product-detail/en/M48Z35Y-70PC1/497-2885-5-ND/606308

32Kx8 non-volatile SRAM with 70 nS access time. A little pricey, but worth it IMHO if you want to keep things simple.

_________________
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?

Hi jmp(FFFA),
Thanks -but maybe you noticed that I'd like to use the other method : powering my RAM with some battery... not as simple, indeed !
Thanks anyway !
Marc

Hi Garth,

Thank you very much this really is a nice trick ! (and well in the 80's spirit).
I'll soon give it a try !

Regards,
Marc

How about using a supercap?

_________________
Andrew Jacobs
6502 & PIC Stuff - http://www.obelisk.me.uk/
Cross-Platform 6502/65C02/65816 Macro Assembler - http://www.obelisk.me.uk/dev65/
Open Source Projects - https://github.com/andrew-jacobs

To BitWise :

Thanks I had that solution in my mind, but if I'm not wrong, it brings us back to the question : "How to cleanly switch supply sources at power down / up ?".
BTW, I had also found another implementation, in the DX-7 Yamaha synthesiser ; please see the first link in http://sci.tech-archive.net/Archive/sci ... 00108.html

Regards,
Marc

My elf membership card just seems to have a 0.1F supercap between VDD and GND with 5.1V Zenor. See here - D11 and C5 beside the 1802 processor on the schematic.

http://www.retrotechnology.com/memship/mem_revh2sch1.pdf

_________________
Andrew Jacobs
6502 & PIC Stuff - http://www.obelisk.me.uk/
Cross-Platform 6502/65C02/65816 Macro Assembler - http://www.obelisk.me.uk/dev65/
Open Source Projects - https://github.com/andrew-jacobs

OK Andrew, understood ! I have to apologize for not having mentioned that my SBC isn't fully CMOS : the 1.8432 MHz oscillator clocking the 65C51 eats up to 22 mA, and there's also the MAX232 on the I/O side. I still need to find a CMOS-based oscillator, and no doubt that there's a better solution for RS-232 level translation... When I began to build this board, one of my goals was to use 1980's technology only
Marc

What makes you think the oscillator isn't CMOS? A draw of 22 mA sounds about right for one of those things.

As for the MAX232 type transceivers, their inputs are TTL-compatible, which will work fine being driven by a CMOS UART, since the latter's outputs swing from near zero to near Vcc. The MAX232's outputs swing from about 0.2 volts to Vcc - 0.2 volts, which is compatible with CMOS inputs.


If you can find a better way of interfacing TIA-232 to your system please let us know about it. The MAX232 family has been around since 1987 and no one has really come up with a more convenient solution. In fact, Maxim makes these devices with up to eight receivers and eight transmitters (the MAX248). There are multiple second sources for these devices.

_________________
x86?  We ain't got no x86.  We don't NEED no stinking x86!

Hello BDD,

What makes me say that the 1.8432 MHz oscillator isn't CMOS is that when I tested it and the 1-MHz (CPU) upon receiving them, the latter draws less than 5 mA... ! I can easily redo the tests/check my notes to give you a more precise figure, but I remember having been surprised by such a difference -hence the "CMOS or not CMOS?" question.

My concern about the MAX232 wasn't related to the logic levels which I know are CMOS-compliant, I wondered about the existence of a newer generation (from the power consumption point of view) circuit -which as you wisely tell, is still to come...

Regards,
Marc

_________________
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?

One of the future generations of POC will be designed so the 20 pin connector of an ATX power supply can be directly plugged into it, which means that ±12 volts will be available. I could use the 14C88 and 14C89 on that unit, since the necessary voltages would be available to operate them.

However, as the number of TIA-232 channels increases, the disadvantages of the MAX2xx series disappear. POC V1 was designed from the beginning to use the MAX238, which is two MAX232s in one 24 pin package (POC V1.1 uses the SOIC version, which consumes little space), hence the functional equivalent of a 14C88 and 14C89 in a 24 pin package. The 14C88 and 14C89 combination would have used much more PCB real estate, even after accounting for the charge pump caps. The MAX238 has been available since 1990, which is when I first used it in a project, a UART board that connected to the cartridge port on a Commodore 128.

The MAX248 that I am using in POC V2 encapsulates eight transmitters and eight receivers in a single PLCC44 package, that being enough to handle four TIA-232 channels with CTS/RTS hardware handshaking. Only one set of charge pump caps is required for all four channels. I would need two each of the 14C88 and 14C89 to get the same functionality—as well as the need for ±12 volts.

_________________
x86?  We ain't got no x86.  We don't NEED no stinking x86!

Hi BDD,

I've been away for a few days but didn't forget about that "CMOS or not" question, speaking of oscillators. Well, I've got an answer -not THE answer of course. I just received a few 1.8432 MHz oscillators I ordered from Conrad (still waiting for the 2-MHz ones I ordered from Mouser), and was wondering what their brand would be since the Conrad site doesn't give the information.

They're from Qantek, I tested them and measured a supply current ranging from 5 to 7 mA depending on the sample. It's well within the specifications (which says that their output logic is HCMOS).

On the other side the ocscillator I initially plugged on my board is an old one from Kony ; I remember having saved it from a board crowded with TTL chips, dating back from 1988.

So by now my little beast runs 84% faster, and the supply current increase induced by the higher frequency is balanced by the less power-hungry oscillator feeding the 65C51 -a nice trade !

BTW : the Qantek oscillator's datasheet : http://www.qantek.com/tl_files/products ... rs/QX8.pdf

The output symmetry is not as good as some of the other manufacturers' products of the same type. Output symmetry becomes important as you approach the maximum Ø2 rate supported by the MPU—timing violations can occur if one of the clock phases is too short. You can easily fix symmetry by running the oscillator's output through a flop, such as a 74AC74.

_________________
x86?  We ain't got no x86.  We don't NEED no stinking x86!