Casio CFX-9850GB Plus 32KB WE (White Edition): 64KByte "upgrade" reconfiguration

[jebem]

I believe this model was released in 2002 by Casio, but can't find any positive authoritative information about this.

I was looking to have this "White Edition" model in my collection since long time ago. It is a extremely popular machine, at least in my country, so there are plenty of used specimens for sale around here, although the asking prices are not that cheap.

Finally got this machine yesterday and it is in pristine working condition.
For a change, there is no need to clean battery electrolytes on this one. Still I wanted to see the internals.

General parts view.




Processor PCA and Keyboard/Display PCA.




Contrast preset potentiometer underneath the Processor PCA.




Processor PCA. Main components:
- Hitachi HCD62121 Processor as per Wikipedia, Martin Poupe, CasioCalc.org, Omnimaga, among others (the reference not visible as it is sealed & hidden away)
- BSI BS62LV1024-70 128KByte SRAM 70nS
- OKI MR27V802D 1MByte PROM 100nS
- MC74HC4066A CMOS Quad Analog Switch
- 4.3MHz xtal with embedded caps set

I can't find the Service Manual for this model, although it is easy to find it but for the previous CFX-9850G PLUS(ZX-935A) but this one uses different memory design. Apparently the Processor would be the same, but so far there is no positive proof.




SRAM reconfiguration to 64KByte.

The CFX-9850GB Plus WE has a SRAM factory set to 32KByte.
However the SRAM chip has 128KByte capacity.

Public information at several sources (for instance, here at CasioCalc.org) states that it is possible to "upgrade" the SRAM size by changing the RAM setup jumpers.

I can confirm that it worked fine on my machine as well:
- Factory set to 32KByte: R20 jumper out, R21 jumper in.
. Enable 64KByte: R20 jumper in, R21 jumper out.

It took me around two minutes to switch the jumper after spending one hour digging in the net.
Other machines may experience issues if the SRAM chip is defective.
Therefore it is essential to run the self test a few times to validate the new SRAM size for good operation.

Basically I have unsoldered the jumper from R21 and soldered at R22 position.
This jumper looks like a SMD resistor but it has a zero Ohm value.




Assembling the calculator.




The Reset push button membrane.



(continue)

[jebem]

The PCA is fixed using a few screws. No snap in / latches / rivets are used here.
Really a pleasure to work with when maintenance is required.




Idle current consumption: 2.5mA from 4 x AAA battery cells (6VDC).




Reconfirming that the new 64KByte SRAM sizing is recognized.
There are 61396 Bytes available to the user after a memory clear.




Running again the self test by pressing F6 + a b/c + AC/ON.
This calculator shows this model code: GY357




Testing the 8Mbit PROM.
ROM version 1.0. Four byte checksum = 0642FEA8




Testing the new 64KByte SRAM configuration.
RAMaddress = F7FF
Write/Read pattern: AA (just a basic h/w test using a 10101010 binary combination).




Color LCD from the 90's by CASIO.
Changing the contrast changes the temperature of each color. To see the Orange, Blue and Green colors will require a specific contrast setting, despite being possible to adjust the three individual colors as well.

[jebem]

Forensics test: 9.00000000733343

Got the same value as reported in the rskey.org site for the Casio CFX-9850GA PLUS and CFX-9850GB PLUS among other models.
However there is no record for this specific CFX-9850GB Plus 32KB WE model.
But considering that the result is the same as for the previous model, it is reasonable to assume that the Processor would be the same.

[Mark]

Hi Jose,
Nice find ! Thank you for this reconfiguration trick ; it made me pull out one of my Casios, a Graph-80 a.k.a. CFX-9990GT. I launched the self-test (didn't know of the procedure) and saw that the RAM test reports a 128kB chip. Since the machine is labeled "64K", it is very likely that such reconfiguration could be done... some day I'll open it to check that :-)

Regards,
Marc

[jebem]

(11-23-2016 07:37 AM)Mark Wrote:  Hi Jose,
it made me pull out one of my Casios, a Graph-80 a.k.a. CFX-9990GT. I launched the self-test (didn't know of the procedure) and saw that the RAM test reports a 128kB chip. Since the machine is labeled "64K", it is very likely that such reconfiguration could be done... some day I'll open it to check that :-)

Thanks for your comments, Marc.

I have here a CFX-9970GT which seems to be identical to your CFX-9990GT is this info here is true.

Interesting!

My CFX-9970GT also shows 128KB in the RAM test.
However, when going to normal operation and select System from the Menu, it tells me that it has just 63443 Bytes free after a clear memory operation.

As the model code ZX936 is different and the test options are also different from the CFX-9850GB Plus series, probably the RAM test is checking here the full RAM size and not the configured RAM value to be available for end users.

Another difference to the CFX-9850GB Plus WE is the ROM size.
This CFX-9970GT shows 4MBit Version 1 + 8MBit Version 2.
I guess one of these ROM chips holds the CAS code.

But both gives the same forensics result, so the processor is probably the same.

Cheers,
jose

Edited to correct a typo (in bold above)

[Mark]

You're welcome, Jose :-)

I was in a hurry when I wrote my first reply ; you're right, the CFX-9970GT seems to be mostly identical from a software point of view. I just did another self-test on my 9990GT :
- The model code is ZX936 (same as yours).
- The RAM test displays "RAMSIZE 128K byte"
- The next 2 lines display "RAM OK (64)" and "RAMaddress FFFF".

I may be wrong, but I suppose that :
- The firmware surely has been designed to manage a 128K RAM.
- If the machine is configured for 64K, the self-test only covers the 64K half it can access.
It will be interesting, shall I eventually succeed in configuring my machine for 128K, to check if the test actually runs through the whole 128K space.

ROM side :
- My machine displays "4Mbit ROM Ver.2.0" and "8MbitROM Ver.2.0"

I get the very same forensics test result, which means that the math routines are the same. I'd also say that the processor core is the same, simply because of the high costs usually associated to porting whatever software from a CPU architecture to a different one (costs mainly induced by the tests).

On the hardware side, I remember that it doesn't look the same as the CFX-9850GB, it is more complex : 6 LSI chips versus 4. I only opened it for a few minutes to check for corrosion, the day I got it for a few bucks. I suppose that the 9850 is a more integrated iteration of the same architecture...

A last note : the "HCD62121" reference puzzles me : In the last 40 years, each Hitachi custom LSI chip I encountered was labeled "HDxxxx" or "HDxxxxx" -but no HCDxxxx. Strange...
EDIT : I read the service manual... it could be the designation for Hitachi bare chips, to be mounted as "chip on board"... ?

Best regards,
Marc

[jebem]

(11-23-2016 04:16 PM)Mark Wrote:  (...) the CFX-9970GT seems to be mostly identical from a software point of view. I just did another self-test on my 9990GT.

It will be interesting, shall I eventually succeed in configuring my machine for 128K, to check if the test actually runs through the whole 128K space.

On the hardware side, I remember that it doesn't look the same as the CFX-9850GB, it is more complex : 6 LSI chips versus 4. I only opened it for a few minutes to check for corrosion, the day I got it for a few bucks. I suppose that the 9850 is a more integrated iteration of the same architecture...

I have opened my CFX-9970G and yes, the mainboard has 6 chips but also have one additional chip in a module inserted in a slot on the other side.

1) Chip in a module inserted in the slot:
8 Mbit PROM module, using a NEC µPD23C8000XGX.
Accordingly to the IC manufacturer, it can be configure as 1M 8-bit words, or 512K 16-bit words.

This module have configuration jumpers (zero Ohm resistors).
Default setup:
R402 - Open
R403 - Closed
R404 - Open
R405 - Closed

Probably it can be possible to configure the operating system firmware to enable above the factory default of 64KByte of SRAM, by changing the above setup. More information needed and/or additional time to do testing.

2. Chips in the mainboard:
- LSI101: unidentified IC
- LSI102: NEC µPD23C4001E 4Mbit with 512K 8-bit words PROM.
- LSI103: SANYO LC3564SM 64KByte SRAM.
- LSI104: unidentified IC.
- LSI105: CASIO F3660-ZD - Can't find information
- LSI106: EPSON SRM20V100LLMX7 128KByte SRAM.

The PROM chips are correctly listed in the sef test.
Interestingly, there are two SRAM chips, one of 64KByte and another with 128KByte.

I don't have the service guide and do not time now to investigate as I have a business trip coming and need to prepare my work.
If you have it, p,ease let me know.

The photos are published at my blog here, as well as the links to the chips datasheets and catalogs.

Thanks.
Cheers,
Jose

[Mark]

Interesting, indeed !

You've done a great job Jose, what a lot of photos... I didn't remember the additional ROM module, neither did I about the 3 crystals and resonators on the board, X101 for LSI101 and X102/X103 for LSI104. By the way, X103 has the cylindrical form factor often associated with real-time clocks.

The architecture looks quite complex for such a calculator ; of course I may be wrong, but it clearly seems that we've got here a 2-processor machine (LSI101 : main + LSI104 : I/O ?), each with its own RAM + ROM. The Casio-labeled chip F3660-ZD / LSI105 probably acts as a communication bridge between both CPUs.
A service manual (which I still don't find) would allow us to be sure about what is very hypothetical...

Cheers,
Marc