Some more pictures.

Case picture stolen online.

I totally forgot to write about the keyboard.

This seems to be quite rare case with red and clicky Alps. It's rebranded Alps Bigfoot with red switches. Personally I like the feel more than IBM Ms. Anyway, that has been documented now here...

Thread @Deskthority https://deskthority.net/viewtopic.php?t=20339

fool wrote on 2023-01-03, 13:34:

I totally forgot to write about the keyboard.

This seems to be quite rare case with red and clicky Alps. It's rebranded Alps Bigfoot with red switches. Personally I like the feel more than IBM Ms. Anyway, that has been documented now here...

Thread @Deskthority https://deskthority.net/viewtopic.php?t=20339

Ok what can I do to separate you from this keyboard?

fool wrote on 2023-01-02, 19:34:

I'll build new 4+4MB memory modules if I can get this thing running.

Build two modules for me too, my T5200 really needs more memory 😀

And that number those cache chips means nanoseconds, so 20 nanoseconds is faster than 25.

Tried to overclock my T5200 from 20MHz to 25MHz (by changing oscillator from 40MHz to 50MHz) and it didn't even give any picture.
But maybe this works at 25MHz when there's 25MHz model of the whole computer.

EDIT: kappas, täällähän on tuttuja 😉

dj_pirtu wrote on 2023-01-04, 11:07:

And that number those cache chips means nanoseconds, so 20 nanoseconds is faster than 25.

Yes that's usually the case. These "Cache controllers" are Toshiba's own or rebranded chip and no datasheet. It' just strange why this 20MHz machine has -20, but only -25 ones are available online. Where the hell did they use even slower -25 if -20 is used in slowest Model 20? That's why I would like to see inside of Model 25, but has been desperate attempt so far.

One step further. Memory IC order to be postponed until CNY ends.

fool wrote on 2023-01-12, 16:25:

One step further. Memory IC order to be postponed until CNY ends.

Are you sure you have it configured for parity for these PCB made? I didn't see any spots for parity bits. T8500 takes parity SIMMs.

Cheers,

Thanks for asking. Yes, there's parity bits (4) included. Toshiba surely didn't make this RAM configuration easy. Maybe one of the reasons why it never took off.

How expensive those PCBs and yet to come memory chips were? Anyway, if you have lots of those I'll buy two sticks from you for my T5200.

dj_pirtu wrote on 2023-01-13, 07:29:

How expensive those PCBs and yet to come memory chips were? Anyway, if you have lots of those I'll buy two sticks from you for my T5200.

Memory chips are so expensive that I will only produce minimum amount to verify the design. If these are functional we can discuss more via email. But there is many variables to deal before that step.

Same memory chips that are used in old PCI display cards and Gravis Ultrasound MAX? 512KB SOJ?

I have those maybe 10 pcs so I can solder my own sticks if I can get those PCBs somewhere.

EDIT: forgot the parity...

Pinout, size and memory organization are sure different. There aren't much compatible ones in this world. But don't worry. I just checked today and my supplier has some stock left.

I have another design with more common cheaper implementation but thats only 1MB/stick. I can order PCBs if there is some kind of need...

Jumper on this 4MB stick allows to cut down the size to 1MB if mandatory due to weird memory layout in Toshiba. Haven't checked it really, it was easy to implement just in case.

Six 1MB sticks should be enough for everyone.

dj_pirtu wrote on 2023-01-13, 17:03:

Six 1MB sticks should be enough for everyone.

Wise man once said: 640k is enough for everyone 😁
It puzzles me how max memory is organized as 2MB+8MB+4MB banks. I want to see how and in what size 8MB kit appear in those 2MB and 4MB banks.

fool wrote on 2023-01-13, 04:30:

Thanks for asking. Yes, there's parity bits (4) included. Toshiba surely didn't make this RAM configuration easy. Maybe one of the reasons why it never took off.

What I mean do you have place for four 1mb x 1bit IC for solder pads for parity? This way these is easier to get chips than the single IC with 4 cas memory which is impossible to find other than salvaging 3 chip parity modules that has the 4 cas IC. Or buy 72pin modules with 3 chip and transfer these chips to your 40 pin simms.

Also you cannot use single 1mb x 4 bit as parity as parity substitute and memory design requires 4 cas lines. Just look at the pinout of 72 pin simm with parity included. Same idea.

Both 4 x 30 pin module and single 72 pin modules are same. Ditto to these 40 pin simms, since all these three I mentioned are equivalent to "4 bank of 8 bits with one bit parity array".

Also there is no room to place chips on back side of the 40 pin simm module due to simm slots so close together.

Cheers,

What I mean do you have place for four 1mb x 1bit IC for solder pads for parity? This way these is easier to get chips than the single IC with 4 cas memory which is impossible to find other than salvaging 3 chip parity modules that has the 4 cas IC. Or buy 72pin modules with 3 chip and transfer these chips to your 40 pin simms.

No, I didn't need place for parity because the chip includes also parity bits in the array.

When I made my calculations this 2-chip design was actually cheaper than using 4-bit data + 1-bit parity chips. And even more important was the ease of soldering process and availability from my supplier. I would like to add that I don't like to "destroy" old memory modules, especially not FPM. I thought about that option as well, but it would have caused too much extra work. Some very generic 2MB EDO chips I have salvaged for AWE64 memory upgrades, but thats about it.

The main reason selecting that particular chip is the lack of room in this 40-pin design. Thats why they implemented that narrow socket for parity module. Original Toshiba 8x 256kx4 + 4x 256kx1 (1MB) designs are very tight and almost impossible to hand solder. Even with reflow some isolation like kapton tape possibly needed.

So it's all about optimization in different areas. But what the heck, I don't even know yet if this even works.

Is those 40 pin simms made up of 16 bits plus parity plus address lines? If it was actually 32bits plus parity per simm, that would take 36 bits total and 4 pins for power. No room for addressing?! Lol.

Because, these 40pin simm contains 8 4 bit wide 256k chips which is 32bit wide which is impossible, since I'm confused unless these chips is tied together as 16 bits with total of 512K split into 16 bit plus 2 bits of parity wide of 2 banks of 256K plus address select to select either 256K or second bank 256K.

Still, I'm not aware of any memory chips IC package with odd numbered data lines to include parity. Usually 1bit, 4bits, 8bits or 16bits wide IC packages.

There are 4mb package with 16 bit wide found on 2 (4MB) or 4 chip (8MB) 72 pin simms which allows you to reuse one 16 bits IC plus two 1 bit parity ICs and have room to spare to create 4MB 40 pin simms, still needs parity.

Anyway, the soldered 8 chips on the daughter board is for other uses not for main memory, is I'm correct?

Cheers,

Yes, these chips are 16bit + 2bit parity (2CAS) each. In SIMM module there is 2 RAS lines to select chip. So there are two double 8+1-bit column chips in 2 rows.
In the 8+4 chip module there is... ok here is hopefully illustrative picture. So that one 42-pin chip is the same thing as all green RAS1 chips combined together, except with one more address input.

That 8 RAM chips on that card are main memory. There are 8 chips more in the main board. I assume those are 1+1MB total.

About the computer itself... I have been working with the power supply. Finally found burned 5W resistor with internal thermal fuse.
After replacing that with the closest I had (6.8ohm, no fuse) this supply made just terrible oscillation noise. First I thought it must come from PWM controller, but no.
Reason was one open circuit 2SC2655 NPN driving a transformer. Replaced that one with exaggerated BD535. Noise was now gone and voltage levels were OK.
I gave it some load by connecting old HDD. Removed my 100W series safety lamp because it resisted too much in this point. When I rapidly altered AC mains voltage, it gave up... It kept stange "thumb" noise and it's totally dead again.
That poor 6.8ohm resistor broke down. Maybe it was just too low resistance...

Now that I think of it, it would be better just use ATX power supply there. At least smaller ones should fit. Needs some soldering work to adapt ATX connector to 2.54mm JST or something, but maybe it will save one fire and a lot of time.

That's interesting layout of dram layout, Could be the routing tracks easier this way?

Cheers,