I installed 512K RAM internally in my msx2 Hb-F9s, works...

Why sockets for the original DRAMs, if you're not using those anymore? (and why remove those capacitors next to them, an extra capacitor across Vcc never hurts...)

In case anyone wonders: that transistor / diode construct makes S1985 output a mapper address line on the pin it shares with KBDIR signal (see datasheet). Probably would have found a simpler construction myself, but who cares if it does what it must do.

Hello everyones, I remodeled the 512K RAM installation to release all z80 wires connections, i placed a 40pin socket for the z80 for easy future changes, for the RAM new install i used some of the IC13 (calcutator firmware sony) original connections, i placed an adapted socket for the RAM in the IC13 and placed the HCT´s also in sockets near the RAM. Now the S1985 wires are all wire in alternative points less the pin4. I resoldered the original RAM 128K capacitors that had been released in the first install.

When i did the first install i could only see 128K of installed RAM because i had the original ROM IC´s, the only way to know the ram I had installed was to use the program testram.com and this shows the 512K when i did the test. Now with the help of NYYRIKKI ,a user of this forum, i can see all RAM that i have installed in boot, this is 512K, NYYRIKKI prepared for me the roms needed to update to msx2+ my hb-f9s, thanks eternally NYYRIKKI from here.

The problem is that for some reason sometimes i see 256K only on boot, this happens sometimes when i RESET the machine, and if i do a hard turn OFF (turn off about 1 minutes aprox.) the 512K see again on boot up, anyways it seems that after new tests the problem is something random because this happens also if i boot the msx from a hard turn OFF, i think that will be problem with the X5 circuit, i tried to install a NPN BS170 (DGS=CBE wires) and i tried conect this NPN i another ways, but if i follow the correct CBE (corresponds to DGS) wires i also see 256K on boot, if i use a NPN 2N2219A (i have this installed now, and i tried someones of this) i can tell that the 70% of times i can see 512K on boot, it seems that i have some hardware problem. If the boot shows up 256K and then i test the RAM using testram.com i only can access to 256K, if in the boot i see 512K the testram access to 512K, it seems with this that the problem is not in the eproms that i burned, the boot up seems that is correct, something will be wrong in the hardware. Another test that i did was turn ON the msx with the keyboard no plugged, and seems that this way i see the 512K 90% of the times, then i desolder the original cable of the keyboard and i changed it in the motherboard for pinouts to can plug a ide cable ( each ide cable was soldered in the keyboard pcb) to try avoid some bad conection on it, but the problem persists. Also i tried some new diodes. I think that the problem is the NPN, any ideas? any way to fix the X5 or make it another way?

In the pictures you´ll see my install and X5 original diagram and notes in my NPN install:

Regards.

Are there any other diodes going to Xx pins of S1985? Can you attach picture of the schematic diagram please?

Just checked the connections shown in that 512K expansion PDF, and concluded what I suspected earlier: those 2 logic IC's (74HCT02 and 74HCT32) are totally unnecessary! Just rip them out: only S1985 pin 55 (#SLT32) to SRAM pin 22 (#CE), Z80 pin 21 (#RD) to SRAM pin 24 (#OE), Z80 pin 22 (#WR) to SRAM pin 29 (#WE), and I predict you'll see no difference in function. The tests with my CPLD-based memory mapper already proved it works fine that way, no extra logic needed on the SRAM control inputs.

As for your 256K/512K issue: very likely that's result of what happens at S1985's #X5 keyboard pin. Datasheet is very clear about it: to enable MA18 output signal, #X5 pin must be low at the time that S1985's RESET signal goes from active (high) to non-active/normal (low). May I suggest:

• Remove that transistor/diode combo you have now.
• Remove/unsolder one side of R77 (so that IC22 pin 9 gets disconnected).
• Connect IC22 pins 9 & 10 (so that this OR gate becomes a simple buffer / short delay).
• Put a diode between IC22 pin 8 and S1985 #X5 signal (pin 78), such that diode conducts in #X5 -> IC22 pin 8 direction (marked side of diode to IC22).
• If that doesn't fix 256K/512K issue: add a small capacitor between IC22 pin 8 & ground (a couple of nF's should be enough I think). That would cause this output to respond a bit slower, such that diode can keep #X5 low until after S1985 reset input has gone to normal state.

Dhampird e-mailed me that above connections didn't work for him. Somehow I doubted that what works for me as a cartridge, on 7 different MSX models (including HB-F9P), built around different MSX engines (or none), wouldn't work internally on HB-F9S.

So I checked datasheet for the part he's using, and guess what? Hitachi HM658512 is not a 512K x 8 static RAM, it's a 512K x 8 pseudo static RAM. Read: it has non-multiplexed address lines with SRAM-like pinout, but needs refresh like any other DRAM. And consequently its control signals work a little different. It's a mystery to me why anyone would think a pseudo static RAM + 2 logic IC's + extra wiring is easier to use than a 512K x 8 SRAM... maybe somebody found a dirt cheap stack of them somewhere (or it was cost-effective solution 10 years ago)? At today's SRAM prices (around 5-7 Euro's for a 512K x 8 in most online shops) it's clearly not worth the trouble. At least we know what those 2 74HCT IC's were for (you can still simplify the logic in those btw, including Z80 #BUSACK signal in memory control is bullshit). If somebody sold that to you as an SRAM, get your money back - this is a DRAM!

Which leaves you with a simple choice, Dhampird: reconnect those HCT02 & HCT32 IC's, or replace with a real 512K x 8 SRAM... guess what I'd do...