pcdata76
Experienced Member
Hello everybody,
After reading the threads which are explaining the methodology of adding memory into UMB area and using it to load device drivers, I decided to build a simple memory card to use in my 5155, Turbo XT and PS/2 model-30 computers. Actually, I’ve decided to buy Lotech’s 1 MB RAM card PCB first, but then realized that they don’t provide shipping option to my country and decided to design a circuit having in similar concept with Lotech’s board.
I also made some design changes according to my needs, which are:
1- I used only one 512 kB SRAM chip since all my computers have at least 512k of memory.
2- Instead of 64k blocks, memory can be enabled or disabled in 32kB blocks (to be able to use possible empty spaces especially in A, B and C region).
3- Smaller design to be able to keep the card in slot without the need of slot bracket.
Since I wanted to source the components from local suppliers, had to buy some of the chips in SMD and some others in through hole package. I also couldn’t source 74LS05 open collector inverter so I’ve used 74HCT04 inverter instead, omitting the pull-up resistor.
You can find the schematic and PCB layout from the following links (sorry about the complexity of the schematic, I didn’t used labels but direct wiring):
http://i1376.photobucket.com/albums/ah14/pcdata76/Vintage PC/RAMCARD PCB_zpsz60ysbes.png
http://i1376.photobucket.com/albums/ah14/pcdata76/Vintage PC/RAMCARD SCHEMATIC_zpse0jnfeu2.png
And it comes to the story:
I’ve produced the pcb (using cnc engraving machine), soldered components and checked for possible shorts and continuity, everything seems to be OK.
To test the functionality of the card, first I installed it to a slot except slot 8 of 5155. This computer had 256k on board and 384k on AST 6packplus, total 640 k of memory. I reduced the memory on the AST to 256k, and set the switches on my card 8000 to 9FFF ON, rest OFF, powered the computer ON expecting a memory count till 640 kB, but finished at 512 with no errors on the screen, so something was wrong.
Then, I intentionally turned switches ON for A, B and F ranges one by one at each retry and saw that the computer didn’t boot as I expected due to the address conflicts. So I understand that the logic circuit on the card was (at least partially) operating. If I remove the SRAM chip, computer behaves like card was not installed independent of the position of the switches, as expected.
I exchanged the SRAM chip and 74HCT30’s (as these were the only through hole components installed on a socket) with different ones of the same model, but did nothing. I didn’t touch SMD IC’s yet as I don’t have hot air station for easier removal.
I’ve done some more test as follows:
-Enabling all 384k on AST and enabling 8000h-87FFh switch on the board: Memory counts till 512k and throws an error.
-Enabling all 384k on AST and enabling 8800h-8FFFh switch on the board: Memory counts till 546k and throws an error.
continued to test in this sequence till 9FFF and saw that memory count is increased 32k at each step as expected.
After, I only enabled D and E range and check for possible umbs with no luck :/ Then, I’ve made a last test in slot8 of the 5155 by enabling slot8 jumper and everything was the same.
I've decided to test the card in PS/2 model 30 with 640k onboard memory. If I enable switches of 8 and 9 ranges, it gives parity error as expected. If I enable D and E ranges, I couldn’t managed to get umbs.
PS: I’ve used both original and Krille’s modded versions of USE!UMBS.SYS
Anybody has a comment about the source of the problem? Did I miss something do something wrong? Is there any obvious error in the circuit design or the layout?
Thank you for your help,
Onur
After reading the threads which are explaining the methodology of adding memory into UMB area and using it to load device drivers, I decided to build a simple memory card to use in my 5155, Turbo XT and PS/2 model-30 computers. Actually, I’ve decided to buy Lotech’s 1 MB RAM card PCB first, but then realized that they don’t provide shipping option to my country and decided to design a circuit having in similar concept with Lotech’s board.
I also made some design changes according to my needs, which are:
1- I used only one 512 kB SRAM chip since all my computers have at least 512k of memory.
2- Instead of 64k blocks, memory can be enabled or disabled in 32kB blocks (to be able to use possible empty spaces especially in A, B and C region).
3- Smaller design to be able to keep the card in slot without the need of slot bracket.
Since I wanted to source the components from local suppliers, had to buy some of the chips in SMD and some others in through hole package. I also couldn’t source 74LS05 open collector inverter so I’ve used 74HCT04 inverter instead, omitting the pull-up resistor.
You can find the schematic and PCB layout from the following links (sorry about the complexity of the schematic, I didn’t used labels but direct wiring):
http://i1376.photobucket.com/albums/ah14/pcdata76/Vintage PC/RAMCARD PCB_zpsz60ysbes.png
http://i1376.photobucket.com/albums/ah14/pcdata76/Vintage PC/RAMCARD SCHEMATIC_zpse0jnfeu2.png
And it comes to the story:
I’ve produced the pcb (using cnc engraving machine), soldered components and checked for possible shorts and continuity, everything seems to be OK.
To test the functionality of the card, first I installed it to a slot except slot 8 of 5155. This computer had 256k on board and 384k on AST 6packplus, total 640 k of memory. I reduced the memory on the AST to 256k, and set the switches on my card 8000 to 9FFF ON, rest OFF, powered the computer ON expecting a memory count till 640 kB, but finished at 512 with no errors on the screen, so something was wrong.
Then, I intentionally turned switches ON for A, B and F ranges one by one at each retry and saw that the computer didn’t boot as I expected due to the address conflicts. So I understand that the logic circuit on the card was (at least partially) operating. If I remove the SRAM chip, computer behaves like card was not installed independent of the position of the switches, as expected.
I exchanged the SRAM chip and 74HCT30’s (as these were the only through hole components installed on a socket) with different ones of the same model, but did nothing. I didn’t touch SMD IC’s yet as I don’t have hot air station for easier removal.
I’ve done some more test as follows:
-Enabling all 384k on AST and enabling 8000h-87FFh switch on the board: Memory counts till 512k and throws an error.
-Enabling all 384k on AST and enabling 8800h-8FFFh switch on the board: Memory counts till 546k and throws an error.
continued to test in this sequence till 9FFF and saw that memory count is increased 32k at each step as expected.
After, I only enabled D and E range and check for possible umbs with no luck :/ Then, I’ve made a last test in slot8 of the 5155 by enabling slot8 jumper and everything was the same.
I've decided to test the card in PS/2 model 30 with 640k onboard memory. If I enable switches of 8 and 9 ranges, it gives parity error as expected. If I enable D and E ranges, I couldn’t managed to get umbs.
PS: I’ve used both original and Krille’s modded versions of USE!UMBS.SYS
Anybody has a comment about the source of the problem? Did I miss something do something wrong? Is there any obvious error in the circuit design or the layout?
Thank you for your help,
Onur
You can see the problematic junction on the attached picture. Thin yellow lines are toolpaths. 