8051 memory access hardware question

[Bob F]

I have a homebuilt home control system based on an 80c51FA processor. I recently
had a memory failure on it. It currently uses 32K x 8 bit mmemories, sharing the
space for code and variables. I would like to replace the 2 memories with one
128k x 8 bit memory.

It has been a long time since I did much hardware design. I am trying to figure
out how I can connect one 128k byte memory to map to 64 KB of program and 64 KB
of data memory, which would mean I need to select half the chip depending on
whether I am reading program code or data. Looking at the hardware manual, I
don't see an easy way to do this, as it seems that the RD\ or PSEN\ are normally
used to access data or program memory, and I need a high address bit to
differentiate them.

The chip only handles a 64 KB address space. I would jumper the program memory
to be data memory while I loaded the program, then switch the jumper back.

Can anyone offer me any suggestions of how to do this? I would really like to
increase the available space in both memories. I have one 32 pin memory socket
and one 28 pin socket available with a fair amount of flexibility of signals to
the high address and chip select bits.

 

[William W Martin]

Bob F wrote:

I have a homebuilt home control system based on an 80c51FA processor. I recently
had a memory failure on it. It currently uses 32K x 8 bit mmemories, sharing the
space for code and variables. I would like to replace the 2 memories with one
128k x 8 bit memory.

It has been a long time since I did much hardware design. I am trying to figure
out how I can connect one 128k byte memory to map to 64 KB of program and 64 KB
of data memory, which would mean I need to select half the chip depending on
whether I am reading program code or data. Looking at the hardware manual, I
don't see an easy way to do this, as it seems that the RD\ or PSEN\ are normally
used to access data or program memory, and I need a high address bit to
differentiate them.

The chip only handles a 64 KB address space. I would jumper the program memory
to be data memory while I loaded the program, then switch the jumper back.

Can anyone offer me any suggestions of how to do this? I would really like to
increase the available space in both memories. I have one 32 pin memory socket
and one 28 pin socket available with a fair amount of flexibility of signals to
the high address and chip select bits.

How about RD\ | PSEN\ -> CS, PSEN\ -> high address bit?

bill

 

[Eeyore]

Bob F wrote:

I have a homebuilt home control system based on an 80c51FA processor. I recently
had a memory failure on it. It currently uses 32K x 8 bit mmemories, sharing the
space for code and variables. I would like to replace the 2 memories with one
128k x 8 bit memory.

Why not have the code in ROM ? That's the norm.

Graham

 

[Franc Zabkar]

On Mon, 10 Nov 2008 21:15:43 -0800, "Bob F" <bobnospam@gmail.com> put
finger to keyboard and composed:

I have a homebuilt home control system based on an 80c51FA processor. I recently
had a memory failure on it. It currently uses 32K x 8 bit mmemories, sharing the
space for code and variables. I would like to replace the 2 memories with one
128k x 8 bit memory.

It has been a long time since I did much hardware design. I am trying to figure
out how I can connect one 128k byte memory to map to 64 KB of program and 64 KB
of data memory, which would mean I need to select half the chip depending on
whether I am reading program code or data. Looking at the hardware manual, I
don't see an easy way to do this, as it seems that the RD\ or PSEN\ are normally
used to access data or program memory, and I need a high address bit to
differentiate them.

The chip only handles a 64 KB address space. I would jumper the program memory
to be data memory while I loaded the program, then switch the jumper back.

Can anyone offer me any suggestions of how to do this? I would really like to
increase the available space in both memories. I have one 32 pin memory socket
and one 28 pin socket available with a fair amount of flexibility of signals to
the high address and chip select bits.

I suspect that you need only look at the problem from the perspective
of the RAMs, not the CPU.

I don't remember much about the 8051, but I would think that your two
RAMs would have OE*, CS*, and WE* pins.

If the WE* pins are wired together, then wire them to your 128KB RAM.

The other two pins could be wired as follows:

|----
OE1* ----| \ |\
| |o---| |o--- OE*
OE2* ----| / |/
|---/

|----
CS1* ----| \ |\
| |o---| |o--- CS*
CS2* ----| / |/
----/

Address lines A0 and A15 of your 128KB RAM would be wired to your CPU,
and A16 could be wired to either CS1* or CS2*.

The inverters could be constructed from the remaining two NAND gates
in a quad package.

- Franc Zabkar
--
Please remove one 'i' from my address when replying by email.

 

[Bob F]

"William W Martin" <wwm@cox.net> wrote in message
news:jpkSk.3390$lo6.196@newsfe01.iad...

Bob F wrote:
I have a homebuilt home control system based on an 80c51FA processor. I
recently had a memory failure on it. It currently uses 32K x 8 bit mmemories,
sharing the space for code and variables. I would like to replace the 2
memories with one 128k x 8 bit memory.

It has been a long time since I did much hardware design. I am trying to
figure out how I can connect one 128k byte memory to map to 64 KB of program
and 64 KB of data memory, which would mean I need to select half the chip
depending on whether I am reading program code or data. Looking at the
hardware manual, I don't see an easy way to do this, as it seems that the RD\
or PSEN\ are normally used to access data or program memory, and I need a
high address bit to differentiate them.

The chip only handles a 64 KB address space. I would jumper the program
memory to be data memory while I loaded the program, then switch the jumper
back.

Can anyone offer me any suggestions of how to do this? I would really like to
increase the available space in both memories. I have one 32 pin memory
socket and one 28 pin socket available with a fair amount of flexibility of
signals to the high address and chip select bits.

How about RD\ | PSEN\ -> CS, PSEN\ -> high address bit?

As I see it, that would mean the high address bit would go low as PSEN\ goes
low. Looking at an HM628128LP, I see no spec for Address stable to PSEN\ (CS)
low, so maybe that would work. I could throw in a couple gate delays. for the
PSEN\ compared to the high address bit. Should that work?

 

[Bob F]

"Eeyore" <rabbitsfriendsandrelations@hotmail.com> wrote in message
news:491A1D9D.C833AF7E@hotmail.com...

Bob F wrote:

I have a homebuilt home control system based on an 80c51FA processor. I
recently
had a memory failure on it. It currently uses 32K x 8 bit mmemories, sharing
the
space for code and variables. I would like to replace the 2 memories with one
128k x 8 bit memory.

Why not have the code in ROM ? That's the norm.

This unit uses an evauation board monitor program in the 80C51 to load programs
into RAM from a computer. I have used it this way for the last 15 or so years
with few problems. I have no easy way to burn PROMs. It is nice for quick
program modifications.

 

[Bob F]

"Franc Zabkar" <fzabkar@iinternode.on.net> wrote in message
news:e26lh4h6496qafb1euom10gi01lm3q4kni@4ax.com...

On Mon, 10 Nov 2008 21:15:43 -0800, "Bob F" <bobnospam@gmail.com> put
finger to keyboard and composed:

I have a homebuilt home control system based on an 80c51FA processor. I
recently
had a memory failure on it. It currently uses 32K x 8 bit mmemories, sharing
the
space for code and variables. I would like to replace the 2 memories with one
128k x 8 bit memory.

It has been a long time since I did much hardware design. I am trying to
figure
out how I can connect one 128k byte memory to map to 64 KB of program and 64
KB
of data memory, which would mean I need to select half the chip depending on
whether I am reading program code or data. Looking at the hardware manual, I
don't see an easy way to do this, as it seems that the RD\ or PSEN\ are
normally
used to access data or program memory, and I need a high address bit to
differentiate them.

The chip only handles a 64 KB address space. I would jumper the program memory
to be data memory while I loaded the program, then switch the jumper back.

Can anyone offer me any suggestions of how to do this? I would really like to
increase the available space in both memories. I have one 32 pin memory socket
and one 28 pin socket available with a fair amount of flexibility of signals
to
the high address and chip select bits.

I suspect that you need only look at the problem from the perspective
of the RAMs, not the CPU.

I don't remember much about the 8051, but I would think that your two
RAMs would have OE*, CS*, and WE* pins.

If the WE* pins are wired together, then wire them to your 128KB RAM.

The other two pins could be wired as follows:

|----
OE1* ----| \ |\
| |o---| |o--- OE*
OE2* ----| / |/
|---/

|----
CS1* ----| \ |\
| |o---| |o--- CS*
CS2* ----| / |/
----/

Address lines A0 and A15 of your 128KB RAM would be wired to your CPU,
and A16 could be wired to either CS1* or CS2*.

The inverters could be constructed from the remaining two NAND gates
in a quad package.

As I look at this from the memory point of view, I begin to think it will indeed
work. I was worried that the address setup time requirement would prevent it,
but that may not be the case.

If anyone has actually done this, I'd love to hear about your success.

 

[Franc Zabkar]

On Wed, 12 Nov 2008 22:44:11 -0800, "Bob F" <bobnospam@gmail.com> put
finger to keyboard and composed:

"Franc Zabkar" <fzabkar@iinternode.on.net> wrote in message
news:e26lh4h6496qafb1euom10gi01lm3q4kni@4ax.com...
On Mon, 10 Nov 2008 21:15:43 -0800, "Bob F" <bobnospam@gmail.com> put
finger to keyboard and composed:

I have a homebuilt home control system based on an 80c51FA processor. I
recently
had a memory failure on it. It currently uses 32K x 8 bit mmemories, sharing
the
space for code and variables. I would like to replace the 2 memories with one
128k x 8 bit memory.

It has been a long time since I did much hardware design. I am trying to
figure
out how I can connect one 128k byte memory to map to 64 KB of program and 64
KB
of data memory, which would mean I need to select half the chip depending on
whether I am reading program code or data. Looking at the hardware manual, I
don't see an easy way to do this, as it seems that the RD\ or PSEN\ are
normally
used to access data or program memory, and I need a high address bit to
differentiate them.

The chip only handles a 64 KB address space. I would jumper the program memory
to be data memory while I loaded the program, then switch the jumper back.

Can anyone offer me any suggestions of how to do this? I would really like to
increase the available space in both memories. I have one 32 pin memory socket
and one 28 pin socket available with a fair amount of flexibility of signals
to
the high address and chip select bits.

I suspect that you need only look at the problem from the perspective
of the RAMs, not the CPU.

I don't remember much about the 8051, but I would think that your two
RAMs would have OE*, CS*, and WE* pins.

If the WE* pins are wired together, then wire them to your 128KB RAM.

The other two pins could be wired as follows:

|----
OE1* ----| \ |\
| |o---| |o--- OE*
OE2* ----| / |/
|---/

|----
CS1* ----| \ |\
| |o---| |o--- CS*
CS2* ----| / |/
----/

Address lines A0 and A15 of your 128KB RAM would be wired to your CPU,
and A16 could be wired to either CS1* or CS2*.

The inverters could be constructed from the remaining two NAND gates
in a quad package.


As I look at this from the memory point of view, I begin to think it will indeed
work. I was worried that the address setup time requirement would prevent it,
but that may not be the case.

If anyone has actually done this, I'd love to hear about your success.

I have used a similar technique when repairing a Commodore 5-1/4"
floppy disc drive. It had a faulty 2332 (?) ROM, so I replaced all 3
ROMs with a single EPROM and used a scheme such as the above to decode
the additional address lines. I had to make an adapter for my device
programmer so that it could read a 2332 ROM in a 2732 EPROM socket,
and I had to provides jumpers to allow for active low or active high
OE and/or CS pins. I removed the 3x 2332 IC sockets and replaced them
with a daughter PCB connected to the main PCB via a single wirewrap
socket and a few fly wires. IIRC the decoding was done using a single
74LS00 or 74S00 quad NAND. The address setup time may be accounted for
by the delays in the two gates.

- Franc Zabkar
--
Please remove one 'i' from my address when replying by email.

 

[Bob F]

"Franc Zabkar" <fzabkar@iinternode.on.net> wrote in message
news:j2knh4dslvbuk8s3tjuoqi8r3k54n9jfq0@4ax.com...

On Wed, 12 Nov 2008 22:44:11 -0800, "Bob F" <bobnospam@gmail.com> put
finger to keyboard and composed:


"Franc Zabkar" <fzabkar@iinternode.on.net> wrote in message
news:e26lh4h6496qafb1euom10gi01lm3q4kni@4ax.com...
On Mon, 10 Nov 2008 21:15:43 -0800, "Bob F" <bobnospam@gmail.com> put
finger to keyboard and composed:

I have a homebuilt home control system based on an 80c51FA processor. I
recently
had a memory failure on it. It currently uses 32K x 8 bit mmemories, sharing
the
space for code and variables. I would like to replace the 2 memories with
one
128k x 8 bit memory.

It has been a long time since I did much hardware design. I am trying to
figure
out how I can connect one 128k byte memory to map to 64 KB of program and 64
KB
of data memory, which would mean I need to select half the chip depending on
whether I am reading program code or data. Looking at the hardware manual, I
don't see an easy way to do this, as it seems that the RD\ or PSEN\ are
normally
used to access data or program memory, and I need a high address bit to
differentiate them.

The chip only handles a 64 KB address space. I would jumper the program
memory
to be data memory while I loaded the program, then switch the jumper back.

Can anyone offer me any suggestions of how to do this? I would really like
to
increase the available space in both memories. I have one 32 pin memory
socket
and one 28 pin socket available with a fair amount of flexibility of signals
to
the high address and chip select bits.

I suspect that you need only look at the problem from the perspective
of the RAMs, not the CPU.

I don't remember much about the 8051, but I would think that your two
RAMs would have OE*, CS*, and WE* pins.

If the WE* pins are wired together, then wire them to your 128KB RAM.

The other two pins could be wired as follows:

|----
OE1* ----| \ |\
| |o---| |o--- OE*
OE2* ----| / |/
|---/

|----
CS1* ----| \ |\
| |o---| |o--- CS*
CS2* ----| / |/
----/

Address lines A0 and A15 of your 128KB RAM would be wired to your CPU,
and A16 could be wired to either CS1* or CS2*.

The inverters could be constructed from the remaining two NAND gates
in a quad package.


As I look at this from the memory point of view, I begin to think it will
indeed
work. I was worried that the address setup time requirement would prevent it,
but that may not be the case.

If anyone has actually done this, I'd love to hear about your success.

I have used a similar technique when repairing a Commodore 5-1/4"
floppy disc drive. It had a faulty 2332 (?) ROM, so I replaced all 3
ROMs with a single EPROM and used a scheme such as the above to decode
the additional address lines. I had to make an adapter for my device
programmer so that it could read a 2332 ROM in a 2732 EPROM socket,
and I had to provides jumpers to allow for active low or active high
OE and/or CS pins. I removed the 3x 2332 IC sockets and replaced them
with a daughter PCB connected to the main PCB via a single wirewrap
socket and a few fly wires. IIRC the decoding was done using a single
74LS00 or 74S00 quad NAND. The address setup time may be accounted for
by the delays in the two gates.

Sounds good.

Interestingly, I just found somw SRAMs on ebay which seem to use a completely
different ordering on the address lines than the convention my sockets are wired
to.

http://download.siliconexpert.com/pdfs/quickdata/mot/mcm6226a.rev4.pdf
as compared to
http://www.jameco.com/Jameco/Products/ProdDS/103982.pdf

I'm guessing that the difference in the address pin numbering would not keep it
from operating properly. Thet each address is distinct, and it would give
appropriate data for each access, even if they were not coming out of the memory
matrix in order. Does this sound right.

Also, would you expect that using a significantly faster memory would cause any
problem. My current memory is 55 nS, I believe, and I found one source for 15 nS
chips. My board is a hand-wired device (Wires soldered pin-to-pin.)
http://rocky.digikey.com/weblib/Cypress/Web%20Data/CY7C109B,%20CY7C1009B.pdf

 

[Franc Zabkar]

On Wed, 12 Nov 2008 23:46:42 -0800, "Bob F" <bobnospam@gmail.com> put
finger to keyboard and composed:

Interestingly, I just found somw SRAMs on ebay which seem to use a completely
different ordering on the address lines than the convention my sockets are wired
to.

http://download.siliconexpert.com/pdfs/quickdata/mot/mcm6226a.rev4.pdf
as compared to
http://www.jameco.com/Jameco/Products/ProdDS/103982.pdf

I'm guessing that the difference in the address pin numbering would not keep it
from operating properly. Thet each address is distinct, and it would give
appropriate data for each access, even if they were not coming out of the memory
matrix in order. Does this sound right.

I've seen this difference, too, but, like you, I can't see a problem
with it. It may be confusing, though, when trying to read a ROM whose
address lines don't match up with convention.

Also, would you expect that using a significantly faster memory would cause any
problem. My current memory is 55 nS, I believe, and I found one source for 15 nS
chips. My board is a hand-wired device (Wires soldered pin-to-pin.)
http://rocky.digikey.com/weblib/Cypress/Web%20Data/CY7C109B,%20CY7C1009B.pdf

I haven't yet had a problem substituting a faster RAM for a slower
one, but I've only tried my hack with ROMs, which are relatively slow
devices.

- Franc Zabkar
--
Please remove one 'i' from my address when replying by email.

 

[Eeyore]

Bob F wrote:

I have no easy way to burn PROMs.

Get a FLASH based 8051 like the 89s52 ?? check their FLASH memory sizes and program
it in circuit using the parallel port adaptor lead supplied by Atmel along with
their programming tool. It seems a bit flakey at times but does work. If the port
pins required conflict with your circuit you only need power xtal and reset to
program it.

I actually made my own adaptor from Atmel's app notes not realising Farnell sold
them and tagged it onto another PCB.

Graham

 

[Eeyore]

Franc Zabkar wrote:

On Wed, 12 Nov 2008 23:46:42 -0800, "Bob F" <bobnospam@gmail.com> put
finger to keyboard and composed:

Interestingly, I just found somw SRAMs on ebay which seem to use a completely
different ordering on the address lines than the convention my sockets are wired
to.

http://download.siliconexpert.com/pdfs/quickdata/mot/mcm6226a.rev4.pdf
as compared to
http://www.jameco.com/Jameco/Products/ProdDS/103982.pdf

I'm guessing that the difference in the address pin numbering would not keep it
from operating properly. Thet each address is distinct, and it would give
appropriate data for each access, even if they were not coming out of the memory
matrix in order. Does this sound right.

I've seen this difference, too, but, like you, I can't see a problem
with it. It may be confusing, though, when trying to read a ROM whose
address lines don't match up with convention.

Also, would you expect that using a significantly faster memory would cause any
problem. My current memory is 55 nS, I believe, and I found one source for 15 nS
chips. My board is a hand-wired device (Wires soldered pin-to-pin.)
http://rocky.digikey.com/weblib/Cypress/Web%20Data/CY7C109B,%20CY7C1009B.pdf

I haven't yet had a problem substituting a faster RAM for a slower
one, but I've only tried my hack with ROMs, which are relatively slow
devices.

I doubt an ancient version of the 8051 will reach enough speed to worry about unless
you're using those darn new-fangled accelerated ones !

Graham

 

[Bob F]

"Eeyore" <rabbitsfriendsandrelations@hotmail.com> wrote in message
news:491C8C50.648919D2@hotmail.com...

Bob F wrote:

I have no easy way to burn PROMs.

Get a FLASH based 8051 like the 89s52 ?? check their FLASH memory sizes and
program
it in circuit using the parallel port adaptor lead supplied by Atmel along
with
their programming tool. It seems a bit flakey at times but does work. If the
port
pins required conflict with your circuit you only need power xtal and reset to
program it.

I actually made my own adaptor from Atmel's app notes not realising Farnell
sold
them and tagged it onto another PCB.

Thanks for the suggestion. This would at least allow me to modify the monitor
program within the CPU to be able to program external flash memory.

 

[Eeyore]

Bob F wrote:

"Eeyore" wrote
Bob F wrote:

I have no easy way to burn PROMs.

Get a FLASH based 8051 like the 89s52 ?? check their FLASH memory sizes and
program it in circuit using the parallel port adaptor lead supplied by Atmel
along
with their programming tool. It seems a bit flakey at times but does work. If
the
port pins required conflict with your circuit you only need power xtal and reset
to
program it.

I actually made my own adaptor from Atmel's app notes not realising Farnell
sold them and tagged it onto another PCB.

Thanks for the suggestion. This would at least allow me to modify the monitor
program within the CPU to be able to program external flash memory.

Farnell's online search engine refuses to identify the part (which is why I never
originally found it), but it wasn't expensive, something like Ł25 bundled with the
software when I found it in the paper catalogue. Atmel's site proves equally
unhelpful but I have some files on it here.

Philips (oops NXP) do something similar but unhelpfully different it seems.

If you use ExpressPCB I could send you the design files for it. It's easy to tag it
onto another board, you can even make them 'snap apart', and at ~ 4cm sq approx
won't cost anything if you have a design going through already. It's all through
hole for simple assembly and fits in a typical 25D sub to 25D sub adaptor housing.

Graham

 

[Eeyore]

Eeyore wrote:

Bob F wrote:
"Eeyore" wrote
Bob F wrote:

I have no easy way to burn PROMs.

Get a FLASH based 8051 like the 89s52 ?? check their FLASH memory sizes and
program it in circuit using the parallel port adaptor lead supplied by Atmel
along with their programming tool. It seems a bit flakey at times but does work.
If
the port pins required conflict with your circuit you only need power xtal and
reset
to program it.

I actually made my own adaptor from Atmel's app notes not realising Farnell
sold them and tagged it onto another PCB.

Thanks for the suggestion. This would at least allow me to modify the monitor
program within the CPU to be able to program external flash memory.

Farnell's online search engine refuses to identify the part (which is why I never
originally found it), but it wasn't expensive, something like Ł25 bundled with the
software when I found it in the paper catalogue. Atmel's site proves equally
unhelpful but I have some files on it here.

Philips (oops NXP) do something similar but unhelpfully different it seems.

If you use ExpressPCB I could send you the design files for it. It's easy to tag it
onto another board, you can even make them 'snap apart', and at ~ 4cm sq approx
won't cost anything if you have a design going through already. It's all through
hole for simple assembly and fits in a typical 25D sub to 25D sub adaptor housing.

That's how Atmel do it too.

The search term was 'AT89ISP'
http://uk.farnell.com/atmel/at89isp/programmer-cable-kit-for-isp-at89s/dp/1095733?_requestid=432321

The 'technical data sheet' gives partial info but no schematic. But I have that. There
have been a couple of revisions to add functionality for newer processors. I think I
incorprated those into my last 'spin' of the board. It is all there on the Atmel site
but you have to dig deep.

Graham