HP48GX RAM card design

[pdo]

Hello All,

For anyone interested in making their own RAM cards for HP48GX calculators, I have just uploaded a KiCAD design on which I've recently been working. The design files are hosted in a git repository at:-

https://bitbucket.org/cloudycat/rc48-512/overview

along with a small amount of documentation. As you can see from the README, I have built two of the cards myself and have been using them for the past few weeks with no major problems so far.

The files are released under a CC-BY-SA licence which essentially means that any use can be made of them as long as modifications are also released under the same licence.

Obviously I cannot give any guarantees as to the correctness of the design or fitness for purpose (particularly as the information I found on the net about this was incomplete and sometimes inconsistent). Nevertheless, I had enough faith in the design to consider it worth the risk with my one and only HP48GX calculator!

Any comments welcome,
Paul

[EugeneNine]

I was going to ask about SX but see you mentioned it. I was thinking the SX can't address that much so I'd need to use a smaller ram chip.

[pdo]

(06-12-2017 01:08 AM)EugeneNine Wrote:  I was going to ask about SX but see you mentioned it. I was thinking the SX can't address that much so I'd need to use a smaller ram chip.

Yes, I hoped that the addition of the A17 and A18 switches would allow operation in an SX as well as operation in either slot of a GX but, as I described in the README, something I don't understand is going on there, so now I can't be confident of it working in an SX.

When I have some more time I'll write a bit of software and look at the signals on a scope to try to figure it out (unless someone out there knows what is happening and can educate us first!).

Paul

[pdo]

(06-12-2017 12:29 PM)pdo Wrote:  Yes, I hoped that the addition of the A17 and A18 switches would allow operation in an SX as well as operation in either slot of a GX but, as I described in the README, something I don't understand is going on there, so now I can't be confident of it working in an SX.

Further testing shows that there is an intermittent problem with one of the two boards I have assembled: it can randomly cause bad card data errors. I need to investigate further, but I suspect a duff solder joint as my soldering is not the best!

However, the other card works fine. It can be used in slot 1, switching between 4 different port 1 configurations via SW17/18's '0' & '1' settings, or in slot 2, mapping the 4 banks to ports 2 to 5 with SW17/18 set to 'A'.

Given this turn of events, I feel more confident that the card should also work in an SX too. I will feel even more confident when I have fixed the problem with the other card!

Paul

[EugeneNine]

Looks like a bit of empty space on the board, maybe move everything toward the connector a bit and then sw3 could move over where the battery is leaving space between it and sw2 to make a larger pull hole.

The ram looms to be only $5 so I'll order a couple next time I make a digikey order and once I have them make a board and try it in my SX.

[mfleming]

(06-22-2017 12:15 AM)EugeneNine Wrote:  The ram looms to be only $5 so I'll order a couple next time I make a digikey order and once I have them make a board and try it in my SX.

I'd love to see if this design would work in an SX! I pulled mine from mothballs a while back, along with a TDS memory card. It had a CR2016 holder that allowed the coin cell to be replaced while the card was in the SX so that RAM content was not lost. Perhaps in lieu of programmatic bank switching, the address switches could be moved to the interior to make room for the battery?

~Mark

[EugeneNine]

It just may be a while before my next order, I recently got an order of parts and don't have much time to work with electronics now a days so I don't make orders too often.

[pdo]

(06-22-2017 03:04 AM)mfleming Wrote:  I'd love to see if this design would work in an SX! I pulled mine from mothballs a while back, along with a TDS memory card. It had a CR2016 holder that allowed the coin cell to be replaced while the card was in the SX so that RAM content was not lost. Perhaps in lieu of programmatic bank switching, the address switches could be moved to the interior to make room for the battery?

Well, it may not be as much of a problem as you might think. I realized that my testing of the C1 & C4 battery change capacitors was done with the faulty board, and that the fault was probably invalidating my battery change tests. So today I retouched some of the solder joints on the misbehaving board and this evening was able to confirm that the board seems to behave correctly now :-)

I then redid my battery change tests:-
1. Take the board out of the calculator.
2. Remove the battery from the board, hold for 10 seconds, then reinsert battery.
3. Put the board back in the calculator and check objects in all 4 ports are still okay.

I did this three times and am happy to report that the test passed each time. Not exhaustive I know, but encouraging non-the-less.

I now feel an edit of the README coming on...

Paul

[wan]

I made a piece, but there are some modifications:
gold fingers flip to the back;
rounded four corners of the card;
1.6mm is too thick, use 0.6mm board instead；
eliminate all the pull-up resistors of we and oe lines;
no bank swiching.

It works well so far.

[Eric Rechlin]

(01-18-2021 03:51 AM)wan Wrote:  I made a piece, but there are some modifications:
gold fingers flip to the back;
add rounded around the card;
1.6mm is too thick, use 0.6mm board instead；
eliminate all the pull-up resistors of we and oe lines.
no bank swiching.

it works well so far.

In his Readme he wrote that 1.2 mm was too thin and to use 1.6 mm instead. How does 0.6 mm work then? Wouldn't that be way too thin for the connector?

The connector on the edge of my RAM cards is actually 2.0 mm thick, with the total card thickness being about 2.5 mm, so I would think a better approach would be to start with something approximately like what you did (a 0.6 mm board with the gold fingers flipped to the other side), and then add a second bare 1.2 mm board glued to the top, with holes for the components to go through it, to add thickness (for the edge connector) and provide some protection for the components on the board.

Also, what was the reason for removing the pull-up resistors on WE and OE?

Finally, might you be able to please share your modified layout?

[wan]

(01-18-2021 03:09 PM)Eric Rechlin Wrote:  The connector on the edge of my RAM cards is actually 2.0 mm thick, with the total card thickness being about 2.5 mm, so I would think a better approach would be to start with something approximately like what you did (a 0.6 mm board with the gold fingers flipped to the other side), and then add a second bare 1.2 mm board glued to the top, with holes for the components to go through it, to add thickness (for the edge connector) and provide some protection for the components on the board.

Yes, that's what I did. A bare 1.2 mm board on top is necessary. An optional short 0.6mm board on back can help(2.4mm total then), leaving goldfingers uncoverd. Pay attention to the location of the batteryholer to avoid scratching in card slot 2(no worry in slot 1). You can get a smooth insertion by this layout, even neither a handle nor a hole is needed for pulling out.

(01-18-2021 03:09 PM)Eric Rechlin Wrote:  Also, what was the reason for removing the pull-up resistors on WE and OE?

According to this post
https://www.hpmuseum.org/forum/thread-13070.html
Mfleming mentioned two pullups on OE/ and WE/, and at 47K they could sink tens of uA. So they are removed for power saving.
I think the 330 ohm series resistors for data lines can also be cancelled.

(01-18-2021 03:09 PM)Eric Rechlin Wrote:  Finally, might you be able to please share your modified layout?

The picture of the actual board is in the attatchment.

[mfleming]

Very nice arrangement from your description. I'd love to see a picture of the other side as well. The PCB you've labeled Middle Board isn't just a spacer, there are tracks and unpopulated component pads. Was it meant to be part of the design? The cutouts look great. You might cement a plastic sheet (preferably clear!) over the top surface to keep fingers out, and with a notch to pull the battery out from the side. HP would probably have used a metal sheet for EMI shielding, a little overkill for this project

The series resistors in the data lines can be dropped. I didn't use them in my own board. They were probably added to the original design for static protection or to isolate a defective memory chip.

Nice work!
~Mark

[wan]

(01-19-2021 02:03 AM)mfleming Wrote:  I'd love to see a picture of the other side as well.

No component on back side. The backboard is not attatched in this one. See attatchment 1. With backboard see attatchment 2.

(01-19-2021 02:03 AM)mfleming Wrote:  The PCB you've labeled Middle Board isn't just a spacer, there are tracks and unpopulated component pads. Was it meant to be part of the design?

No, they have no use at all. They are used to misleading board manufacturer into thinking this is a "useful" board.

(01-19-2021 02:03 AM)mfleming Wrote:  You might cement a plastic sheet (preferably clear!) over the top surface to keep fingers out, and with a notch to pull the battery out from the side.

Nice idea!

[mfleming]

(01-19-2021 02:03 AM)mfleming Wrote:  The PCB you've labeled Middle Board isn't just a spacer, there are tracks and unpopulated component pads. Was it meant to be part of the design?

(01-19-2021 02:36 AM)wan Wrote:  No, they are no use at all. They are used to misleading board manufacturer into thinking this is a "useful" board.

[pdo]

Hi Wan,

Nice work!

For the record, the 330 ohm resistors in the data lines were put there to limit the current flow in case of contention between the processor and RAM chip. i.e. if something was very wrong in the control or timing of the chip and the processor and RAM were both trying to drive opposite levels onto the data lines at the same time.

An abundance of caution, you might say, because I had no proper specification for the processor chip and didn't want to damage it if my design was wrong.

Paul

[wan]

According to the 48gx inside board photo and circuit schematic, the built-in sram chip is connected directly to Saturn chip. That's why I think the series resistors can be cancelled.

About the bank switching card, I have some questions.
1 Can the high-level signal of the switches be connected to vccon instead of vcc, for power saving?
2 Can write protect switch be connected to the we line instead of the cdet line, for hardware level protection?

[pdo]

(01-20-2021 03:48 AM)wan Wrote:  About the bank switching card, I have some questions.
1 Can the high-level signal of the switches be connected to vccon instead of vcc, for power saving?
2 Can write protect switch be connected to the we line instead of the cdet line, for hardware level protection?

Hmm, interesting questions, but I can't answer them with any certainty because I don't know in detail how the processor chip works.

Perhaps there is some design documentation from HP somewhere that would enlighten us in this area, but unfortunately I'm not aware of any.

Paul

[Jonathan Busby]

(01-20-2021 09:18 PM)pdo Wrote:  

(01-20-2021 03:48 AM)wan Wrote:  About the bank switching card, I have some questions.
1 Can the high-level signal of the switches be connected to vccon instead of vcc, for power saving?
2 Can write protect switch be connected to the we line instead of the cdet line, for hardware level protection?

Hmm, interesting questions, but I can't answer them with any certainty because I don't know in detail how the processor chip works.

Perhaps there is some design documentation from HP somewhere that would enlighten us in this area, but unfortunately I'm not aware of any.

Paul

Well, I have the Seiko-Epson official catalog of SRAM cards ( and other memory cards ) with all the technical specs, but, unfortunately, I've moved to Colorado from Texas and most of my possessions are in storage ATM. I scanned the catalog and I have it saved on an external HDD but it is in storage as well. I also wrote down, a long time ago, all the salient information regarding HP48GX compatible memory cards, but, it again is in storage :/

Regards,

Jonathan

[Marco Polo]

(01-19-2021 12:54 AM)wan Wrote:  

(01-18-2021 03:09 PM)Eric Rechlin Wrote:  The connector on the edge of my RAM cards is actually 2.0 mm thick, with the total card thickness being about 2.5 mm, so I would think a better approach would be to start with something approximately like what you did (a 0.6 mm board with the gold fingers flipped to the other side), and then add a second bare 1.2 mm board glued to the top, with holes for the components to go through it, to add thickness (for the edge connector) and provide some protection for the components on the board.

Yes, that's what I did. A bare 1.2 mm board on top is necessary. An optional short 0.6mm board on back can help(2.4mm total then), leaving goldfingers uncoverd. Pay attention to the location of the batteryholer to avoid scratching in card slot 2(no worry in slot 1). You can get a smooth insertion by this layout, even neither a handle nor a hole is needed for pulling out.

(01-18-2021 03:09 PM)Eric Rechlin Wrote:  Also, what was the reason for removing the pull-up resistors on WE and OE?

According to this post
https://www.hpmuseum.org/forum/thread-13070.html
Mfleming mentioned two pullups on OE/ and WE/, and at 47K they could sink tens of uA. So they are removed for power saving.
I think the 330 ohm series resistors for data lines can also be cancelled.

(01-18-2021 03:09 PM)Eric Rechlin Wrote:  Finally, might you be able to please share your modified layout?

The picture of the actual board is in the attatchment.

How you manage to change the battery? It should be changed with card installed in calculator, if I remember correctly, in order to avoid memory clear.
In the original design the battery is located on top side (the one accessible when card is installed ). Cynox design is similar.

[wan]

(01-21-2021 06:27 AM)Marco Polo Wrote:  How you manage to change the battery? It should be changed with card installed in calculator, if I remember correctly, in order to avoid memory clear.
In the original design the battery is located on top side (the one accessible when card is installed ). Cynox design is similar.

There‘re two gold colored capacitors on the bottom right of the board, these two 440uf capacitors will provide power when changing the battery. They are larger than Paul's original design and provide longer holding time. I think they can hold for more than 20 seconds.

The reason for not placing the batteryholder on top is to avoid touching it when pulling the card out. Pulling the card with batteryholder will apply shear force on it, which may cause the batteryholder to desolder over time.