Astonishing !

[Michael de Estrada]

A few days ago I tried turning on my HP 42S, only to discover that the batteries had gone completely dead and there was no display. I removed them and the total combined voltage was less than 2 vdc, or less than half the normal 4.5 vdc for the three 1.5 vdc cells. I didn't have any spares, and had to wait for nearly a day before I was able to buy new ones and install them. When I turned on the calculator, the display showed the expected "Machine Reset" message. However, I was astonished to find that none of my programs had been lost or even the custom menus. They had all been retained for nearly a day without any battery power. I am amazed at the persistence of the memory in this machine.

[Mike Morrow]

In 15 years of frequent use, I've never lost memory on my HP-42S.

[Dan W]

Reminds me of a problem I had with my first 42S. Bought in Feb 1989 but when I changed the batteries in November that year, I lost everything. So I tested it with new batteries and sure enough, the instant they were removed, I lost everything.

So I returned it and HP replaced it free of charge. Never had a problem since. It has been continuously powered since 1989, though I don't use it as much these days. Couldn't say how many sets of batteries I have run through it, probably around 8 or so.

[db (martinez, ca.)]

Dan; i wish i would have thought of returning it for one with a better memory. when mine dumped 6k of programs i dumped it.

[Luiz C. Vieira (Brazil)]

Those guys knew how to do things back then... 8^)

[uhmgawa]

Quote:

---

Those guys knew how to do things back then... 8^)

---

I suspect there is a little more to the story than
meets the eye. The original 64Kb static ram
was only characterized down to 4.5V and that was
about the maximum voltage achieved by a new stack
of LR44s. So apparently there was some sort of
HP special characterization provided for operation
at the lower VDD encountered in a Saturn system.
Not too far fetched as the bus cycle generated by
the Saturn CPU is considerably relaxed relative
to what the SRAM could provide at/over 4.5V. IIRC
I'd measured a bus cycle time of ~2us.

The data retention voltage for the SRM2264LM10 is
guaranteed down to 2.0V so assuming low leakage
silicon and tantalum capacitors, having a power
rail held up over night isn't unreasonable.

[Mike Morrow]

Quote:

---

The original 64Kb static ram...

---

For the 42S??

[Marc Ferrer (France)]

Yes, simply add "it" and you get 64Kbit, or 8 K bytes
:-)

[Mike Morrow]

Ah so. Thanks.

[Luiz C. Vieira (Brazil)]

Hi.

No questions about SRAM data retention under low-voltage condition (is it the equivalent to the brown-out term currently used in some u-controllers?), what actually amuses me is the way to deal with that. If the processor is not designed to work under the same circumstances, considering also that no operation was attempted while the batteries where 'down', then a [ Memory Lost ] might also be 'ordered' even if memory contents were OK. So, not even the memory chips work with low level voltage, the HP42S processor have also been designed to follow it. And that's what I was wondering about.

You mentioned the low leakage tantalum and silicon capacitors: chances are that defective components or cold soldering might also be the cause of the memory loss while changing batteries in some units.

Cheers.

Luiz (Brazil)

[uhmgawa]

Quote:

---

No questions about SRAM data retention under low-voltage condition (is it the equivalent to the brown-out term currently used in some u-controllers?)

---

It's just the minimum voltage where a guarantee can be
comfortably made,
the state of a memory cell will be retained. Actually that
2.0V is rather pessimistic and I'd expect dropout is probably
considerably lower.
Recall the shameless "HP over-engineering to a fault",
where empirically NUT silicon has such a
low dropout voltage it manifests as somewhat of a comic
problem. Intentionally trying to trigger a power cycle
reset via cell removal often requires a coin, paper clip,
or similar unholy procedure to short the power rail.

Quote:

---

what actually amuses me is the way to deal with that. If the processor is not designed to work under the same circumstances, considering also that no operation was attempted while the batteries where 'down', then a [ Memory Lost ] might also be 'ordered' even if memory contents were OK. So, not even the memory chips work with low level voltage, the HP42S processor have also been designed to follow it. And that's what I was wondering about.

---

I'm uncertain exactly what is used to ensure integrity in this
case.
But a systems approach will take the processor into reset
upon entering a brownout, a reasonable margin before erratic
operation could occur. And when in reset, electrically the
memory control (essentially chip select and/or write enables)
of the processor will be designed to remain inactive well
past the minimum dropout voltage of the associated memory.

The firmware could detect the CPU has come out of an
electrical reset and flag the condition as "Memory Lost".
The integrity of RWM data however would best be validated
with a wide CRC. I don't believe the latter occurs, but
I haven't examined the firmware specifically for this either.

Quote:

---

You mentioned the low leakage tantalum and silicon capacitors: chances are that defective components or cold soldering might also be the cause of the memory loss while changing batteries in some units.

---

Given the claim battery removal causes a rapid power reset,
I'd expect a bad solder joint and/or bad bypass capacitor.
Otherwise if the leakage leakage load was that great I'd expect
short battery life to be a more obvious problem.

[Randy]

The most common cause of memory loss in pioneers occurs when installing the new cells. It has to be done in such a way that you don't short the terminals or the edges of the new cells. There was a thread here recently about this with a good method to minimize memory loss.