HP 9100A/B: Effects of AC power interruption?

[Kiwi Engr]

In the 9100A/B support and public-facing document describing the design and operation, much is made of the non-volatility of the coincident-current magnetic core memory. The core memory retains the program steps, the X, Y, and Z registers, and the storage registers (1-9, and in the B also - 1 to 9, a to f), even after the machine's power is turned off or otherwise interrupted, and indeed will retain this data indefinitely. Moreover, when the power is turned on, the state of these locations is the same as before the power went down. Have I understood all of that correctly?

Is anything else relating to the core memory also stored and recalled? E.g., Error state, carry flag, zero flag, sign flag for the last or current ALU operation?

What happens if a program is running when the power is switched off? When the power is restored, can the program continue automatically, or must it be continued manually by pressing RUN, or must the program be restarted? Is there a risk to data integrity if I cycle the power? Is there power-fail detector hardware in the calculator that hurriedly saves additional status flags before the power fully drains away? How reliable is the storage of machine state across power outages?

Lots of questions! ;-) However, inquiring minds need to know. I may have most of the 9100 documentation but no physical 9100 working or non-working. However, despite searching, I've been unable to find much more than non-volatile memory is a valuable calculator feature.

Are these issues discussed in any of the support documentation? One place I looked at that showed some promise but did not discuss this was the "Pseudo-Hardware Problems" list in the official service manual.

[John Garza (3665)]

The memory and entire machine state are preserved.

I have a 9100B and I tested this by writing a simple counting program that paused the incremented number to the display. I switched it off, counted to ten, and switched it back on. The numbers kept incrementing on the screen after power was restored. Not even a hiccup.

Also, I recall reading in the manual, that if the machine appears locked up on power on, to hit the STOP key, as the machine may be executing a program. So this capability appears to be known to HP.

Still, my gut tells me this is iffy as some parts may achieve 'wake-up-voltage' a split second before others, which could do god knows what to a running program. So I play it safe and only use this as a novelty.

-J

[Kiwi Engr]

Quote:my gut tells me this is iffy as some parts may achieve 'wake-up-voltage' a split second before others, which could do god knows what to a running program.

The implication is that if your machine is old (as they all are!), things like power supply voltage and transistor hFE, etc., might no longer be in spec, changing noise margins and set-up times for various circuits, especially during powering up.

Thank you, John! As you point out, HP mentions pressing the STOP key if the machine is on, the CRT has warmed up, and the screen is blank (meaning it's running a program). This minor mention suggests that they regarded this behavior as entirely usual and not worth the printer's ink. ;-)

[teenix]

(12-21-2024 02:45 AM)John Garza (3665) Wrote:  Still, my gut tells me this is iffy as some parts may achieve 'wake-up-voltage' a split second before others, which could do god knows what to a running program.

I wonder if the same can happen on the Continuous Memory models. The data is pushed into memory as a serial bit stream, so cutting the power at the right time might interrupt it. This might result in errors the next time it is used as the data might not be BCD but have some HEX values instead. The [Clear] buttons would fix it though.

[John Garza (3665)]

(12-21-2024 04:34 AM)Kiwi Engr Wrote:  

Quote:my gut tells me this is iffy as some parts may achieve 'wake-up-voltage' a split second before others, which could do god knows what to a running program.

The implication is that if your machine is old (as they all are!), things like power supply voltage and transistor hFE, etc., might no longer be in spec, changing noise margins and set-up times for various circuits, especially during powering up.

Although in my case, it always works perfectly. This is just a (so far) unfounded fear of mine.
Another possibility is that this behavior is not by design but by serendipity.
Meaning it's not guaranteed to work, but does or usually does. NOMAS, I think it was called.
Maybe we could hear results from other HP-9100 owners out there?

-J

[AnnoyedOne]

I think I last saw "core memory" in a glass case in the early 1980's.

Core memory was/is non-volatile. That said it might be re-initialised at startup by firmware/software.

https://en.wikipedia.org/wiki/Magnetic-core_memory

I'd go by what any HP documentation says and perhaps do your own tests (if/when you have the hardware) to confirm.

A1

[rickb]

I have tested this on both the 9100A and 9100B, with various different programs on both machines, and have always experienced that the programs resume right where they left off when the power was interrupted, without causing any errors or other problems with the programs.
That would imply that all of the state of the calculator is stored in the core memory, including various bits of information that are not accessible to the user. I don't know if this information is stored in an 'off limits' area in the core memory array that isn't accessible programmatically, but is accessible by the microcode. The calculator microcode must sequence the operations in such a way that the state information is maintained in such a way that if the operation of the machine is interrupted, it can be restarted where it left off. Pressing the [STOP] key has this effect, and I suspect that when the power is sensed to be dropping that the microcode essentially immediately executes the equivalent of the [STOP] key quickly enough that everything is static before the power drops off below operating levels. That is made possible by the large capacitors in the power supply. All supposition on my part...I haven't looked at the microcode to see what it does when the power drops, but I'm just guessing.