WP 34S and 31S bugs and fixes

[Ivan Rancati]

(01-18-2016 06:11 PM)pascal_meheut Wrote:  Works fine on my Ubuntu install. Which liinux version do you use?

Kubuntu 15.10. It looks like a Kubuntu specific issue in starting a browser instance. Nothing to worry about, the calculator itself works very well.

Code:

ivan@jannacci:~/bin$ wp-34s/WP-34s &
[1] 8972
ivan@jannacci:~/bin$ KRun(0x7b2720) ERROR (stat): 111   "The file or folder http://wp34s.sourceforge.net/ does not exist."

thanks and best regards

[pascal_meheut]

(01-18-2016 09:03 PM)Ivan Rancati Wrote:  Kubuntu 15.10. It looks like a Kubuntu specific issue in starting a browser instance.

This is indeed what it looks like because I'm using very standard code to launch the browser and open this page and it works almost everywhere.

[Ivan Rancati]

(01-19-2016 06:49 AM)pascal_meheut Wrote:  

(01-18-2016 09:03 PM)Ivan Rancati Wrote:  Kubuntu 15.10. It looks like a Kubuntu specific issue in starting a browser instance.

This is indeed what it looks like because I'm using very standard code to launch the browser and open this page and it works almost everywhere.

Hallo Pascal,
just an idea. I see that
http://wp34s.sourceforge.net/
redirects to
http://sourceforge.net/projects/wp34s/

perhaps the Kubuntu library defaults to not following redirects, so it interprets the http status code "301 - Moved permanently" an error (i.e. anything else than http status code "200 - OK" is treated as an error)

At the next update round, if possible, it might be interesting to replace the url, so it goes right away to http://sourceforge.net/projects/wp34s.

Best regards
Ivan

[efroymson]

This thread looks inactive, but is maybe the right place for this inquiry? I've recently received a new WP-34S and am enjoying playing with it very much, but I notice some weird behaviors. There are certain commands that will cause it to shut itself off. In particular, one that is reproducible is using the inverse Gamma function from the catalog. After reseting the calculator, by the reset command, I tried calculating the Inverse Gamma of 66. It said "Calculating" briefly, then shut off.

I've gotten it to eventually give an answer by doing this enough times (calling the function I mean, without reset), and sometimes it seems like when I switch it back on there is an intermediate value sitting in the X register ...

Does anyone else see this? I can't reproduce it on my iPhone app, FWIW. That works fine.

Thanks,

Robert

[Leonid]

(02-03-2017 11:07 PM)efroymson Wrote:  There are certain commands that will cause it to shut itself off. In particular, one that is reproducible is using the inverse Gamma function from the catalog. After reseting the calculator, by the reset command, I tried calculating the Inverse Gamma of 66. It said "Calculating" briefly, then shut off.

Try with fresh batteries (2*CR2032)

[Paul Dale]

Try two things:

• New batteries
• Set SLOW mode (from the MODE catalogue)
  

The symptoms you mention are likely related to low batteries. Ignore the reading from the BATT command, it is an instantaneous reading generally after a low power mode sleep. A continuing/long computation needs more power and lowers the available voltage. Also, purchase quality brand name batteries, rather than eBay cheapies.

For confirmation, try some of the inverse statistical distribution functions (i.e. the quantile functions). They are probably the most power intensive operations on the device. Inverse Gamma is a better on this front than these. The F distribution is probably the worst -- it should work fine with decent fresh batteries.


- Pauli

[efroymson]

(02-04-2017 08:12 AM)Paul Dale Wrote:  The symptoms you mention are likely related to low batteries. Ignore the reading from the BATT command, it is an instantaneous reading generally after a low power mode sleep. A continuing/long computation needs more power and lowers the available voltage. Also, purchase quality brand name batteries, rather than eBay cheapies.

It did turn out to be the batteries. Since you mentioned it, I did use the BATT command, which reported 2.9V, which seemed close enough to 3V to be ok. That caused me to look for other causes, fruitlessly.

May I make the humble suggestion that the BATT command be altered to actually provide some useful information? As it stands, it is worse than useless. Perhaps it could check the voltage it sees on wakeup, then run a few calculations to get the processor going, and check again? If it notices a drop in voltage, it could report "Replace Batteries" even, since there is an alphanumeric display. Alternatively (additionally?) is there any way for the calculator to know if it has shut down due to a low voltage condition? I thought I saw it say "Bye" at least once when it had the weak batteries. If so, that could be checked by the BATT command, and the same message given.

I don't know enough about the interaction of battery chemistry and processor sleep states to fully specify this, but somebody must. I would classify this as a "bug" and not an "enhancement request" because the information currently provided is actively misleading.

Lest I seem overly grumpy, let me close by thanking you for all your work on this amazing piece of technology. I've been showing it off to everyone lately. Not only has no one called for the men in white coats, but I've gotten some interested responses!

Best,

Robert

[vido]

About bitwise operations, this is not a bug, but would be a useful enhancement for the 21st century : could it be possible to increase the max window from 64 bits to much more than that ? 256, 512, 1024 bits or more could go a long way, nowadays.

[emece67]

(02-07-2017 08:00 AM)vido Wrote:  About bitwise operations, this is not a bug, but would be a useful enhancement for the 21st century : could it be possible to increase the max window from 64 bits to much more than that ? 256, 512, 1024 bits or more could go a long way, nowadays.

Just curious. Why do you need such big operands in a handheld calculator?

Regards.

[Nick]

(02-06-2017 10:41 PM)efroymson Wrote:  May I make the humble suggestion that the BATT command be altered to actually provide some useful information? As it stands, it is worse than useless. Perhaps it could check the voltage it sees on wakeup, then run a few calculations to get the processor going, and check again? If it notices a drop in voltage, it could report "Replace Batteries" even, since there is an alphanumeric display. Alternatively (additionally?) is there any way for the calculator to know if it has shut down due to a low voltage condition? I thought I saw it say "Bye" at least once when it had the weak batteries. If so, that could be checked by the BATT command, and the same message given.

Such a test would consume power and a low battery indicator is built-in (see below).

You can perform the test you're asking for by inserting the BATT command into an intensive program / loop and have it periodically (not continually so that it doesn't pause the workload) update the display with the voltage. That will output the voltage "under load".

Normally, the calculator goes into SLOW mode if the voltage gets low and display a battery icon (not sure if they're tied together). You'll notice this first during intensive calculation then the calculator will return to normal. [Advanced: You can add a check for SLOW in a battery test routine to observe the same. It's not a built-in function like BATT, you'll have to test bit 59 of STOM if you'd like to explore this behavior.]

The calculator remains stable as the batteries go weak and can run in SLOW long before shut-down, so this could just have been bad batteries -- I wouldn't be concerned with it happening again.

[vido]

(02-07-2017 09:38 AM)emece67 Wrote:  Just curious. Why do you need such big operands in a handheld calculator?

How about curiosity ? There are lots of applications for large numbers, why wouldn't it be possible to manipulate such numbers with a handheld calculator ?

[Paul Dale]

Cryptography is the main use for large sized numbers that I know of.

The 34S's binary integers would be ill suited in this role -- they are intended for computer programmers not number theorists or cryptographers. The algorithms used don't scale and the carry/overflow support is quite a burden. The different formats could also become a burden as the size increases.


Pauli

[emece67]

I know what uses such number have. But I'm curious about its use in a machine with an output device as small as the 20b/30b LCD screen.

[efroymson]

I am not concerned about it happening again, now that I know what to look for. I will only point out that what you are saying here seems to contradict what Paul Dale said, about the BATT command not being accurate. Also, I don't recall seeing a low battery indicator on my screen while this behavior was occurring.

It seems to me that I saw somewhere that the voltage the calculator is looking for 2.7V? Mine read 2.9V and still had the problem.

I agree that having the BATT command work as I suggest would consume some power, but it would only do so on the request of the user ... I wouldn't think it a good idea to check automatically at power-on, as it might cause it to immediately switch off again (though that would presumably trigger the immediate thought "dead battery" in the user who could not get it to switch on).

Is this a unique problem to the WP-34S? Is it related to the combination of a wide variety of operations, some of which put a much higher burden on the processor than others? I'm thinking that this essentially could not have happened on an old HP-35 from the 1970s, since it used so much power to operate the display, and the computing part was capped at the level of the trig operations ...

I ramble.

(02-07-2017 02:47 PM)Nick Wrote:  

(02-06-2017 10:41 PM)efroymson Wrote:  May I make the humble suggestion that the BATT command be altered to actually provide some useful information? As it stands, it is worse than useless. Perhaps it could check the voltage it sees on wakeup, then run a few calculations to get the processor going, and check again? If it notices a drop in voltage, it could report "Replace Batteries" even, since there is an alphanumeric display. Alternatively (additionally?) is there any way for the calculator to know if it has shut down due to a low voltage condition? I thought I saw it say "Bye" at least once when it had the weak batteries. If so, that could be checked by the BATT command, and the same message given.

Such a test would consume power and a low battery indicator is built-in (see below).

You can perform the test you're asking for by inserting the BATT command into an intensive program / loop and have it periodically (not continually so that it doesn't pause the workload) update the display with the voltage. That will output the voltage "under load".

Normally, the calculator goes into SLOW mode if the voltage gets low and display a battery icon (not sure if they're tied together). You'll notice this first during intensive calculation then the calculator will return to normal. [Advanced: You can add a check for SLOW in a battery test routine to observe the same. It's not a built-in function like BATT, you'll have to test bit 59 of STOM if you'd like to explore this behavior.]

The calculator remains stable as the batteries go weak and can run in SLOW long before shut-down, so this could just have been bad batteries -- I wouldn't be concerned with it happening again.

[Paul Dale]

Marcus is the battery voltage level expert from the 34S team.
He is also really busy and preoccupied.

Pauli

[Leonid]

(02-08-2017 04:49 AM)efroymson Wrote:  about the BATT command not being accurate. Also, I don't recall seeing a low battery indicator on my screen while this behavior was occurring.

The reason may be due to rising internal resistance battery, so in idle mode voltage is normal >2.7V but in calculation mode critical low and has no time for BATT indication
You can check this (quick!) by current measurement multimeter in short-circuit mode for new and used batteries.

[grsbanks]

(02-04-2017 08:12 AM)Paul Dale Wrote:  purchase quality brand name batteries, rather than eBay cheapies.

This is something that people often don't bother considering.

Battery life isn't the only thing that varies between different battery manufacturers and the quality of their products. Another factor that sensitive electronic devices (such as calculators running at low voltages!) can have an adverse reaction to is the internal resistance of the battery.

Cheapies often have a higher internal resistance than batteries of an established brand because the manufacturers have cut corners in the manufacturing process in order to cut costs. What is the effect of this? It means that the voltage being output by the battery when under load (eg. when performing a computationally intensive calculation) drops more than with a quality battery. This makes them look to the calculator like they're going flat when they're not yet.

Another side effect of this is a direct result of Joule's law: P=UI; U=RI => P=RI²

The power dissipated (in the form of heat) by a resistor is proportional to the resistance (in Ohms) for a given intensity (current, in amperes).

More resistance = more voltage drop and more heat dissipated. Not good for a calculator.

Buy good batteries!

[Nick]

(02-08-2017 04:49 AM)efroymson Wrote:  ...about the BATT command not being accurate.

By itself, that's true as you observed (but not the whole story). The voltage was "enough" yet the calculator "shut down". Let me explain how both can be true at the same time:

Batteries are not only a voltage source, they have an internal resistance. If you were to measure the battery voltage outside of any load, you'll have the open circuit voltage (OCV). That's the highest voltage the battery can obtain under a 0 or near 0 load. As a calculator during idle is near-0, the voltage you read can appear high or sufficient, especially after the calculator has been off for a while.

Under load, the battery's voltage will drop. This is a combination of factors... internal resistance (which is really a net effect measure), temperature, discharge state, recent use, chemistry, abuse, defect, tolerance..... in short, a battery is not a perfect conductor and has limits to how much energy it can output under varying conditions. That's why a battery test routine gives you another measure of the battery's voltage that can be even more useful in determining the state of discharge. It tells you what is the voltage of this battery when the calculator is doing some work (consuming energy). That's a more critical measure of the battery as it indicates relatively how close the battery is to tripping SLOW/low-batt or cut-off due to voltage drop when the calculator is doing something other than idle or sleep.

(02-08-2017 04:49 AM)efroymson Wrote:  Also, I don't recall seeing a low battery indicator on my screen while this behavior was occurring.

Yep, that can also be true at the same time. If the battery had sufficient near-OCV to power on the device (basically just a residual / static charge) but the battery was still "bad", the voltage could drop so rapidly when you used the calculator under sufficient load that it apparently skipped past SLOW and low-batt and went directly to shut-down (it happened before the screen could refresh or just briefly flickered).

(02-08-2017 04:49 AM)efroymson Wrote:  Is this a unique problem to the WP-34S?

No, this behavior is not unique to the WP-34S. We observe it in laptops, cellular phones, and digital cameras. This is not the same thing as firmware / battery management behavior in these devices -- the "battery icon". If you recall early digital cameras that took 4-AA batteries, they would run perhaps 10 minutes on a set of alkaline batteries vs 1 hour on a set of NiMH batteries even though the alkaline batteries had a higher OCV and more energy within them. The critical difference was the lower internal resistance of the NiMH batteries that allowed the device to stay above the cut-off voltage for longer under load.

(02-08-2017 04:49 AM)efroymson Wrote:  I agree that having the BATT command work as I suggest would consume some power, but it would only do so on the request of the user ... I wouldn't think it a good idea to check automatically at power-on, as it might cause it to immediately switch off again (though that would presumably trigger the immediate thought "dead battery" in the user who could not get it to switch on).

Such a test could easily consume more than 1,000x the energy one would have consumed by performing a few interactive calculations. That would substantially shorten the battery life if automated. This is something you can easily explore yourself with a short program -- I've done it. You can store the program in your flash area so that it's fairly permanent. But it's not something I've ever repeated because the calculator is stable. It gives plenty of warning as-is when the batteries are becoming weak during long calculations and appears more frequently as they begin to fail.

That's why I say it's likely that your experience was just the batteries, that it won't repeat with good batteries, and that next time you'll probably observe the power-saving / low battery conditions the unit is already designed with. If the calculator sits unused until the batteries fail (like the new-in-box units or if left in a drawer) and then you turn it on, you might get the same behavior without prior warning. (a built-in or DIY load test couldn't even run to tell you that)

It's not anything I'd be concerned with. The battery management is stable and documented in the 3.1 PDF.

One good spare battery is sufficient to power it in a pinch (or flash it if it's a new, good battery -- I flash without a cable; the flashing procedure can accidentally discharge your batteries as well -- it happens).

Two like batteries in parallel operate with half the internal resistance and you'll be able to extract more energy from the batteries (more than twice what a single battery could provide) and be more stable under a wider range of conditions (ex. temperature). [Another benchmark -- test your batteries independently, one per slot; one could be "bad" or have a dirty connection (which increases resistance)]

[efroymson]

Thanks for the detailed replies. I'm guessing that it may have been the situation discussed where the batteries sit unused for a considerable time, as the HP-30b has been discontinued for a while, and the one used to build mine could have sat unused for a good while. It came with the original packaging ...

I did find that this was discussed a couple years back. That thread also links to a video on the subject of battery life and resistance, for those interested: http://www.hpmuseum.org/cgi-sys/cgiwrap/...ead=250336

(02-08-2017 08:41 AM)Nick Wrote:  

(02-08-2017 04:49 AM)efroymson Wrote:  ...about the BATT command not being accurate.

By itself, that's true as you observed (but not the whole story). The voltage was "enough" yet the calculator "shut down". Let me explain how both can be true at the same time:
*** deleted ***

P.S. In other WP-34S news, I was prompted by the functions under the 4 key to learn some statistics. I watched a video about solving elementary problems with the tables in the back of a stat book, and that helped me understand a bit about the Standard Normal Distribution. I got up to Linear Algebra and Set Theory in Math, but never did take Statistics ...

[Briancady413]

I calculated e^(i*Pi) and got a real component of -1(as should be) plus a tiny imaginary component, which I believe shouldn't be there. This is on an emulator running on ubuntu 16.04.2