Re-calibrating Prime's Battery?

[TheLastMillennial]

Hi, as many of us know, the Prime's battery can be upgraded if you get a higher capacity Galaxy S3 battery. I've had a 2,300mAh battery in my G2 for several months now and the battery life is really great, however the calculator doesn't seem to know how to read the battery level. Nearly every time I turn it on, the calculator 'gains charge' as I use it (i.e. jumps from 25% to 50%). I know many devices have a way to re-calibrate the battery so the reading is more accurate, does the HP Prime G2 have a similar feature?

[HP User]

Until someone else has a better answer, I'd say this: don't count on it.

[Tim Wessman]

There is not any calibration. Also, as far as I know the measurement is based on averaged voltage readings not capacity of any kind so it should be OK i'd think.

It is possible for the measurement to flicker upwards on occasion. Considering the battery will be running you for upwards of a month or two for most people, are you really concerned about this? Just kind of curious.

[TheLastMillennial]

Thanks Tim, it's not my biggest concern, just a slight annoyance so I was just wondering if there was a built in fix. If it's not unexpected behavior then I can live with it.

[eried]

(12-09-2019 03:05 AM)TheLastMillennial Wrote:  Thanks Tim, it's not my biggest concern, just a slight annoyance so I was just wondering if there was a built in fix. If it's not unexpected behavior then I can live with it.

The electric potential does not change with a bigger capacity battery, so there is nothing to calibrate. Laptops and phones keep some statistical health status as the boundaries of that "electric potential" change when the battery wears down (i.e. voltage drops quickier) but remember that the prime only has a 2 bit resolution for the remaining battery

I can plug my battery test video here for you:

[TheLastMillennial]

Thanks for clearing that up eried! I completely forgot to add that the battery wasn't in the best health when I got it. I thought the issue would be because of the bigger battery since my TI-84 Plus CE'S battery is in poor health but it rarely ever underestimates the level. I guess there's something else going on instead.

Have you considered redoing that test for the G2? It would be neat to see how the new battery and CPU effect the battery life!

[eried]

(12-09-2019 04:50 PM)TheLastMillennial Wrote:  Have you considered redoing that test for the G2? It would be neat to see how the new battery and CPU effect the battery life!

Yes, but I haven't bought a G2

[cyrille de brébisson]

Hello,

Battery charge determination is HARD...
In spec sheets, you see nice curves of battery discharges, but the reality is that they are LOTS of battery voltage fluctuations, and not all of them are due to discharge.
Battery load will affect the voltage a lot, so will temperature, so it is "normal" to see battery voltage going up and down as you use the device, especially when the device draw is non constant (as in the case for all devices due to sleep modes)...

Furthermore, the battery discharge is a very flat curve, and CPU ADC are not that good (but, hey, they are cheap, well they have no costs as they are build in the CPU!)... This adds a lot more errors...

I do not know how battery reading is done on the G1, but on the G2, I take 10 ADC readings per second. They are converted in voltages and stored in a ring buffer...
I then remove the min and max readings from these and take the average of the rest. This is my battery reading for the current 100ms... This is then divided into 5 groups, from 100% to Critical battery level and reported upward...
The 'dividing values' for this groups was created using a small program that would display the battery voltage reading every 30 minutes and letting the calc run at full bore (max power use) until it died while filming it. I did that 10 times in a row (if I remember well). This would give me the battery life expectancy, and the ADC readings over time. I picked the ADC reading at 25%, 50%, 75% and 20minutes of the total run time as my jalons.

Furthermore, at a higher level, an anti jiter system ensures that a move from one value to the next one can not occur at less than a given interval. This avoids rapid oscilation from 75% to 100% and the like...

I do not know how computers and other devices report "good" battery level, I would assume that they have a specialized chip which is dedicated to continuously measuring the energy extracted from the battery (not just the voltage), and doing a subtraction with what it "thinks" or "knows" was injected in the battery at charge time... This is of course something that I can not do with just an ADC, especially when said ADC will only work when the device is turned on...
Hence the limited battery level reporting of the Primes (and indeed all the other calcs like the 12C, 10BII+...)

Cyrille

[eried]

(12-10-2019 07:30 AM)cyrille de brébisson Wrote:  This is of course something that I can not do with just an ADC, especially when said ADC will only work when the device is turned on...

I am now a little afraid of my stored Prime home I though there was some kind of BMS inside the galaxy batteries:
   

I think most of the cheap devices use just voltage, and it works perfectly on the Prime. But if the CPU was able to crazy ramp up consumption, that would be another story.

[Jean-Baptiste Boric]

The NumWorks calculator has a similar battery management system. There's a self-contained Li-Ion battery charger chip to prevent over-charging, but the measured battery level is just the MCU's ADC running over a high-impedance voltage divisor. Unlike the HP Prime, the mechanism is extremely straightforward as the result (measured once every ~1/3 seconds) is directly projected onto 4 static battery levels and drawn on the status bar. Some people did notice a slight flickering right at the threshold, but the battery level indicator doesn't stand out and the screen's backlight is the main power consumer by far, so putting a hysteresis between battery levels wasn't a priority.

I believe most calculators share this architecture for power management since it's cheap and usually good enough. The downside is that it's quite imprecise, especially when current draw fluctuates quite a lot (which I assume the HP Prime G2 exhibits more than a calculator with a STM32F7).

[John Colvin]

(12-10-2019 07:30 AM)cyrille de brébisson Wrote:  Hello,

Battery charge determination is HARD...
In spec sheets, you see nice curves of battery discharges, but the reality is that they are LOTS of battery voltage fluctuations, and not all of them are due to discharge.
Battery load will affect the voltage a lot, so will temperature, so it is "normal" to see battery voltage going up and down as you use the device, especially when the device draw is non constant (as in the case for all devices due to sleep modes)...

Furthermore, the battery discharge is a very flat curve, and CPU ADC are not that good (but, hey, they are cheap, well they have no costs as they are build in the CPU!)... This adds a lot more errors...

I do not know how battery reading is done on the G1, but on the G2, I take 10 ADC readings per second. They are converted in voltages and stored in a ring buffer...
I then remove the min and max readings from these and take the average of the rest. This is my battery reading for the current 100ms... This is then divided into 5 groups, from 100% to Critical battery level and reported upward...
The 'dividing values' for this groups was created using a small program that would display the battery voltage reading every 30 minutes and letting the calc run at full bore (max power use) until it died while filming it. I did that 10 times in a row (if I remember well). This would give me the battery life expectancy, and the ADC readings over time. I picked the ADC reading at 25%, 50%, 75% and 20minutes of the total run time as my jalons.

Furthermore, at a higher level, an anti jiter system ensures that a move from one value to the next one can not occur at less than a given interval. This avoids rapid oscilation from 75% to 100% and the like...

I do not know how computers and other devices report "good" battery level, I would assume that they have a specialized chip which is dedicated to continuously measuring the energy extracted from the battery (not just the voltage), and doing a subtraction with what it "thinks" or "knows" was injected in the battery at charge time... This is of course something that I can not do with just an ADC, especially when said ADC will only work when the device is turned on...
Hence the limited battery level reporting of the Primes (and indeed all the other calcs like the 12C, 10BII+...)

Cyrille

I worked on a Coloumb counter design for a cell phone chip when I was an
engineer at Cadence. This circuit measures the charge and discharge currents which then control the rate of count of an up/down counter circuit. This is the
best way of monitoring the true battery charge state that I know of. I suspect that most laptops and cell phones use this method.

[DrDarius]

At present there are chips available that measure the charge flow, and, on top of it, have a battery model built in so they can estimate SOC with reasonable error immediately after the circuit is powered for the first time.
Those are very common and not expensive chips. Maxim makes them, for instance (MAX17048/49, MAX17205, etc.). I used them in many designs with very good results.
Disclaimer: I am in no way affiliated with Maxim. I don't own their shares. I'm just a user of their products.
Measuring voltage of the Lithium Ion family of batteries ans trying to derive SOC out of it is a very inaccurate method because of the shape of the voltage curve.
If you are interested in details, look for "gas gauge". This is a colloquial name for this feature.
Given the price, availability, and accuracy of measurement of those chips, I am surprised that they were not used in Prime. (That said, I have not met a calculator yet that would have precise battery SOC system in place...)

Darius

[cyrille de brébisson]

Hello,

Yep, the chips exist, but as has been pointed out, the calc does not really need that much precision... so they are not used...

What surprised me the most, honestly, is that the power chip associated with the CPU does not have this function/feature. I was kind of annoyed when I found out.... (the CPU relies on an external chip to handle all the different charge pumps and voltage generators used by the LCD, CPU itself, with power coming from USB and/or battery, but also manage batter charge)

Cyrille

[DrDarius]

Cyrille,
You're talking about PF1550... I looked closer at the data sheet, and, indeed, SOC monitor is not there, let alone the more simplistic readout system. Weird.
Well, it is what it is.
As for the accurate battery measurement: indeed, a calculator is not a battery powered medical device, for which the physician has to know exactly how much juice is left in the battery, because if the device suddenly died on him in the operating room, the patient safety would be compromised. But let's have look at the problem from broader perspective.
Calculators, like every other devices, evolve. They get more complex, and this complexity allows for using them for tasks which were unimaginable 20 years ago. Prime, for instance, allows for data collection and wireless communication. (On the very rudimentary level, and mostly for educational purpose, but it may change in the future.) This may (and will) trigger modus operandi which is unconventional from the today's usage standpoint. I already used Prime for tests lasting more than 12 hours during which the calculator was running full speed all the time. Not being sure what to expect, I kept it plugged to the charger most of the time, which was inconvenient.
Have a look at the phones and the cameras. Most of them have very accurate battery measurement systems; at least those from the top shelf, but Prime is considered to be a forerunner in the calculator's world. More on, the battery readout for phones and cameras is always available.
From this standpoint, I believe it would be nice to have a more accurate SOC readout system in Prime. Or at least a built-in function returning battery voltage. (My subjective opinion.)
But this opinion also applies to other Prime features which, as I see it, could be designed differently. And, as I said before, it is what it is.
Disclaimer: Don't get me wrong: I like this product.
Darius

[KeithB]

My HP-71 can be used for data collection, and has no battery warning at all. Neither did the HP41 for that matter. 8^)

[cdmackay]

(12-11-2019 09:35 PM)KeithB Wrote:  My HP-71 can be used for data collection, and has no battery warning at all. Neither did the HP41 for that matter. 8^)

The 71B has a BAT annunciator, which I thought worked (although mine has never come on, I've still on my first set of batteries); doesn't it? I don't believe the AC annunciator comes on though.

And doesn't the 41 also have one? Mine is dead at the moment, so can't check.

[rprosperi]

Both the 71B and 41C family have a BAT annunciator, to give you warning before batteries are fully drained, often leading to data loss.

The 41C is reasonably resistant to data loss unless you blatantly ignore it and keep draining the batteries (especially using the card reader - IIRC, the reader could pass at most 1 card once the annunciator came on). Most often, if you change the batteries shortly after seeing BAT, you will not lose memory.

The 71B is not quite as nice - you will often lose main RAM (though typically not IRAM in ports) fairly quickly if you ignore the annunciator. Once you see BAT, shut it down and do NOT turn it back on until new cells are installed. Plug in the AC adapter and change the 4 x AAA cells; the adapter will maintain memory during the swap. NiCd or NiMH cells can be used in a 71B (but NOT charged there) but the time between BAT and MEM LOST is even shorter with these due to the discharge rates at end of charge life.

So the Prime's system of saving state to flash before each session is really nice! I don't know competitive products well at all, however I'd assume Prime is the only machine in this class that does this, else word probably would have gotten around.

[grsbanks]

(12-12-2019 01:02 AM)rprosperi Wrote:  Plug in the AC adapter...

Speaking of which... One didn't come with my 71B but I do have a rechargeable pack and the adapter that goes with it for HP-41C. The plug on it looks like it should fit. Is it the same adapter? I didn't want to try it just in case it could damage the 71B.

Regarding changing the batteries in a 41C, I have three such machines so there's always a spare battery holder lying around. I make up a battery pack with a spare holder and then all I have to do is whip the pack with depleted batteries out of the 41C and stuff in a pack with new batteries. The operation takes seconds, literally, and no data is lost, even on the 41CL.

[rprosperi]

Yes, the 41 and 71B use the same power adapter. In US, the model is 82059A/B/D or 82040A, the UK flavors are 82067A/B. Unlike most of the previous adapters, this same adapter is used for many HP machines and accessories: 41, 71B, 75C/D, 19C, 97/91/92, IL Printer, IL Tape Drive, IL Interfaces, TopCat battery charger and others.

For the 41, I do the same thing, keeping a spare (actually 3 spare) 41 battery holders, loaded with N cells, ready to swap into a 41 when needed. When changing batteries on a 41CL, I've never lost data, however the first couple times I thought the CL configuration had been lost, and was not sure why. In actuality, all config info had been retained but the MMU had simply been disabled by the battery swap, so a simple MMUEN command restores everything immediately.

[KeithB]

The best low-battery indicator I have on the HP-71 is that it starts registering multiple keystrokes per keypress. 8^)