11 replies to this topic

[MJim]

A question occurred to me, and that is with very low currents in uA's, does the electrolytic capacitor's current leakage while charged divert much needed current away from the CPU at low light levels?

 

I recently bought a Classwiz fx-991EX (I liked it so much, I'm purchasing more of them).  I've done some basic experiments on the Solar cell power characteristics (removed the 991EX's 22uF electrolytic buffer before starting), in which unsurprisingly I found a large amount of light is needed to do difficult and long calculations purely on Solar energy (~600+ lux, which is about the light from being near a window on a sunny day).  Interestingly though, the fx-991EX seems to have some type of power management control built in, so while the display disappears completely when the lux is lower than about 300 (with an indoor 9W LED, it seems to do better with natural light), both the history and variable storage seems to be retained down to about 5-6 lux (~3.5 metres from a 9W LED lamp at night), so while not usable at lower light levels, memory is retained at much lower levels than any other solar cell scientific I have (usually once I lose the display due to insufficient light, quite often the memory contents go, but this is not the case with the fx-991EX).

 

It makes sense to me for a purely solar powered calculator like the Casio fx-260II Solar to have that buffer, or dual powered calculators that retain the history contents on power off (eg Sharp EL-W516X), or programmable calculators (eg Casio fx-3650P), to give a bit of time to replace the battery.  But since the fx-991EX doesn't really keep it's history (well that isn't strictly true, statistics, vectors, matrices and equation variables in equation mode are retained on power off), it isn't clear how necessary it is.

 

I have an older fx-3600PV and fx-50F, both of which don't have electrolytic capacitor buffers, but are functional even down to about 27lux or so (though the fx-3600PV isn't really readable at this level of light).  Still it is interesting that these 2 programmable models from the late 1980's didn't bother with a capacitor buffer.

 

One unintended advantage to removing the electrolytic capacitor on the fx-991EX, is that the solar power indicator seems much more accurate and responsive.  For example ~600+ lux (No battery, just on solar power) it can perform a long calculation, but the display does fade (and calculation time slows);  With a battery installed and under the same light conditions I get a slow flickering on/off solar power indicator during calculation, so that seems to fit with what I see under solar power only.  Before with the electrolytic capacitor, the solar power indicator would be solidly lit even under 27 lux which is completely insufficient to power the display, and only enough to maintain the memory contents, while now under similar light conditions the solar power indicator is almost always off.

 

Anyone know whether electrolytic capacitor leakage current could draw a significant portion of the miniscule current supplied by solar cells at low light levels, or in other words, does removing the electrolytic capacitor buffer allow a solar powered calculator to work under even lower light levels than it would with the capacitor?

 

I'm not an electronics person (I just know some basics, though I must admit even some of those basics I know could be wrong), so if what I bought up isn't right for a number of reasons, it would much appreciate an explanation to correct my misunderstandings!

 

I love solar powered scientific calculators, which is why this post is so long, obsessive and wordy

Edited by MJim, 03 December 2020 - 07:15 AM.

[KamimuraCal]

One of the most underestimated problems of the energy supply is the power that two of the most important components draw, namely the display and the CPU or SoC. This fact becomes all the more sensitive in a low power circuit. An LR44 alkali-manganese cell with only 1.5 V ensures that the memory contents remain secure in any case. The buffer capacitor plays a double role here, on one hand it should hold the operating voltage as long as possible and on the other hand it should provide an electrical charge for a short time as a bridging. The entire circuit in the fx-991 series, as in similar models, was designed for a permanent dual power supply. It was therefore essential to create an intelligent control system for the two hungriest components. In the event of an undersupply, the display is switched off first, then the clock frequency of the CPU is gradually reduced, because this also draws a lot of current, depending on the clock rate. This also means that in the extreme case, a running process would be stopped in order to protect the memory content.

 

The device wasn't designed to operate in the worst light conditions, so this doesn't correspond to any natural use. Casio tries to protect the environment more or less with almost all models, these are the compromises that a manufacturer has to make.

 

As a side note:

A comparison with the HP 50g shows why design philosophy matters. This device was designed across the board for data security and robustness. This is also shown by the power source, which consists of 4 × AAA + 1 × CR2032 batteries. The power consumption is correspondingly high, which allows some conclusions to be drawn about the hardware used. This makes it one of the most powerful devices ever built. This comparison is only intended to illustrate how different concepts and objectives will ultimately work out.

Edited by KamimuraCal, 03 December 2020 - 05:17 PM.

[Hlib2]

@ MJim

Anyone know whether electrolytic capacitor leakage current could draw a significant portion of the miniscule current supplied by solar cells at low light levels, or in other words, does removing the electrolytic capacitor buffer allow a solar powered calculator to work under even lower light levels than it would with the capacitor?

In fx-991DE_X, this capacitor allows you to keep the operating mode for about 30 minutes in the absence of the main battery and in complete darkness. Since this is less than the 991DE_X auto-shutdown time, it is sufficient to resume lighting to continue working without data loss. The voltage from the battery goes to the " + " of the capacitor via a Schottky diode, and the capacitor itself is charged from the solar cell up to 3.3 volts max. (depending on the light source). Therefore, even in low light, the calculator no longer consumes the main battery.
I had an idea that if you power the calculator at the point where the capacitor is connected, removing the main power, the auto-off function will stop working. But I haven`t had time to check it out yet.

[Hlib2]

@ KamimuraCal

A comparison with the HP 50g shows why design philosophy matters. This device was designed across the board for data security and robustness.

In terms of data security and reliability, the hp-50g is not the best solution. Remove the main batteries from it for a few seconds, and the system time and date are erased in the calculator, and CR2032 does not save the situation.
Try to simultaneously remove the main batteries and the CR2032 backup cell from the afx-2.0 calculator while running the program. If the power is resumed even after five minutes, the calculator will continue the program from the interrupted position.

[MJim]

One of the most underestimated problems of the energy supply is the power that two of the most important components draw, namely the display and the CPU or SoC. This fact becomes all the more sensitive in a low power circuit. An LR44 alkali-manganese cell with only 1.5 V ensures that the memory contents remain secure in any case. The buffer capacitor plays a double role here, on one hand it should hold the operating voltage as long as possible and on the other hand it should provide an electrical charge for a short time as a bridging. The entire circuit in the fx-991 series, as in similar models, was designed for a permanent dual power supply. It was therefore essential to create an intelligent control system for the two hungriest components. In the event of an undersupply, the display is switched off first, then the clock frequency of the CPU is gradually reduced, because this also draws a lot of current, depending on the clock rate. This also means that in the extreme case, a running process would be stopped in order to protect the memory content.

 

The device wasn't designed to operate in the worst light conditions, so this doesn't correspond to any natural use. Casio tries to protect the environment more or less with almost all models, these are the compromises that a manufacturer has to make.

It sounds like you have an even better idea about how the power management works than I do (I only found out about it via testing myself).  I think though that Casio did put some effort into designing a power system to deal with worst case conditions, otherwise I don't see why they would bother with designing a power management system at all if that was the case, just have a simple switch (normally 2 diodes I think) like the one they used with their past solar models. 

 

Do you have any more information on how this power management works...I'm tempted to perhaps put my Oscilloscope to the task, but I heard others suggesting that the clock is set internally, so I wouldn't be able to measure it.

 

@ MJim
In fx-991DE_X, this capacitor allows you to keep the operating mode for about 30 minutes in the absence of the main battery and in complete darkness. Since this is less than the 991DE_X auto-shutdown time, it is sufficient to resume lighting to continue working without data loss. The voltage from the battery goes to the " + " of the capacitor via a Schottky diode, and the capacitor itself is charged from the solar cell up to 3.3 volts max. (depending on the light source). Therefore, even in low light, the calculator no longer consumes the main battery.
I had an idea that if you power the calculator at the point where the capacitor is connected, removing the main power, the auto-off function will stop working. But I haven`t had time to check it out yet.

Thanks for this, this actually points out something I wasn't aware of.  I grabbed the specs of the Trony SC1535I solar cell from a datasheet, but the voltage was considerably lower (~2V), but that was tested at 200 lux, but it didn't click that the voltage could be a lot higher than this under better light conditions.  It sounds like you tested it yourself and it goes up to 3.3 volts.

 

In addition it helps to have a bit of a reality check.  It could be possible that removing the capacitor would mean retaining the contents under lower light conditions, but having 30 minutes buffer at 0 lux is much more useful; at least for the fx-991EX which needs a much larger amount of light to operate, so the current leakage likely doesn't matter for this calculator.  Though I would be interested to see how it might affect something like the fx-260 II solar which can still operate at much lower light levels (~20 lux looking at the EEVBlog video review).

 

I was wondering about reducing the power myself to disable the auto-off function as well (I didn't test to confirm that auto-power off was disabled, but I did hypothesise that it did), but it looks like your way ahead of me there

 

Thanks KamimuraCal and Hlib2 for your comments, they have both been helpful.

 

EDIT: By the way Hlib2, does the fx-991DEX allow solving for variables in integrals/differentials?  I heard the Japanese version doesn't have this feature in their solver, but I was wondering if the German version doesn't as well.

Edited by MJim, 04 December 2020 - 12:29 AM.

[KamimuraCal]

@ KamimuraCal
In terms of data security and reliability, the hp-50g is not the best solution. Remove the main batteries from it for a few seconds, and the system time and date are erased in the calculator, and CR2032 does not save the situation.
Try to simultaneously remove the main batteries and the CR2032 backup cell from the afx-2.0 calculator while running the program. If the power is resumed even after five minutes, the calculator will continue the program from the interrupted position.

 

The system time and date are considered to be an ongoing task and can easily be restored by entering them manually. It should be understandable that HP has waived further measures in the circuit technology here and the purpose of a memory backup battery should be clear. If you read and follow the relevant information in the operating manual, you shouldn't have any problems. This point should be seen as a strength rather than a weakness!

 

As far as the Algebra FX 2.0 is concerned, it's a completely different animal compared to the HP 50g, the entire circuit, including all components, is significantly smaller and much more economical. In this respect it's no wonder that the buffer capacitor still holds the voltage even after 5 minutes. This means that the Algebra FX 2.0 is neither safer nor more robust. But here, too, exactly the same rules apply as for the HP 50g, the batteries should never be removed from the calculator at the same time. (Please read hier on page 25!)

 

While the energy consumption of the Algebra FX 2.0 is comparatively low at 0.2 W, it's 0.4 W - 0.6 W for the HP 50g, depending on the firmware. A comparison on a technical level therefore makes no sense, Casio pursues a completely different objective than HP!

Edited by KamimuraCal, 04 December 2020 - 05:04 AM.

[Hlib2]

@ MJim

...does the fx-991DEX allow solving for variables in integrals/differentials? I heard the Japanese version doesn`t have this feature in their solver, but I was wondering if the German version doesn`t as well.

Yes, it allows solving. For example, 1÷A + S(√x,dx,2,B) = C SOLVE, it can calculate the value of B when the constants A, C and the variable x are set. As well as d/dx(ln(x+B)÷2) = 0.04 | x=5, SOLVE. The answer is: B=7.5
@ KamimuraCal

As far as the Algebra FX 2.0 is concerned, it`s a completely different animal compared to the HP 50g, the entire circuit, including all components, is significantly smaller and much more economical. In this respect it`s no wonder that the buffer capacitor still holds the voltage even after 5 minutes. This means that the Algebra FX 2.0 is neither safer nor more robust.

I had in mind other aspect: an emergency power failure in afx-2.0 does not lead to any negative consequences, and does not even disrupt calculations. This indicates the reliability of the system. In some electronic products CASIO applies excellent solutions in the field of power supply and data protection from various failures. In this regard, the super-economical Classwiz series is very well executed, in which dual supply is not a marketing gimmick, but a very important addition.

Edited by Hlib2, 04 December 2020 - 08:28 PM.

[MJim]

@ MJim
Yes, it allows solving. For example, 1÷A + S(√x,dx,2,B) = C SOLVE, it can calculate the value of B when the constants A, C and the variable x are set. As well as d/dx(ln(x+B)÷2) = 0.04 | x=5, SOLVE. The answer is: B=7.5

So the German version is even more impressive than the Japanese version

 

I would love to get hold of one, but they are very expensive, ie the cheapest 991DEX on ebay will cost me ~77 NZD, while the 991EX on Amazon costs ~23 NZD (both excluding freight costs), and realistically I only would want one for it's 1 hour auto-power off feature (although I think it supports RREF function for matrices which is also a nice feature to have). 

Edited by MJim, 04 December 2020 - 11:52 PM.

[KamimuraCal]

@ KamimuraCal
I had in mind other aspect: an emergency power failure in afx-2.0 does not lead to any negative consequences, and does not even disrupt calculations. This indicates the reliability of the system. In some electronic products CASIO applies excellent solutions in the field of power supply and data protection from various failures. In this regard, the super-economical Classwiz series is very well executed, in which dual supply is not a marketing gimmick, but a very important addition.

 

I know what you mean, Casio has a different approach to the circuit design than some other companies. That's also the reason why the fx-CG50 was equipped with only one set of batteries despite its energy consumption of 0.6 W. Unfortunately, without certain rules, calculators are still more or less vulnerable to possible data loss.

[MJim]

I removed the solar cell and measured the current and voltage draw under different conditions with just the battery (all tested for 5 minutes for the average unless otherwise stated):

 

Calculate Mode (IDLE - Nothing typed on display, what you see after pressing the ON button in Calculate mode):

Voltage: 1.577v

Current: 29.5uA (min: 22.1uA, max: 87.1uA, avg: 29.5uA)

 

Menu Screen (IDLE):

Voltage: 1.574v

Current: 51.1uA (min:39.9uA, max: 71.2uA, avg: 51.1uA)

 

Calculate Mode (Navigation - scrolling back and forward through an equation, only done for 1 minute):

Voltage: 1.549v -> 1.574v

Current: 120.1uA (min:26.4uA, max: 285.3uA, avg: 120.1uA)

 

Calculate Mode (Running the SOLVER function that contains an integral):

Voltage: 1.547v -> 1.549v

Current: 235.9uA (min:225.2uA, max: 245.4uA, avg: 235.9uA)

 

I have recorded occasional peaks of around 300uA (which didn't come up during the actual tests, but during navigation and test setup), which at around 1.55v, corresponds to ~465uW, which is well under Casio's stated maximum power draw of ~600uW; however it isn't likely you will be consuming this much power even under a heavy calculation load (~250uA seems a bit more likely which is ~400uW).

 

Power off current draw is low as expected (around 40x less then the idle calculate mode current draw):

 

Calculator off:

Voltage: 1.580v

Current: 0.7uA (min: 0, max: 2.5uA, avg: 0.7uA)

 

I have just started testing the solar cell (daylight from window), and have come to understand the additional importance of the power management on the fx-991EX. I started off by simulating the fx-991EX load while under a heavy calculation load (I used a 5.6k ohm resistor, which should be roughly right for the maximum performance draw 1.55V/300uA ~5.2k ohm).  These are some very rough initial results using my Galaxy S7's built in lux meter:

 

5.6k ohms in series with solar cell:

650 Lux, 1.04v, 185uA

2100 Lux, 1.72v, 305uA

 

From these first two results, it looks like for maximum performance without clocking down, requires ~2000 Lux on the solar cell.  I then tried reducing the light even further:

 

5.6k ohms in series with solar cell:

600 Lux, 0.89v, 161uA

400 Lux, 0.59v, 104uA

200 Lux, 0.26v, 48uA

100 Lux, 0.11v, 20uA

 

This is when I added an additional 150k ohm resistor in series (~156k ohms total), as I was wondering if the reason for the voltage being so low despite the specs on the datasheet was due to a good deal of energy being dissipated inside the solar cell due to it's increasing series resistance under low light.

 

150k + 5.6k ohms in series (~156k ohms) with solar cell:

100 Lux, 2.05v, 13.8uA

 

That is a much better result (much more power delivered to the load); so my understanding is that by reducing the power consumption by clocking down the CPU and turning off the display, the increase in series resistance of the calculator circuit means that more of the low light solar energy is delivered to it, rather than wasted by being dissipated in the solar cell.

 

Anyway my phone lux meter isn't likely accurate, but I am interested in testing to see how it deals with the 9W LED light at night.  I think this should be easier to control (clouds covering the sun make daylight readings fluctuate to much).

 

EDIT:

I decided to simplify the testing for an indoor light, by setting up the solar panel at horizontal distances of 250cm and 300cm from the 9W 806 Lumen LED light.  I had tested these locations before (when I removed the electrolytic) and had found that at the 250cm point, both history and variable memory were retained, while at 300cm only variables were retained and the history was lost.  I also increased the dummy load to 2M ohms (corresponds to roughly 1.55v/0.7uA or the off state on the calculator).

 

The LED light is marked as a 9W @ 806 lumens at a temperature of 2700K; I assumed that it was emitting light in a hemisphere (none above the lamp, just below).  I also kept the solar cell parallel with the floor since this is pretty much how it sits in the calculator.  This also means that the panel is not perpindicular with the light source, so receives less light as a result, but I tried to calculate roughly the lux taking that into consideration.

 

250cm Horizontal distance (335.67cm radial distance, ~5.07 lux) - Variables and History retained:

1.239v, 0.73uA (2M ohm dummy load)

 

300cm Horizontal distance (374.40cm radial distance, ~3.28 lux) - Variables retained, History LOST:

1.15v, 0.66uA (2M ohm dummy load)

 

Considering the power off voltage and current is 1.55v @ 0.7uA, it looks like Casio could of possibly retained calculation history in the power off state with barely any change to the standby power drain.  Even if Casio upped the standby current drain to 2uA to make sure that calculation history was retained, that would still get ~9 years on the battery (Maxell SR44).

Edited by MJim, 06 December 2020 - 08:49 AM.

[Hlib2]

@ MJim

...so my understanding is that by reducing the power consumption by clocking down the CPU and turning off the display, the increase in series resistance of the calculator circuit means that more of the low light solar energy is delivered to it, rather than wasted by being dissipated in the solar cell.

Indeed, this is how the calculator works. I removed the main battery and measured the calculation time of the expression S(e^(1÷x);1;5)+Ran# depending on the amount of illumination. Ran# allows you to see if a new calculation has occurred after the display is turned off for the duration of the operation. The calculator showed results within 0.5...5 seconds, which confirms the dependence of the CPU frequency on the solar cell output impedance.

[MJim]

@ MJim
Indeed, this is how the calculator works. I removed the main battery and measured the calculation time of the expression S(e^(1÷x);1;5)+Ran# depending on the amount of illumination. Ran# allows you to see if a new calculation has occurred after the display is turned off for the duration of the operation. The calculator showed results within 0.5...5 seconds, which confirms the dependence of the CPU frequency on the solar cell output impedance.

I didn't fully test myself if calculations were still running, or whether at a low enough light level they simply halted waiting for more light before completing, did you find that at all light levels it would still finish given enough time?

(I'm being lazy here, as I should test this myself, but if you have already got the results then I don't need to look at it myself).

 

I think I could still use the keyboard once the screen turned off, though once light got low enough even that wouldn't work anymore.