At Lyuka's suggestion I tested the current draw on the 15C LE and found a problem.

The current draw is identical to the 12C+ in the following states:

OFF - 0.004 ma

ON, idle no keys presses - 0.047 ma

ON, running tight loop - 21 ma

so far so good, but there is a critical difference.

On the 12C+ the current draw when pressing and holding a key averages about 0.4 ma, while on the 15C LE it's a full 21 ma! This is going to substantially shorten the battery life on the 15C LE compared to the 12C+ if you use the calculator a lot.

If you assume that the linger time on a key is 1/10 of a second this roughly means that every 10 key presses is equal to running a program for 1 second and I've guesstimated that you have around 5 hours of long program run time on a set of batteries. That's only 18,000 key presses. A lot less if your fingers are not so nimble.

(Before DaveJ jumps on me for this, I realize that short pulses like keystrokes are not as damaging to the cells as a continuous current draw. However, in the best case I can't imagine cells will last for more than twice my guess.)

I have no idea how hard this is to fix, but it was done in the 12C+ so it should be doable on the 15C LE too. I'm hopeful that this fix will get rolled into a new firmware version that addresses the PSE issue and the self test problem too.

-Katie

Edited: 13 Sept 2011, 11:40 p.m.

Hi Katie,

Thank you for your quick response.

It's interesting, but such amount of current draw is not acceptable for just a calculator.

However, performance / mA is about five times better than that of original 15C.

80x / 21mA : 1x / 1.2mA = 4.6

On the other hand, the current draw while key pressed is more than eight times worse,

so, you must type eight times faster than ever :-)

Regards,

Lyuka

Edited: 14 Sept 2011, 12:16 a.m.

So I guess the batteries won't last 20+ years like they did on my old 15C ? :(

Hi Michael,

If you never pressed the key of the 15C LE, it will last 10 years or so :-)

200mAH * 2 / 0.004mA = 100kH = 4167days = 11.4years

Lyuka

I'm interested in the corresponding current draw in the beta version of HP-15C LE.

I am interested how to measure the current draw. some diagrams or photos would be much appreciated.

Oh... I might have to prepare many batteries.

I don't have a beta test version nor an HHC2010 one. It's all dependent on the emulator and it was probably the same, or potential worse.

true, if both cells are connected in parallel. are they?



hans

It's very simple. Just put a power supply in series with a current meter in series with the battery contacts in the calculator. like this:

The only important thing is to use a current meter with a low voltage voltage burden -- the voltage that drops across its internal current sense resistor.

Edited: 14 Sept 2011, 1:57 a.m. after one or more responses were posted

Thanks, Katie. You're a gem :-)

hpnut in Malaysia.

Quote:

---

true, if both cells are connected in parallel. are they?

---

Yes. This is so on all the ARM-based machines made so far (20b, 30b, 10bii+, 12C+, 12C AE, 15C LE).

Edited: 14 Sept 2011, 2:06 a.m.

That's an impressive desk! Are you an EE, Katie?

I'd like to see the same measurements for the 30b and a WP 34S. My hobby equipment is far inferior to yours.

Edited: 14 Sept 2011, 2:24 a.m.

I'm not an EE but I play one on the internet :)

I'm a computer consultant working for wall street places now, but I started off doing a lot of electronics and early microcontroller work years ago which is when I started accumulating these things. They seem to last forever so you end up with a lot of test equipment (there's more stuff that I pulled off the bench and that you don't see in the picture).

Somewhere on this forum I think I posted current measurements for the 30b, but I have yet to do so for the 34s -- I thought that you had, I guess not.

Edited: 14 Sept 2011, 3:49 a.m.

Interesting.

The next generation of calculators will come with rechargeable cells you will have to recharge every few days. So welcome back in the seventies!

That's technological progress!

Quote:

---

So I guess the batteries won't last 20+ years like they did on my old 15C ? :(

---

Quote:

---

The only important thing is to use a current meter with a low voltage voltage burden

---

Yes, that is VERY IMPORTANT. If you use a typical cheap multimeter, the burden voltage of the meter will result in the voltage to the calculator being too low, and either the calculator won't work at all, or you'll get a possibly inaccurate measurement. The inaccuracy of the measurement is caused by running the calculator at less than the expected voltage. If the calculator does not use a switching regulator, it will probably draw less current at lower voltage (almost proportionate). If the calculator uses a switching regulator in boost configuration, it will draw more current at lower voltage. Either way you're not getting a representative sample for the intended supply voltage.

Note that the burden voltage depends on the current range used on the meter, even with autoranging meters. Rich and I have run into that problem with our calculator prototypes on a handheld DMM. On some ranges the burden is low enough for everything to work OK, but at the lowest current range, necessary for accurate measurements of sleep current (i.e., under 20 uA), the calculator will not work *except* in sleep mode.

I'm now using an Agilent 34410A. The burden voltage on the 100uA and 10mA scales is <0.03V, which is great, but on the 1mA and 100mA scales it is <0.3V, which is barely acceptable.

That looks like an HP 34401A on Katie's bench. It has a burden voltage of <0.1V for the 10mA range, which is reasonably decent, but <0.6V on the 100mA range, which I would consider unacceptable. I use the 34410A rather than a 34401A partly because I wanted lower current ranges than 10mA, and partly for the lower burden voltage.

Inexpensive multimeters often don't even provide specifications for burden voltage.

To really do a good job of measuring very low currents, you need something like DaveJ's µCurrent.

I usually use a calibrated shunt (with max 100mV drop @ full scale of wahtever range I'm interested in) with a storage scope for short pulses (a DMM cannot be trusted for this either). I have never used a DMM for current measurement, as I ususally remove the fuse (having worked with battery installations of 50V to 500V & up to 1000Ah an accidental connection to the current inputs of a DMM can be disastrous).

I often use a 25 years old I-V converter that is made of AD515A to measure leakage current of 100nA-10uA full-scale range.

The most important things to do is a thorough cleaning of the feedback resistors and the summing point terminal.

Lyuka

You could also work around the ammeter voltage burden by measuring voltage with a DVM "after" the ammeter, and adjust the power supply so that the calculator gets 3.3V. In this configuration the error term is equal to the current running through the voltmeter, which can be neglected in this case.

FWIW!
Joel

Lithium cells? My 15C runs on "normal" LR44 batteries; i don't understand the 15CLE picture of the battery compartment...

Quote:

---

...i don't understand the 15CLE picture of the battery compartment...

---

The 15c LE (and 12C+, and 20b, and 30b, and 10bii+) runs on two CR2032 coin cells.

Edited: 14 Sept 2011, 8:51 a.m.

Katie, with my limited equipment it's just not the most fun and accuracy could be better. So I kindly ask you to help me in this regard. Be aware that 34S has an "active idle" mode where interrupts are enabled. This mode is entered with a PSE statement or after the control has returned to the user for half a second before going to deep sleep.

Joel, there is a problem with the varying load having an impact on the actual voltage. An amp meter typically measures the voltage drop over a shunt. With varying current draw this drop varies and you may end up with too high a voltage at the input of your device.

Katie:

Interesting results.

If I recall, CR2032 cells have an internal series resistance of about 15 ohms.

I hope to do some measurements later today on both (15CLE and 15CHHC2010) units.

TomC

I guess HP designers don't use such methods for determining power consumption any more. What would Dave and Bill think?

John

That's quite true if you have a really poor ammeter... But if the power supply is set to a constant 3.3V, then the voltage will never be higher than that and the calculator will be safe. All you have to do is switch down the range on the ammeter when calculator consumption is low.

Joel

Quote:

---

I have no idea how hard this is to fix, but it was done in the 12C+ so it should be doable on the 15C LE too. I'm hopeful that this fix will get rolled into a new firmware version that addresses the PSE issue and the self test problem too.

---

It should be a relatively easy fix for most cases where the
firmware is fixated on keyscan by just dropping the cpu
clock rate during an approximate window containing key i/o
instructions. As the firmware assumes a 4KHz ballpark NUT
instruction execution rate, doing so also helps preserve the
original timing in loop scenarios containing key access
instructions. While a nit to be cleaned up I'm not sure how
serious of a issue it is in practice given the pressumed
key-down duty cycle.

However one interesting related scenario to test would be the
current draw of the unit in OFF mode when a key other than ON
is depressed. Eg, the unit is closed in a drawer or briefcase
in OFF mode but someting depresses a non-ON key over a long
time interval. This is similar to the "remote control falls
between the couch cushions" battery preservation logic.

Quote:

---

I often use a 25 years old I-V converter that is made of AD515A to measure leakage current of 100nA-10uA full-scale range.

---

That's the right way to do it. Although I noticed you weren't
using the offset null. I'm guessing that measurement error was
buried in the noise given the target resolution/gain? Can't
quite make out the voltage divider.

Quote:

---

The most important things to do is a thorough cleaning of the feedback resistors and the summing point terminal.

---

True. :) But even less esoteric is the fact you can easily swamp
a 100nA measurement through a misplaced finger tip on a probe.
For those of us guilty of the occasional hasty measurement..

Quote:

---

That looks like an HP 34401A on Katie's bench. It has a burden voltage of <0.1V for the 10mA range, which is reasonably decent, but <0.6V on the 100mA range, which I would consider unacceptable. I use the 34410A rather than a 34401A partly because I wanted lower current ranges than 10mA, and partly for the lower burden voltage.

---

Good eyes, it is a 34401A. I also have a Fluke 867B and get similar readings on it. The 867B has burden voltage of 0.3V on the 30ma range (the manual says 0.03 volts, I think it's wrong).

Quote:

---

That's impossible, I'm afraid. Lithium cells tend to expire in some years. So even with a ridiculously low current draw, the useful life of the cells would be limited.

---

I've heard opposite claims for shelf life due to the naturally
occurring insulating oxide layer which forms during periods
of light/no use. This reduces cell self-discharge rate and
[reportedly] extends their usable life considerably. But the
only practical issue is whether you can reasonably expect to
find whatever cell you need in the next 5-10-15-20
years.

Quote:

---

If I recall, CR2032 cells have an internal series resistance of about 15 ohms.

---

Herein lies the problem with running these cells at high current. The initial internal resistance is around 15 ohms or lower and you can easily pull 20ma out of them two of them parallel without a voltage drop. But at this current level the internal resistance skyrockets so that at 20ma you start seeing a large voltage drop in short order.

The are no specs for this because these cells are being run so far beyond the manufacturers stated nominal current draw of 0.2ma. They are a terrible choice for a power source in these ARM-based calculators, especially the programmable ones.

Quote:

---

However one interesting related scenario to test would be the current draw of the unit in OFF mode when a key other than ON is depressed. .

---

No problem with this. When the calculator is off the current draw when any key(s) are pressed (other then ON) is just the off state current of 0.004ma.

Quote:

---

Note that the burden voltage depends on the current range used on the meter, even with autoranging meters. Rich and I have run into that problem with our calculator prototypes on a handheld DMM. On some ranges the burden is low enough for everything to work OK, but at the lowest current range, necessary for accurate measurements of sleep current (i.e., under 20 uA), the calculator will not work *except* in sleep mode.

---

I've been guilty of this namely leaving a DUT connected to a meter
set to a 100uA scale, long forgetting about it, and then wondering
why the supply rail is dipping wildly at unpredictable times.
As a bandaid stopgap, a low forward voltage diode across
the meter may clamp the voltage drop to an acceptable level.
You won't get a meaningful measurement when it does but the
meter will be overscale at that point anyway and have your
attention.

The better solution as pointed out is a precision shunt and offset
nulled instrumentation amp. I might toss a uC with A/D in
there such that I can collect continuous data synchronized with
events of interest. With a bit of finesse the gain of the opamp
could be adjustable during acquisition to keep the sample
within digitization range.
Most SoCs have suitable precision & range A/Ds
available but self-sampling above/below its own vdd/gnd is
involved.

Another back burner project to join all of the others..

That's just massively disappointing, and bad engineering.

Cyrille at HP knows about the problem after I bought it to his attention >2 years ago, and he said he would not make the same mistake again in future calculators based on the same platform. Looks like he either forgot, or there is some other reason for them doing it.

I've posted it countless times, but I'll post again, technical details for those interested:

http://www.eevblog.com/2009/04/16/eevblog-4-low-power-calculator-design-and-fpgas/

Dave.

Quote:

---

That's just massively disappointing, and bad engineering.

---

I think that's a bit harsh of an external assessment on the developers.
Yes there are bugs to be stomped out as those here have
found. And there appears to be some regression in power management
optimization. But I suspect the developers may as well be
victim in part of a less than comprehensive beta program,
given the series of issues quickly surfacing in this forum.

I'm unsure what HP's business aspirations are for this product.
Perhaps it is diabolically strategic and they are out to reseed
the remaining voyager crony market with 10K units of the 15C LE,
after which a cosmetically modified 15C+ LE pops up and they
continue to sell that until the market is saturated.
Or perhaps this was a largely altruistic effort where they
decided to give something back to their loyal fan base who have
been forever banging on the table for the 15c to be reintroduced.

If the latter, I believe we risk teaching HP the lesson that
no good deed goes unpunished and assuring they don't forget the
payback received for the 15C LE.

I get the feeling that your last point will ring true.

Unlike the "faceless nameless" period when HP shut down all calculator operations, this project, the 15C LE, involves people whom we actually know--and even in many cases know personally, face to face!

I'd say this is an *enormous* turnaround for such a huge corporation. Really, really have to be pleased.

Take your mind back to early 2000s when they pulled the plug. Remember what you were thinking then? Remember scrambling to find and NOS you could? (I bought 4 20s machines from Samson cables for $20 a piece--just to have pioneer keyboard quality backup!)

This is, really, a Dream Come True. Really.

Rather extraordinary.