accurately measuring integrated power consumption

[whit3rd]

On Friday, May 31, 2019 at 2:30:35 AM UTC-7, Winfield Hill wrote:

What's a good instrument (need make & model number)
to measure the integrated power consumption of my
complex micro-controller controlled bee-hive monitor?
It rapidly goes in and out of multiple modes, turns
74 sensors on/off, pulses LEDs to 50mA, draws 65mA
to heat micro-miniature hot plates, etc.

Well, one way is to buy lots of batteries. The old Tadiran lithium cells for computers (Macintosh 6100)
all went flat about three years from purchase date, and from the manufacturer's
data I estimated the clock drain at 36 uA. With some sheets of brass, I shunted the current through
a VOM, and sure enough, it read... 36 uA.

So, within a few percent, you could drain a fresh battery to calibrate, then after a month or year of
hive instrument operation, yank the used battery and drain it, noting the mAh difference.

The battery chemistry really IS a good, reliable indicator of current drawn No little spikes or
slow leaks will get past the chemical change record.

 

[Winfield Hill]

Jasen Betts wrote...

On the bench. Thanks for the suggestion, George,
but with rapidly-changing, and pulsing currents,
we need an integrating ADC, rather than SAR type,
with millions of samples, some missing the pulses.

https://en.wikipedia.org/wiki/Source_measure_unit

I have several nice Keithley SMUs, maybe the answer
is to code a sampling-integrating mode (they connect
to your computer using Java). But then, why not use
an HP 34970A logger? Hey, maybe I should ask Legg
about that, he's the expert with those machines.


--
Thanks,
- Win

 

[Winfield Hill]

Tom Del Rosso wrote...

Winfield Hill wrote:
What's a good instrument (need make & model number)
to measure the integrated power consumption of my
complex micro-controller controlled bee-hive monitor?
It rapidly goes in and out of multiple modes, turns
74 sensors on/off, pulses LEDs to 50mA, draws 65mA
to heat micro-miniature hot plates, etc. Average
current is from 0.5 to 3mA, but is hard to measure.
It spends much of its time drawing microamps, and
this must also be accurately integrated over time.

What's wrong with the coulomb meter in the book?

Nothing, it's very good, but I'd have to build one.


--
Thanks,
- Win

 

[Winfield Hill]

bloggs.fredbloggs.fred@gmail.com wrote...

On Friday, May 31, 2019 at 2:32:46 PM UTC-4, Winfield Hill wrote:
bloggs.fredbloggs.fred@gmail.com wrote...

Cheapest and fastest would be an eval board for one
of numerous coulomb counter ICs. They all seem to use
a voltage-frequency-converter (VFC) with each pulse
representing a discrete unit of charge ...

Sounds good, have a part number, to get me started?

The eval board and software for the DS2740K looks like the
simplest and most self-contained, but as usual Maxim fails
on ordering information.

Thanks, Fred! Aha, so the secret was "coulomb counter", this
unleashes what I was looking for. Maxim suggests the MAX17211
as a replacement, but I prefer the MAX17201 I2C version. Eval
versions for both are available.

Analog sells the eval board for their LTC4150 as p/n DC756
in stock at $50, but it's an I2C interface requiring a DC590B
USB controller for another $50 with no availability until July.

This looks better, although the 100uV offset spec is a huge
0.2% of 50mV full-scale, a real problem with my high dynamic
range. But I guess it can be trimmed. Too bad they don't
give an offset drift spec. Hopefully it's 100x lower.

Arrow has DC590B in stock. I can hook the output pulses up
to a counter. Yeah, no software or programming required!


--
Thanks,
- Win

 

[Winfield Hill]

Klaus Kragelund wrote...

This one seems more advanced and easier to use with any target:

STM32_power-shield, p/n X-NUCLEO-LPM01A $70 in stock.
Yes, that looks very good as well, ready to go.


--
Thanks,
- Win

 

[Winfield Hill]

whit3rd wrote...

On Friday, May 31, 2019 at 2:30:35 AM UTC-7, Winfield Hill wrote:
What's a good instrument (need make & model number)
to measure the integrated power consumption of my
complex micro-controller controlled bee-hive monitor?
It rapidly goes in and out of multiple modes, turns
74 sensors on/off, pulses LEDs to 50mA, draws 65mA
to heat micro-miniature hot plates, etc.

Well, one way is to buy lots of batteries. The old
Tadiran lithium cells for computers (Macintosh 6100)
all went flat about three years from purchase date,
and from the manufacturer's data I estimated the
clock drain at 36 uA. With some sheets of brass, I
shunted the current through a VOM, and sure enough,
it read... 36 uA.

So, within a few percent, you could drain a fresh
battery to calibrate, then after a month or year of
hive instrument operation, yank the used battery
and drain it, noting the mAh difference.

The battery chemistry really IS a good, reliable
indicator of current drawn No little spikes or
slow leaks will get past the chemical change record.

Yes, a good suggestion indeed.


--
Thanks,
- Win

 

On Saturday, 1 June 2019 11:41:12 UTC+1, Winfield Hill wrote:

How about an ultracapacitor (bypassed with something smaller) in the
feedback loop of an op-amp to make a precision integrator?

John

 

[Winfield Hill]

jrwalliker@gmail.com wrote...

On Saturday, 1 June 2019 11:41:12 UTC+1, Winfield Hill wrote:

How about an ultracapacitor (bypassed with something smaller) in
the feedback loop of an op-amp to make a precision integrator?

Yes, John, right. So many easy ways to do it by designing
and building something. I was hoping for an instrument we
could purchase, with a few controls and a mAh display.


--
Thanks,
- Win

 

[Clemens Ladisch]

Winfield Hill wrote:

It spends much of its time drawing microamps, and
this must also be accurately integrated over time.

Most of TI's MSP430 eval boards have "EnergyTrace",
a software-controlled DC/DC converter that counts
charging pulses. Google says there is a standalone
mode where you do not need to debug an MSP program.


Regards,
Clemens

 

[Sjouke Burry]

On 01.06.19 18:13, jrwalliker@gmail.com wrote:

On Saturday, 1 June 2019 11:41:12 UTC+1, Winfield Hill wrote:

How about an ultracapacitor (bypassed with something smaller) in the
feedback loop of an op-amp to make a precision integrator?

John

Those caps have an awful memory effect.
They can store a lot of charge, but with a huge error because of memory.

 

On Sunday, 2 June 2019 04:23:58 UTC+1, Sjouke Burry wrote:

On 01.06.19 18:13, jrwalliker@gmail.com wrote:
On Saturday, 1 June 2019 11:41:12 UTC+1, Winfield Hill wrote:

How about an ultracapacitor (bypassed with something smaller) in the
feedback loop of an op-amp to make a precision integrator?

John

Those caps have an awful memory effect.
They can store a lot of charge, but with a huge error because of memory.

Thanks. That's good to know.

John

 

[Chris Jones]

On 31/05/2019 19:30, Winfield Hill wrote:

What's a good instrument (need make & model number)
to measure the integrated power consumption of my
complex micro-controller controlled bee-hive monitor?
It rapidly goes in and out of multiple modes, turns
74 sensors on/off, pulses LEDs to 50mA, draws 65mA
to heat micro-miniature hot plates, etc. Average
current is from 0.5 to 3mA, but is hard to measure.
It spends much of its time drawing microamps, and
this must also be accurately integrated over time.

Going through my considerable collection of smart,
highly-capable instruments, there's not one that
will make continuous, accurate integrated-current
power-consumption measurements. The closest are
awesome 6.5-digit bench multimeters, that can be
set to measurement integration times of 100 PLC
(about 1.6 seconds), but these are discontinuous
measurements, with gaps, not suited for the job.

Yes, there are fuel-gauge ICs available, but we
need to purchase an instrument, with a wide
dynamic range / peak handling capability, plug it
in, and take readings, so we can experiment as we
work, programming the beehive-monitor's software.

Not even my new awesome-looking Keysight E36312A
data-logging bench supply can do this simple task.
The word, "integrate" doesn't appear in the manual.
Nor does the word, "average" appear.

OK, I have this little $15 USB stick, with a cute
display, that includes a running mA-hr value. If
I open it up and do some cut-jump hacking, maybe
it could be integrated into my bee-hive monitor's
power system. What's its peak-handling capability,
who knows. Sheesh!

Maybe my 66321B "Mobile Communications DC Source",
although the word, "integrate" doesn't appear in
its manual either. OK, it will "average" discrete
measurements. Not very re-assuring. It can fill
up a buffer with 2048 data points... Arrghh!

I'm not sure whether this thing is any good, but might be worth a look:
https://www.joulescope.com/
Reviewed here:
http://www.ganssle.com/tem/tem368.html#article4

other alternatives:
http://www.ganssle.com/tools.htm#current

You could low-pass filter the current waveform (e.g. put a big capacitor
across the meter), then it would not be necessary for a DMM to have an
integrating characteristic, it would just have to sample the waveform
faster than twice the highest frequency component present in the
filtered waveform.

Another idea: use a fast log converter circuit (e.g. based on p-n
junctions) and digitize the output of that with a low resolution but
fast ADC such as a DSO. Then you can figure out the current at each time
point with very high dynamic range, though only modest accuracy. That
should still get you the total coulombs within a few percent, over pA to
mA with no range switching.