Low power consumption

[Eugeneo]

Chip: 16LF88

I've search the forum for low power consumption with limited results.

I've got my whole circuit running at about 12-20 micro amps in sleep. I've ditched the LDO regulator since it consumes too much power. I've noticed by setting the pins to float it takes more current than outputting them low. A P chan fet is used to power all the extras when the chip wakes up. The fet takes less than 1ua with the gate drive high. I've turned of the hardware usart, timers and everthing that consumes power. Only 2 internal pullups have been enabled for interrupt to wake (since it takes power to enable the pullups external ones maybe an option). Every component is low voltage tolerant. What kind of life can I expect to see if I use 3 x 700ma/h Nimh batteries?(considering the device never powers up). Should a regulator even be used since most of the energy in the battery is gone by the time it reaches 3v. Is it a good idea to go to a single cell boost regulator setup?

[Gerrit]

Hi Eugeneo,

How long do you want it to be alive?

(self discharge battery)

http://en.wikipedia.org/wiki/Nickel_metal_hydride

"Nickel metal hydride batteries have a high self-discharge rate of approximately 30% per month and more"



Gerrit

[C Turner]

When trying to minimize power, NEVER set ANY pin to "input" mode if there is even the slightest chance that it can float: Unless it is tied high or low through a resistor or due to being strapped to Vss or Vdd, Murphy says that it will find an indeterminate state and hang there, causing extra current to be pulled. (An pin connecting to a CMOS, FET or other insulated gate device is, for all practical purposes, floating!)

A friend of mine found this out the hard way when trying to build a device to deploy a model rocket chute: He'd used a FET to trigger the explosive - and normally had the PIC asleep with pins in the Hi-Z state and was wondering why the chute would randomly deploy. Adding a pulldown resistor (a few hundred k) or having simply used a bipolar transistor (which can usually be guaranteed to be "off" if there is no base drive:-) would have prevented grief.


I would avoid use of rechargable cells (either NiCd or NiMH) for devices with long-term idle states: Alkalines are dirt cheap and, nowadays, have ridiculously long shelf life at reasonable temperatures.

A general rule of thumb when using AA alkaline cells:

- If your average current is 100 uA per cell, you can, conservatively, expect at least a years' lifetime if you can tolerate operating down to 1 volt/cell.


This means that if your devices will tolerate operation from 4.8 down to 2.7 volts (1.6 to 0.9 volts per cell - a range inclusive of a "very fresh" battery to one that has reached its 90% depletion voltage) *PLUS* a dramatic increase in internal resistance of the cell - usually to a few ohms per cell (at 0.9 volts/cell) then you are in good shape.

A single-cell boost circuit will be even more-affected by voltage loss/internal resistance increases so unless you can justify the added cost of such a circuit (with size/weight reduction of having only one cell rather than three) then I can't see that there would be any benefit. Note that most single-cell switchers simply don't function much below 0.9 volts/cell whereas a low-voltage PIC could manage to chunk along at 2 volts, or 0.66 volts/cell before you start violating operating conditions. (Most PICs that I have seen actually operate down to 1.5 volts - assuming low clock frequency, room temperature, etc. - but that's something that I wouldn't base my design on:-) Another consideration, however, is that you may not be able to turn on the P-channel FET adequately at very low voltages - an advantage of a bipolar device in this particular instance (if you can handle having several % of your power consumption just being base current to turn transistor on.)

CT

[Eugeneo]

[C Turner]

Hi again,

>So is keeping the pin on float and pulling it low better than setting the pin low in the program?

My practice has been to set all "un-connected" pins to output mode rather than to tie them to Vss or Vdd either directly or through a resistor - or, in the case of pullups, let the pullups do their job.

Now, the obvious caveat with using a pullup/pulldown (internal or external) is that you NEVER want, in the low-power state, to pull against the pullup/down.

Now your post as caused myself a question to which I don't know the answer - and a quick persual of the datasheets didn't answer: What pulls less current - an Input pin pulled decisively high/low, or a floating (not-connected) output pin? (We are, for the moment, ignoring a "special case" I/O pin that is open-drain, as is found on some devices like the 'F84, etc.) My experience is that the current consumption in each case is about the same but I've never really tried to absolutely minimize current consumption in a PIC (that is, trying to get it below a few uA.)

Suffice it to say, in your application, I'd suspect that setting to output and setting to the power-minimal state would be perfectly adequate - remembering, of course, that if the PIC resets, ALL pins are INPUT until told otherwise. (This may dictate the strategic use of pullup/down resistor to avoid floating C/MOS gates...)

>That's good to hear since the pic only wakes up for 2-5 mins once a month, where do you find specs like this?

When you asked the question, I looked again at the web sites of battery manufacturers and noticed that the datasheets had been updated since the last time I'd really looked at them. Gone seemed to be the handy chart that showed internal resistance versus cell voltage. Years ago, when this first came up (back when the 16C54 was the newest PIC and before the public internet) we dug up this info in one of the manufacturers' paper databooks.

To be sure, one can infer what the internal resistance must be from the charts that *are* still there. From those charts, one can infer that a 100 uA load will easily be borne by a fresh AA cell over the course of a year, as nearly all of the 2800 mAh will be available under such a small load if you can tolerate 0.8-0.9 volts/cell. (Theoretically, one could expect a couple of years - but one is conservative in these situations.)

>You're right, at 3.0 volts, the fet was only supplying 1.3 - 1.8 and its a 160 mil ohm rds part... I've changed it to a pnp that has a gain of about 100 and it works all the way down to 1.5v.

That's the downside of FETs. There are some FETs with *really* low turn-on voltages, but they tend to be slightly more expensive. Now, if insisted on using a FET you could always construct a capacitor-based charge-pump to create a negative and/or a higher positive voltage, but that increases complexity and will probably increase current consumption overall. I've used this trick (using one of the PIC's I/O pins and toggling it rapidly) to generate a voltage >Vdd - or more often, a negative voltage - such as one that might be needed for an op-amp or RS-232 driver.

CT

[Eugeneo]

Hi CT,