Chicken Door Opener / 4 AA cells direct to PIC?

[NivagSwerdna]

To be fair the device I am currently using is 18F1320 (which I had lying around)...

Voltage on VDD with respect to VSS ... -0.3V to +5.5V

[NivagSwerdna]

Lets see a schematic of what you have so far.

Good plan.  I'll draw something up.

[NivagSwerdna]

OK... posting my circuit (from memory)... only on the basis that you all promise to play nicely.... (I know I should probably use a LDO and sleep mode)... The gadget will be on for approx 3 hours per month.

[Ian.M]

The kindest thing I can say about that schematic, is that the design was obviously an evolutionary process.   Your choice of an optocoupler for non-isolated output is unusual and rather power-hungry.

You could futz around with another 2N7000 MOSFET to disconnect the IN signal from RB1, with a pullup to Vss on the RB1 side, or even with a diode so IN can only pull the pin low, but IMHO the best option would be to ditch the relay and start over with the buttons and IN signal on interrupt capable pins so they can wake the PIC from sleep, and have it always powered and 'in the loop'.  IN would need to be on an IOC capable pin (RB4 or RB5 assuming you want to keep the ICSP/ICD pins free) so you could wake on both edges, but the buttons could be on ordinary INT pins (RB0-3) as you only need to wake when they are pressed.   If the wakeup was a button push it would perform whatever action you require, but if it was an IOC from the IN signal and it wasn't already performing a button action, it would simply repeat that to the OUT.

We'd need a few more details about the IN and OUT signals to decide on the best interface for each . . . .

[NivagSwerdna]

The kindest thing I can say about that schematic, is that the design was obviously an evolutionary process.

   Your choice of an optocoupler for non-isolated output is unusual and rather power-hungry.

Hm.  The intention was that when active the output would be isolated. When the latch is latched the OUT is driven by the Opto and not connected to anything else but they do share ground.  On reflection I guess that could just be a 2N7000.  I didn't want to sink the current directly at the PIC since it could be >5V.  I also don't know the lower bound of the door terminal so if it was very low... that might not be good for VDS in a FET solution? Ideally I would not have the shared ground.. but I wasn't sure how to do that without more switching!

You could futz around with another 2N7000 MOSFET to disconnect the IN signal from RB1, with a pullup to Vss on the RB1 side, or even with a diode so IN can only pull the pin low, but

Yeap. I was wondering that.

IMHO the best option would be to ditch the relay and start over with the buttons and IN signal on interrupt capable pins so they can wake the PIC from sleep, and have it always powered and 'in the loop'.  IN would need to be on an IOC capable pin (RB4 or RB5 assuming you want to keep the ICSP/ICD pins free) so you could wake on both edges, but the buttons could be on ordinary INT pins (RB0-3) as you only need to wake when they are pressed.   If the wakeup was a button push it would perform whatever action you require, but if it was an IOC from the IN signal and it wasn't already performing a button action, it would simply repeat that to the OUT.

Definitely the simplest.. but I'm not sure exactly how to drive the outputs...

We'd need a few more details about the IN and OUT signals to decide on the best interface for each . . . .

OK... it's like this... the existing door opener has two terminals which connect to a timer.  The timer is normally-open but goes to closed during the door closed interval.  The terminals don't have any definition of their polarity and/or relation to ground or battery.  With a multimeter it appears that one of the terminals is directly connected to the same terminal that the battery -ve arrives from. i.e. the door closes then the door terminal is shorted to ground.
So for my gadget I have two input terminals for the timer... this is normally open but closed when the timer demands door close.  On the output it seems I need only one terminal since ground is really shared... on that terminal I need to be able to short it when I want the door closed and leave it open when I want the door open.
The idea of the gadget is that it holds the door open/closed as per user request until the timer matches the desired state at which point it goes to sleep.
A look at the door opener circuit itself suggests that there is a positive potential on the door timer terminal which when grounded either charges up the gate of a FET or allows a current to flow and turns on a BJT... not sure which... it's a black box to me.
The gadget is very rarely used... hence the desire for low to zero quiescent power demand.

Thanks for the interest 

[Ian.M]

I think I get that - the existing door motor box has an input that is continuously monitored.
If it transitions to grounded, the door closes then the motor shuts off.  If it transitions to open, the door opens then the motor shuts off.   The existing timer has volt-free contacts that shorts that input to ground when the door is scheduled to be closed.



As is, even with the relay, it has to maintain its output until the timer 'catches up'.  The drive current for the 4N24 is killing your power consumption.  Even at only 1mA (and I suspect you are using far more,) that's equivalent to a weeks power for a PIC in sleep mode, worst case, and probably something like 6 weeks to 2 months power if the typical figures are to be believed.

Interfacing to the door motor without extra quiescent current  isn't a problem - a N-MOSFET can do that easily.  The problem is the timer output - its grounded for the entire door closed period, so if you simply use a pullup resistor it will be drawing current whenever the door is closed.  You can mitigate that by using a high value pullup, but that increases the sensitivity to noise, and although you can use a debouncing algorithm to ignore the noise, the increased average current consumption due to unwanted wakeups during the door open period could be worse than the pullup resistor consumption during door closed.
If the pullup is under about 400K you'd do better to enable the watchdog timer (using SWDTEN), and use it to provide a  wakeup every minute or two, during which you enable the pullup (by setting an output pin driving the other end of the pullup to '1'), poll the IN signal and then turn the pullup off (actually turn it into a pulldown, by setting its control pin to '0'), all within a few ms.

[NivagSwerdna]

The existing timer has volt-free contacts that shorts that input to ground when the door is scheduled to be closed.

This is actually an assumption that has always bothered me.  The existing timer shorts two contacts.  The definition that one is ground was by observation... ideally I would not assume that and just short or non-short two contacts without any particular reference to voltage or polarity but I don't know how to do that... I'm just ignorant in that respect.  I thought an Optocoupler might be a place to start but then ended up with common grounds since I wanted the gadget off state to be pass-through and only one side was switched. (And the opto has a specific polarity).

The drive current for the 4N24 is killing your power consumption.

It's a few mA which is BIG in the scheme of things but I'm also flashing the LEDs on the buttons so it is reasonably power hungry when operating.  e.g. You press DOWN and the DOWN button flashes intermittently until DOWN matches timer DOWN and then it turns off.  It does look nice though.    I also have a stay open/shut mode.. after a long-press... which leaves the door open/shut with the respective button permanently illuminated (until another press).

[Ian.M]

The existing timer has volt-free contacts that shorts that input to ground when the door is scheduled to be closed.

This is actually an assumption that has always bothered me.  The existing timer shorts two contacts.  The definition that one is ground was by observation... ideally I would not assume that and just short or non-short two contacts without any particular reference to voltage or polarity but I don't know how to do that... I'm just ignorant in that respect.  I thought an Optocoupler might be a place to start but then ended up with common grounds since I wanted the gadget off state to be pass-through and only one side was switched. (And the opto has a specific polarity).

If you really need a volt-free contact output with minimal power consumption, we are back to latching relays.  You couuld use a photo-MOSFET optoisolator like the TLP175A to provide a volt-free output that doesn't mind the polarity, but it needs 5mA of drive current for its LED, not exactly ideal when it could be required to sustain it for hours.

However, once you have confirmed that one terminal of the door motor box timer input is actually ground there is absolutely no reason to go the the extra expense and complexity compared to using a simple transistor output. 

You should measure the input current with the timer disconnected and the input terminals shorted via a DMM on mA, and also check the terminal voltage relative to battey 0V with the input open.  That would give you enough to be certain of your output requirements.

The drive current for the 4N24 is killing your power consumption.

It's a few mA which is BIG in the scheme of things but I'm also flashing the LEDs on the buttons so it is reasonably power hungry when operating.  e.g. You press DOWN and the DOWN button flashes intermittently until DOWN matches timer DOWN and then it turns off.  It does look nice though.    I also have a stay open/shut mode.. after a long-press... which leaves the door open/shut with the respective button permanently illuminated (until another press).

I'd use low duty-cyle flash pattens for both.  1/10 second on every ten seconds is still highly visible, but it reduces the average current by a factor of 100.   Even if you use a double flash to indicate the stay open/shut mode, that's still a factor of 50 better than just leaving the LED on all the time. 

[Zero999]

To be fair the device I am currently using is 18F1320 (which I had lying around)...

Voltage on VDD with respect to VSS ... -0.3V to +5.5V

But that's the range of voltages the MCU is supposed to survive. There's no guarantee it'll work over that range. It's specified to work from 2V to 5.5V. This is because under abnormal conditions, e.g. voltage spikes on the power supply, the voltage may be outside the design range, which may cause it to fail meet its specifications, not destruction of the device.

There's an error in the data sheet for the 18F1320. The absolute maximum voltage range is always wider than the operating voltage range. For example if we look at a data sheet for the 18F1330 which is very similar to the 18F1320 and manufactured using the same process, the voltage range on VDD with respect to VSS is -0.3 to +7.5V. In fact the maximum ratings for temperature, currents, power dissipation and other voltages are exactly the same as the 18F1320, the only exception being the upper bound on VDD.

http://ww1.microchip.com/downloads/en/DeviceDoc/39758D.pdf

So what does this mean? Your MCU will be able to withstand 7.5V for a short period of time, even though its only designed to work up to 5.5V. It may not meet all its specifications when run above 5.5V and isn't good design practise to run it outside its specification but it won't blow up when connected to four fresh AAs via a diode.

Sorry for going off on a tangent. You're currently running it of 5V, which is well within the specification.

[NivagSwerdna]

You could futz around...or even with a diode so IN can only pull the pin low

I was out tonight but a few minutes tinkering before bed.... I added a diode at RB1 and that solves the original problem, in fact functionally it actually seems to work now.
When I have time I will swap out the 4N25 for a 2N7000.  Power consumption when driving ON is currently up to 25mA (there are three LEDs, one on the button, one to indicate the door and one for luck I added earlier), using a FET will probably halve that and then I can lose the extra LED at some point.
I also need to reduce the duty flash cycle for forced up/down.
I will probably call that Mk 1 design and give it a test in situ and see if it actually works.
One issue I can foresee is power supply sag when the motor is lifting the door.

[Ian.M]

One issue I can foresee is power supply sag when the motor is lifting the door.

Yes, that might not be so good.   You could add 5.5V supercap after the diodes to keep Vdd up while the motor's running - just keep the LEDs off for long enough for full door movement so you only have to power the PIC through the brownout.

[NivagSwerdna]

Just as well I added the diode... when in straighthrough mode the RB1 input gets connected to OUT.  OUT has unknown potential...
Might have to re-think this with much more isolation and perhaps a seperate supply for the gadget.

[NivagSwerdna]

Back late... and out to the shed...

I measured the voltage across the battery terminals... 6.37V... when the door is lowering it drops to 6.28V and when raising it drops to 6.20V so the sag isn't too bad.

I measured across the timer outputs... it had 0.565V across it... didn't expect that.   Resistance measures 27kOhms in one direction, infinity in the other... I think that is telling me something..   Open Collector Output Transistor.

When the timer is across the door contacts it has a constant 0.012mA current flowing through it.

I think the Mk1 door opener should work... it might have to wait until the weekend to try it in anger.... I'm already considering Mk2... perhaps a 74HC4053 and a LDOed PIC.... any other suggestions or do you think Mk1 was really optimal 

[Ian.M]

That current and measured voltages would imply the door opener probably has a 470K pullup.     I think MkII can be simplified to one PIC that's in the loop all the time + a 2N7000 to interface to the door opener, with a 470K pullup on the PIC's input from the timer + any LEDs you want on separate pins.  You'll be hard-put to beat the two diodes to drop the voltage as it can be difficult to find LDO regulators that have less than 10uA quiescent current even when their input voltage isn't enough to maintain the output.   Its probably worth staying at a nominal 5V Vdd as going down to 3.3V would mean you'd need to replace the 2N7000 with a low Vgs threshold MOSFET.

A total quiescent current of under 20uA should be possible, which would have a negligible impact on AA battery life.

[NivagSwerdna]

...but the buttons could be on ordinary INT pins (RB0-3) as you only need to wake when they are pressed.
...
A total quiescent current of under 20uA should be possible, which would have a negligible impact on AA battery life.

The buttons would need pull-ups (or downs)... looking at the datasheet the power budget for PORTB pull-ups is somewhere between 50 to 400 uA.... which is >>20uA... are you proposing sampling rather than wake on interrupt?

[Ian.M]

No. because you cant control the internal pullups individually on this PIC, use external pullup resistors.  Try 470K with 1nF capacitance to Vss at each pin to make them less sensitive to noise.   The only pullup that will draw current continuously is the timer input one, when the timer is commanding door closed.  The button pullups wont draw current unless the button is pressed so they could be 100K with no issues.

Some comments on the power budget:
An Energiser E91 (AA) Alkaline cell has an approximate capacity of 2800mAH, and a shelf life of 7 years to 80% remaining capacity.   The self-discharge is therefore 80maH/year, which is equivalent to a 9uA drain.     An extra drain that left the batteries at 50% capacity after 7 years would have negligible impact on their life in any application so you've got about 14uA to play with, or far more if you replace them at least annually with freshly purchased ones.

[Howardlong]

Not withstanding others' valuable contributions, and for fear of flogging a dead horse ;-)

As I recommended above, LDO regulator MC78LC00 series. 1.1uA Iq, Vin up to 12V (note many low Iq regulators do not have a reasonably wide Vin).

Using an unregulated power supply, especially relatively high internal resistance dry batteries with intermittent inductive loads and high current draw, is almost certainly going to take your PIC outside of its operating envelope (4.5-5.5V).

Use the LF version and a 3.3V regulator such as the MC78LC33. This will give you sufficient headroom for voltage drop. No need for any super caps. Regulators aren't just there for voltage drop, they're there to remove excesive power input fluctuations.

[NivagSwerdna]

Not withstanding others' valuable contributions, and for fear of flogging a dead horse ;-)

Your horse is well alive, and your suggestions are well taken and much appreciated. 
I have a LM2936BM-3.3 sitting next to me, MC78LC00 looks interesting. (I also have some MCP1703A somewhere).
I don't have any PIC LF types knocking about.... Might try a PIC12LF1571 although the AT Tinys (for which I don't have an ICD) look pretty interesting for this application. I think my 2N7000 will operate at the lower voltage... I've got a few of those.

[Howardlong]

You need to look at LDOs considering multiple facets, but in your application you are looking for:

o Low Iq (say <2uA) for prolonged battery life. LM2936 looks about 15uA :-(

o Low drop out voltage (say <200mV for your max PIC operating current)

o Reasonably high Vin max (say >7V) especially as ISTR you may have an inductive load on the unregulated side

o Sufficient current (>=50mA)

o Low enough output voltage to allow for operation with partially depleted batteries even when batteries under load.

o Output voltage within PIC spec

Note that some more modern PICs in the standard F versions operate from about 2.5V to 5V. The PIC18F1320 is quite old in this respect, only operating from a fairly narrow range of 4.2 to 5.5V.

The PIC12F1571 is 2.3 to 5.5V, while the PIC12LF1571 is 1.8 to 3.6V, both with a slightly higher Vmin of 2.5V for Fosc>16MHz.

I would still recommend a low Iq LDO even with the PIC's extended voltage range, to mitigate voltage transient effects on an unregulated rail. Such transient effects are often very difficult to track down, you might as well deal with them ahead of time.

There is a negative about using a lower PIC voltage such as 3.3V and that is that the Vgs to switch on the MOSFET might not be sufficient to fully switch on a 2N7000 depending on its load. At this point, it's time to choose an alternative MOSFET or use a bipolar transistor instead (Si bipolars switch on at ~0.7V, but check the DS for the collector/emitter leakage current when off!).

(Edit: just saw Ian.M already addressed the 2N7000 Vgs issue).

[Howardlong]

Don't forget back EMF diodes across your inductive loads (i.e., relay or solenoid coils).

[NivagSwerdna]

At this point, it's time to choose an alternative MOSFET or use a bipolar transistor instead (Si bipolars switch on at ~0.7V, but check the DS for the collector/emitter leakage current when off!).
(Edit: just saw Ian.M already addressed the 2N7000 Vgs issue).

You're not going to like this... According to the datasheet I actually think the 2N7000 would continue to provide sufficient drive even at the lower voltage.
I'm open to suggestions and not fussy about through-hole versus SMD. 2N7002? BSS138?

Don't forget back EMF diodes across your inductive loads (i.e., relay or solenoid coils).

I'm rapidly running out of diodes. 

[Ian.M]

You definitely need a PIC that can operate well below 4.5V.   However a trap for the unwary with older PICs that can run below 4.5V, as well as at 5V, is that many of them require >=4.5V for chip erase, so if you run them from a 3V battery or from a 3.3V regulator that cant tolerate its output being taken to 5V while programming (or with other non-5V tolerant devices on the same rail), then you can only program them in circuit *ONCE*.   That's OK for production, but not so much fun for development.   Younger players who aren't obsessive about reading the datasheet  often discover this the hard way with much wailing and gnashing of teeth when something expensive on their 3.3V rail lets the holy smoke out when they attempt a chip erase with the programmer supplying target power.

I would question the selection of PIC12 parts for a one-off project that isn't space or weight constrained.   Although its possible to use the ICSP pins for general I/O, doing so makes it impossible to debug in-circuit (if the PIC12 in question even has built-in debug support) without an expensive debug header, so if you need more than 3 I/O pins then there is a clear case for using a debug capable PIC16 with enough pins to avoid the need to use the ICSP pins in the application circuit.

@NivagSwerdna: What PICs do you actually have in stock (ignoring >28 pin ones)?  You may have one with a better operatig voltage range . . . .

[Howardlong]

Younger players who aren't obsessive about reading the datasheet  often discover this the hard way with much wailing and gnashing of teeth

Agreed, I am not sure it's even possible to be over-obsessive about the DS parameters, and also add to that the errata: you should at least have scanned the errata to at least be aware of any gotchas that may affect your design at an early stage.

It's also not unusual to find yourself re-reading portions of the DS several times in an effort to try to make sense of things.

Ignore the DS & errata parameters at your peril, the devil is in the detail.

[NivagSwerdna]

@NivagSwerdna: What PICs do you actually have in stock (ignoring >28 pin ones)?  You may have one with a better operatig voltage range . . . .

These are the LFs I have lying around...

PIC12LF1840-I/P

PIC16LF18313-I/P
PIC16LF1939-I/P
PIC16LF1824-I/P

The PIC16LF1824 looks quite fun.  Has some kind of SR Latch

And the F's

PIC10F322-I/P
PIC12F1572-I/P
PIC16F1509-I/P
PIC18F4450-I/P
PIC16F18323-I/P
PIC24F16KA101-I/P

PIC16F648A-I/P
PIC18F1320-I/P

That's all folks.  (The others have a lot more pins)

I also found a couple of MCP1703A-3302

[Ian.M]

Quite a lot of F parts work down to 3V.   e.g. PIC16F1824 has an operating range of 1.8V-5.5V.   So please don't limit your list of what's on hand to just the LF parts.