Shop OBEX P1 Docs P2 Docs Learn Events
Solar/Micro-power Quickstart board. — Parallax Forums

Solar/Micro-power Quickstart board.

LawsonLawson Posts: 870
edited 2012-10-23 22:03 in Propeller 1
Solar Quick-start.jpg


As shown, the quickstart has been running in a window for weeks. The blink pattern is a 1-D elementary cellular automaton running at rcslow. I think the circuit's total current budget is 20-50uA.

A few hardware modifications are needed. First, a 1F 3.3V super capacitor was soldered between Vcc and Vss. This capacitor stores plenty of power to run the quickstart for hours if the current drain of the circuit is kept below 100uA. Second, the power LED needs to be disabled. It draws a couple of mili-amps on it's own. Either remove the LED or the resistor right next to it. ( for details how, follow the Sparkfun SMD tutorials. ) Third, the regulator needs to be isolated from Vcc with a schottky diode or fet so that it's several mili-amp quiescent current doesn't drain the super capacitor. The regulator is Isolated by lifting the center pin and thermal tab. A diode is then used to allow USB power in. Alternatively, a logic level N-fet (like the IRLML6346) can be used. The FET source is connected to the regulator tab, drain to Vcc, and the gate connected to the Vin pin of the regulator. With just these three mods, the quickstart will charge off of USB and run for hours to days on rcslow depending on what it's doing.

To complete the solar conversion a few more modifications are needed. To save ~3uA and enable low voltage operation disable the brownout by moving the brownout jumper resistor on the quickstart. Since this also disables the internal pullup on the reset pin, an external pullup resistor needs to be added. (circled in red on the image above) Next a solar cell needs to be added. (I used an 8 cell amorphus silicon solar cell off Digikey. 869-1006-ND will work as well) Because the circuit doesn't measure it's supply voltage, the solar cell size/voltage needs to be matched with the software's power draw. If you're really worried about over-voltage, 2x red LED or a red and green LED placed in series from Vcc to Vss will keep Vcc from going too high. I do also have a reverse current blocking diode for the solar cell hidden under the super capacitor. After measuring the dark leakage of a similar solar cell, it's really not needed.

I've also attached the code the board is running. The blink patterns are cool to watch even on a stock quickstart board.

Lawson

P.S. support my Kickstarter for Lumen Electronic Jewelery. I'm fermenting a future Propeller powered piece for our next round of products.

Comments

  • LoopyBytelooseLoopyByteloose Posts: 12,537
    edited 2012-10-20 09:55
    Lately I have been thinking that most micro-controller have voltage regulators provided that are excessively large for the actual use. The Propeller can do quite well with 100ma or less, so if you really want to stretch the low power, using an ultra low quiescent power regulator (maybe only 50ma output) can stretch a backup battery that is being charged by a small solar panel.

    This is one Parallax has used in the past. http://www.ti.com/lit/ds/symlink/lm2936.pdf Just make sure that the solar panel and battery normally avoid the low drop out region as that significantly increases power demand.
  • LawsonLawson Posts: 870
    edited 2012-10-20 10:09
    Lately I have been thinking that most micro-controller have voltage regulators provided that are excessively large for the actual use. The Propeller can do quite well with 100ma or less, so if you really want to stretch the low power, using an ultra low quiescent power regulator (maybe only 50ma output) can stretch a backup battery that is being charged by a small solar panel.

    This is one Parallax has used in the past. http://www.ti.com/lit/ds/symlink/lm2936.pdf Just make sure that the solar panel and battery normally avoid the low drop out region as that significantly increases power demand.

    Yea, the quiescent current draw and dropout power draw are why the energy storage is after the regulator with this mod. In a dedicated circuit, I'd put the super capacitor before the regulator and use a better regulator. The LM2936 does have nice specifications and actually shows the quiescent current when the regulator drops out! (most data sheets leave that off...) It still draws a bit more power than I'd like. My preference right now is the MIC5232. It's ground current is 2uA in all conditions.

    Lawson
  • LeonLeon Posts: 7,620
    edited 2012-10-20 11:48
    Really low power systems don't use regulators.
  • Tracy AllenTracy Allen Posts: 6,664
    edited 2012-10-20 13:52
    Even so, micro-power systems often need a voltage reference, and this new crop of micro-power regulators is a welcome addition to the parts kit. A nano-power alternative is to switch the reference on and off.

    Lawson, thanks for the link to the MIC5232. I've been using the MCP1700 series. It's quiescent current is also less than 2µA, but its maximum current goes up to 250mA. which is nice for use with SD cards, XBees and the like. One nice feature I like about that MIC5232 though is the reverse current limiting. The mcp17xx part uses a standard pmosfet with reverse diode, so there is no reverse current limiting at all, which may or may not be a problem. It definitely precludes using it in a scheme with wired-or power supplies.

    Does the Lumen jewelry use a supercap too, or a rechargeable battery?

    Without the Prop brownout detector, have you seen any problems, lockups etc.? OF course there are external brownout detectors that operate down to lower voltages and with an Iq of around 1µA.
  • LeonLeon Posts: 7,620
    edited 2012-10-20 14:02
    Some of the latest XLP PICs have a built-in reference. They take as little as 9 nA.
  • Tracy AllenTracy Allen Posts: 6,664
    edited 2012-10-20 15:47
    Leon,
    I don't want to hijack Lawson's thread here, but I'm curious which PIC is your favorite for low power. I'm working on a design now where a PIC16F1782 running at 3.3V will serve as an ADC and all around analog input node for a system. The '1782 is distinguished by having a multitude of 12 bit ADC channels that can be muxed with other i/o functions, at far less than the price of a dedicated ADC. I must say, Microchip does a great job of documenting the power drain under all sorts of operating conditions. The FVR (programmable voltage reference for the ADC and comparators) draws 15 µA when it is turned on. BOR draws 7µA. The Prop compares favorably with its BOR at 3µA, and I know Chip took a lot of care with it in the design. The PIC16F1782 according to the data sheet can get down to minimum of 300nA in sleep with everything turned off, while waiting for an external wakeup source. The minimum for the Prop is a few microamps at RCslow in a wait state. I know Microchip has the XLP products, but, so many to choose from, it is pretty daunting to dig through the details of the operating conditions.
  • LawsonLawson Posts: 870
    edited 2012-10-20 19:01
    Even so, micro-power systems often need a voltage reference, and this new crop of micro-power regulators is a welcome addition to the parts kit. A nano-power alternative is to switch the reference on and off.

    Lawson, thanks for the link to the MIC5232. I've been using the MCP1700 series. It's quiescent current is also less than 2µA, but its maximum current goes up to 250mA. which is nice for use with SD cards, XBees and the like. One nice feature I like about that MIC5232 though is the reverse current limiting. The mcp17xx part uses a standard pmosfet with reverse diode, so there is no reverse current limiting at all, which may or may not be a problem. It definitely precludes using it in a scheme with wired-or power supplies.

    Does the Lumen jewelry use a supercap too, or a rechargeable battery?

    Without the Prop brownout detector, have you seen any problems, lockups etc.? OF course there are external brownout detectors that operate down to lower voltages and with an Iq of around 1µA.

    If you just want a low power reference I'd use the Mic5231. It has a 0.65uA quiescent current that is independent of load or drop out. I've gotten burned by drop out current with the MCP1700 series. The quiescent current in drop out is over 20uA for the 2.5v part. (and the 3.3 and 5 volt parts should be even worse) Though as a reference, the 1.2v MCP1700 works well. By the time it drops out, none of it's transistors have the gain left to draw extra current. (similarly the 1.5v part should only get a few uA worse in drop out).

    The Lumen Jewelery pieces use a super capacitor. While a surface mount battery would store ~10x the energy, they all die if discharged to zero volts. So, locking the jewelery in a jewelery box for a few months (or years) would kill any battery powered circuit.

    Haven't had any lockup problems without the brownout detector. I've tested it to run with output at RCSLOW down to 1.1 volts. Memory is retained as low as 0.7 volts, but this is well under the arming voltage for the power on reset circuit so the supply has to be brought up slowly. Either way, any jewelery with a Prop core will have the reset line (and programming lines) accessible in case the core does hang.

    Lawson
  • LeonLeon Posts: 7,620
    edited 2012-10-20 21:51
    Tracy,

    I don't have a favourite PIC, I just select the best one for a particular job. One of the XLP devices would be best as you suggest, you will have to use the product selector and go through the data sheets. I'd probably use something like the PIC16F1786.
  • Tracy AllenTracy Allen Posts: 6,664
    edited 2012-10-20 22:10
    I know what you mean about that issue with the MCP1700 series. That is a fact of life with most regulators as they try as hard as they can to provide the voltage demanded. Are the Micrel parts really different in that respect? I don't see a graph of quiescent current vs input voltage down through dropout in the data sheet. It would take extra circuitry to implement that feature. Here is the same effect from two members of the Linear Tech LT1761 family:
    Attachment not found.
    Screen shot 2012-10-20 at 9.52.58 PM.png
    Screen shot 2012-10-20 at 9.51.21 PM.png


    I like Linear Tech parts because they really do a good job of documentation and they incorporate a lot of bulletproofing. The Micrel part most similar to the LT1761 seems to be the MIC5235. About the same output current of up to 100 or 150mA, quiescent current 20µA, and in a SOT23 with shutdown. But LT parts are bipolar with a PNP pass element, so the wasted ground pin current increases as a % of output current. Micrel uses the Pmosfet pass element, so there is very little load dependence, and they have a nice zero-power shutdown as well as such bulletproofing features such as reverse battery protection (absolute -20V to +38V) and thermal limiting. And considerably cheaper than the LT1761 too. The MIC5231 is interesting too, special purpose.

    I remember that you said something about the Prop memory retention down to quite low voltages. But if it does drop a few bits in the program, it could be dead jewelry. Then it has to draw 3 or 4 mA to run in RCFAST to reload the program from eeprom, and that will have to be triggered by some kind of reset mechanism.
  • Tracy AllenTracy Allen Posts: 6,664
    edited 2012-10-20 22:23
    Leon, Okay, I just thought you might have a particular job you did with a PIC aimed at micropower. The PIC16(L)F1786 is pretty much the same as the PIC16F1782, except the '1786 is a "future product" with more flash program memory and more RAM. Nice to have an upgrade path! Both are listed as "XLP" for extra low power, but as you say, the actual limits have to follow on perusal of the data sheets in the context of the application. This will be solar powered, with a Prop as the main brain, but has nowhere near the constraints of Lawson's project.
  • Tracy AllenTracy Allen Posts: 6,664
    edited 2012-10-21 11:43
    Leon again,
    I found the PIC18F2XJ13 that offers the 9nA you mentioned in deep sleep mode, and also 12 bit ADCs and exactly the same pinout as the '16F1782 I've been looking at. A 558 page data sheet to wade through.
  • LeonLeon Posts: 7,620
    edited 2012-10-21 11:55
    I prefer the 18F parts as they are easier to program and they are better if you are using C.
  • LawsonLawson Posts: 870
    edited 2012-10-21 19:11
    I know what you mean about that issue with the MCP1700 series. That is a fact of life with most regulators as they try as hard as they can to provide the voltage demanded. Are the Micrel parts really different in that respect? I don't see a graph of quiescent current vs input voltage down through dropout in the data sheet. It would take extra circuitry to implement that feature. Here is the same effect from two members of the Linear Tech LT1761 family:
    Attachment not found.
    Screen shot 2012-10-20 at 9.52.58 PM.png
    Screen shot 2012-10-20 at 9.51.21 PM.png


    I like Linear Tech parts because they really do a good job of documentation and they incorporate a lot of bulletproofing. The Micrel part most similar to the LT1761 seems to be the MIC5235. About the same output current of up to 100 or 150mA, quiescent current 20µA, and in a SOT23 with shutdown. But LT parts are bipolar with a PNP pass element, so the wasted ground pin current increases as a % of output current. Micrel uses the Pmosfet pass element, so there is very little load dependence, and they have a nice zero-power shutdown as well as such bulletproofing features such as reverse battery protection (absolute -20V to +38V) and thermal limiting. And considerably cheaper than the LT1761 too. The MIC5231 is interesting too, special purpose.

    I remember that you said something about the Prop memory retention down to quite low voltages. But if it does drop a few bits in the program, it could be dead jewelry. Then it has to draw 3 or 4 mA to run in RCFAST to reload the program from eeprom, and that will have to be triggered by some kind of reset mechanism.

    The MIC5231 datasheet has a graph of ground pin current versus supply voltage that includes dropout. I suspect the MIC5231 is an all FET part so current consumption is independent of saturation. (schottky transistors should be able to pull the same trick too) While I suspect the MCP1700 series uses at least a BJT error amp.

    I don't know why the Mic5232 datasheet lacks a similar graph. I've measured 3.3 volt part's ground pin current with no load to be constant while dropped out. Maybe the ground pin current is bad when some load is applied? Still, the suggested applications listed on the first page include "SRAM Memory Back-up Supply" and "SRAM Memory Back-up Supply" both of which are micro-amp drains likely to operate the regulator in drop out for long times.

    All the jewelery pieces have optional USB quick charging. All the propeller based pieces will have an exposed prop-clip connector. (I'll probably write protect the eeprom as the ram will last months anyway) This will allow an easy two step recovery process in the event of a scrambled ram. Plug it in, then short RST to VSS with a key, coin, paper-clip, prop-clip, etc.

    Lawson
  • Tracy AllenTracy Allen Posts: 6,664
    edited 2012-10-21 22:38
    Feeding is a feature. Care for your pet! An hour of sunlight every day. Avoid resuscitation. A Phoenix, in contrast, would expect multiple reincarnations!
  • LawsonLawson Posts: 870
    edited 2012-10-23 15:00
    Feeding is a feature. Care for your pet! An hour of sunlight every day. Avoid resuscitation. A Phoenix, in contrast, would expect multiple reincarnations!

    Hasn't the electronic pet fad run it's course by now, or is that fad due for a resurgence? Anyway, an electronic pet program isn't planned for any of the Lumen Electronic Jewelery. Though someone else certainly could make one.

    Lawson
  • Tracy AllenTracy Allen Posts: 6,664
    edited 2012-10-23 22:03
    Lawson,

    I did try your program and had fun playing with the timing parameters and the alternative automata. The thing that struck me most was that the PWM and sweep down the row of 8 leds had no hint of flicker. That at RCSLOW, clkfreq=20kHz, one single PASM cog.

    The LEDs are drawing far more power than the Prop!
Sign In or Register to comment.