efficient 12v -> 5v switching regulator

[gregcope]

Hi All,

I want to build a remote, very low power, monitoring solution ... However as I want to sometimes also do GPS tracking, running a GPS pushes me over the power budget of AA batteries as I am presently estimating using about 3Ah a year.  This is for a boat, on a swinging mooring (ie not a marina).  I want to know that it has not moved,  on board batteries are ok, and that it is not filling with water.  If out of bounds it will send an SMS.  I have a raspberry pi prototype running - https://github.com/gregcope/piboatmon.  It will also have a spare battery so that if the main ones fail, it can still work/send SMSs.     

I will have access to 12v system (that I also want to measure voltage on).  Due to sometimes being charged voltages can range from 16v (15.5v, but a little overhead for boost charging) to 11v (flat).

I want the 12v drain to be minimal.  Most regulators I seem to find are only efficient around 1A, when my device will be spending allot of its time asleep drawing uA.  For example some draw upto 25mA without load!!!

So - question - can anyone recommend any decent voltage regulators?

• Must - be very efficient 90% at low mA ranges
• Must - be able to supply 1A as I plan to power up a GPRS modem to send an SMS occasionally
• Must - have low questiant current
• Should - preferably off the shelf - as I am no electronics builder (I can connect things, and do basic soldering)
• May - ideally physically small

Something based on this; http://www.linear.com/product/LT8610 would be good.  However building that circuit is probably beyond me!

Thanks,

[TomWS]

<snip>I am presently estimating using about 3Ah a year.  <snip>

I don't understand what you mean by this statement.  Are you saying that you are expecting to consume a TOTAL of 3AH over a year period?  I don't think you're saying that you are consuming an Average of 3A over a year.  If it's the former, then the average power consumption is 342uA over a year period.  What is your PEAK current and duration?

You do know that AA's are 3.0AH capacity (and therefore could last a year) and C or D cells are 8.35AH and 20.5AH respectively?

If you're only going to need the peak occasionally, it might be best to go with a hybrid approach where you only draw (and step down) from the 12V battery when you need the peak boost.  Otherwise you can run happily on a couple of AAAs for years...

I'd be happy to help you, if you need more detail. 

Tom
UPDATE: fixed typo on AA capacity.

[gregcope]

Thanks Tom for the reply.

I had assumed that it would consume around 3AH over the year (see below).  Basically;

• Standby mode which does some light checking of sensors every 16 secs, and once and hour takes a GPS fix.  Once a day it sends a status SMS.
• There is another mode where it acts like a geo-fence (anchor alarm) where it constantly tracks GPS position vs a fixed location (captured when a button is pressed) and alarms if it moves outside this (by buzzer/sms).  GPS uses 25ma when running.  This is likely to be used for several days in a year

So yes the "average" is probably 342uA

Peak could be around 2A, whilst sending SMS, for a few secs whilst the GSM modem does it bit.  The GSM modem will consume 200mA or so for a minute whilst it connects to the network.  Most GSM modem need 5v to work.

I was working under the assumption AAs where 2.5Ah from here: http://www.batteryshowdown.com/results-lo.html

So I assume you are referring to Energizer Lithium AAs?  http://data.energizer.com/PDFs/l91.pdf which seem to have 3.2AH which is great (if expensive).

Basically my calcs are;

• AA Battery assumed 1200mAh usable
• Assume we want to power for 1 year (365X24 = 8760 hours)
• 1200/8760 = 0.228 mAh "budget" or 228uAh or 5472uAd (day)
• Moteino + PIR motion sensor use about ~78uAh or 1872uAd
• Temp Sensor around 6uAh or 144uAd - this could be turned off most of the time
• Once a day use 200mA for 1 min to send SMS 138uAh a day
• Once an hour use GPS for 1 min @25ma (416uAh.. or 10000uA a day)
• Daily = 10000+138+144+1872 = 12154uA a day or 4.4Ah a year

Sorry if the above logic does not make sense.

[TomWS]

Thanks Tom for the reply.

I had assumed that it would consume around 3AH over the year (see below).  Basically;

• Standby mode which does some light checking of sensors every 16 secs, and once and hour takes a GPS fix.  Once a day it sends a status SMS.
• There is another mode where it acts like a geo-fence (anchor alarm) where it constantly tracks GPS position vs a fixed location (captured when a button is pressed) and alarms if it moves outside this (by buzzer/sms).  GPS uses 25ma when running.  This is likely to be used for several days in a year

So yes the "average" is probably 342uA

Peak could be around 2A, whilst sending SMS, for a few secs whilst the GSM modem does it bit.  The GSM modem will consume 200mA or so for a minute whilst it connects to the network.  Most GSM modem need 5v to work.

I was working under the assumption AAs where 2.5Ah from here: http://www.batteryshowdown.com/results-lo.html

So I assume you are referring to Energizer Lithium AAs?  http://data.energizer.com/PDFs/l91.pdf which seem to have 3.2AH which is great (if expensive).

Basically my calcs are;

• AA Battery assumed 1200mAh usable
• Assume we want to power for 1 year (365X24 = 8760 hours)
• 1200/8760 = 0.228 mAh "budget" or 228uAh or 5472uAd (day)
• Moteino + PIR motion sensor use about ~78uAh or 1872uAd
• Temp Sensor around 6uAh or 144uAd - this could be turned off most of the time
• Once a day use 200mA for 1 min to send SMS 138uAh a day
• Once an hour use GPS for 1 min @25ma (416uAh.. or 10000uA a day)
• Daily = 10000+138+144+1872 = 12154uA a day or 4.4Ah a year

Sorry if the above logic does not make sense.

Good data, thanks!  It's clear you've thought about this. 

Just a couple of comments, the batteryshowdown link used a totally different usage model than you would have, the battery life degrades much more rapidly at 200mA drain.  Designed properly you really can factor the battery consumption average (and life span) using <342uA so you would be on the top line of the Lithium AA batteries you cite (curve attached below) and reliably get +3000mAh from this battery - I know, it's all I've used for years for AA or AAA primary batteries (and I don't work for Energizer      Also take a look at the knee of the curve, the sharp fall off is characteristic of this battery while alkaline and others tend to taper off (there's still 'juice' but the voltage is too low to use) so you get usable life longer with these.

Let me look at your data and I'll get back to you.  Does the GPS sensor run off 3.3V or does that require 5V also?

A couple more questions:
1. what electronics building skills do you have?  Would working with discrete components like MOSFETs, resistors, etc pose a problem for you? 
2. If I were to suggest a power converter, it would most likely be a module or a 'breakout' board implementation.  Would this work for you?
3. When you cite the Moteino & PIR sensor numbers, what duty cycle (sleep vs active) did you assume?


Tom

[gregcope]

Yes noticed the nice knee on the graph - lots of useable power.

Thanks for the post + looking into this.

Some answers;

The GPS is an adafruit module.  Lots of nice features (Hz adjustable, low power, RTC/battery to get warm fixes).  http://www.adafruit.com/products/746.  Yes can run on 3.3v.

1. I can solder / sure I can work out a bread board.  If it where a (really) simple diagram/guide I could follow it.
2. Suggest something and I will have a look. :-)
3. The Moteino & PIR sensor numbers are from Felix (I assume) here; https://github.com/LowPowerLab/RFM69/blob/master/Examples/MotionMote/MotionMote.ino#L5

I can save more power by switching off the sensors/PIR when not in use.

Question: I have read I can power the devices +V from the Digital Pins (Output, High).  For example the GPS is 25mA (max), the PIR sensor 60uA, Temp Sensor 6uA.  Can Anyone see an issue with this?

Doing that I can get the standby (sleep) running at 18uA (Moteino sleep) and only power up the bits when awake every 12 secs ie 66uA, and lets assume we can power those up for 1 sec to get readings.

Moteino (18uA): 432uAh (a day)
Sensors + Moteino awake (assume 100uA for 1 sec in 13 ~ 277 iterations / hr): 8.7uAh (a day - I think my maths is right ...)
or total now 440uAh a day (instead of 1872uAh a day)

The hourly GPS fix still blows the budget @ 10000uAh a day ...

[gregcope]

I am considering using this;

https://learn.adafruit.com/adafruit-fona-mini-gsm-gprs-cellular-phone-module?view=all

Which I think can be powered by 3V or so... need to check as it is not clear.

This PIR also looks good;

http://www.allproducts.com/ee/irtec/Product-200881595949.html

<10uA
2.4 ~ 3.6 VDC
$18, with $30 shipping.

[TomWS]

Yes noticed the nice knee on the graph - lots of useable power.

Thanks for the post + looking into this.

Some answers;

The GPS is an adafruit module.  Lots of nice features (Hz adjustable, low power, RTC/battery to get warm fixes).  http://www.adafruit.com/products/746.  Yes can run on 3.3v.

1. I can solder / sure I can work out a bread board.  If it where a (really) simple diagram/guide I could follow it.
2. Suggest something and I will have a look. :-)
3. The Moteino & PIR sensor numbers are from Felix (I assume) here; https://github.com/LowPowerLab/RFM69/blob/master/Examples/MotionMote/MotionMote.ino#L5

I can save more power by switching off the sensors/PIR when not in use.

Question: I have read I can power the devices +V from the Digital Pins (Output, High).  For example the GPS is 25mA (max), the PIR sensor 60uA, Temp Sensor 6uA.  Can Anyone see an issue with this?

Doing that I can get the standby (sleep) running at 18uA (Moteino sleep) and only power up the bits when awake every 12 secs ie 66uA, and lets assume we can power those up for 1 sec to get readings.

Moteino (18uA): 432uAh (a day)
Sensors + Moteino awake (assume 100uA for 1 sec in 13 ~ 277 iterations / hr): 8.7uAh (a day - I think my maths is right ...)
or total now 440uAh a day (instead of 1872uAh a day)

The hourly GPS fix still blows the budget @ 10000uAh a day ...

I'll take a look at your numbers and proposed hardware tomorrow.  You can power all the devices EXCEPT the GPS module from Digital I/O and this is a good move.  I wouldn't try to drive 25mA from a CPU pin, even if it's rated at that.  You can use a transistor to amplify the Digital I/O current for the GPS and still be able to cut off power when not in use.  Either PNP bipolar or P MOSFET (MOSFET would not consume any power of its own, but you need to be careful on transistor selection) will work.  I can provide some suggestions.

Are there any space constraints that would prevent you from using a larger battery?  Is there any power source on the boat (Power hookup or solar panel)?  LiPo Rechargeables might be an option if you have some power source.

Tom

[gregcope]

Thanks for the reply

I wrote a longer reply during breakfast and must have forgot to push send ...

Can you expand on why not drive the GPS from a digital pin?  Should be well under the 40ma per pin?

Checking again the adafruit module seems to draw around 20mA most of the time.  From the spec;

"MTK3339 Operating current: 25mA tracking, 20 mA current draw during navigation"

Pretty desperate to keep the thing KISS, and hence only use digital where I can.

[TomWS]

Can you expand on why not drive the GPS from a digital pin?  Should be well under the 40ma per pin?

Sure, 40mA output current is an Absolute Maximum parameter, which I, from experience, rename as "Don't EVER do this unless it's an accident".   However, glibness aside, if you look at the 'real' output spec V_OH, you'll see that the voltage is shown as 2.3V for a -10mA load and 3V supply, and 4.2V @-20mA and 5V supply.  Net, the processor will not drive this kind of load to full voltage.  A transistor will.

Pretty desperate to keep the thing KISS, and hence only use digital where I can.

My working assumption is that you want it to work reliably first and then KISS...

I'll be looking at your data shortly, but you didn't answer my question about size constraints.

Tom

[gregcope]

Sorry. That was in the other reply that is sitting on a laptop at home where I have not yet pressed go on.....

Let's assume size is not an issue

Also let's assume that reasonably small yearly AH drain (ie less than 10Ah a year) can use the boat battery as this will need to be charged within a yearly time frame anyway.

[TomWS]

Sorry. That was in the other reply that is sitting on a laptop at home where I have not yet pressed go on.....

Let's assume size is not an issue

Also let's assume that reasonably small yearly AH drain (ie less than 10Ah a year) can use the boat battery as this will need to be charged within a yearly time frame anyway.

Good info, thanks!  I was thinking along the lines of a 6V Sealed Lead Acid battery to reduce the need for high efficiency low power converters to convert from 12V (very hard to do because of capacitive losses on switching components).  SLA is easy to charge, is a safe battery and, if you have the space, can deliver a lot more AH than you'll need.  Also, connection reliability is substantially higher than typical AA battery holders in a nautical environment...

I'll munge on this a bit more and try to get back to you with some 'what I'd do in your case' info.   Two more questions:  1. Do you plan to make a PCB for this or simply wire up modules?  2. if you want a PCB, what is your limit on pin pitch (ie, you'll only deal with 0.100 through hole DIPs or, you might be able to deal with 3 lead SOT-23 transistors and maybe an 8 lead SOIC.)

Tom
"Bonus" question: You didn't mention Moteino radio.  I assume that your only wireless communication path will be the GSM radio?

[gregcope]

Was supposed to send this much earlier today...

I'll take a look at your numbers and proposed hardware tomorrow.  You can power all the devices EXCEPT the GPS module from Digital I/O and this is a good move.  I wouldn't try to drive 25mA from a CPU pin, even if it's rated at that.  You can use a transistor to amplify the Digital I/O current for the GPS and still be able to cut off power when not in use.  Either PNP bipolar or P MOSFET (MOSFET would not consume any power of its own, but you need to be careful on transistor selection) will work.  I can provide some suggestions.

Are there any space constraints that would prevent you from using a larger battery?  Is there any power source on the boat (Power hookup or solar panel)?  LiPo Rechargeables might be an option if you have some power source.

Tom

Thanks.

The GPS has an enable PIN that I should be able to leverage.

Most Modems also have a hardware on/off pin that requires pulsing for 2 secs HIGH.

When "off" not sure if either have parasitic draw to monitor that pin.

I could fit larger batteries.  Although I wanted to run off AAs :-(  The parts are all looking to be quite small bar the battery.  The GPS is tiny, the modem not much bigger.  Ie looking at a stack of PCBs not much bigger than a cube of SD cards or something the size of a cigarette packet (maybe thiner).  Three AAs would double the volume.  C or Ds tripple.

I was originally hoping (hence the post) to run off the boats 12v.  If the power draw is low ie 5Ah/yr then we could assume the boat boat batteries are charged by another source (either solar, or alternator, or trickle charger) as the owner will have to deal with that anyway (ie battery drain/condition over longer periods).

Hence back to the Original idea where;

Boat 12v-> efficient step down reg -> 5v MightyBoost + backup Lipo -> system

However the challengette is that if the Step down switching regulator is not really efficient at the 0.1ma range, we will lose more power there than anywhere else.  ie a 4Ah/hr consumption becomes allot higher.

Thanks again.

[gregcope]

1. Do you plan to make a PCB for this or simply wire up modules?

As I have never made a PCB was just going to wire up.  But have heard it is simple so may give it a go.

2. if you want a PCB, what is your limit on pin pitch (ie, you'll only deal with 0.100 through hole DIPs or, you might be able to deal with 3 lead SOT-23 transistors and maybe an 8 lead SOIC.)

Ok - you've lost me there.  I have assembled / soldered simple kits, but someone else made them.  Sorry!

Tom
"Bonus" question: You didn't mention Moteino radio.  I assume that your only wireless communication path will be the GSM radio?

Ah.  I had kept that quite.  I plan on using a Moteino radio so that I can have more sensors (temp/water level), or ones that are furthur away reporting to the base that will have the GSM, and probably the GPS to keep them simple.

[gregcope]

Ok.

Am I being massively stupid, and I do not need a switching regulator IF the current draw is tiny?  Ie a Linear one might be allot more efficient?

[TomWS]

Ok.

Am I being massively stupid, and I do not need a switching regulator IF the current draw is tiny?  Ie a Linear one might be allot more efficient?

Ok, I am being Massively confusing!   

If, for example, you have to convert from 12V down to 5V (or 3.3V) there is a big drop in voltage that will waste 60% of your power if you use a linear regulator, so it seems to beg for a switching regulator so that you can improve the conversion efficiency.  The problem, as you have noticed, is that the the best efficiency  (85+%) is when the load is measured in the high mA or low Amp range.  That's because there is a fixed overhead (more or less) due to the switching electronics.  If your efficiency is 85% at 100mA (at the input) then 15mA is lost and a significant portion of this is the switching overhead, let's say 1/3 of it (5mA) as a SWAG. 

So 5mA is fixed switching loss.  This is ok if your load is 100mA or more, but, if your load is only 1mA (or less), you still have that 5mA of loss, resulting in 5mW (5Vx1mA) out for approximately 60mW (12Vx5mA) in.  Not a good trade and it gets much worse as the load decreases because the 5mA is constant load on the 12V supply.  By the way, these numbers are just numbers I pulled out of the air to demonstrate the point.

However, if you drop the source voltage to 6V and you're converting to 5V, a linear regulator makes sense.  Here your efficiency (assuming the VR doesn't consume a lot) is above 80% and stays that way until your load drops into the double digit uA region.  The key thing to remember in this case is to optimize for the LONG time case, ie, the device's are sleeping and there is virtually no load.  A pure switching regulator (not a hybrid) would be very hard pressed to win in this case.

There is a LOT more to this story, but this is the thumbnail.

Tom