Arduino PWM Solar Charge Controller v4 - PWM85

I've resurrected a project from four years ago to bring back the ATTiny version of Julian Illet's PWM5 solar charge controller. The project, based on arduino attempts to replicate Julians low power and simple design with a couple of design tweaks.
Code, Gerber, BOM and Schematic: admw.uk/pwm85
Items used in this video:
❗️These links are affiliated and I may earn a small commission if you purchase❗️
Vici VC99 Multimeter: ebay.to/2W6LFEM
AD584 Voltage Reference: ebay.to/32BzxOg
#pwm85 #pwm5 #arduino

[@WildLifeBackyardCamera]

you are probably the only english speaker i don't put at 1.25 or 1.5X playback speed :) Your accent is just too awesome !

[@marcrives279]

Seeing ATTiny85 datasheet "§22 - Typical Characteristics" charts and tables - You can permanently disable USI module in the Power Reduction Register (PRR) to gain up to 0.1mA (100µA@5V-8MHz); - You can disable ADC module when not in use (it is tricky!) and gain up to 0.34mA (340µA@5V-8MHz); - You can use SLEEP_MODE_IDLE instead of delay(10) to gain around 0.6mA (while still using counter/timer), see forum.arduino.cc/index.php?topic=220096.0 Moreover, AT85@1MHz VCC can be as low as 1.8V. It reduces consumption a lot (but you need a minimum voltage in the driver pin). Those numbers may be optimistic, but it worth a try! Great work, that's a really neat board!

[@AdamWelchUK]

Cheers Marc and thank you for the ideas - I’ll definitely look into them.

[@AudreyRobinel]

@@AdamWelchUK Hello, i came here to say basically what he said. However, you can go much lower in power : www.gammon.com.au/power this guy does it with an atmega328, and reaches sub milli-amps power. Further down the line, he demonstrates how deactivating certain parts can go even lower, at some point going from 335µA to 335 nA, with sleep and optimisations. He provides the commands to do so; i'd strongly advise you to read this article. At this point, the consumption of the chip is lower than the self discharge rate of the batteries... For a charge controller, you definitely can afford long periods of sleep (1 or 2s) between wake-ups to monitor voltages...

[@iancollins1051]

gymnastics

[@fstoggi1257]

Adam, It's good to see that you improved on your design and for sticking with ATTiny/Atmel. When Julian announced his new design I was wishing for him to move away from Assembly language and PIC. I can understand the choices when he had a sellable product, where he was the sole hardware and software engineer, but the extra entry cost for a tinkering project is not worth it (referring to PICkit programmer). Julian's current code base leaves little room for other people to contribute bug fixes and/or new features.

[@CarlRecktenwaldJr]

I like the way you did the float voltage code at the beginning of the sketch. Much easier to get accurate battery voltage.

[@AdamWelchUK]

Thanks. I’ve tried to explain the equation as well as ensure it’s calculated at compile time rather than leave it to the mCU complete.

[@deangreenhough3479]

A rather excellent video and giving plenty of credit and an honest appraisal. Well done Adam 🏴󠁧󠁢󠁥󠁮󠁧󠁿👍

[@AdamWelchUK]

Thanks Dean. Great to see your name popping up again on my videos :-)

[@bwack]

Awesome. Looking at this video now after seeing the Solar Charge controller. I love the evolution and interaction with Julian.

[@AdamWelchUK]

Cheers. I do have another little improvement I must make a video about.

[@Tore_Lund]

The Arduino has even a deep sleep mode, where it halts everything, drawing only nA, that can be woken up from by a pin toggle, I.e. when the solar panel produces a voltage. The low draw pwm 5V regulator you use possible also has a nA sleep mode when the draw drops below a mA or so, so this charge controller could even be used in a remote sensor setup, etc. with a small 7AH lead acid battery that then would could save its charge through weeks without sun.

[@johnnodge4327]

I'm liking the design Adam. Nice and neat, with plenty of space. I think you'll find the diode gets pretty warm at 100 Watts. I'm using 10 Amp diodes and have measured them at about 50°C. Hopefully you're thermal design will help keep that in check. Nice evolution of Julian's PWM5.

[@AdamWelchUK]

I hope so. I have been able to test it under a couple of amps and I’m expecting a peak of six. If the pcb heatsink attempt isn’t big enough I have planned that it could be a place to mount something more substantial. Thanks John.

[@NiHaoMike64]

An ideal diode like the SM74611 would barely get warm even with a 100W panel at the optimum time of day. It would be worth adding an ideal diode circuit based on a standard MOSFET.

[@xanataph]

Actually most solar panels these days *do not* have a blocking diode fitted, only bypass diodes. With the extra space on your board you could take the opportunity to get rid of the Schottky diode and replace it with a reverse connected Mosfet. You should be able just to parallel the gate drive with the switching Mosfet. This will rid the design of most of the voltage drop and extra heating.

[@AdamWelchUK]

Thanks for this Xan. I can’t quite see how this would work though. I’ve looked at using an N-Channel mosfet with its source connected to the solar (and a p channel) and I can’t see how I turn it on with the gate driver I have. Julian did do a video discussing this issue here - YouTu.be/qY-6GK873ts It’d be great if you could see a solution Julian missed. I’ve been meaning to contact you about this anyway - any chance you could email me (click about on my channel page for an email address) when you get a minute?

[@renaudfensie3020]

Run the attiny at lower voltages and use deep sleep to get to average consumption levels in the order of 100's of uA

[@BGLENN-dp4tx]

Outstanding video. Thank you.

[@andresj5512]

Nice! I prefer your Atmel version.

[@Chickey]

Great video and design!!

[@AdamWelchUK]

Don’t know if you’ve look at the github, but I took your schematic design advise on board! :-)

[@andymouse]

Nice one Adam !

[@RaithUK]

Adam, I think this is a great little project your doing.. im still new the circuit building but am loving it. i have 2 tiny ebay solar panels that i want to build a small charger for and yours looks like a great place to start. I'l have to see if i can get some money and build one my self. Keep it up!

[@AdamWelchUK]

Many thanks. Send me an email and I’ll see how I can help.

[@sarkybugger5009]

Good on you, Adam.

[@NiHaoMike64]

Perhaps the next evolution of the design would be to make it into a MPPT charge controller?

[@WereCatf]

That would be *exceedingly* interesting, at least to me. If Adam ever decides to tackle that, I hope he'll also include very in-depth explanation on how it works -- even if it required splitting the video into a couple ones.

[@fratermus5502]

@@WereCatf we really need an open source MPPT project!

[@TheEmbeddedHobbyist]

There are quite a few MPPT charge controller chips out there, just can't remember the part numbers. plus all the data sheets i had are now at the company i no longer work at. just looked up the LT8490 which does the whole thing without software. just throw a few big fets at it and away you go.

[@WereCatf]

It's not easy finding them and they're generally pretty expensive. I, personally, only have experience with the CN3791 single-cell Li-Ion MPPT-chip from Consonance Electric. It's cheap and easy enough to work with, but I can't help but wonder if there aren't any other hobbyist-friendly -- ie. cheap -- ones out there. In fact, I'd totes love to find a single MPPT IC designed for 3S-packs, instead of cobbling some together from multiple single-cell ones.

[@fratermus5502]

@@TheEmbeddedHobbyist Thanks for the interesting info. I'll look up the LT8490. I'm more interested in the software side, so I'm imagining hardware hooked up to a Pi or something that controls it.

[@David_11111]

yay liked this

[@slybunda]

where can i get one of these PWM85 controllers from which you have made? it would be ideal for my car solar panel

[@electronic7979]

👍👍👍

[@unperrier5998]

Quick remark: for comparing the 3 boards, would'nt it be preferable to use only one and solder on it the 3 voltage regulators (in turn of course), because the few µA of variation between those 3 boards may be due to differences of the other components on each board. Passive components have some tolerance, somewhere around 1% to 5% presumably, and the ICs (85 and mosfet) have to be different, even if taken from the same lot/batch! I mean the ICs internal resistance, capacitance, leakage, etc... must be (ever so) slightly different between each board that when we add-up the variations of all the components, maybe the difference in current is accounted for.

[@AdamWelchUK]

Possibly that would have been more fair. However the mCU is obviously the most power hungry device and I’ve tested the ATTiny a number of times in versions of this circuit and found little variance.

[@johnnodge4327]

@@AdamWelchUK At 2.6 mA I don't think it's an issue. I know the PWM5 uses about 1 mA, which includes the LED, but all these numbers are tiny, at less than 3/4 of a Wh per day.

[@ahaveland]

@@johnnodge4327 3/4 Wh/day is significant if you want to keep a 10Wh 18650 charged through an Arctic winter... though in practice I don't think it would be a big problem. A timer could perhaps be used to say switch everything off after 1 week of no sun, or if the battery voltage drops below threshold.

[@dapelow]

Good job, and if you replace the mosfet 3205 for irlz44 you can eliminate charge pump and his associate code. The irlz44 give 50 amp for 5v gate.

[@AdamWelchUK]

Thanks. I’m afraid you can’t illuminate the charge pump. The n channel mosfet is on the high side of the circuit - it’s source is attached to the battery positive. That means you need a voltage about 5v higher than the battery to turn the mosfet on. An irlz44n would still need the charge pump so the gate voltage was above the battery voltage.

[@jacksprat7087]

Question - Using 32650 (LiFePo4) in a 4P configuration to drive a 6V game camera and being charged with a 5W solar panel. My idea is to use a small Buck converter for charging and a small boost converter for the load. My box does not allow much room for large boards. This setup will be for field use and visited maybe once every 3 months. The quiescent current for the game camera is quite small < 1ma; that required for the onboard motion detector. The max current is around 600ma and is only when the IR lamps fire for a fraction of a second. The max voltage for the solar panel is about 17-18 volts with max current of about 500ma. When the voltage falls below 4V, the current is essentially 0.0ma. So, what small buck converter will handle this voltage swing and create a constant 3.650 volts? I have seen several boost converters that are adjustable up to 6V. I’m so glad I found your channel!! Thanks, Mike Taylor Texas USA

[@MD4564]

I love it Adam, everyone's suggestions pretty good, I'm still new to electronics and of course circuit design so input from various KZbin given more help. I would like to know if some sort of measuring graph/logging can be done, so it can be read over a display/computer program..

[@AdamWelchUK]

Unfortunately not. That would require more pins on the microcontroller and probably use more power. I’m not sharing it can’t be done, but it would be a fairly big redesign. The cheapest commercial option I’ve used that can do this is the landstar series from EPEver (the more recent models have RS485 ports on them and you can read that data and log it to a database or to the EPEver windows software). I’ve covered all of that in other videos - but sadly it isn’t a solar controller you build yourself. Otherwise another arduino or similar with some modules could log battery voltage and current via serial as a basic logger.

[@MrBrymstond]

Great Job.

[@AdamWelchUK]

Thank you.

[@Minipimmer]

a lot of your effort has been directed to the improvement of the standby current draw but how about the efficiency of the MPPT "at circuit level"? the topology of this circuit is odd because every time you switch the MOSFET you force the voltage of the battery into the PV panel. This should cause the current of the panel to be a series of short impulses followed by constant steps, I think this may break the MOSFET on the long term. Has the current waveform been checked?

[@kkathulgangadharan2262]

Please add the led indication of voltage and pwm

[@sparehead2214]

I like your determination Adam, keep it up! It's very difficult to find ultra power efficient LDO's with a wide input voltage range. Most of the very efficient (low power consumption) devices are limited to no more than 6V peak input. Normally they are marketed with terms such as, 'micropower', ultra-low power' and the like. It might be that you've taken this design as far as is practicable. Maybe consider looking at the ultra low power MSP430 processors. Yes, I know it's not Arduino or PIC, but they have been used in battery powered devices such as multimeters for many years. It was designed for low power battery applications. Maybe something like the MSP430F2013. A development kit starts at just $24 (USD), www.ti.com/tool/EZ430-F2013 Might be worth a look, what can you do during a lock-down!

[@AdamWelchUK]

Thank you for the suggestions. The next exciting instalment which is available to view I make hardware changes which help reduce the consumption further. Since that video I’ve found that code changes have also saved me quite a bit too. Video still to come on those changes. Stay safe!

[@Space_Reptile]

how expensive would these be to buy if pre assembled? their small size and simple design makes it ideal for an ebedded solution (while being a bit more flexible than say the gumstick PWM)

[@richardwayne1990]

considering a far more complex PWM controller costs as little as £5 id imagine £2-£3 each

[@AdamWelchUK]

I’m working on a spreadsheet to see how much they are costing me - doesn’t include my time though and I’m a pretty slow pick and place machine! I’ve not looked into it, but the new building service from jlcpcb might be the best place to find a real commercial price.

[@Space_Reptile]

@@AdamWelchUK it would be a nice product for you to sell, seeing how well the PWM5 did

[@marcvan]

Hi Adam, I have been subscribing for a while (and to Julian's). Good stuff. Question: Why just bring the charge to 13.5V A good charger IMHO should do Bulk, Absorption and then Float. You just bring it to float. For AGM and GEL used in cyclic applications as a suggestion (and to avoid adding a current sensor) can you detect start of day ie solar I/P then charge to say 14.4V, hold at that voltage for a time, say an hour THEN jump to a Float of 13.5V . If the battery isn't cycled a lot then reduce Absorption time. Interested in yours and others feedback. I'll order a few PCB's from JLC and have a play with the code. Thanks again.

[@AdamWelchUK]

The code is Julians - except for the adjustments for the ATTiny. I believe the answer to your question is more about keeping the code as simple as possible so the charge controller is able to adapt to changing conditions quickly. The PWM5 does charge to a higher voltage for a short period on a morning, and that is something I’d like to see in the arduino code. It’d be great to have someone else on board looking closely at the code - thank you.

[@marcvan]

@@AdamWelchUK will have a play

[@marcvan]

@@AdamWelchUK Hi again Adam, I have been playing. I have built what I call PWM328. Barebone using 328p chip. Keeps things very simple and standard. Only get 8ma but I use sleep at low light/night and get micro amps which is the time you don't want to use power anyway. Included also is temp compensation (important) and jumpers to enable temp, wet cell/ dry cell chemistry and cyclic/float charging. I also use a bi led to show all states of charge. If interested how do I send you details? I would love your feedback especially the use of a P chan mosfet rather than a charge pump. It appears to work well but do not have oscilloscope to check.

[@tupai62]

do u have make solar charge controller for 6v SLA battery?. what the video link adress?

[@andreasschindler6077]

The accuracy of the 5V regulator doesn't matter by using the internal 1.1V bandgap reference for ADC.

[@tengelgeer]

But the 1,1V reference isn't accurate either. Although it should be pretty stable across temperature I think.

[@AdamWelchUK]

I did look at this but the 1.1v reference can be anywhere in between 1.0 and 1.2 volts according to the data sheet - that’s like 10%! The 2.56v reference anywhere in between 2.3 and 2.8 volts. I guess in practice they might be far better?

[@pfeerick]

@@AdamWelchUK The trick would be to first calibrate the internal reference (i.e. measure and store in EEPROM), and then the tiny85 could compensate for any drift in the supply voltage... It would be better to go that way as the 1v1 internal reference is stable in the full operating temperature range IIRC.

[@TheEmbeddedHobbyist]

use an external ref. then you can pick the accuracy you want. but they will all suffer from drift over time. even your resistor divider chain will drift. so you need to use high spec resistors! it all depends what tollerance you want on the battery voltage.

[@dartfrogdk]

Hi nice project, you might have said it in the video but what is max input voltage from solar panel?

[@AdamWelchUK]

This is for a typical 36 cell solar panel up to one hundred watts. It’s the sort of panel sold as 12 volt, but one that has a open circuit voltage of about 21 volts and a maximum power pint of around 17-18 volts.

[@DAILYTECHNEED]

Dear Adam, what should be the output voltage of the solar panels that feed energy to this circuit and up to how many watts can I connect?

[@AdamWelchUK]

Hey. It’s designed for 36 cell panels which typically have an open circuit voltage of 21 volts and an MPP of 18 volts. Now Julian who originally designed the circuit rated his PWM5 for 6 amps - or about 100 watts. I however have changed the solar input diode to a completely different package, and haven’t been able to test at six amps yet. I’ve run 3 amps through it for two hours and it’s fine, but I need more sun!

[@DAILYTECHNEED]

@@AdamWelchUK That's great hearing from you. May I suggest you something? There's a channel named "TheDIYGuy999". He upgraded It to another level. You may have a look. Here's the link-kzbin.info/www/bejne/jIS0dmunpb2pjKc

[@jamest.5001]

Is there any mppt controller kits, that offer larger current capabilities? To charge liion batteries? With about 50-60v input, and 29.4v out? I'd like to charge up to 100 amps, crazy huh? Floating liion LIFePo cells at full voltage will destroy them, I heard of a test, and depending on conditions, nearing 1% decline of capacity per cycle! Batteries cost too much for that, I wonder if floating at a lower voltage would have any effect, about 4v/cell, but it shouldn't show and issues if held at nominal voltage, 3.2LiFe, or 3.7v for liion. Possibly, adding a normally closed relay to PV input with a voltage controlled switch, to disconnect pv, when the liion battery is full! To stop float, without having pv in without battery on the controller..

[@milenedejong1400]

look at the victron brand . try searching victron mppt .

[@pirateman1966]

Why not use a logic level gate P-Mosfet like the SQP50P03-07_GE3 (7mOhm @ 30A) and a single 2N3904 to drive its gate for $2 total? Instead of this horror of 3 transistors, a charge pump and a mosfet. Use a TL431 ($0.04) as a 2.5V reference. You can't get a cheaper, temperature compensated ref IC. Whoever designed this, is still living in the 80's.

[@xanataph]

Good points, but what's wrong with living in the 80's...? lol

[@antibrevity]

I'm interested in your recommendation and hope someone looks into it. Adam? Others? This circuit was initially designed by youtuber Julian Illet a number of years ago. He's quite knowledgeable, but you are probably right about the 80's thing as he's slightly older than me and probably did much of his professional training in the 70's and 80's :) .

[@andymouse]

Julian is over 100 years old....quite young for a Timelord.

[@solventtrapdotcom6676]

Aren't the resistances higher on the p-channel? As in, a lot higher? I'm trying to do an IGBT version...

[@UpLateGeek]

Jeez, these product counterfeit makers are getting more and more brazen! Now they're openly publishing videos about cloning other people's designs on KZbin, and asking others to help refine their code! Not to mention publishing their designs on the Internet so anyone else can make their own clones and even improve on them! Is there no end to their audacity! ;)

[@richardtallian5156]

You could try an attiny13 if the code fits on it.

[@pfeerick]

Nice! :) Next things I'd be looking at are (not necessarily in this order): a) lowering the system voltage if possible - you can get to close to half the power consumption... check out chapter 22 of the datasheet, where it shows the relation between clock speed, active current consumption and system voltages. b) seeing if it's possible to use the Internal 128 kHz Oscillator - at 5v that should bring you BELOW 0.5ma ... :-O 6.2.4 and 22-5 - maybe look at using this maintained BSP instead? github.com/SpenceKonde/ATTinyCore c) seeing if can we do away with the charge pump, switch to a logic level mosfet d) what about an ideal diode? e) calibrate the 1v1 internal reference voltage and store it to eeprom to be able to correct variance in the system voltage / regulator output f) ensuring any unused peripherals are disabled, ensure unused pins are floating (inputs), use sleep mode instead of delay, shut down ADC when not in use (and especially before sleeping), etc.

[@nirodper]

the charge pump is to drive a n-channel mosfet on the high side, so the ground is continuous

[@pfeerick]

@@nirodper Don't you mean P-channel? N-channel are ground switching mosfets ;) And parts like the FQP27P06 are good for switching over 10A at around 4.5v... so no charge pump needed.

[@nirodper]

@@pfeerick no, n-channel, and he is driving it on the positive rail so he needs a voltage higher than the supply voltage, which the charge pump provides. The reasons are n-channels are slightly more efficient (difference was higher when the pwm5 was designed I think) and having a continuous ground and switching the high side is nicer with our conventions

[@antibrevity]

@Peter Feerick In regards to: e) calibrate the 1v1 internal band-gap reference While watching some of Julian's videos from a few years ago, I initially thought that using the internal band-gap reference like the one on the ATtiny/ATmega uC's would improve the project, but this doesn't seem to be true in reality as *all* of these internal band-gap references seem to have an error margin of more than 9%, even those on microcontrollers from other manufacturers. Rather than calibrate the internal reference (Vbg), which involves using a separate sketch or special operating mode to output Vbg to a pin for measurement, I suspect that Julian was right to use the actual regulator output (Vcc) as the reference since it appears that the LDO's discussed in this video are actually *more* accurate than the internal reference; most are < 1% or better. I haven't found characterization data for either the AT-series Vbg or the BOM LDO, so I cannot compare temperature drift. Anyone capable of building this PCB is also capable of measuring the regulator output (Vcc) in-circuit, whereas calibrating the internal reference is more complicated and may not even be practical in-circuit. It's also *very* easy to short out the internal reference when it's connected to an external pin; such delicate work is best avoided. Characterizing the float battery output drift over a range of temperatures would include all in-circuit drift and likely improve accuracy substantially, but this would take some effort and may not be entirely applicable to builders who sourced different brands and batches of resistors, etc. It would probably still be an improvement for a module that may spend its life outdoors in sun, rain, and ice, though. The voltage divider resistors can easily be calibrated in the existing sketch if builders have an accurate ohm-meter to test them with. Most cheap multimeters today are quite accurate at this resistance range (

[@pfeerick]

@@antibrevity Yes, it does require measuring the voltage at that Aref pin with the Vbg turned on, but if this were being designed from the start with this in mind, that would not be an issue. You would simply ensure that a test-point is available - no components to populate, nothing to short out! If you can measure the Vcc, you can measure Vbg just as easily, if not easier. The only hassle is *having* to measure the internal reference voltage, enter it, and also measure and enter the self-measured supply voltage for secondary calibration. If you really want to get fancy, you can use the internal temperature sensing to make it even more accurate with temperature compensation. But then again, even with an external reference, depending on how accurate you want it to be, some tuning is needed. The realy annoyance is that Atmel never did give the specs on the Vbg stability, but this is probably as it is a freebie, as it was implemented for the Brown-Out Detection and designed for low power consumption, not absolute stability. Obviously, if this were being done as a production run, you'd be running a test sketch anyway, and would probably automate the measurement and storage of calibration values in EEPROM... (as opposed to the manual entry I currently do)... it really depends. At least with even cheap DMMs being pretty accurate as you say, it's within the realms of the home enthusiast to get the accuracy of these things pretty darn good ;) I can appreciate though as Julian showed in his investigations though that it's isn't just as easy as "turn on 1v1 internal reference and you're done"... but then again, where's the fun in that? As far as float battery voltage drift, better to have an external temperature sensor on that battery like most non-junk solar charge controllers, so you can do temperature compensation tweaking.

[@Luke-san]

First view and first like ... Adam wrong time of the year to start playing with solar

[@AdamWelchUK]

Cheers Luc - it's true, this would have been better in spring!

[@MrBrymstond]

It's never too late if you live in Florida.

[@RobinHilton22367]

Not here in New Zealand, Summer is on its way :) I had 1.5 kW of solar installed in winter and even in early spring I am getting it running at over 75% (inverter only shows % ranges) for 4+ hours which means am generating around 6+ kWh per day even on average days.

[@Luke-san]

@@RobinHilton22367 Depends what you call average days of course :) . Here in the Benelux an average day might be 1 hour of full sunshine in this period.

[@RobinHilton22367]

@@Luke-san Wellington, New Zealand from some historical stats I read somewhere experiences 5.5 hours per day of PRIME sunshine when averaged over the year. Obviously more biased towards the spring/summer months.

[@solventtrapdotcom6676]

400v battery IGBT version...

[@OntarioMiner]

Does anyone build these to sell on eBay or the likes online?

[@HiruS22]

He’ll have a moan at you mispronouncing his name again 😄

[@AdamWelchUK]

Julian? I don’t remember him moaning at me.

[@Graham1904]

You are mixing up mA an up quite badly. You said 7.5mA when you should have said 7.5uA for the original. Your measurement are in mA where the datasheet says a quiescent for the 5Vdc regulator of 2.7uA. You should consider the microcontroller quiescent current as well. Quiet confusing to the novice.

[@AdamWelchUK]

Hello. I did at one point say milliamp rather than microamp. That is true, sorry. You’ll understand that the data sheet can only quote a figure for the regulator itself - because they have no idea how you are going to use it. I on the other hand can only quote figures I’m seeing in a circuit. Further on from this video I have managed to decrease the quiescent current of the whole circuit by redesigning the hardware a little - and there’s more to come with software improvements. Keep watching :-)

[@wynandwolmarans6689]

Plz send me

[@whatthefunction9140]

blasphemy