Building an MPPT booster circuit

[elt]

Okay, I think I finally get it: the problem with using input voltage (only) as a control signal for the boost converter is that for a given voltage input, there's no indication of where the mill is on its power curve. My belief was that starting with the input voltage of the unloaded mill, that the processor could program the PWM to add "a certain amount" of load. Perhaps the "certain amount" could be learned experimentally and then programmed into the microcontroller. The problem is after that, the processor has no clue what's going on with the mill if the voltage changes. Did the mill load up? Did the wind die down? As a work around, the processor could remove the load every now and then, wait for the mill to return to its "wind speed" and then restart the PWM based on the (perhaps different) voltage read. But the mill wouldn't me making power while the processor is waiting for the mill to speed up and since this is a low power boost application, it can't be the best approach to throw away energy...

So I added a current sensor to the circuit. That will allow the processor to optimize amps to the battery. It may still have to wait a little bit for the mill's speed to change to match the load but at least it will be generating power during that time. I have no experience with hall effect sensors; I chose an Allegro part because they're usually pretty good about sending me samples. That particular one has nice beefy pins and is magnetically shielded, I thought that'd be a plus with it in the vicinity of the coil.

And since there were still unused pins and board space, I connected another analog to digital converter input to sense the battery voltage and added a relay driver to implement a shunt load controller... needed one of those anyway.

As always, feedback is greatly appreciated!

- Ed.

[Flux]

It's probably a bit early to worry but with that processor you can easily make sure the pwm has zero pulse width until the dc volts reaches perhaps 8v. This will let the mill start easily.

With a 50% pulse applied there is no big power transfer until the dc reaches about 10v but as a result of recovering energy from the inductor there is some load from below blade stall speed.

I held the dc supply off until 8v but you can easily hold the pulse off.

flux

[BigBreaker]

RMS voltage is basically proportional to the RPM of the turbine, but for MPPT you are better off just sensing the voltage signal's frequency than it's RMS magnitude.  The "input" voltage that you can read will be altered by the load the MPPT presents in a way that will cause feedback.  The voltage signal's frequency is as good as a tachometer without needed an extra sensor.

Compare the current into the batteries to a target level from a lookup table.  The lookup table only needs one other dimension, RPM.  Each RPM has a target current.  If the current is too low, increase the % "on" in the MPPT's duty cycle (when it's pushing current into the batteries).  If the current is too high, decrease it.  Optimization search would be nifty but unnecessarily complicated.

Once the lookup table is calibrated, a search is redundant anyhow - the optimal point is always known.  If the turbine is overspeeding then too little current is being pushed for that RPM - push more current.  If the turbine is stalling from trying to push too much current, ease off the duty cycle so the turbine can speed up.  At your maximum safe RPM you'll want the duty cycle to be 100%, of course.  That's the time for stall!

You'll want to check for full batteries and other boundary conditions, but MPPT doesn't need to be too complicated.  I'd love to see MPPT integrated into an inverter and with diversion load protection for full batteries.  That's where the fun begins.  Ok - time for Amanda to smack us all down with her expertise.

[Nando]

MPPT solution requires to determine the WATTS been generated.

Solar, Wind and Hydro ( small) all have a bell shaped power output.

Therefore one needs to read the voltage and the output current and decide what direction the PWM is operated to decrease or increase the pulse width to insure that the power available is not over extracted to transfer the available energy.

You need to read the PWM duty cycle and the current then modulate the PWM observing the variations to re-direct the PWM duty cycle -- this type of modulation needs to be done at a much lower frequency with the power source dynamic response that for wind mills, the modulation could be done around 0.5 to 1 second, 0.1 to about 0.5 for Solar and greater than 1 second for Hydro.

The simplest way would be with analog MPPT, micro processor though less parts, the software should be able to have good PWM frequency,( 20 + KHZ)with good duty cycle range (0 to 100 )

Nando

[Nando]

I forgot to inform that You can go to my files and down load the MPPT.pdf for a MPPT Buck converter low power that could be a good step for You to see how the MPPT algorithm can be implemented

[elt]

Did you mean the Electronic Design article? I thought that the notion presented of "dithering" the PWM to hunt for the MPPT would be applicable to software implementation. I also think that the TINY45 has enough program space for a complete PID implementation ... once it's built, I'll see whether it matters. I'm hoping to keep enough program space available to add some smarts to the diversion control. Since the processor can track the battery volts and has a built in temperature sensor, I think it ought to be able to be a little bit smarter than just turning on the shunt at some fixed voltage. Before I can make it smarter, though, I'll have to get smarter about battery charging...

Thank you,

- Ed.

[stephent]

What's the return path for the inductor?

[altosack]

Wow, you're going to add diversion control too, with only 6 I/O pins ? I would be truly impressed. However, since you are already getting the battery voltage, you will just need the net battery current (not the wind generator current) and a PWM output for the dump load, just 2 more pins. I think it would be prudent to add a second controller that is a backup dump load just in case (the NEC recommends this, even).

Battery charging, all you need to know:

1. st stage: Full current until the voltage reaches the absorption set point (usually about 2.4V per cell for flooded lead acid).
   
2. nd stage: Reduce net current going into battery gradually to maintain the absorption set point, by either blocking it in the case of solar or sending the excess current to a diversion load. Both methods would use PWM.
   
3. rd stage: After either a set time at the absorption set point (usually 2-4 hours) - this is the way most commercial charge controllers do it - or after the net current has fallen to about C/200 to C/100 (a better way in my opinion), reduce the current further to maintain the float voltage, usually about 2.2-2.25V/cell.
   
   

Your algorithm should go through these 3 stages every time it charges fully, but not more than once per day; after that, if it comes back up to the float set point with a current of less than C/200 in the same 24-hour period, just maintain it at float. If it takes more current than that, go back to the absorption stage and top it off again.

Equalization: about every 30 days (or up to 90 if the batteries aren't worked that hard), after the end of the absorption stage (2nd), raise the voltage up to the equalization set point, usually about 2.5V/cell, and keep it there for 2-3 hours. If your charging source can't keep it there that long, accumulate time at that voltage each time it is possible until 4 hours total in a 24-hour period before ending the stage.

Just because no commercial controller (that I'm aware of) does true 3-stage charging with a dump load doesn't mean that we can't.

Best Regards,

Dave

[elt]

Hi Stephen,

The five pin connector on the booster will be wired to a high current 3-phase rectifier so the return path from the inductor is through "ground" which is connected to the "minus" connection on the external rectifier ... I didn't give any systems information in this thread; briefly: The mill is designed with a higher RPM cut-in (for 24 volts) to better match the power curve in higher winds. This will result in a higher stall speed and less stator heat. But because the 24 volt cut-in speed is so high (about 12 mph for my mill,) there's significant power available at lower wind speeds. The booster circuit will kick in below 24 volts and turn itself off when the alternator reaches battery voltage on its own.

- Ed.

[SamoaPower]

Sorry Dave, can't agree that this is "all you need to know" about battery charging.

Note to Ed: can't resist getting on my soapbox again here, no hijack intended.

Dave, you seem to be quoting the party line (commercial mfgrs) for charge algorithms and it can work reasonably well (but not accurately) for fixed charge sources such as mains and perhaps hydro. However, for variable sources, like wind and solar, it falls quite a bit short.

The basic problem is that we haven't yet come up with a reasonable way to dynamiclly, electrically determine SOC (state of charge) of batteries.

As these representative curves (varies with chemistry) show, using fixed voltage set points to switch algorithm phases can't work very well when the available current varies. Using time and/or a current percentage also doesn't make good sense.

I disagree with the generally accepted equalization interval of 30 days. I believe that equalization should be done much more frequently to help prevent the sulphate transition to the hard-to-remove type.

There are other points of your comment that I take issue with, but I don't want to risk Ed's ire in making this too long.

By the way, a C-40 does implement the full three stage algorithm in both series and diversion modes.

[elt]

Note to Ed: can't resist getting on my soapbox again here, no hijack intended"

Ir'a inevitable that I will come to you crawling on my knees asking for advice on charging! But at first, I'll only be able to implement something simple; eventually maybe something better that is stil short of "best."

- Ed.

[Nando]

I did not pay good attention to the circuit before.

I question the connections that seem to be misplaced, though there is not good info to really say what changes should be made.

The rectification of the 3 phase leads seem to be half way and not full wave, the three negative diodes are missed in the drawing, the negative of the diodes should be the "Ground" of the booster.

FAST current diode with good voltage break down ( peak generator voltage plus battery voltage)in parallel with the power choke.

Analyze the MPPT article, Yes You will need dithering as defined in the article.

My suggestion is to read and understand well the reasoning for the searching mode that is required to attain the best MPPT solution.

Shunt regulation can be done with most of the software for the MPPT, in this case the loading is operated in the opposite mode. -- basically --

Also do not shunt the battery, shunt the wind mill, this way the battery may last longer since is not cycled so often -- You are doing high PWM with duty cycle variation which can maintain the generator at a constant output voltage below the battery voltage

Nando

[altosack]

SamoaPower,

You're right; I misworded it badly. I should have said "all you need to know to get started." Anyway, using the simple things that I said will get you a charge controller that works at least as well (and perhaps a bit better) as any commerical one.

By the way, I agree with nearly everything you said about charging (not about the C40; both it and the TriStar implement 3-stage charging in a different - and inferior - way in diversion mode vs. series).  What is particularly humorous to me is that someone accused me of repeating the party line of commercial charge controllers ! Reading back what I wrote, it almost sounds true, but it's really not my intention at all.

I'm currently (sort of currently) designing my own microcontroller-based battery monitor + MPPT solar + diversion charge controller that I hope will fix most of these problems (and yes, I'm working on a way to dynamically keep track of SOC for varying charging sources); once I get a bit further on it, I'll present it to the board, and I'm sure (and I hope) you'll have comments.

Dave

[elt]

Hi Nando,

"The rectification of the 3 phase leads seem to be half way and not full wave, the three negative diodes are missed in the drawing, the negative of the diodes should be the "Ground" of the booster."

I've written about that in the text but I realize that I run on a bit and am not surprised that some folk skip over it... here's a picture.

This circuit the the "low power MPPT booster" and doesn't include the high power rectifier or battery... The MPPT boost circuit shares the low-side diodes in the "high current" rectifier. The booster is only operating when the alternator voltage is below battery volts. It turns itself off when the voltage rises and high-side diodes in the high current rectifier begin to conduct current directly to the battery.

"FAST current diode with good voltage break down ( peak generator voltage plus battery voltage)in parallel with the power choke."

I've never seen a diode used in parallel with the inductor; what does it do?

Thank you,

- Ed.

[Nando]

I was assuming that the rectification should be full bridge 3 phase for the choke section -- did not read all --

The diode in parallel with the coil is connected with the anode to the positive rectified terminal, same place the Choke should go then the Cathode is connected to the battery, since the choke has a diode to dump the energy when the MosFet is turned off ,

This diode is to allow the generator to charge the battery without going through the choke when the Booster is not longer working,

The diode does not need to be high speed, regular diode will do.

I WROTE too fast and did not edit anything.

I have used them for several decades when certain Booster required them (when the choke internal resistance that has to be taken in consideration).

The way you have will ( extra diodes) work OK and No parallel diode will be needed with the large 3 phase rectifier bridge.

[SamoaPower]

Dave,

Looking forward to seeing your controller design. It would be nice to see someone finally do it right.

Still disagree about the C-40 (have two), but will let it go.

[elt]

This is a question about inductor values, switching frequencies and current ripple:

I got some .75mm Litz wire to use to wind the inductor for the MPPT booster circuit. I had figured 80 turns for 220uH and that four in hand would fit on my spool. Lightly twisting the four wires together made it fairly bulky and I only got 63 turns on the spool or about 150uH by my calculations. (I don't have an L-meter or stuff to do an AC test circuit right now....)

As an aside, I'm glad I found some info on soldering litz wire; even with info on the web, I had still had to experiment a little to get the four wires soldered together.

I'd picked 220uH based on info on the coilcraft website. It indicated that 100uH of less would be fine for the little boost needed to get 20+ volts at 7 amps up to battery voltage but that a lot more inductance would be needed to boost 10 or 12 volts with a reasonable amount of current ripple. Do I care about "current ripple"? Probably not in terms of charging the battery but since I'll be using feedback from the output current to match the alternator (MPPT approach) I thought it'd be a bit easier if the current wasn't all over the place.

Then it dawned on me, just switch faster... From the coilcraft web site, it appears that switching 8 or 10 volts at 100KHz will give about the same current ripple as 20+ volts at 30KHz. I'm not an analog guy but I have looked at converter data sheets and app note and it seems to me that 100KHz isn't that fast. Are there some additional considerations or other plans that I should make if I raise the switching speed from the 30KHz that Flux originally proposed for the booster?

The PWM frequency is programmable so running it faster isn't an issue; I could even vary the switching frequency - running it faster at higher boosts and slower at lower boosts if it made a difference.

BTW: I found a 10 amp 100uH coil from hammond available at digikey for $6 USD. I spent $8 for the litz wire on ebay and wouldn't be surprised if my coil was "off" in some way.... If there's no problem upping the switching frequency, then there's a ready source of a high current coil for anyone else that might want to try something like this.

- Ed.

[Flux]

You may be able to run this at 100kHz but at the higher frequency the surprise factors increase rapidly. With your printed circuit layout you may need a double sided board with a ground plane.

I have a feeling that you are playing with tiny inductor cores, I realise that some people have difficulty finding ferrite cores but it is far easier to start with a good big core. Your 80 turns seems incredibly high. I used about 20 and that gave an inductance far higher ( about .5mH I seem to remember). The cores I used had an unusually large air gap and with a normal gap about half those turns would have done.

I have just found a core, here are the rough dimensions to give you an idea. Something smaller would have done.

two E cores 70mm wide, 35mm legs to make 70 x 70 outside. Thickness 16mm.

In these the centre leg is round, 16mm diameter and one leg is cut back 5mm short to give a 5mm air gap.

I have a feeling that these are about equivalent to ETD 70. Probably ETD 39 would be adequate or E42/21/15. With conventional cores you need a non magnetic (and non conducting) shim to produce the air gap (2mm?).

The first converter ran at audio frequency ( 6k ?) and was wound with single strand wire. The layout was a mess. When I changed to 30k I had considerable trouble with dead mosfets and it took some time to find a suitable layout. If you go to 100k you have to seriously think, any wire can be an inductor, capacitor or resonant circuit depending on its shape and position. Capacitors are an inductor roughly equivalent to a wire of the same length, so you have to pick tiny ones with leads close together ( ok for 5v computer supplies but more of an issue at power levels).

Flux

[elt]

I have a few EC 35-17-10 cores with a 1mm air gap. Two together make a core about half the size of yours with a 2mm air gap. The issue I have using them is that the EPCOS data sheet gives numbers and formula that they give for calculating saturation per 100 turns. I imagine that there must be a way to scale that to other turn values but I can't find it. Using their formulae I get an AsubL of 77 for a 2mm gap, 53 turns for 220uH and reduction in L of 10% at 3.6 amp per 100 turns... but I don't know to scale the L-reduction to see how the coil would handle 10 amps.

- Ed.

[Flux]

My guess is that at 50 turns you will need 7A for the 10% reduction ( same AT)

By the time you get to 10A the converter will only be boosting a volt or two so I think you will be ok. With a 2mm gap you will have little danger of saturating.

If you have plenty use 4, 2 each side.

Flux

[Nando]

ED:

You need to calculate based on the project itself.

First: How much power do you expect at peak power, before the generator starts producing enough voltage to directly charge the battery bank.

Second: You need to assume the maximum Duty Cycle, in this case, You need to define the max input voltage ( Vin) and peak battery voltage ( Vout), then define the duty cycle ratio Vout / Vin.

Choose a duty cycle below this value to avoid possible erratic behavior of the loop of the controller 0.85 of it will be OK.

At this point, You need to define the generator WATTS output.

Then calculate the peak current that can be generated and knowing the PWM and the duty cycle or ratio one calculate the MOSFET ON time then use the formula

Vin=( L * dI) / dT to get the L= choke

Or L = Vin * dT / dI NOW you have the choke and also you have the peak current which You need to multiply by 2 to get the Choke Saturated Current = Isat .

For proper work, to avoid stalling the Mill, YOU NEED to make sure that the peak current needs to follow the power curve of the mill as I noted in a previous message.

Knowing the Inductance and the Peak current, then one can start choosing a core that does not saturate.

Nando

[elt]

Hi Nando,

" L = Vin * dT / dI "

That's the formula that I've been using... well, that's the formula that the web app I've been using claims to use (http://www.coilcraft.com/apps/selector/selector_1.cfm)

The issue isn't the 7 amp max power before cut-out case, it's the .7 amp min power at cut-in case. Using the same switching frequency (but appropriately different PWM values,) the formula will say that the booser needs over 500uH at cut-in but only 60uH or less at cut-out.

So I guess I'm concerned about two things at once... I'm thinking that raising the switching frequency will lower the L value needed at lower power, higher boosts. That L is still going to be higher than the 60uH value needed for the higher-power, lower boost case. I guess that since I'm going to have a big L value in there anyway, the circuit probably has some headroom if higher currents lowers the inductance value of the coil.

So what I'm thinking now is to vary both switching frequency and duty cycle with current rather than just using a fixed, higher speed switching frequency; I'm hoping that at higher currents using the lower switching speed won't make as much noise. Do you think that's true?

- Ed.

[Nando]

ED:

You are using two points were one is needed. ( 0.7 and 7 Amps)

The peak power at cut-in is your booster maximum current after that the booster circuit is OFF, and the power diodes start charging the battery directly and not via the booster.

You choose the PWM and then as I noted before You do the steps to define the Choke

The standard voltage that Coilcraft has may not be the right set up for You since You are NOT Making certain output power, you are setting a maximum voltage and the current at the peak voltage and below it You need to limit the current to a lower values following the generator power curve.

So the Booster will have a maximum defined Ton time and following the power curve toward Zero voltage, the Ton will be lower

If at cut-in you have 0.7 amps then just below that is the peak current of the booster circuit

Define the duty cycle and make the peak current 0.7 amps * 1 / duty-cycle

Using Ton + Toff = T of the PWM frequency, get the time of one Hertz and use for the formula as You indicate above ( as well as I).

The 7 amps is the peak of the generator at peak power and wind ( cut-out) or am I reading You wrong ?.

I am assuming that the booster will operate up to Cut-in and from there to cut-out of the wind mill the diodes will charge the battery WITH THE BOOSTER OFF.

On these bases not need to have a second and lower value inductor -- unless you something else in mind.

Nando

[elt]

Hi Nando,

I think I used non-standard terminology... when I wrote cut-in and cut-out, I meant for the booster circuit. Design parms are that the booster will turn on when the mill is making about 10 volts and 20 watts. Not sure what efficiency will be but I expect to be about .7 amps at 24 volts. I expect that alternator will be making about 175 watts when it makes 24 volts on it's own so that at the time the booster turns itself off, the output current is about 7 amps.

While using the coilcraft site, enter both different input voltages and different output currents that will give watts on a cubic curve between the two endpoints (like the blade will produce.) The results gives what I'll call a "nominal" PWM duty cycle along the curve. I expect that the maximum power point will be near those values but I won't hard-code them into the microcontroller, I'll use a search alogrithm to find them dynamically.

"Using Ton + Toff = T of the PWM frequency, get the time of one Hertz [...]"

fine, but

" L = Vin * dT / dI "

Considering that Vin and dI are determined by alternator and the wind, that leaves two variables (L and dT) in one equation... that isn't enough to generate a single number as "the answer." And considering that Vin and dI change with the wind, it appears that different L values are needed at different wind speeds.

As I understand the formula, the L value calculated is actually a minimum inductance required; higher values are okay so the higher inductor value calculated for the lower-power end of the curve will be more than enough for the higher-power end of the curve (and that's okay.) ... so a single L value is okay, it should be picked from the low end of the power curve, not the high end... but is has to be able to handle the amps at the the high end of the curve though it seems likely that if the wire doesn't melt, that even a derated inductor (due to high current) will still have enough inductance to work at the high end of the curve.

Flux uses about 500uH; that matches practically perfectly the results of the coilcraft app for low voltage in, low current out at 30KHz. IIRC, he got the core out of an old television. It's unlikely that I'll get that much inductance without winding a giant air core inductor. With what I've got, I can wind a high power inductor with about half that value. For less than cost of the wire I can buy a 100uH 10 amp inductor... so that's why (inductor requirements at the low end of the curve) I am asking about making up for a lack of inductance by raising the switching speed.

- Ed.

[elt]

Hi,

Just a couple of pictures of the board for the booster circuit.

I use the "toner transfer method" for putting the resist on the board. Briefly, you use a laser printer to print the mirrored circuit on a special paper that has a water solluable coating, then run the board and print through a hot roller laminator. The laminator fuses the toner to the coper, then you put it in water to dissolve the coating on the paper. The toner sticks to the copper and the paper floats away. If you're just going to soak it in etchant, that's enough. I do another step with a plastic "transfer film" that allows you to rub the board with cotton balls soaked in etchant ... this is about twice as fast and uses about 1/4 of the chemical as soaking the board in a tank.

Here's a picture of the board after I've transfered the toner and then run it through the laminator with the plastic transfer film. (You can see the black toner through the translucent film.)

When you peel the transfer film off, it only sticks to the toner...

... the plastic film provides additional protection for the toner; otherwise the toner would come off when you rub the board with the the cotton balls. Once the board is etched, acetone is used to dissolve the plastic transfer film and clean off the toner underneath. Here's a picture of the etched board.

I did make a partial ground plan for the top of the board. I opened up the the holes for the pads quite bit because I'm not that good at aligning the two layers. The instructions suggest etching one side and drilling, then using the drill holes to align the second side; basically doing two etching steps. I'll do that if the board has two signal layers but I thought that doing both at once would be good enough for this one... the ground plane alignment is "okay"; looks like I'll have to trim around two component pins but that's not too bad.

It'll probably be a week or so before I have all the components; sometime between now and then I'll drill the holes...

- Ed.

[commanda]

If you (pilot) drill out the 4 mounting holes at the corner of the blank board, you can use these as alignment marks to get the two pieces of transfer film aligned.

Amanda

[elt]

Thanks, I'll give that a try next time.

 - Ed.

[elt]

"If you have plenty use 4, 2 each side."

I do have enough to do that... Will that change inductance (change A-sub-L)?

Thank you,

- Ed.

[elt]

Will that change inductance (change A-sub-L)?

I don't know... so I wound one and then built this circuit

http://cappels.org/dproj/Lmeter/lmet.htm

to measure the inductance.

What I got was that the inductance increased by a factor of about 1.5x _ but _ that may not be accurate. I wrapped two layers of electrical tap over the gaps but one of the wire turns indented at the tape and some wire was partially in the gap... I didn't think much of it at the time but now wonder it didn't effectively narrow the gap (and therefore raise the inductance.) I also read later that flux density is highest in the gap and that there is a higher potential to burn out wires near the gap so I don't think that it's a good idea so have one of my little wires (#22, four in hand) in the gap...

 - Ed.

[elt]

I stuffed the parts on the board. Wow is that crowded! I said that I didn't know big heat sinks would be. Now I know that they are bigger than I expected. Here's a picture -

The little microcontroller chip is in my development system; I won't plug it in until the software is working (more or less.)

 - Ed.

[mrpackethead]

You've certainly done quite a bit of work on this!  How much power are you expecting to run through your circuit>

[elt]

Hi Andrew,

I expect that this circuit is good for about 250 watts of wind. That limit is set by the width and thickness (thinness) of the copper traces on the board. I don't think it's cut and dry but I figure that copper is only good for about 10 amps. If things go according to plan, though, the mill should reach cut-in at about 175 watts so the booster should cut out at that point, about seven amps @ 24 volts, giving me a little safety margin.

I'll know better once my mill is finished. I have the software almost ready to test so I plan to hook it up to a 12 volt solar cell and 24 volt battery. I only have about 20 watts of solar panels but figure that the circuit would be good for about 150 watts of solar.

- Ed.

[mrpackethead]

I'm about to start work on a microcontroller MPPT, i've started putting together some ideas, i'll be needing something that will run at 2kW. My windmill will produce in execss of 1kW and i want a bit of head room.

I've decided i'm going to use a Linksys WRT54GS running openwrt ( www.openwrt.org ), effectively giving me a low powered linux machine to write my code on.   Its easy enough to interface the wrt to various devices via 1wire or even SPI.  One of the bonus's of using the WRT will be that i'll be able to have an IP connection to my windmill.. so at least in theory, i can have some kind of monitoring and alarming ( if somehtings going wrong! ).. My windmill will be approx 1km from the house, ( at the top of the hill ), so i'll hook up a bit of a patch antenna to it..  

Im just looking at your drive circuitry to see how i might borrow the design and merge it into what i'm doing..