Small scale solar MPPT wins 21%

[elt]

I did some testing with the 36 cell, 12 volt panel I made and it wasn't as good as I hoped; I expected around 20 watts but it suffers heat fade badly ... without any load it quickly drops from 20 volts open circuit to about 15 volts.





Connected to a fully charged 12v 200 AH battery bank it does make float volts producing .51 amps at 13.4 volts ... 6.834 watts

I connected the panel to the MPPT booster I made and a 24 volt 200 AH battery bank:

I used a serial LCD for debug and testing. The LCD isn't need to run the booster and the power usage given for the booster (.015mA@ 7 volts) is without the LCD.

In use, controller draws .78 amps at 12.16 volts = 9.485 watts out of the solar panel. At that time the battery bank was at 25.4 volts and taking in .33 amps = 8.382 watts. 8.382/9.485 = 88.4% efficiency. In this case, the open circuit voltage is 15.1 volts and the controller is programed to load the panel to 80% of the open circuit voltage.

The little micro running the booster is power hungry and consumes 90 milliwatts to the net power produced is 8.292 watts.

Compared to the 6.834 watts going into the 12 volt bank, boosting it and getting 8.292 watts into the 24 volts bank represents a 8.292/6.834 = 21.3% gain in power.

Whoopee!

I was surprised to see that over time the battery volts raised to 26.8 volts, I thought more power would be needed to reach float volts.

Anyway, it's just a basic booster circuit (cause I don't know how to do better!) The logic FET is driven by one of the processors PWM pins at 72KHz. It adjusts the duty cycle to load down the panel to about 80% of the open circuit voltage. Every two seconds it unloads the panel for 15 ms and checks the open circuit voltage again.





Hmm... maybe using a Schottky instead of that SB540 would get a little better efficiency.

- Ed.

[commanda]

The logic FET is driven by one of the processors PWM pins at 72MHz.

I trust that is a typo?

If the current going into the battery was an analog input to the micro, you could dither the pwm, see which gives the greater current, then move the average pwm in that direction.

Amanda

[elt]

Hi Amanda,

That sure is a typo! But it's only off by thirty ... db.

The micro switches the FET at 72KHz (k_i_l_o_h_e_r_t_z), thanks for pointing that out.

... but the whole idea is to not sample the output current. Turns out that a plot of maximum I*V at different levels of Sun is practically a straight line, at the point where the loaded voltage is somewhere between 70 and 80 percent of the open circuit voltage. Where I could find specs, it looks like the single datum given by panel makers is about 70%. So (as usual) I allow the parameters to be set, changed and stored in EEPROM.

I got the idea from a wikpedia image. Maybe it only implements "pretty good" power point tracking, but with the simplicity I'm hoping that's good enough.

What I'm really jazzed about, though, it the thought of taking the idea to a buck converter. I thinking that a low power converter could handle something like the three panel $200 USD Harbor Freight panels in series (nominally 36 volts) and do MPP feeding to either a 12 or 24 volt system and, while it's at it, doing the charge control as well.

 - Ed.

[TomW]

Ed;

Fixed the MHZ typo to be KHZ. Just so it was in there correctly for anyone researching and not reading the replies.

Tom

[tecker]

  I don't understand the blue glazing . The circuit is interesting C3 can go away it's just there to keep the ripple shunted . batteries like ripple just fine As a matter of fact the caps won't react at all across a battery . looks like your wasting power at Qi   Sb540 is a 5A schottky . Looks like fun

 

[elt]

In hindsight, it was probably a mistake - I thought that painting the inside of the box blue would make it look more like a "real" panel at a distance. Now I think that it adds to the heating and voltage loss. At any rate, the glazing is clear Lexan.

> [...] C3 can go away

C3 is a .22 uF polypropylene cap; IIRC, it was suggested that I put one in a different design to protect the FET from switching spikes and I just copied it into this one. That cap is big and expensive, I'd be happy to put it back in the cap pile if you think it's not needed here.

Thank you,

 - Ed.

[tecker]

There's no spikes the battery doesn't obsorb here and no doubt the caps is retaining battery voltage . Keep working on it

[scottsAI]

Hello elt,

Good job.

Back ground of panel should be white or silver to reflect light between cells. Reduces heating.

Cooling panel is an option, many ways to do this. Some are even practical.

Large panel active cooling system can be worth doing. Stay above dew point.

MPPT

Max power as you state is I*V, important to know which V and I are used.

The V is the load (or battery in this case) once translated by the inductor.

The battery V is fixed (more or less between samples).

Thus maximizing current will maximize Power! Keep it simple.

Rshunt introduces an undesirable power loss... why else not use it?

There are other ways, interesting is a current sense MosFet. About 0.5% loss.

Current sense MosFets are not know for being accurate, here accuracy is not needed.

Hall sensors are another, maybe the best.

Let me know if your interested.

Have fun,

Scott.

[oztules]

Would it be useful to sample the voltage drop over the switching fet when on to gain the current info (opposite to the sampling the voltage open circuit of mthe panel, sample the voltage drop over the fet when it shorts the inductor). As it is a necessary resistance, you might as well use it.

.......oztules

[DamonHD]

I'd have thought that it could be a very strange and difficult-to-gauge sort of 'resistance', with possible negative/transconductance/tunnel/bizarre elements to it, at least potentially.  But I've never tried, so I don't know.

Nonetheless, it could possibly be calibrated once with known currents (and gate voltage, etc) and simply looked up in a table thereafter (with interpolation).

Rgds

Damon

[oztules]

I have used this technique for current monitoring for DC motor speed control. (analog though) I sync it so that I measure the voltage drop during on periods..... seems to work ok for over current monitoring.

You may be right for sensitive realtime things.... although worth a thought.

.........oztules

[GlutealCleft]

Very, very cool stuff - terrific work!

A 21% increase is nothing to sneeze at.  It would be interesting, though, to see if the addition of a current sensor could get you any significant boost over what you've got already.  Perhaps a hall-effect sensor, so you're not dealing with resistive drops?  5-A DC units from Allegro are only a few dollars.

[elt]

Hi Scott,

Regarding my panel, I should be able to slip some silver mylar between the gaps in the rows to cut down on the heating. I have some left over so I will give that a try.

I didn't really care about current sensing in this build; the idea was to test the simple solar cell model: V(maximum power from the cell) = V(open circuit) * 80% or perhaps some slightly different tunable ratio. So far, this looks very promising.

Hi Scott, oztules, Damon,

Current sensing MosFets???

!!!

I am very interested in trying this idea. Not sure about measuring the voltage drop across a regular FET though, other than thinking that's what I'm already doing to measure the panel voltage. I haven't tried syncing the PWM and ADC to see what's happening at the microsecond level but I could do that if you guys think it would tell me something...

I did google current sense MosFets and got hits from NXP and IRF but I haven't found where to get less than 400x quantities.

What I'm really thinking is that a current sense mosfet would be very nice for my windmill MPPT because I am having sensing issues there and doing the equivalent to the solar technique of unloading the mill every now and then doesn't sound like a good path (though I think I've read of one commercial controller that does exactly that.)

So, where can I get one of them current sensing mosfets?

Thank you very much,

 - Ed.

[elt]

I do have a few 50 amp Allegro sensors in project box; I think I'll go to their website and ask for samples of smaller units..

When the display is attached, I can manually step the PWM up or down and see shat happens. Inside or when cloudy, the power dropped and rose quickly with small changes in duty cycle. More then enough milliamps to power the board but not enough to get excited about. In full light, +/- 10% is barely detectable on my cheap meters. With a lot more power, it might matter more. The interesting range is about 45% to 70% duty cycle so I did see that is beneficial to change the boost versus a simpler circuit using a fixed boost... another question I had.

 - Ed.

[scottsAI]

Oztules, DamonHD,

Yes, this is possible. Might be more difficult than your expecting.

Remember this MPPT is boosting the panels output voltage, more commonly MPPT reduce voltage to match the battery.

The current in the Fet is ever changing with the inductor:

http://www.ngsir.netfirms.com/englishhtm/RL.htm

The current peak will be controlled by how long the Fet is on.

So as the on time is varied the current will vary.

Fets resistance changes over temperature... effect should be slow enough to not be a problem.

Fets on time is short 72khz, will require a sample and hold to capture reading.

Goal is to hit the Power Nee, current can peak higher than desired. Making control more difficult.

More may be required, my first look at doing it this way.

Over all an interesting possibility. Will have to think about it some more.

Have fun,

Scott.

[DamonHD]

I think we're both just talking about measuring the voltage across the (switched-on) FET, and knowing what current that that implies, either by knowing the on-resistance of the FET for your gate voltage OR by having seen X voltage across the FET when Y amps is flowing for enough values of X and Y to make a little lookup table.

No magic 'current-sensing' FET.

Rgds

Damon

[tecker]

I 've been messin around with the buck, boost, Cuk for a while. I won't steel the tread you have going .I got some stuff to put up later. Mosty about buckin down X10 limited current voltages to get the current for lower voltage devices and storage .

A thing to remember here is  battery impedance. They have much slower return voltage times if you hold that voltage two to 4 volts above ,the electrolyte stores the charge without a lot of heat(if the batteries are good)factor batteries or the battery bank into your circuit where the bias is good for switching I ran across this .In this PDF are a good explanations of switching  copper to  do this and more. They talk about a charge pump .It's more static than coil functions but looks interesting . Food for clarity anyway.

   http://www.jaycar.com.au/images_uploaded/dcdcconv.pdf I also uploaded a Doc file with buck boost calculations.

http://www.otherpower.com/images/scimages/1431/buck_boost_basics.pdf

Good stuff hang in there.

[commanda]

Ed,

What I was getting at with current monitoring, is then you can use the same circuit for your windmill. Unloading the mill really won't work.

I typically use commercial 1 milliohm shunts with an op-amp amplifier. This cct without the analog meter. Scale the gain to suit adc input on micro.

http://www.fieldlines.com/story/2004/9/4/193922/2839

Amanda

[elt]

Hi Damon,

I've some studying today and here's where I'm at:

I don't see how a single measurement would be enough to do MPPT but I think that I can sync the ADC with the PWM and digitize the charge/discharge curve of the coil. If I did the math right then that sum of squares of the voltages measured at any point is proportional to the power in the coil (because the voltage was measure over the sampling time) and that the part of the curve where the voltage is greater than the battery volts plus V(f) of the diode identifies the power that goes to the battery side..

I think that the right hand side of the pic on http://www.4qdtec.com/mircl.html

shows what happens inside a magic current sensing FET in a discrete component kind of way... other than perhaps being a better impedance match for the ADC, I don't see the advantage over a simple voltage divider. (So I suppose that I need to keep reading!)

Anyway, from what little I've able dig into it, it seems complicated but doable...

Thanks to all,

 - Ed.

[elt]

Hi Amanda,

Yes, I'm learning that you need the full-boat to control wind. I used a hall effect sensor in my first booster, then used the battery cable as the shunt in the second. The first approach seemed more reliable than the second, at least as far as my abilities could take it. This booster is pretty much a cut and paste from the advice I got building those two, though the "control section" is different in all three.

So far, I haven't used a commercial shunt but one thing that I like from looking at them is the real terminals for attaching real cables. Yes, the Allegro current sensors have really big fat pins but that sticks me with having to implement the part between the cables and the pins and I just don't have that much experience with more than digital power levels. Seeing an experienced engineer like you using a commercial shunt is very persuasive.

For wind, I'm hopeful that the sensing the FET voltage idea suggested elsewhere in this thread works. I'll build a testbed to see if I like it; if not I'll dig out the $10 and get a shunt like yours.

Thanks again,

 - Ed.

[s4w2099]

Just a comment about C3 and C2. I dont want to start a flame war about this but I would not remove those capacitors from the circuit and this is why.

Removing those will enable track and wire inductance to act as an open circuit for very small time. Hard to explain but here it goes.

Current in an inductor increases with time. At t=0, when mosfet switches off  (t=time) the inductor will be an open circuit. This would be very bad for your mosfet as the voltage spike created by the boost converter will have no place to go, thus nothing to clamp that voltage down. That situation might enable that voltage to exceed your mosfet's Vgs.

Another thing is that with such a fast dV/dt (rate of change of voltage) the parasitic capacitor between the drain and the mosfet gate will create a voltage in the gate. This is very very bad as this voltage will turn on the mosfet when it should not be on.

If you make the path from your mosfet drain-diode-C3 very very short you will be running on the safe side as this huge spike's energy will just be stored in C3. C3 will not let the voltage spike so high then.

It is extremely important for the survival of the booster that the output inductance it sees is minimal, and that is achieved by shortening tracks and adding capacitance.

I am not great at explaining things but I hope you get the idea.

[elt]

I'm going to have to figure out whether a regular buck or a cuk is better (or if it matters) for  converting high current and voltage levels down to battery volts on my windmill ... any thoughts on that?

Thank you,

- Ed.

[boB]

>>http://www.otherpower.com/images/scimages/1431/buck_boost_basics.pdf

Some interesting stuff in there, but they got the output polarity of the Buck-Boost

backwards (it inverts) and the output diode is labeled as an "L" on more than one schematic.  (cut and paste errors I'm sure)

boB

[tecker]

I think I got some stuff that will be easy enough to setup .I'm back on the axil for a week or so .

[scottsAI]

Elt,

Had a chance to think about this some more.

Not sure where your at. (on to something else?)

The load is a fixed voltage (battery), thus controlling the FET to maximize current will be maximizing Power. The inductor is an energy transferring device, by measuring current in it represents overall energy stored cycle by cycle on a PWM system. The solar panel has a fixed power out it can deliver, if the output load is exceeded the current can continue to go up a bit, over all power delivered will drop due to panels output voltage dropping. When charging an inductor as an energy device if the panel is over loaded the overall energy delivered into the inductor will decrease, measuring the current will tell you if it was more or less than the last time!

The power converter has a low side switch, current sensing FET is not needed. Like DamonHD said measure the voltage across the FET at the end of the PWM on time for peak current. By adjusting the on time for max current will be the max power point.

Simulation may be a place to start.

Found IV and PV spice model of solar cell:

http://chuck-wright.com/apppv/pv_spice.html

Simulation of the boost converter is possible without building anything!

Watch how the energy changes based on the power out with different on times.

Let me know how its going.

Have fun,

Scott.

[elt]

Hi Scott,

Actually, this is perfect timing as I was just about to etch a circuit as a testbed to see what I can see using this technique.

I have to admit that try to learning Spice exceeded my educational time budget and I've given up on it for now. I am still using the same toy simulator package that I've had for a while.

> Like DamonHD said measure the voltage across the FET at the end of the PWM on time for peak current.

I'm still not seeing it. I may not be simulating it properly, but it looks to me that a higher duty cycle always means a higher initial voltage at the point the FET is turned off ...which is just what I'd expect from a boost circuit. But I notice that you've written to adjust the "on-time", not the duty cycle. Are you suggesting a fixed off-time? I think that off-time is variable too. It looks to me that off-time should end as soon as the coil voltage drops below battery voltage and that is easily determined.

I see varying amounts of the decay curve in the simulator but without integrating the entire curve, I'm thinking that averaging the voltages at the start of the FET off-time with the voltage at the end of the off time may be a fair approximation

If I didn't wander off track, then I think that the eval function would be (Vwhenthefetisturnedoff + Vbatt)/2 * Tfetoff ... times (1-dutycycle) since power is only transferred during the off time. So I guess there's a question; is that correct? Overly complicated?

If that idea survives the walk through, here's a plan

1. Set Ton
   
2. use the ADC to measure coil voltage just after when the FET is turned off
   
3. use the analog comparator to determine Toff
   
4. evaluate and direct change in Ton
   
   

What do you think of that?

Well, not having an oscilloscope doesn't help so I'm putting together a little board with a serial port so that it can run an experiment and then dump the results to a PC. What protection would you recommend putting on the ADC input other than perhaps diodes to the rails?

Thank you very much,

 - Ed.

[scottsAI]

OK elt,

Darn was hoping you would run the simulations and show how it worked!.

Seriously:

Basic, inductors store energy = power * time.

P = iv = Li * di/dt current looks like the only factor. Not true.

http://hyperphysics.phy-astr.gsu.edu/Hbase/electric/indeng.html

Inductance is Henry = volt second/amp

http://hyperphysics.phy-astr.gsu.edu/Hbase/electric/induct.html#c1

Take a look at Faraday's Law:

http://hyperphysics.phy-astr.gsu.edu/Hbase/electric/farlaw.html#c1

Key is the Volt sec, need voltage across the inductor to increase the energy stored.

Going back to the spice model link:

http://chuck-wright.com/apppv/pv_spice.html

Look at the Power into a load plotted on panels output. (page or two down on the web page)

Goal is to Match load to panel to maximize power transferred.

Back to the boost converter and FET voltage. Must determine on the fly the proper loading.

FET voltage while on is NOT panels voltage. FET is a resistive on/off switch.

The voltage developed across the FET when on is V = IR based on the current flowing through the FET.

Inductor connected across a voltage source:

http://hyperphysics.phy-astr.gsu.edu/Hbase/electric/indtra.html#c1

Notice how the current keeps going up until DC conditions are met. Little energy in the inductor at the end point, even with the largest current!

Back to solar panel, all that above to get to this point:

As the on time increased the current will increase to a point.

Once the energy absorbed by the inductor exceeds panels maximum power the current will stop rising and even decrease. Think of it as the transient conditions i = V/R, and the V has come crashing down! Thus by monitoring the FETs cycle by cycle voltage, you can determine the max power point. The FETs on time should be varied about this point to continuously determine if max power point has moved. Adjust to the edge of time for max current. If kept on for couple micro sec longer with same current then back off. (shorten on time).

A FETs resistance changes with temperature, so not considered a good Rshunt. Here looking for a relative maximum, absolute value is Not important. FET shunt should work fine.

Just to be clear measuring the FET voltage while the FET is ON and just before its turned OFF.

72Khz PWM rate snap shot at end of time is faster than most Micro ADC can handle. High speed sample and hold will be necessary. Protection from voltage generated while FET is off is needed.

Assumptions. PWM is generally done at a fixed frequency. Varying the on time, off time completes the fixed cycle time.

Yes, Off time can be fixed with varying on time and many other combinations.

Most inductors have a range of power they can handle, about 10x as a rule (not a steed fast rule).

Have fun,

Scott.

[elt]

Hi Scott,

Detecting the solar MPP on a cycle by cycle basis sounds like an interesting idea. If I understood, this sounds like a simple two state circuit. I quote "simple" because it sounds simple conceptually but I really have no idea how it would be implemented:

1. With fet is turned off, sample and hold the voltage across the FET (which will be Vopencircuit.)
   
2. Turn on the FET to charge the coil
   
3. When voltage drops to .6 times Voc (".6" is discussed below) turn off the FET to discharge the coil.
   
4. When the voltage drops below Vbatteryplusdiodedrop, sample and hold. (This is the new Voc). Go to step 2.
   
   

The factor ".6" might not be correct. The idea is to average .8 time Voc since that is as close enough to Vmaximumpowerpoint as to not matter to try to get better.

I believe that this approach not only adjusts the duty cycle to achieve the MPP boost but will also increase the PWM frequency at lower power levels (which I've taken to see as a good thing.)

How's that sound?

- Ed.

[scottsAI]

Working on a reply,

Decided I could use this, so designing it.

Maybe by tomorrow I will be done.

Have fun,

Scott.

[elt]

Can't wait to see it!

 - Ed.

[scottsAI]

Well another day. Two problems came along Sunday.

First, lost power for 20 hours did not get to finish it last night as expected.

My backup power covers lighting and Water... no computer or Internet.

Design problems, Inductor has min and peak current, battery will see average current. Panel will deliver average current so inductor does not get "charged" up to peak current. By adding Cap on input fixes this. Problem inductor can 'take' more current for a while than the panel can deliver, making the software more complex. For a while did not think it had a solution. Figured it out... I think!-)

Hold on for another day or two OK?

Have fun,

Scott.

[elt]

I'm glad to hear that you had some backup power and that everything is okay. We were lucky this time, just a few small branches and leaves.

But software? I thought you could do one this with an op-amp and couple of caps ... :-)

- Ed.

[scottsAI]

Elt,

It will work. Monitoring only MosFET current.

Intelligence is required to control for MPPT.

Using your above circuit connected to a Microprocessor.

Current monitoring

Analog switch to capture peak current in FET.

Don't forget must Handle Boosted voltage!

Input capacitor to Inductor is required to supply peak current.

Size the cap to limit voltage change to 30mv or less.

C = i * delta(t)/delta(v)

Delta(t) is PWM on time, delta(v) is the allowed voltage change.

Battery side cap should be as large. On booster circuit board to limit EMI.

Please note large cap's have leakage currents losses, don't make larger than needed.

ADC input

Signal from MosFET is small, amplify it some.

Target half ADC range for normal operations. 10 bits may be needed.

How to make it work?

You have the math down for the boost converter. So not covered here.

ADC must sample at the LC time constant or slower. Input Cap.

Set Ton time, watch current, going up, same, decreasing.

Create a decision tree based on power curve and the Panels effects by being over loaded or under loaded. Code not looking too bad.

Due to input Cap, the inductor can draw extra current for a few cycles, that is why sample current at LC time constant rate, giving it time to respond to the new conditions.

If Ton is too long, the current will not keep going up.

The driving force into the inductor is the voltage across it. See my above long posting.

When the panel is over loaded the output voltage will drop, with same Ton will end with less current!

Vary Ton to find Max current point. Constantly vary Ton to see if Max point has changed.

I sent this figuring you may want to get yours going while I work on mine!

Have fun,

Scott.