Building a voltage booster

[elt]

I'm going to build boost curcuit based on the ideas in Flux's post-

http://www.fieldlines.com/comments/2006/3/17/185646/194/25#25

This isn't the "all in one" curcuit that Flux has posted most recently but the one that operates in parallel with the full power rectifier that's normally connects to the battery.

... this is for a 10 foot mill with a 250 rpm, 12.5 mph cut-in for 24 volts. I expect about 175 watts, 7 amps, at 250 rpm so that's a reasonably small current to layout on a home brew pcb.

I've tried to do my homework but I'm not an analog guy. I lack at the practical matters that real engineers understand so all feedback will be greatly appreciated. (I'm willing to accept that the entire thing may be 100% wrong!)

Here's the schematic -

The little microprocessor connects to everything...

I power the processor from the battery side. A voltage regulator that take up to 30 volts input is required; not being sure how high the battery voltage might peak, I put a diode in front of it to give a little bit of voltage drop in front of it.

I'm not sure if the disconnect relay is needed or not. It's conceivable to me that the two schottkys might have less of a voltage drop than the diode in the high power rectifier and conduct higher currents than I want when the curcuit isn't boosting. (Though I suspect that the coil will add enough voltage drop to limit the current flow.)  At any rate, if the relay isn't needed, it can be left off (and the the power requirement of the processor itself is so low that it could be powered from a 5.1v zener diode.)

The relay coil needs 40ma at 5v. That's the "absolute maximum" for a single microcontroller port pin so I paralleled two ports with current leveling resistors to play it safe. The ports are capable of both sourcing and sinking voltage but the voltage level shift is about half as much when sinking as it is when sourcing higher currents.

The chips supports bipolar differential input on its analog to digital convertor. The output on one coil is run through a voltage divider to keep it within range off the adc input. The generator rpm is slow enough that the processor can compute peak to peak voltage and rpm.

Duty cycle to the FET can be computed from input voltage. I'm not sure that I understand the gate-source curves on the spec sheet but it appears to me that the FET will conduct somewhere between 20 to 40 amps with a 5 volt gate voltage so I think that it can be used without a high voltage driver.

Coil - I got some ferrite e-cores with a 1mm gap. Put together to make an EE core, that'd be a 2mm gap. I can compute the turns to make a 200uH coil but the data sheet doesn't give the info in a way that I know how to calculate saturation current. So as a fallback, I've done the math for an air core inductor and found what I think is the perfect sized bobbin in the form of a small wire spool from RadioShack. The spool is a little more than 2 inchs diameter, a little more than half an inch wide. About  8 to 10 meters of #21 wire four in hand will fit on it make a coil between 160 and 220uH.

I did a test layout to see if it all fits on one side. Things are a bit loose in this layout because I don't know how large the heatsinks are or need to be. The wide trace widths should keep temp rise to 30c with 1 ounce copper, half that with 2 ounce.

 

 - Ed.

[elt]

Scaling the image size on the page causes some the horizontal lines in the scheamtic to disappear (in my browser.) If it looks whacky, you can view the full scale image directly in my files section (or "right click" on it if your brower supports that.) - Ed.

[Flux]

Sorry I can't help with the microprocessor bit, but you will have to do something about the mosfet gate drive.

You will kill it with a high impedance drive. During rapid switching, miller effect will induce spikes in the gate from the drain voltage and kill it. Keep gate drive impedance as low as possible ( 100 ohms absolute max ) . If you must use a high impedance drive then connect a 15v zener directly to the gate, but even then it will switch slowly with high losses and sometimes the zener causes ringing and the gate may still be damaged.

Add a pulse capacitor about .22uF between source of fet and cathodes of D4 as close as you can get it. You may need a few hundred uF of electrolytic across the dc out.

You probably need a few hundred uF across the output of your schottky input rectifier to ground with a .1uf across it, otherwise you will have mess all over the leads from the alternator.

Otherwise things look fine.

[Ungrounded Lightning Rod]

I'd hang a cap to ground just upstream of the relay - electrolytic plus a ceramic for high frequency bypass.  That will keep the high frequencies from showing up in the wiring to the mill.

I'd also hang another set of caps on the + "output", and bring + and - from the battery side of the bridge in on the same connector as the alternator, to keep the high frequency stuff off the battery and house wiring (unless you intend to use the booster as an impromptu desulphator).  While the battery makes a capacitor unnecessary for rectification, it has too much impedence to keep electrical noise down.

I'd also fuse the + "output" line near the connector.  Otherwise a component failure or defect on this device could pull heavy current from the battery and start a fire.  (Having a fuse here also makes it more important to have output filtering.  Otherwise the output pulses could, perhaps over months, cause fuse failure.)

[SamoaPower]

Flux is right about the gate drive. The problem is that in order to keep the switching losses low, the FET has to be switched quickly. The input capacitance of the FET is substantial, typically 1-6 nF. Charging and discharging this C quickly is the issue. You have to supply and sink peak currents of 1-3 Amps to do it so low impedance drive is necessary. Higher gate voltage also helps. Here's a simple driver that will do it:

You may want to put in a few ohms between the emitters and the gate.

I don't see that you have any provision for current feedback to control PWM.

[Ungrounded Lightning Rod]

You need a full bridge of schotkeys, to feed the negative side of the switcher (and the caps I described earlier) if you want to avoid the voltage drop of the negative side of the main bridge.  That also implies another schotkey on the negaitve side of the output, with the negative side of the switch floating about a schotkey drop above the battery, if you don't want the schotkey on THAT side bypassing the other half of your main bridge.

[Flux]

The low inductance high pulse capability capacitor from the mosfet source to cathode of D4 is vital. It needs to be a high pulse duty polypropylene or a ceramic.

Without it, the circuit will have so much stray inductance that when the mosfet turns off the energy in the inductor will cause a monster voltage spike across the mosfet.

The capacitor will limit the spike.  The loop consisting of the mosfet, D4 and the capacitor should be as small as possible to keep the inductance down and if you use a double sided circuit board, an earthed ground plane under this area will lower the inductance. This is the critical region, other capacitors can be placed in convenient places.

Samoa's driver circuit should be fine, but if you only have 5v drive then you will need a logic level mosfet.

I have no idea how your digital bit works, you obviously have no current feedback, I can only assume that the ac voltage fed to the processor is in some way developing a speed signal.

I used conventional diodes for the auxiliary bridge. Schottkys may make a marginal improvement but at 24v I doubt if you will see much difference. Not having Shottkys in the negative arm of the bridge will make little difference.

If you start introducing Shottkys in the converter supply bridge you may need the relay as the drop via the converter may become lower than via the main bridge, but a tiny bit of resistance would likely solve the problem with negligible loss. I personally think the relay will be a hindrance rather than a help in the long run.

[ibedonc]

hmm two diodes in parallel will not work , one will conduct before the other, so the other diode is a waste

I have a design that I did that I have been using for some time now and it is a dual phase uses 2 coils and 2 fets and a tl494 chip (what I had around at the time) TI makes

better ones for this job and will be using those in next rev

I have some blank boards for the one I am using now

http://www.redevices.com/projects.html

I have had it upto 330v dc ouput and run CLF and soild state ballast - T8 lamps off it @ 130v dc

[elt]

Hi ibedonc, you wrote:

"hmm two diodes in parallel will not work , one will conduct before the other, so the other diode is a waste"

Well, that's why I wanted to put real part numbers on a circuit, to move the designed from theory to implementation. (Though I realize that Flux has been using something like this with analog control for quite some time.)

The MBR6045 data sheet says "Terminals 1 and 3 May Be Connected for Parallel Operation at Full Rating"

I think someone posted on the board why it does work during a thread on using bridge rectifiers as legs of a three phase rectifier...

 - Ed.

[Ungrounded Lightning Rod]

Diodes have trouble in parallel because their current isn't linear, but a sharp "knee" up at the bandgap voltage.  Even a very slight difference in the bandgap (due to very small process or temperature variations) can cause a large imbalance in the division of the current among the paralleled diodes.

But if the diode are from the same lot - or better yet, on the same chip and not cut apart - and are on the same heatsink, they balance just fine.

This dual-diode is probably just such a device:  An integrated circuit composed of two diodes.  Completely balanced and capable of full rated current if the two diodes are paralleled - just as if they were the right and left half of a single diode.

[elt]

Hi folks,

I really, really appreciate all the feedback. The FET drive and filtering are areas where I am particularly inexperienced. I picked components mostly because I already have them sitting on the shelf. I can see the wisdom of not using schottkys on the input rectifier so that I can get rid of the relay and I think it may be worth spending a buck on a logic level FET rather than putting a second DC supply on board. (Also think that a fuse and connectors on the sames edge of the board are pretty good ideas as well.) I'll revise and repost ...

Samoa wrote:

"I don't see that you have any provision for current feedback to control PWM."

Flux wrote:

I have no idea how your digital bit works, you obviously have no current feedback, I can only assume that the ac voltage fed to the processor is in some way developing a speed signal.

Maybe I'm wrong but I figured that I could calculate the rectified voltage by measuring the peak to peak voltage on one of the input legs. (Perhaps something like PTV volts/2 x .9 ?) From the input voltage, the processor can calculate (or look up in a data table) the duty cycle needed to boost the input voltage to battery voltage and then to shut down the PWM when the input voltage is high enough.

I could measure RPM but then the processor would need to know how to convert RPM to volts ... seemed to me that the processor could just measure the input volts directly.

Originally I had the ADC measure the volts on the booster side of the rectifier but figured that that side would be noisy from switching. I had a low pass RC filter on the analog input but didn't have the filtering capacitors from rectifier positive to ground that folks have recommended... Maybe with the caps in there now I can switch back to measuring the rectifier output directly.

I could measure current at the battery but figured measuring input voltage saved using a shunt or other current sensor... hope that makes sense!

 - Ed.

[altosack]

Hello Ed,

I follow this with great interest because I am currently designing a system/battery monitor / solar charge controller w/MPPT / dump load controller using an AVR (ATmega48). After I get that done, I'll go for the wind generator controller using the same chip. I've studied the AVR pretty thoroughly now (including the ATtiny45), and have quite a bit of experience programming (though new to AVR), so if you have any questions, I'll be glad to try to help.

I really like the tiny45, it has PLL to run the PWM at up to 64MHz (8-bit resolution at 250 kHz), differential input and 20X gain on the ADC (very useful for the current input), and it's cheap, although the mega48 is pretty much just as cheap and has a lot more pins (28) and a 16-bit timer.

So, what speed are you planning on running the PWM ? Are you using C or AVR assembly ?

For your relay, I would just use a MOSFET; it has much lower drive requirements, and since you are not turning it on and off fast, the gate capacitance is of no importance. For the design that is currently in my head and partly in computer files, I originally thought it would have a relay to enable the controller, but later decided that the buck MOSFET would do just fine. Even if I were just doing a boost controller, I might just skip it.

Flux (and SamoaPower) is right about the turn-on and turn-off of the MOSFET. At the very least, follows Flux' recommendation about the parallel capacitor, and if you are ambitious, you might want to consider soft-switching (zero-voltage turn-on and zero current turn-off). This takes another MOSFET and another (quite small) coil, but it greatly reduces the stress on your primary MOSFET and increases the efficiency of the boost converter a bit.

For those that commented about feedback for the PWM, things are different here from hardware PWM chips. By inputting the battery voltage and wind genny RPM (you don't really need the wind generator voltage; you'll know that close enough from the RPM), a microcontroller can easily compute what the duty cycle needs to be and have the other 98% of it's processing time to sleep.

By the way, when I first starting investigating this, I thought it would be great to have the microcontroller update the duty cycle really fast, and feedback on the output current to get maximum power, but this is unnecessary. The inertia of the rotor also causes problems with MPPT since an increase in the duty cycle (buck or boost) will instantaneously give more power, but it is just bleeding off some energy from the rotor inertia. Now I'm thinking about 2 to 5 Hz or so, without current feedback. Since the TSR plateau is pretty broad (about +/-10% from optimum RPM), it does not have to be tracked that closely.

Keep us posted and best regards,

Dave

[terry5732]

All that said, maybe it's simpler to rewind your alternator coils for higher voltage.

[Flux]

Take this with a pinch of salt, some digital things work when I can't see why.

The peak ac volts of the alternator off load are proportional to speed. When the diodes are conducting the peaks are bitten off first so peak voltage will be a mess.

Mean volts may be good enough to tell you the speed.

Boost converters are energy converters and the relationship between pulse width and voltage out is very load dependent. It is probably not adequate to choose pulse width to match battery volts. You need to match it to the power level at each prop speed. You will need somehow to find the pulse width at each speed to provide the power match. If you can do this then you can use the look up table.

I suspect that is the data that Windy Boy and other inverters need to have programmed into them. Once set up for an alternator it's ok.

I think this is why the current feedback works, the volts are fixed by the battery so current gives a reasonable indication of power. If you can measure current and calculate the required power for each point for your table then I think you can do it.

Flux

[elt]

There's no magic in the digital circuit, if the approach is wrong, it won't work...

I've been referencing this online boost converter design app at coilcraft -

http://www.coilcraft.com/apps/selector/selector_1.cfm

It says that duty cycle is ((Vout+Vdiode)-Vin)/Vin .... there's no reference to current in determining duty cycle.

I don't know what's going on underneath the application but I was thinking that it's assuming a fixed output voltage because it has current ripple at the output as a design parameter, not voltage ripple.

I really think that current monitoring and voltage monitoring will product equivalent results because there is a one to one relationship between both power available at a given RPM and alternator voltage at a given RPM.

Okay. So switching currents will bugger the voltage reading. The max sample time on the analog converter is 260uS so the processor can just turn off the FET for a millisecond, take the voltage reading and then go back to switching. Since the mill can't change speed that quickly, I imagine that the sampling frequency could be quite low.

The main reason that I want to look at input voltage and not rpm or current is that I think both of those approaches require the circuit to know something about the alternator and mill in order to be "tuned." Input voltage boosted to output voltage would not; it would be "plug and play" and might be useful to other folks less technically inclined like me... the circuit would just need to be soldered together and it only takes four pins wired to the parallel port of your PC to program the chip.

As always: Nothing I've written should be taken as an indication that I know what I'm talking about! ...if all that's wrong then the circuit can be made to measure RPMs or battery current.

- Ed.

[Flux]

You seem to be basing your ideas on a boost converter power supply that is regulated at constant voltage.  The load determines the current and the pwm changes to maintain constant volts.

In this case the wind determines the available current and you need to track the pwm to maintain the available current at a voltage ( prop speed ) that lets the prop develop the current to run at the chosen speed ( top of Cp/tsr curve).

You have to know something about the machine characteristic. You can try to calculate it at each point which will be very difficult or you can learn its characteristic over a series of measurements and store the characteristic in a look up table. That seems to me to be more likely to succeed.

In theory you could use wind speed and you ought to track alternator speed with wind speed, but wind changes so rapidly that it would take weeks of data to sort out the characteristic. Once you have found the curve then wind input is not required. It is so difficult to use that wind speed is probably best avoided.

Get the converter working and try it, I am sure you will learn so much quicker from practical results.

I have used speed as a control signal so you should be able to do it from volts into the converter, but I don't see it as being a one fits all situation.

Flux

[elt]

Hi Dave,

I'm a big fan of the atmega168; in quantity 25 it's about as inexpensive as any of the smaller chips and has enough memory that you're generally not cramped programing it. I wanted an 8 pin chip for this app with hopes that it would increase the odds of a simple one sided PCB that other folks might be able to make on their own. The tiny45 was appealing because it has enough ram for a stack so it can be programed in C. (I don't generally like programming more than a line or two in assembler...)

I appreciate the offer of help and will take you up on it if I get stuck. In any event, I'll post the code under GNU because I think the notion of free software goes well with free (or renewable) energy and I wouldn't get very far on this project without all the help on this board so I'd also like to make what small contribution that I can in return...

- Ed.

[elt]

Hi Terry,

I haven't wound any coils yet; I'm planning on doing this on purpose.

Flux has shown that the combination of booster and lower voltage stator is more efficient at extracting power from the wind ...

My belief is that it will be simpler to build the booster circuit and a lower voltage stator than to remake a higher voltage stator a second time (and a third time and a fourth time) after it's melted from high RPMs...

 - Ed.

[elt]

Flux,

I am confused but I think I'm getting closer and I really appriecate you hanging in there with me... I posted way back when that I was doing this to learn and I've been learning a lot!

I guess that I've been assuming that if one "24 volt" mill produced 1 amp at 12 volts and other produced 2 amps at 12 volts that the booster PWM duty cycle for both would be the same. I considered the booster as an unregulated power supply that would make somewhat more than battery voltage and that the battery would draw down the voltage and pull whatever amps it could...

You wrote: "...you need to track the pwm to maintain the available current at a voltage ( prop speed ) that lets the prop develop the current to run at the chosen speed ( top of Cp/tsr curve)."

Are you saying that changing the PWM duty cycle will make the prop RPM speed up or slow down at a particular wind speed? Is that significant at low wind speeds? I realize that we're keeping the mill out of stall at higher wind speeds by having a stator with fewer turns but at that point the booster circuit is out of the loop...

- Ed.

[Flux]

I think a different explanation may help.

If we make the alternator super efficient it will have a very steep curve, a few rpm increase will result in a large increase in current.

We have aimed to have an efficient alternator and even with the losses from running at half volts and twice current and with the converter losses it will still be very efficient.

The cube law power curve has a very low slope at cut in and gradually the curve becomes steeper as the wind speed rises. At near full load the game is to make the alternator slope match the top end of the power curve.

When you boost from low volts the alternator curve is far too steep and yes indeed you will go rapidly into stall. The conventional way to reduce the alternator slope is by making the thing inefficient. We don't want to do this, so the converter must start phasing back right from cut in and the input voltage will rise from about 10V at cut in to the battery volts at the point when the main bridge takes over.

If you haven't looked recently I have added some data to the earlier post about the 3 phase boost unit. If you look at these I think it will help you follow what is needed.

This is one reason why star/delta doesn't really work too well. If the alternator is designed to match the prop in delta, it has a characteristic that is too steep in star. We just need a way to alter slope without adding resistance. The converter effectively recovers what would be lost in the resistive case.

flux

[SamoaPower]

If I may, perhaps I can add a little to Flux's excellent explanation from a little different perspective.

An air rotor can be stalled at any wind speed. Indeed, this is the issue at start-up, when the wind is too low to generate enough lift and drag to overcome the rotor inertia and frictional losses - the blades are in stall.

As wind increases, and rotation starts, the blades slowly accelerate to bring them out of stall and acceleration then increases rapidly with no load. The no load speed will quickly get up to much higher than design TSR. This has been observed often.

Now, we load the rotor via the alternator to extract power and RPM decreases. Too much load and we bring the rotor back into stall for a given wind speed and power drops. Ideally, we would like to keep the rotor at the design TSR at all wind speeds. There are some second-order effects such as changing Reynolds number that may modify this somewhat.

The limiting cases for the alternator load, which are passed on to the rotor, are the open and short circuit conditions. Open circuit, the rotor speed is too high. Short circuit, it's zero. Obviously, there's no power at either limit.

The boost converter attempts to seek the optimum load line to match the curves at low wind speeds. So, yes indeed, the PWM duty cycle will control the rotor speed.

I think one way to have a converter fit a variety of machines is to use something akin to MPPT to control PWM. This would be an interesting development project.

[elt]

I think that I've been able to incorporate every change suggested... Thank you all very much!

I also put a low pass filter on the input of the analog to digital converter.

I changed a few part numbers - made the input diodes non-schottky (but fast) diodes to get rid of the need for a relay and I changed the power FET to a logic level FET to get rid of deriving another (higher) DC voltage from the battery input. I'm not experienced with the part numbers referenced; I pretty much just went to digikey.com and looked for parts in stock. (The diodes could be any "good" choice in a TO-247 or TO-3P package, the FET any good choice in a TO-220 package.)

The new layout is a few tenths wider (4 inches by 3 inches) but gets the switching components closer together.

Flux wrote: "I have used speed as a control signal so you should be able to do it from volts into the converter, but I don't see it as being a one fits all situation."

Sorry, I didn't mean to imply that it was universal when I said "plug and play" but I did think that it has a wider applicability than my one mill. I got the notion when I noticed that the 50 turns of #15 wire, two in hand, that I'm using is the same thing that Hugh P. recommends for the 24 volt version of his 8 foot mill. I was thinking that, perhaps, if other folks who built the 24 volt version of 8 foot mill wanted to get a little more power that they could rebuild a 10' footer with their existing alternator and use a boost circuit like this...

In any event, give or take additional feedback, I'll make a booster and see what it does.

Thanks again,

- Ed.

[stephent]

Nice project and good looking artwork.

But I've been wondering if your stator winding is a star?

I can see the 3 phase wires going to the input diodes--but where is the common/ground (dc-) being developed from the genny?

[elt]

Hi Stephen,

This is a "low power" circuit that operates in parallel with the normal high current rectifier. The five inputs connect to the five terminals on the the 3-phase bridge (three input phases, and plus and minus) and it only operates when the voltage is less than the battery cut-in voltage. It uses the dc-/ground through the main rectifier and gets the path through the "ground" input.

This is seen in this diagram by flux -

 - Ed.

[stephent]

Oh--ok..

Was cyphering on doing something like that myself.

Only using a Picaxe micro.

Didn't see total picture I guess with your schematic.

Still scratching my head on the voltage reading part without having a noise pickup problem on the same type boost controller. Guess shutting off the boost fet for a few pieces of a second will work better then trying to figure out a filter for the crud most of the components will generate.

Looking at doing load dump/diversion control in the same box.

Micro's sure keep the total parts count down.

[altosack]

For getting rid of noise when reading the ADC, it depends on the microcontroller.

Most of them with a few more pins than the ATtiny45 (it has have a separate power and ground pin for the analog plane (i.e., all ADC and associated pins). You can either put small caps in to isolate it from the digital plane, or even better is to just put in a separate 7805 (or whatever you're using) supplying the analog plane.

At about US$0.25 each, this seems like the simplest, cheapest, and most effective solution to me.

Another possible is to use oversampling. Since the AVRs do 15,000 conversions per second (most other microcontrollers do more; the PIC16 - and probably the picaxe too - does 50,000), you can do 256X oversampling and still have lots of free time. This should remove most of the noise problem and has the side-effect of even increasing the effective ADC resolution (theoretically from 10 to 14 bits). More programming, but if you've only got 8 pins (and no separate analog plane), that's the way to go.

I've just implemented this on the AVR design I'm working on, so if you want more info, I've got it.

Best Regards,

Dave

[altosack]

SamoaPower,

I haven't actually implemented it yet, but everything I've seen in looking at the design for MPPT for a wind generator controller says that MPPT using direct feedback is problematic.

As I wrote in another post in this thread, whenever you increase duty cycle (in either a boost of buck converter), you get an immediate boost in power that is really stealing inertia from the rotor (and if you decrease duty cycle, you likewise get an instant decrease since the rotor can't speed up fast enough to compensate). You have to be very careful that you've reached steady state again before you take another current reading, but are still changing fast enough to track the wind. Of course, this is hard to do with a generic solution since this stabilization time would obviously change greatly from rotor to rotor.

So, I wrote a great big long equation that lets the microcontroller (again, hypothetical, I haven't implemented it yet) compute the duty cycle that would work given RPM and battery volts (everything else can be calculated if you know details of your machine). The solution was iterative, but the microcontroller had plenty of time to do it and update > 20 times per second (absolutely unnecessary to do it this fast). Note that this is a indirect feedback design that does not use power as an input.

Anyway, the upshot of all this was that I could write a much simpler double linear equation (one linear line with a linear correction below a certain point, like a hockey stick) that did just as good a job considering that the optimum TSR is a plateau. I even put in errors in the input (hmmm, maybe my rotor Cp will only be 0.32 instead of 0.37 at such and such RPM and wind speed), and it had very little effect, the duty cycle was still very effective at tracking close to the MPP.

For a boost controller only at low wind speeds, it's sufficient to write a simple line based only on RPM and assume a constant battery volts and you will be close enough. Of course, you do have to test your wind generator to know how many volts per RPM it does, what your design TSR is, and your generator and line losses, but I believe anyone who is considering implementing such a beast can (and should) do this.

I am sort of working on your "interesting development project", but I have yet to build a wind generator !

Best Regards,

Dave

[stephent]

Even with a lot of oversampling, the small glitches and etc get figured in.

It will certainly be better  than just reading the voltage a time or two each second.

But I was figuring a lot slower "setting" rate for the voltage boost or control.

Possibly every 3 to 5 seconds. Guess oversampling would give that micro something to do and keep it out of trouble with "idle" time like me.

But responding too fast would/could put the alternator ahead of where it needs to be in gusty winds.

But I know those little 780X regulators have a definate isolation limit --about 60 db I think....and at 60 db down..ain't much left--but the response time ain't the greatest.

 Maybe I will just stick to a short wait state and then read. Simpler code to debug later or document.

Those 'short" waits are really just a few milliseconds long anyway for a stable bus read.

Awww--I ain't even etched the board yet and I'm worrying about code...

Will have to etch and then get it running and just hook up the scope for the real readings of any noise. If there ain't much--it ain't a problem. If there is, a wideband noise filter for that low a frequency could introduce a problem or two of it's own with capacitance. peak volts vs average, etc.. but a swamping resistor should handle that.

But then that's added into the equation and the filter is out of whack---aaaahhhhh--never mind--I'm going to play Mahjongg--this makes my head hurt.

OOOOHHH--before I go --that place in AU that makes the kits (Oatley Electronics) has a "new" wind genny controller that looks "kinda" beefy and close to what you are doing. The K241. Maybe not with the micro and reprogrammable and all.

Cheap enough--but the international postage would eat your lunch more then likely.

[Flux]

Your voltage signal filter to pin 3 seems to have got mixed up, the resistor is in the wrong place.

The input diodes don't need to be fast but it will not hurt. They do need to be fast in the later version I mentioned boosting through the rectifier.

Do you have a test rig to try it on or are you going to try directly with wind?

Whatever you do, don't try the converter on a non current limited power supply, it relies on the alternator for current limit.

Look forward to results.

Flux

[commanda]

Dave,

Consider this and let me know what you think. You've obviously put a lot of mental effort into this beast like I have.

Primary input to mppt controller is power out of alternator (volts times amps). We regularly change pwm (about once a second to pluck a number out of thin air) and see if the power goes up or down, and adjust pwm accordingly.

Now, the control voltage which sets the pwm is analog. We have an analog mixer circuit (op-amp and 2 resistors); the second input to this mixer is the analog output voltage from an LM2917 tacho.

Pwm goes up, power goes up, but rpm starts to drop. The falling rpm backs off the pwm control signal.

A potentiometer to adjust the transfer function between tacho output and mppt control; set the gain too high and it will hunt and oscillate all over the place.

Amanda

[SamoaPower]

Dave,

Looks like you're getting into some good stuff.

I'm not sure that the perturb and observe technique would be the best approach for wind MPPT. Perhaps, the curve following technique has more promise. Have you yet looked at:

http://www.elecdesign.com/Articles/Print.cfm?ArticleID=6262

It would take some modification for wind as the author points out.

As you say, the calculated/table look-up approach requires prior knowledge of both rotor and alternator characteristics. Alternator characterization is certainly doable, although a lot of work.

However, characterization of the rotor is less than straight forward and probably not practical without a wind tunnel. I suspect that second-order effects, particularly at the low speed end, may have more influence than generally assumed.

Perhaps, the approach adopted by Flux is the most straight-forward practical solution. Controlling PWM with integrated current in a low-gain loop is certainly easy to do, although I do see a fair period of tweaking to get it right.

Keep on truckin'.

[elt]

No, I don't have a mill yet; I am building the alternator now but at this point I don't think I'll actually get it into the air until Spring. I figured I'd use a resistive load and a variable power supply at first just to get the PWM running and to make sure that it doesn't smoke.

Re current limiting: The manual says that when it hits the current limit that it automatically goes into constant current mode and starts folding back the voltage. Well, I can't visualize what will happen if the PWM circuit keeps trying to boost the voltage...

Thanks for for catching my mistake with the low pass filter!

- Ed.

[altosack]

Amanda,

I think that while the potentiometer-controlled transfer function will work, it may only work optimally at one wind speed for each adjustment.

The problem is that rotor inertia varies as the square of RPM, while wind power (and the ability to adjust the rotor RPM) varies with the cube. This leads me to believe that direct feedback may work better at higher RPM (which is what we want), but also realize that this is at steady-state; when you have an acceleration of wind, the rotor inertia/wind power balance will be different yet again.

Will there be only a narrow band of adjustment at each wind speed that keeps it from working at the other ?  I don't know. I do know that it will take a lot of trial and error, and you may still have wind conditions from time to time that were not in your trial period. If we use direct feedback, and the oscillations go crazy, we may break the machine (and a few other nearby entities); if we use calculated indirect feedback and we're a little off, we just lose a bit of efficiency that we can improve upon later.

The way I plan on doing it (whenever that will be) is to pre-calculate the optimum RPM band for each duty cycle step, to be stored in a look-up table, and then use a correction factor for the actual battery voltage. The only calculation the microcontroller will do in real-time is the battery voltage correction. Of course, since the TSR plateau is pretty broad, the RPM bands will overlap with many adjacent duty cycle values. The advantage to this is that once it passes out of the optimum plateau, the adjustment will be rather rapid, but as the wind speed hunts around a little, the duty cycle will stay put (less stress and fatigue on the machine). Also, this is not that sensitive to update speed; too slow and we lose a little efficiency; too fast and ? - I can't really think of a problem.

We are of course free to take measurements from the microcontroller and compare it to the anemometer to calculate the TSR and power and whatnot and see how close we are to "optimum". With this we can refine the lookup table and download a better one once we have enough data (since this is indirect feedback, we could even do this without shutting down the machine). I suspect that this won't really be necessary to obtain the large part of the gains we're looking for, though.

Best Regards,

Dave