Toasted my new MPPT booster

[scottsAI]

Hello elt,

Been several days, figured I should post what I have been working on.

A bit more rusty on inductors than I thought!

Use them just enough to know they are different from caps yet I have made this mistake before.

Frustrating, I try to be soo perfect, yet not.

Yes, two identical inductors in parallel the inductance is half or 50uH.

Assuming their fields do not interact.

Before when I was talking about 12v 10a I was just talking about a booster circuit.

Does not apply to your power limited situation. I thought you would understand, sorry to confuse.

Spent several hours trying to find how the RPM changes from unloaded to loaded. Found a bunch of theoretical stuff, good read, but hours later no easy answer to the question. This was interesting:

The unloaded wind turbine is transparent to the wind, the loaded turbine decelerates the wind. A change in loading implies a change in "induced" velocity. Thus, the velocity at the disk is the average of the upstream and downstream velocities. Betz limit defines the downstream velocity to be 1/3 of the upstream velocities, therefore the blade is at 2/3 of upstream velocities.

When I see a blades Cp over wind speed I assume it's loaded for this test? Else how to get Cp!

Many plots out there with Cp vs TSR. Not wind speed.

Theory is nice for a perfect blade, I assume it's not.

PMA wind with furling (not using a boost circuit)

Unloaded blade will spin based on its TSR with the wind, slowed down little by its wind drag, if the wind speed becomes great the furling system will not protect the blades. Therefore the blades must be loaded for the furling to work. I assume at this point some much higher wind speed it may still work, most likely well past their rated RPM.

PMA output voltage is proportional to RPM. Most will agree on this.

At cut in a small current is flowing into the battery. The blade is lightly loaded, so the Cp is very low.

As the wind speed increases the power generated increases. The RPM will rise slowly due to the higher current flowing requiring a higher generated voltage to over come the system resistance. The blades loading has increased thus the will Cp improve.

To use some numbers loosely, if the cut in is 12mph, then by 18mph the Cp is now maxed. (assuming 2/3 is the max Cp based on above assumptions) Again due to system resistance the max Cp will occur at a higher wind speed.

What will the power generated vs wind speed look like?

What will happen as the wind speed increases?

Will this is as far as I have gotten, will keep working on it.

I think this will help understand how the MPPT controller should work.

What do you think?

Have fun,

Scott.

[elt]

Hi Scott, I hope all is well.

You wrote:

 I was just talking about a booster circuit [...] sorry to confuse.

I didn't know whether I'd given you bad data or what, but it was the SMPS simulators that were confusing me. They first left me with the impression that duty cycle was the be all and end all of concerns but now I think of the circuits in terms of power and energy transfer and I think I understand them a lot better.

[...] therefore the blade is at 2/3 of upstream velocities.

!!! I was surprised to see the huge difference. We has some moderate wind this morning and when I turned on the MPPT circuit the unload blades were spinning at 150 RPM. When the booster settled down, they were running at 100 RPM... that's (within my ability to measure) exactly the ratio you cite!

It does not appear to me that the mill/booster is very efficient at 100 RPM. It only makes about 10 watts (vs 15 or so predicted) and with the slower PWM I'm using now I can't control the mill finely enough to extract the 3 watts that the circuit running at 64 Khz could. (Of course, that circuit toasted itself so I'll give up the 3 watts to keep that from happening again!) Once the loaded RPM gets up to about 50 watts, though, the efficiency seems much better and the booster gets the power suggested by the wind-power calculator.

Of course since I don't know what the wind speed really is I don't really know what's going on. So this morning I bit the bullet and ordered an anemometer cup set from our host. (I am still excited about the solid state anemometer but I know I'll get the spinning one up and measuring sooner.) ...it's not like I and many other newbies weren't told early on that you have to know your wind speed right at your mill to know what's going on. So, I guess I'm at that point...

I think this will help understand how the MPPT controller should work.

What do you think?

Well, I think that will help with half the problem. Having the MPP curve for your mill helps with charging the coil but the state of charge and load on the batteries will effect the discharge time. Optimum may vary by 25% but I assume that the MPPT would optimize the difference and the actual error could be half that and might not be worth worrying about.

BTW: In "manual mode" I've seen that a fixed setting optimized for about 75% cut in power is "pretty good" all by itself. I'm greedy and will do better but I have alread scavenged an old PC power supply for parts for "the $5 (non) MPPT power booster" ;-) That's farther down the project list though...

To use some numbers loosely, if the cut in is 12mph, then by 18mph the Cp is now maxed.

I get the same number; that seems right to a first order. As the current raises, so do the losses in the alternator coils and diodes. I only bring that up because I've seen from your other posts that you'd like to make a mathematical model. I think that you have to add a few lower order considerations to the "2/3" number depending on how precise you want to be.

BTW: With regard to running at slower PWM speeds, I wrote:

>[...] that sounds audible

It is quite amusing to hear the booster play a little tune as it hunts for the MPP and then reach a steady tone as it finds it (until the wind changes again.

The inductors have bolt holes through them and I did drill the PCB for the bolts but I haven't put them in. Should I bolt them down, maybe put some silicone under them too or just let them sing?

Thanks again,

- Ed.

[scottsAI]

Hello Ed,

terms of power and energy transfer

Yes! Yes!

I was surprised to see the huge difference

I had heard it was half, 2/3 is higher speed. I wanted to make sure I had it right.

Remember those numbers are by definition or by Law the best that you can do. The downstream velocities may be higher due to the blade and system not being perfect.

state of charge and load on the batteries will effect the discharge time

Looks like you should measure the battery voltage to predict the discharge time correctly.

...under them too or just let them sing

I like heat sinks on all power MosFETs for myself. Production then I design to max and live without.

If you like the noise then let them sing. Do they get warm?

At the lower frequency the system could be more efficient, if not then speed it up. Your inductance is lower than what I had messed with so the higher freq may be necessary. That is why I like the simulators, what did you calculate out? I think continuous current in the inductor is the way to go.

I just bought a WMR968 weather data system. Got busted one cheap. Have been bidding for weeks.

Very excited to get it, priority shipping, maybe tomorrow.

You did not take a stab at how the system responds with higher wind speeds? Like 24mph.

I am very curious to see what you think happens next!

I have the system impedance included in the model, very complete electrical model, it's the blades and wind which I have been lacking in understanding, now have a better understanding.

This is working as I expected, the details are a little different, if you look through my post you will see several have disagreed with me, so beware and decide for yourself!

Have fun,

Scott.

[elt]

>Do they get warm? [...] the higher freq may be necessary.

No, nothing gets warm at all, makes me think I'm not pushing hard enough!

I going to make the change to separate the low voltage turn on from the low voltage turn off number first; I think that will let me study what it's doing at low power levels better... after all the heavy mechanical work building the mill, it does seem a little weird going out to the millshed with a little 8-pin dip to do a "software upgrade" ...

> I just bought a [busted] WMR968 weather data system

I google'd that; looks very nice! Are you figuring it's fixable or using it for parts?

>You did not take a stab at how the system responds with higher wind speeds? Like 24mph.

I hope not. My main concern starting this project was the reports of thrown magnets and melted stators. I bought magnets with holes and used #10 stainless screws to peg them in place then used fewer than "the usual" turns and a healthy rotor spacing for a high cut in to keep the system voltage down at higher RPMs.

I did go overboard making the tail a little shorter with a little less surface area to make it furl quicker; no one here suggested that to me and that was a mistake... or at least I went too far. I've only seen the mill furl once and it was already pushed far off the wind and only making about 400 watts. I added some more tail and it's tracking nicely. I've seen 750 watts into the battery from inside the mill shed but it wasn't steady wind; when I went outside the mill was bit off the wind but not furled, when I went back in the power down to about 600 watts so I don't know what was going on when. I'm happy if 750 watts in max but don't really know if that's the max or not. If my calculations are right, with 750 watts in, there'd be about 500 watts of stator heat for a total of about 1.25 K out of the wind. And if it's furling there and if it follows the wind power calculator than that would be about 22 mph of wind. Too many if's but if that's what happening then that's the top limit of the range I was shooting for and I might even consider taking a little bit of tail off.

[The model] is working as I expected,[...] several have disagreed with me, so beware and decide for yourself!

I'm enough of a physicist ("optics" is just E&M and mechanics is a prerequisite of E&M) that I believe that the formulas can be written. I'm just not sure where the numbers that get plugged into the formula come from. I imagine to get going that you're modeling a data set generator for your model and will then move on o measuring physical mills... hoq can I help?

- Ed.

[scottsAI]

Ed,

WMR968 - busted display. Has touch screen concerned it may mess things up. I hope the electronics are fine, only need the serial link to PC to collect data. Do not need to see display.

If interested in helping check: http://www.fieldlines.com/story/2007/4/6/21137/25078

If you can supply any data I would be happy.

I hope to model and answer questions:

If I shorten up the tail what will happen? Will it furl sooner, will it still track?

If I replace the rectifiers with Schottky will it be worth it?

I have 6 awg wire, 100 foot run, how much power will it save if I use 4 awg?

I want to use 47 turns of 13 awg wire, what generator output will I get?

If I used pitch control how much more power will I get?

I live at 42.5'N 83.8'W, how much power from 13 foot blade PMA?

Have fun,

Scott.

[elt]

> If you can supply any data I would be happy.

I'll send what I have once I start logging. That's part of my "charge controller" and I have a head start on that since I'm using an existing hand held controller to do the job.





http://www.otherpower.com/images/scimages/6527/ccITW_01.jpg

I've had to make a few mods to the board. The base design used a graphic display and used all the ADC lines as data ports to talk to it. I swapped that out for a text display that can be controlled with 4 bits; you can see at the top of the display were I sliced the 4 ADC lines out of the ribbon cable (and a couple of unneeded control lines as well.)

The board has a lot more features than I would have put in "the $5 charge controller" but they were already there and it's nice, for once, to be starting with proven electronics!

I haven't decided yet whether I'll be using the battery cables as shuts or do current sensing with hall effect sensors but so far the shunts are working great on the MPPT controller so I'm leaning that way.

- Ed.

[elt]

I'm beginning to feel inadequate...

We had moderate wind this morning and went out the the mill shed to see that the booster was doing. As soon as I saw the mill though, I could see it was only making a few dozen RPM and I knew that was way to slow. I disconnected the booster and it sped up...

As near as I can tell, only the switching FET is dead. I reads 10 ohms both ways across the source and drain.

I've seen it cut in and cut out enough that I don't think my bad programing killed it (this time!) The only clue I have is that my dump controller said that it dumped for about 15 minutes during the night so I do know that mill pumping out good power for at least a little bit of the time. (But no clue whether the FET blew before, during, or after that...)

BTW: I did replace the cap on the output of the FET (C7) with a 250v .22uF polypropylene pulse cap. That tested fine.

I don't have any snubbers or clippers in the circuit. Googling the web, I found reference to them; googling the board I didn't see anything like that in other's booster circuits. (I did see a comment by Flux recommending them but didn't see anything in an implementation.)

And/or: Should I be using a higher voltage FET? The ones I've been using (RFP30N06LE) are rated for 60 volts... I'm out of them so I have to order something, whatever it is.

I realize that I have no real info to go on so I'm just looking for ways to make the circuit more robust.

As always, thank you very much!

- Ed.

[commanda]

At the risking of stating the obvious..............

Invest in some higher voltage fets. The incremental cost is negligible.

But you already know this. You just want someone to reassure you that this is a good idea.

Designing snubbers is a serious black art. You need a really good cro and a lot of trial and error.

If it ran for 15 minutes dumping before failure, it's not thermal melt down is it? How big is your heatsink? Got a temperature probe for your multimeter?

Amanda

[elt]

Thanks Amanda, I was hoping it would be obvious to someone.

But without experience, I see picking FETs voltages like taking prunes: "Is two enough? Is three too many?..."

I've been pulling parts out of an old PC power supply and I pulled a 900v FET ... yikes! Do I need to go that high? (Is two hundred volts enough, is nine hundred volts too many?) Also, I see HV FETs with fairly low threshold voltages but if the manufacturer's data sheet or distributor's part selector doesn't say "logic level" I'm not so experienced to be certain that it's an appropriate choice. (Though I don't suffer from paralysis. Not knowing, I would try something rather than do nothing, but I certainly do appreciate input before hand.) Your experience may distance you from it but the endless choices are daunting to an analog-newbie like me.

If it ran for 15 minutes dumping before failure, it's not thermal melt down is it? How big is your heatsink?

The dump controller ran for 15 minutes. That's a separate and independent module.  If there's enough power for the dumper to run, it's almost certain that there was enough input voltage that the booster was (or was supposed to have been) turned off. Current would still flow through the diodes in the booster booster, I questamate up to about 20 amps when the mill is at full power, so I did put big heat sinks on it (from above mentioned PC power supply.)

http://www.otherpower.com/images/scimages/6527/mppt2_1.jpg

I've never felt the heatsink get warm since I increased the minimum pulse width to the FET (I was going sub microsecond, now never less than 14uS) but I haven't been in the mill shed during a sustained period of high output. (The windy season is coming, we're in transition now.)

But the dumper would have been putting 50 amp pulses on the battery cable thought I don't see how switching noise from that would get to the FET. It's just a shot in the dark, my EE-vision isn't close to 20-20.

Thanks again,

- Ed.

[elt]

... and I blew yet another one!

I found a 100v FET in the project box and but that into the booster. It wasn't a logic level FET and I did notice that it got a little warm, and warmer and warmer... it lasted about an hour though and I did get to see it fail.

The wind wasn't steady and ran from about 1/3 cut in (8 volts or so) to well above cut in.

With the wind speed changing, I did get to see a problem with the MPPT tracking. In particular, it didn't seem to unload fast enough sometimes and I'd hear a mechanical stutter as the mill stalled. (I haven't been able to implement Amanda's variable speed unloading yet.) However, min and max duty cycles are stored in eeprom so I raised the minimum duty cycle. This made cut in at low speeds jerky; however, once the booster wasn't starving itself, the power output was fairly impressive; sometimes as much as 25% over the Betz limit :-(

So, I think that the booster is (was) maximizing power out of the input cap and blade inertia more than the wind. Perhaps my control loop is too fast and perhaps my caps are bigger than they should be. That in itself doesn't might not be a problem but it may be related to the failure.

The wind picked up and the mill was producing 25 amps at 28 volts (about 280 RPM) and the booster was still running. BTW: all these readings are with external meters, not from the controller itself. The mill is at full cut in at 28 volts and makes about 10 amps (right on the money as predicted by the wind power calculator) 25 amps is about twice what I normally see at that speed and the booster is (supposedly) programmed to turn off at either 21 volts or 8 amps, whatever comes first.

I suspect that the FET blew during this period because if it were already blown then the amps would have been going down the drain, and not sending 2x the normal power to the battery? I think that the current and voltage sensing was somehow messed up and the booster was "optimizing" by sucking the inertia out of the blades. Eventually, (within a few seconds,) the FET blew. Now it measures 12 ohms in both directions, a failure nearly identical to the previous post.

- Ed.

[commanda]

Logic-level fets. I suspect you'll find that high voltage versions are very thin on the ground.

A regular fet will not be turning fully on with only 5 volts drive; hence will run warm and eventually melt down under sustained pressure, as you discovered.

I also suspect that your major problem at this stage is explained in your last post:

The wind picked up and the mill was producing 25 amps at 28 volts (about 280 RPM) and the booster was still running.

How clean (well filtered) are the current & voltage sense inputs to the micro. Do you have an oscilloscope so you can confirm this.

In software, is there a lockout time after the shut-off has been triggered?

Is there hysteresis? If it shuts off at 21 volts, how low does it need to fall before it turns back on. And when it does come back on, does it have a programmed soft-start.

Amanda

[elt]

Hi Amanda,

Logic-level fets. I suspect you'll find that high voltage versions are very thin on the ground.

I've looked and it seems I had about as good as I can find. I do have some TO-5 shaped FET drivers that I can dead-bug onto the board to drive some HV FETs but now that I've seen it fail I want to give the logic FETs one more chance.

I've seen the booster turn off plenty of times in gradual wind changes. I don't know what happened the second time but there was a sudden, dramatic increase in power the first and third times. And I know from th dump controller's logging that it was running at times during the night that the second one blew ... the same failure condition seems likely.

How clean (well filtered) are the current & voltage sense inputs to the micro. Do you have an oscilloscope so you can confirm this.

I don't have an o-scope. I have simple RC filters on both the mill volts and mill amps  ADC inputs. (.22uF with 15k and 6.8k resistors respectively.) Both inputs would have to be in error as either one should have caused the the booster to turn off.

I do have hysteresis for low voltage booster turn on but I thought it was "built in" for turning off: Since the turn off volts is set about 4 volts below main rectifier cut in, I figured that the blades would speed up when the booster unloaded them. When the wind dies and the volts drop below the cut-off point (and the booster turns on again) the booster will load the blades and slow them down even more. I realize that that isn't optimum use of power in the transition range but I thought it was bullet proof.

> when it does come back on, does it have a programmed soft-start.

Yes, I always start boost at the lowest duty cycle and load up from there.

... I'm at a loss as to what's going on but have two things I can think of to try: First, I'll put the booster back in without its switching FET and see if the PWM turns off in gusty winds and "good" power out from the main rectifiers. I'm not sure what I'll do with the piece of knowledge but it's something. The second think I'll do is switch the mill-voltage reading to an RPM reading; other folk have written that they've had trouble using volts... I haven't thought too much about that but am thinking that a zener diode and a RC filter connected to one of the phases will give me some pulses that I can count. If that works, it may address the high voltage problem (by avoiding it, even if I don't understand it). It doesn't address the failure of the high current cut out but may avoid that one too.

I don't mind a bit that it's not easy but it bugs the heck out me not knowing what's going on!

Thanks again,

- Ed.

[commanda]

I figured that the blades would speed up when the booster unloaded them.

The point I'm trying to make, is that micro-processors work at micro-second time. To talk to the real world, you need to forcefully set delays in software. When the booster drops out, do you forcefully set a delay before checking whether it should come back on or not? ie; give the blades sufficient time to speed up.

And you're a braver soul than I. You're basically working on a live patient, blind-folded. I'd want that alternator on the ground, driven by a variable speed drive, and access to a decent cro and multiple multimeters so I could see what's going on, and spin it up in a controlled environment. Or even simulate the alternator, with a variac and a variable series resistance.

Amanda

[elt]

Yes, everything runs on a millisecond clock; it's a real time multitasking app. There's a minimum 100mS delay before the PWM will turn on again after it's disabled. A range check occurs every 20mS so there's at least 5 opportunities to disable the PWM if volts or amps increase before the PWM can turn on again.

I'm doing "what if"s and am wondering if that could cause what I saw.

I'm wondering what if the current sense messed up under load. Then, (perhaps?) the 10 amps or so on the input cap might have sagged the voltage enough so that the voltage checks passed. Current was actually increasing as the booster sucked the energy out of the blade inertia and the alternator would only have had to drop from 280 RPM to 210 RPM for a valid "turn-on" condition based on voltage... that's what I'm thinking and why I'm looking at the ADC inputs.

Thanks again,

- Ed.

[jeffbirkle]

Hi Elt.

Great post,and lots of valuble info.

I see you built a Watt meter.

Do you have any info-schem for it?

Thanx

J

[elt]

Hi Jeff,

There's no schematic, I just wired four LEDs and an Allegro ACS752 to the tiny AVR. The Allegro chip is a 100 amp hall effect current sensor; they say that the newer ACS756 is preferred now.

http://www.allegromicro.com/en/Products/Categories/Sensors/currentsensor.asp

... cost about $5 USD from Digikey.