More Buck Converter Question

[martin1]

English isn't my language. So I better draw some pictures.

Would this work as a speed signal? The idea is to capture the peak voltage at the end of the T(off) cycle when the input current is zero. Using a very small capacitor and a diode.

A more detailed view. The idea with the zener is to provide for a cut in. The voltage divider is there as voltage divider and as high resistance load for the capacitor.

I don't know if it is an idea worth pursuing or if it is completely stupid. Any comments would be appreciated.

[Flux]

I take it that you are trying to extract a speed signal from the ripple of the dc from the 3 phase bridge rectifier. Under normal conditions this could be possible but tricky, I suspect that in the presence of the noise from the converter chopping it will prove completely impossible.

I use the voltage across one pair of ac leads from the alternator fed through a little transformer and filtered then fed to a 2917 tacho chip. Even this needs a lot of care to keep the chopper noise from interfering with the speed signal but it can be done.

A better way may be to use a hall current sensor to look at the current in one of the alternator leads to the bridge but I haven't tried it since I managed to get the voltage signal to do it in the end.

Flux

[martin1]

I never considered the noise problem. Thank you Flux for pointing that out. The reason why I don't want to use the frequency of the AC line is that my system is Jerry rigged with the rectifier in the tower. Maybe I have to run two small wires down just to get a signal.

I like the idea of using a hall sensor or perhaps a hall switch is even better. One problem though is that you only get a signal when the generator is loaded.

[martin1]

No the idea is not to extract a speed signal from the DC ripple but to get a direct measurement of the emf.

[elt]

If you're going to run two wire up there, why not put a little wire loop (perhaps an inductor you have lying around) up near the magnets on the back rotor? Won't matter if the genny is load or not that way. The hub I used has an ABS sensor in it, though I haven't yet tried to see what, if anything, comes out of it.

Or if you want to do the sensing at ground level, you only have to run one wire down from one of the phases; you can use your ground as the other input to your sensor.

Hope that helps,

- Ed.

[Flux]

I don't see much chance with emf, to get at it you need to open circuit and with wind that is most undesirable.

I see your problem with Jerry rig. Elt's idea of a sensor direct from the magnets seems a good one if you have to run a separate cable.

Whether it will cause confusion about speed being zero up to cut in depends on your control scheme, my analogue scheme wouldn't be bothered by it as the converter is phased fully forward at cut in. With a processor looking at time interval it may be a problem.

Flux

[martin1]

I haven't completely given up on my first idea of using the input voltage as speed signal. My thinking is that in a buck converter operating in a steady state the input current must be zero at the end of the OFF state. At this moment the generator is an open circuit and if it's possible to catch this moment using a peak-hold circuit you will have a speed signal.

Elt's idea of an inductor is great but I'm wondering if there's enough flux on the back of the rotors. The thickness of my rotors are 6mm and the thickness of the magnets are 10mm.

[Flux]

If you stick a small coil to the stator as close to the magnet disc as possible I think you will pick up a signal, even if you have to mount the coil at right angles to the stator surface, there is lots of leakage flux I wouldn't try at the back of the disc.

I am convinced you will not get emf from the dc as you will need a sensible input capacitor to the converter.

Flux

[martin1]

Thank you Flux and Elt for the suggestions. I will put a pick-up coil on the stator very close to the side of the rotor and hopefully it will pick up a signal.

Yes you're right. The input capacitor will probably smooth everything out. Another way to get the emf could be to measure the input current and add it to the input voltage. It could all be done with just a hall sensor and an op amp chip.

[Flux]

That was an old trick to maintain constant speed on motor drives without using a tacho.

Use armature voltage as a reference and introduce IR drop compensation from the current. If you got it right it worked quite well. I hadn't thought of trying in reverse to get a speed reference from a loaded generator but yes indeed it would work especially with axial machines where the impedance drop is almost entirely resistive.

Flux

[Slingshot]

This would be a little problematic, as the voltage out of the generator would be a function of both RPM and the electrical load.  The more current flowing, the lower the voltage would drop.  So you'd need to monitor both voltage and current in order to compute RPM.

[tanner0441]

Hi

A quick question aimed at Flux. If a small signal coil was added when the stator is being cast the frequency in this coil would be directly related to the rotor speed, and the voltage in it would be related to the load on the stator.  So the speed and load could be displayed, and as it is a seperate coil it would not be dependent on the integrity of the main winding. The signal could even be used to sound a warning in the event of burn out or over speed. Or dump the battery into a solenoid or motor operated brake.

The electronics could be a simple frequency counter and comparator for speed, and amplifier driven meter for current.  If the impedence is kept low with a bit of decoupling then there would not be a problem with cross talk from the main cables or noise from the controler circuits.

Simplified I know but posible?

Brian

[scottsAI]

Martin1,

What your building is a MPPT for wind systems.

The control is more complicated than knowing the RPM.

Wind turbine without a load will spin faster. Ideally loaded blade speed should be half non-loaded speed.

Most designs lock the generators output to a battery. Once this point has been reached the RPMs do not go up much as the wind speed increases. Blades are now Stalled, efficiencies drops. The more stalled they are the lower the efficiencies. Most DIY systems plotting power output vs wind speed goes up linearly when it should be a cubic function. DIY systems, for the most part are interested in the lower wind speeds so this is not a problem.

BTW, what are your goals? Trying to get power below cut in?

Have fun,

Scott.

[Flux]

The sensor coil will give you frequency ( hence speed). The induced voltage will be a function of the flux linkage. With neo magnets and large air gaps the armature reaction will be very small and I think you can reasonably assume that the voltage induced will be linear with speed and you could use voltage as the speed signal. ( it virtually gives you emf as you don't have the resistive drop of the main windings carrying load)

The voltage will almost entirely be dominated by the speed and there will be very little effect with load. You could certainly use it as an overspeed signal but I can't see it being any use for sensing load. You could compute loss from emf, battery volts and current but you can do this knowing speed and winding resistance without the sense coil. There may be enough armature reaction from a burnt out winding to show on the sense coil but it is an imaginary concept as the loading from a shorted coil bad enough to show on armature reaction would drag the blades down into stall or probably stop them.

Flux

[martin1]

Yes I realize that the control is more complicated than just knowing the speed but I like to deal with one thing at the time. Right now it is getting the speed signal right. Next I will be working on the cubic function but I already have some ideas on that part.

Maybe if you skip the low part of the wind curve it is possible to use a square function instead. If that is true than perhaps it's possible to use a square root function on the current signal going to the batteries and than use the speed signal as it is as feed forward to the PWM circuit.

The goal is not to lower cut in. It's more like getting a higher cut in. The stator was made for 48 V but I want to use it for 12 V.

[Flux]

In the perfect world you need a cubic function. In real life for the input power to track the cube of speed then with the typical losses of a small machine it does work out quite close to tracking the output power to speed squared.

That is that way I do it. I take speed and use a multiplier to get a voltage proportional to speed squared. I assume battery volts is constant so output power is proportional to current.

I use a servo loop to track the current to the speed squared reference so that my output power tracks the square of the speed.

I also find that my props work best with falling tsr with wind speed but other props may not behave the same way. I am inclined to think that most as constructed here do behave that way and it is one of the reasons why the direct matching scheme works better than it otherwise would. In any case props are very tolerant and any reasonable match to the ideal loading will be so much better than no attempt so absolute perfection in the tracking is not very necessary.

Flux.

[martin1]

Thank you Flux. It's nice to know that a square function is good enough since it is so much easier to make than a cubic function.

I don't understand what you mean by servo loop. Do you have some additional circuitry between the current feedback and the PWM?

[scottsAI]

Martin1,

Recently posted on Solar MPPT controller design. Function ended up maxing only current. Controller Cycle time on solar panel could be relatively fast, as in fraction of a second.

With wind maxing out only current should work also. Have not spent much time on the thinking. Realizing the delay loop is much longer than solar. At Lower power, longer control loop delays.

Since your speed is based on loading I wold skip it. Maximize on current going into the battery. Forget everything else (as a MPPT function, pay attention to other things for battery charging controller:).

My back ground is Micro, so I would recommend using intelligence in a micro to control this wind MPPT.

Have fun,

Scott.

[martin1]

I'm not prepared to use a micro in this project. I've just recently picked up PIC programing so this project will be all analog or perhaps a mix. It make sense what you are saying about the delay being longer for wind than solar. It must take some time for the turbine to respond to load changes.

Now I'm thinking of using a digital potentiometer to deal with the delay. It could go up or down almost like a regulator with an integrating search.

[scottsAI]

Martin1,

Not surprised you say that.

After the fact the software is much easier to change vs hardware.

Intelligence in micro is superior to analog logic.

Everything has a price. Micro is typically cheaper than analog logic.

In the mid 70's fist used a micro.

This summer for a project selected a "New one".

Figured it would take two weeks to learn and write the page of code. No biggie.

A month later still learning, little writing. Documentation is honorable!-(

6 weeks coding completed. Wow, way off on that one.

Have fun,

Scott.

[Flux]

I sum the speed squared voltage and the voltage proportional to current in an error amplifier and feed the error signal to the pwm.  Perhaps this was what you had in mind, it is just a servo loop as used for any analogue feedback control.

This way the current is proportional to the reference signal.

Interesting things to watch for are that if the "speed squared" signal limits then the mill goes to constant current ( run away and avoid it). If the current signal limits then it aims for constant speed and that is nice to hold the things under control and induce a measure of stall at the furling point.

For those trying to follow this but not completely understanding, it is no way a mppt controller. It is a controller to force an alternator power out to follow speed cubed at the input. It has to be set up to match a given machine. The set up is not difficult, you have to set the pwm at cut in ( error amplifier offset) and set the current gain of the current amplifier to get the correct pwm at full load.

Flux

[martin1]

Yes I will do it like that. It's far more simple than constantly searching for the maximum current output. I don't know if that is even possible.