AC Sensing - Generator control

Hi folks,
This isn’t going to be quick … Please bare with me.

I have, or at least had, a small single phase generator, circa 2.3KVA with an electronic field control circuit ( AVR - Automatic Voltage Regulator)
The AVR module is dead, short, so the genny output is way up and uncontrolled.

To test the system I applied a field current from a bench supply, it draws about 1.2A DC at around 30V when the genny has a 500W load.

All the windings are OK, the main stator has a tap at aprox 10% which was connected to the AVR, I assume for voltage sensing, and there is a secondary stator winding that outputs about 100VAC, during the above test, which was also connected to the AVR unit and must be the source of the field current, via whatever control circuit used to be in place.

The broken AVR unit is completely potted, giving only the merest indication of a substantial bridge rectifier and one big electrolytic cap, everything else is hidden in solid opaque epoxy.

I have built a small circuit around a FET capable of driving the field from a PWN signal (It probably needs additional work on the gate drive to reduce switching losses but will do for now.)

I have a sawtooth generator, supplied from the power control board above, which is stable even when the generator is under load, and can control the field adaquatly by applying this signal and a DC level to a simple comparator ......

My first problem, more RE fet gate drives later, is getting a stable DC level that is representative of the AC output of the genny.

I currently have a half wave rectifier, on the 10% tap, feeding a cap via a simple resistive bridge. The problem's are:-

Large cap = smooth DC level but way too slow to be of any use and the whole system oscillates slowly.

Small cap = large ripple on the DC level and the PWM becomes wildly unstable with a very noisy gate signal. Overall however the final regulation is 'better', high frequency oscillations, but the fet would fry if driven in this way for long.

In addition the type of load, (chopping PSU verses water pump), makes a huge difference to relationship between the average DC level and the generator output. I think this is due to the 10% tap behaving partially as a current shunt and thus superimposing a voltage proportional to the ofset current waveform that the above reactive loads induce in the main winding.

Sorry that's long but it seemed reasonable to provide all available info before asking questions.
SO .....

Q1 Do you think my 'shunt' theory is correct, if so it is obviously intentional but why?

Q2 How can I remove / cancel the ripple on my existing DC level?

Q3 How would I get a DC level proportional to the peak of the AC.

Q4 How would I get a DC level proportional to the Average of the AC.

Q5 Can I measure RMS without a microprocessor given that the generator is highly unlikely to produce a clean sinusoidal output and will almost always be exhibiting a poor power factor.

Q6 Am I just going about this all wrongly, if so why?

Any help would be much appreciated, 'O' and dont pull any punches I am quite happy to look silly but have a working generator again.

Thanks for reading all this
Al
　
　

Hello Al,
Posting a schematic, preferably in .PNG format, would be a big help in getting your thread going. Schematics eliminate many questions that would take verbose, complex wording to adequately describe.

It also provides a record of "what is"; so that you can see a logical progression to "what it should be" as a result.

If you don't have schematic capture software, you could download LTSpice, a free schematic capture and SPICE program, from Linear Technology. Google will help you find it quickly.

OK Sounds like a plan ....
Its all on breadboard right now, well other than the power board, and sketched on paper.
I will do something 'postable' ...
I never tried LTSpice in fact havnt used any simulation since EWB way back in the 80's
Would you have an opinion on my fat gate drive issue .. 'a huge potential rail swing'
Off to find LTSpice right now, will add schematics fro both.
Thanks this far,
Al

Well, you need some type of circuit that will wait until the cap is charged sufficiently before attempting to turning on the MOSFET gate, or your MOSFET will be subjected to high power dissipation due to operation in the linear (resistive) region.

However, since I don't see a schematic yet, it's difficult to make progress.

Tips for creating a .PNG image file:
1) Display your LTSpice schematic on the screen.
2) Press Ctrl+PrintScreen.
3) Start MSPaint, either by clicking on it's icon in Start -> Programs -> Accessories -> Paint, or click Start -> Run -> enter: MSPAINT
4) Press Ctrl+V or click Edit -> Paste to insert your image.
5) Click File, then Save As...
6) Give it a file name, and select .PNG in the scroll box below.

In order to attach it to the Forums, click the "Go Advanced" button at the bottom of the text box, and then click the "Manage Attachments" button on the next page.
You'll need to allow popups from these forums in order to get the Manage Attachments dialog to display.
The rest is somewhat self-explanatory. File attachment sizes are limited, as shown in the dialog. .PNG format files are very compact, not "lossy" as .jpg files are, and don't require any other software than an Internet browser to view.

This is what I have now.
I had a quick look at spice, looks like it will be useful but I will have to learn how to use the thing and find / add the models I need.
I have drawn this with open office draw, please forgive the quality and or out of date symbols.

With the values shown U1a is running at about 200Hz
The output of U1b is far from correct because there is way too much ripple on the 'DC' level at its inverting input.
The 5.1V and 10.2V rails are stable.
the 4 47K resistors are noticably hot
The fet gate is not right... noisy obviously, too low - it only swings to about 8V, and occasionally needs to be disconnected from the opamp to initally switch on at startup.

I did try a bypolar totem-pole to buffer the opamp,
it just made smoke
A single NPN turnoff transistor worked better but always prevented initial startup.

There is another thread asking about the gate drive issue. I want to solve the control problem first if possible.

Spaciffically how to get a smooth level at U1b - that is representative of the generators output?

Please bear in mind this is simply an experimentation board ... when I find out how it really needs to be done I will redesign accordingly.
Any input will be much appreciated.
Thanks
Al

I am a little surprised that no-one has an opinion .. :-(

There are obviously some seriously clever people contributing here.

Am I asking the wrong questions or am I just so far off the mark as to sound silly?

I have no intention of shoving these posts continually back to the top, that’s just rude but I hope you will all forgive this one indulgence.

We live off grid, have little money and this stuff is really important to me.
Thanks folks
Al

Dyslexicbloke said:

I am a little surprised that no-one has an opinion .. :-(

There are obviously some seriously clever people contributing here.

Am I asking the wrong questions or am I just so far off the mark as to sound silly?

I have no intention of shoving these posts continually back to the top, that’s just rude but I hope you will all forgive this one indulgence.

We live off grid, have little money and this stuff is really important to me.
Thanks folks
Al

Click to expand...

Truth is most of us aren't blessed with unlimited time on our hands, we have families, jobs and other things we have to share our time with. Then there's the fact the person(s) with the definitive answer haven't been online this weekend due to vacation, or illness, or who knows?

Most all questions eventually get answered in one way or another, in forums like this there are people like me that occasionally look for threads that have no answers yet and do our best to get something as far as a response written even if we don't know the answer.

Don't despair, you haven't been forgotten about. I'm considering a number of ideas that involve PWM controller ICs.

Rather than use an ordinary rectifier bridge/cap, I'm considering a precision rectifier in conjunction with a peak detector to provide the output sense level.

[eta - this paragraph is in error, as the schematic does indeed show an N-ch power MOSFET; however it was placed in the schematic inverted which caused me to misinterpret it as a P-ch] You're using a P-channel MOSFET to supply the rotor current. In order to more easily use one of these PWM controller ICs, you'd need to use an N-ch power MOSFET instead; going to an N-ch gives you a benefit in a considerable reduction of gate charge, thus less power dissipation in the MOSFET. N-ch MOSFETs are much more common due to this characteristic. A P-ch MOSFET with the same Vdss, Id ratings of an N-ch MOSFET has approximately 2.5 times the gate charge.

Since you appear to be using this genset as a primary power source, it's very important that it is as efficient as possible (within reason of course).

Rather than a simple PWM applied to the rotor, I suggest what would be a rather large improvement is a synchronous buck type of arrangement. This uses one N-ch MOSFET to switch current on one side of the inductor (your rotor winding), and a second N-ch MOSFET to act as an "ideal diode" to provide an alternate current path when the inductor is not being charged. This is why high-power rectifiers are getting hard to find nowadays, as they waste a lot of power in comparison to using a MOSFET as an ideal diode.

For a somewhat simplified look at the concept, have a read of Linear Technology's LT3800 datasheet, a HV synchronous current mode step-down controller.
Link: http://www.linear.com/pc/productDetail.jsp?navId=LT3800
The datasheet is available under "Documentation" on the right side of the page.

Note the high efficiency (>90%) over a fair range of output current and input voltage.

I am not a professional power controls designer - however, I think this type of approach has a good chance of success.

I'll estimate that you might need as much as 8A rotor current to provide your rated 2.3kW output.

Do you have an oscilloscope available? This type of circuit would be nearly impossible to troubleshoot without one.

Hi,
That all sounds both interesting and peommising.
The fet that I am currently supplying the field with is N-ch .. did I draw i incorrectly?
It was salvaged from an old inverter .... I have a couple more.

If the drawing is wrong I will update it as that will definatly help.

I am about to go and look at the concepts you have mentioned to get my head arround them.

I have a small Vellaman LCD scope which will be quick enough for anything here I think.

The field current figure I gave earlier was not accurate it is closer to 750mA with a 500W load on the generator, the voltage 30VDC was correct.... Oooops

I really apriciate your time and effort ..
Thanks thus far
Al

The LT3800 looks like a great piece of kit but I thought converters like that, working at high freequencies, needed careful PCB design because of stray capacitance and inductance ... cant imagine how I would go about building that on stripboard.

The other nagging dought I have is an adiquate supply at startup, remember I only have about 12VAC until I can get some current into the field. After that it comes up rapidly. Sort of hoping you are hatching a plan there

I tested secondary stator rectified, smoothed and applied directly to the field, no pwm. the generator output into an incandecent 60W lamp in that config is around 340V

Al

Hi Folks, I'm desperate for some help here ....

Atached is what I have now, the bits in red are almost cirtainly not correct. Several other threads and a huge amount of other reading have contributed to this. It's time to pull it all together I think.

I am working on the asumption that no smoke thus far is a good thing .....!

Q1 (Alu heatsink on the board) is running HOT, its source is at 25V with the drain at 120V. At startup 10-12V rail it's offset is lower at about 2V.

The 50Ω resistor shown in the gate leg isnt actually fitted because it didnt seem to make any difference.

Ripple on the inverting input of U1b is causing problems, although the gate signal looks relativly clean the pulse width is jumping around all over the place. This is far worse when the load on the generator is my inverter based welder, which I use for charging the batterys.
(And yes that a whole other story and topic)

The final generator output is reasonably stable, for any particular load, and will probably benifit from an increse in PWM base freequency once the reference at U1b is stable.

When a load is removed, or even decreesed to a single 60W lamp for example, the output is rising changing significantly. Mucking about with diferent capacitor values on the ref at U1b changes things but not in any way I am able to predict.

Please throw anything into the pot you think may be of interest, I am doing this because I have to not because I think mucking about with high power circuits is a good idea with my level of knolidge and ability.

I believe the big issues are:-
A power supply capable of working from a verry wide input.
A clean DC reference that accuratly represents my generators output.

(The Power board is built whilst the PWM circuit is still on breadboard)

Al

What does the out put of the 317 look like on the scope? Could some trash on this DC level be doing something undesired to your PWM circuit? If you find some noise on it, perhaps using some more bypassing on the opamps will help.

I'm trying but not seeing much there to give me ideas. I'd be better if I could 'poke' around that circuit in person.

maybe a choke coil could smooth things out as well. Maybe.

???

I'l check the DC but I think its OK, I cant see it being worse with the 317 than it was without.

The noise on the PWM is mostly due to the ripple on the U1b ref.
I have proved this by applying a test level from my bench supply ...
I can swamp the ripple with a big cap easily enough but when I do that the whole system starts to oscilate slowly which obviously isnt any good.

The rectified supply from the secondry stator is at 125V with the generator under load and this circuit supplying the field.

I have removed the LED, power I dont need to disipate.

I think reducing the 330R pulup will probably help, with power consumption, but I am unsure how far I can go befoer the totem-pole output stops driving the fet gate effectivly .... One of the things I need advice on.

I had wondered about using J-FET's instead of BJT's to drive the MOSFET would that work? I'm not sure because they are depleation devices arnt they?

Al

Ugh - I suggested a TL783 in the other thread: http://forum.allaboutcircuits.com/showthread.php?t=42922&page=3
however, it has a dropout of anywhere from 14v to 20v, so that won't work for your startup requirements (12v in -14v = 0v out)

OK, back to the Darlington circuit - you could use an MJE340 instead; they're rated for 20w, 300v, and Farnell UK stocks them for only £0.097 each. Cheap at twice the price, right? Buy several.

http://uk.farnell.com/multicomp/tmj...406?matchedProduct=MJE340&whydiditmatch=rel_3
MULTICOMP - TMJE340 - TRANSISTOR, NPN, TO-126

Now that sound just peachy ..... I will order some ASAP Thanks.

I assume the heat I am seeing now is power disipation rather than poor use of a FET though isnt it?
that means I would still need to deal with that right?

I is a pity there isnt some way to use a chopping supply ... how is that done commertially?

Could I go with something depletion based and simply turn it off when the output rail was high enough and back on again when it dropped by some margin.
I supose I am talking about a hysterric oscilator that would be allowed to free run?
I say depletion because it would be on by default meaning by the time I needed to turn it of I would have a supply to do it with.

I have liffted the 330R to to 1K3 which has definatly cooled things down. The DC has a maximum ripple of about 80mV but that is only the occational spike, the ripple at gate freequency is only circa 8mV.

The gate on the output FET is looking fairly clean with an inductive load on the generator, my pump, and is swinging nicely by 10.1V I forgot to check the minimum but it must be low enough to turn off properly because the thing is running nice and cool.

http://cpc.farnell.com/jsp/search/b...ions=false&ref=globalsearch&_requestid=634287

I have an account with CPC .... which of these would be better?
(and why if you dont mind so I can learn)

I am a little concirnrd about the max power disipation there dosnt seem to be much headroom. The 317 requireing 10mA min load dosnt help I probably need to work out the min current of my circuit anf then change the 317's devider.
Al

Dyslexicbloke said:

Now that sound just peachy ..... I will order some ASAP Thanks.

Click to expand...

Sure.

I assume the heat I am seeing now is power disipation rather than poor use of a FET though isnt it?
that means I would still need to deal with that right?

Click to expand...

With a linear regulator dropping the voltage from ~120v down to ~12v, your power dissipation is going to be (120v-12v)*current. There isn't a good way around that, except to go to a different type of supply.

I is a pity there isnt some way to use a chopping supply ... how is that done commercially?

Click to expand...

The problem here is that you have a REALLY wide input voltage variation. ~12v to ~120v is quite a stretch for a switching supply.

Could I go with something depletion based and simply turn it off when the output rail was high enough and back on again when it dropped by some margin. I supose I am talking about a hysterric oscilator that would be allowed to free run?
I say depletion because it would be on by default meaning by the time I needed to turn it of I would have a supply to do it with.

Click to expand...

Depletion mode MOSFETs are seldom used nowadays, and they are very low power. Most of what you see is enhanced mode MOSFETs.

I have liffted the 330R to to 1K3 which has definatly cooled things down. The DC has a maximum ripple of about 80mV but that is only the occasional spike, the ripple at gate freequency is only circa 8mV.

Click to expand...

Anything to reduce power consumption while keeping the supply coming is a good thing.

The gate on the output FET is looking fairly clean with an inductive load on the generator, my pump, and is swinging nicely by 10.1V I forgot to check the minimum but it must be low enough to turn off properly because the thing is running nice and cool.

Click to expand...

That's good.

Dyslexicbloke said:

http://cpc.farnell.com/jsp/search/b...ions=false&ref=globalsearch&_requestid=634287

I have an account with CPC .... which of these would be better?
(and why if you dont mind so I can learn)

Click to expand...

They're equivalent. If you want a name-brand component, then go with the ST Microelectronics. If you want to save a few pence, go with the other. If they both say "MJE340" on them, they need to at least conform to the requirements of the device. Usually you are better off to stick with name-brand components from manufacturer-authorized distributors; as that way you can be assured that the components will meet the specifications.

Always buy some spares, as you will likely have to make repairs in the future.

I am a little concernrd about the max power disipation there dosnt seem to be much headroom. The 317 requireing 10mA min load doesn't help. I probably need to work out the min current of my circuit and then change the 317's divider.

Click to expand...

You can avoid overshoot during the LM317's start-up by placing a 1uF to 10uF cap from the ADJ terminal to ground. This is documented in National Semiconductors' datasheet for the LM117/LM317.

The 10mA minimum current is to provide the guaranteed regulation. If your output load is less than that, the output voltage will rise somewhat; exactly how much depends on a number of things.

However, since your regulation really isn't that critical (within a volt should be OK, right? 10v to 15v should be just fine) you might just go with 1k from OUT to ADJ, and an 8.5k resistor from ADJ to GND, with a 1uF cap in parallel with it.

OK so staying with a linear aproach to power supply, using a BJT in place of the MOSFET, I need to reduce the overall current consumption so ....
How about a totem-pole from JFET's, or even smaller MOSFETS.
Any thoughts ... I will be ordering from somewhere anyway so a couple of extra componants makes no diference.

Al