Problem with Shunt Regulator

[VinceB]

Hello guys;

Last summer I built a Shunt regulator for my wind generator using Chris Greacen's design. I actually tried it twice, and it worked for a while and then eventually failed both times. Here is the problem:

I initialy wanted to have a 40 Amps dump load through the regulator instead of 15 Amps, so I chose a IRFZ3205Z instead of a IRFZ40. According to the datasheet, the 3205Z can tolerate up to 75 Amps of continuous current, which is more than enough!

For testing purposes, I only plugged about 8 Amps of resistors. After operation for a couple minutes in dump mode, the lead of the tranny pins melted and the controller failed.

I guess soldering the tranny pins to 12 AWG gauge solid wire wasn't the best idea, because the second time around, I had the same results. Maybe the contact area of the solder is just too small.

So I am thinking soldering is not the solution, but instead, using a junction box or something mechanical like that would work. What did you guys use to connect the FET to your dump load...?

Any suggestion would be greatly appreciated, as I need to have it done as soon as possible and I know the controller would work otherwise. The electronics is fine!

If you can provide me with a part # or whatever relevant information would be awesome!

Thank you!

Vincent

[dinges]

I can't find the datasheet online for that FET, but it seems the Rds=0.008 ohm and Vds = 55V.

First of all, what system voltage are you using? I assume 12V.

Secondly, it's good use to post the schematic of the circuit you use; I assume this is the one you mean?

http://www.anotherpower.com/gallery/dinges/dumpload_controller_homepower_modified_as_built

(I can't be bothered to find the original circuit; this is the one I built, mostly based on the homepower design). It works, btw.

Apart from the obvious wiring errors (which I assume you have already carefully checked for), an issue might be that the FET doesn't turn fully 'on'. IIRC, most (non-logic level) FETs need about 14V or so on their gate to turn hard 'on' for that specified Rds=0.008 ohm; when there's less drive on the gate, the Rds could be significantly higher, leading to more power dissipation in the FET (as opposed to dissipation in the dumpload, where we want it). I know this is a possible issue with the homepower design.

Not a problem for me since my homepower controller doesn't have to dump more than 2A, so overheating of the FET is not much of an issue if it doesn't turn hard on, but when dissipating 40+ A as in your case, you definitely want the Rds of the FET to be as low as possible. This can be achieved by putting the right voltage on the gate (see the datasheet for that) to achieve the specified Rds=0.008 ohm.

Just the first thing that comes to my mind.

[dinges]

It may be less of an issue in your case than in mine, as I used an extra voltage regulator, I remember now (78L09). I used an extra voltage regulator because the NE555 can't stand more than 16V; a lead-acid battery can get dangerously close to 16V when no dumpload is connected and it is being charged.

I'd advice to measure the voltage on the gate of the FET; if it's much below that what is specified in the datasheet, that could be the issue.

If you want to really dump that 40A I'd definitely go for a type of FET (like an ISOTOP, e.g.) where the wires can be mounted using bolts and lugs.

Succes,

[ghurd]

Guessing thats a IRF3205.

Sounds to me like the fet and solder is not the problem.

See if the 555 output is getting up over 8V.  See if it's getting to the fet gate too.

See if the 555 is warm.

I ran a IRFZ42 with 6A, #16 wire, open air, no heat sink, for half an hour in a 74'F room and it stayed under 90 degrees F.

G-

[RP]

Yes that TO-220 case with its skinny leads isn't really designed for continuous 75amps.  The max rating on that is theoretical based on internal ability to conduct heat out through the tab.  Although the semiconductor will handle high currents, they need to be in pulses (like 50 micro seconds) to stand up to it.

I think you'd be better off using the original specified device and consider paralleling 2 or more FETs together to spread the load across multiple devices (and lead wires).

[Nando]

I have analyzed your problem.

First the 3205 MosFet though it is good for 75 amps ( pulse conditions), the static current has different parameters, at 40 amps the RDson the saturation voltage may go as high as 0.35 volts at 25 degrees, once the temperature of the Fet starts to rise, the RDson increases in value up to 1.7 times at around 75 - 90  degrees C.

Also, the voltage regulator is providing the wrong voltage to the Gate which should be at least 12 volts, max 14 volts.

Change the regulator to a 12 volts ( maximum 13.8 volts) then add a resistor of 1.2 to 1.5 K ohms in parallel with the positive supply pin and the output pin of the 555, then added another resistor of 39 K from the gate to the Source ( to insure that the gate is off when the device regulator is OFF.

The 1.5 K resistor forces the 55 pin to go to the positive volts supply level.

Under normal conditions the device needs a Heat Sink capable of about 25 watts for 20 degrees C rise,

My recommendation is to put at least 2 ,(better 4 ) MosFet devices in parallel, driven by the same 555, this way you may need a small Heat Sink for about 5 to 8 watts.

[VinceB]

Hi again,

Thanks for sharing your thoughts with me...

I realize now I didn't provide much information to start with.

The FET is a IRF3205Z. Here is the datasheet:

http://s21.quicksharing.com/v/9227832/IRF3205z.pdf.html

Here is the exact schematic representation of what I built:

http://i179.photobucket.com/albums/w294/thrice24/ShuntReg.jpg

The system voltage is indeed 12V.

According to the datasheet, the minimum threshold voltage VGS = 4 V (max).

I'm assuming the 555 provides more than 4V, so the FET must be turned ON.

And finally, here's the heat sink I used:

http://i179.photobucket.com/albums/w294/thrice24/Heatsink.jpg

I guess I could go the way of paralleling multiple FETs and loads, but with only 8 Amps load, the tranny melted, which is still odd.

[Nando]

IN PLAIN ENGLISH : You screwed up.

I am familiar with the circuit and in the past I have offered circuit changes for it.

The heat sink is good for 4 - 5 watts at 35 to 40 degrees Celsius temperature rise.

The initial turn on for the FET is 4 volts, but for the saturation turn ON, the device needs at least 12 volts, 14 to 15 volts is best.

Read my previous suggestion is the right solution.

Nando

 

[RP]

Vince,

The thing about your FETs is at 4 volts they "start" to turn on.  Being partly on means it acts like a resistor that lets some current through.  All the current going through this resistance creates heat in the FET and will burn it up fairly fast.

For this kind of application you want it to be "all the way" on so the internal resistance is almost zero.  That way almost no heat gets dissipated in the FET itself.

For another example of this consider making a high current switch out of a piece of steel and a 1/8" welding rod.  If you jamb the rod into the steel hard, a lot of current will flow and things will get warm.  If you just kind of barely touch it to the steel it'll have a lot of resistance and generate a lot of heat at the contact point and melt the welding rod.

Using the full 12volts as Nando said on the gate makes the FET turn on hard and very little power will be dissipated inside the device.

Using multiple FETs in parallel will have the further benefit of spreading all the load over several solder joints and wire leads making everything run cooler.

[kell]

mosfet datasheet

http://www.irf.com/product-info/datasheets/data/irf3205z.pdf

"threshold voltage" doesn't get you diddly squat.  

I didn't look at the irf3205 datasheet very long but most mosfets cite threshhold voltage at less than a milliamp.  Not forty amps!

So take a more careful look at the specs.

also look at figures 1 and 2 of the datasheet, what do you get for four or five volts Vgs?  Not forty amps, that's for sure.  What do you have on the gate?  For a particular gate voltage, there's a limit to how much the mosfet will conduct.  You need to see how much voltage it needs on the gate to conduct forty amps (figs 1 and 2) and then apply significantly more than that to make sure the thing turns on.  (This is not a constant current circuit, you want a high current switch with minimum dissipation.)  Your Rds will get lower the higher the gate voltage gets.

[Opera House]

Personally, I think anything with a 555 in it is just junk and can't understand why circuits like that are perpetuated.  That said, I think the FET is doing just what it should do.  I noticed that this circuit does not have a capacitor on the power supply for the integrated circuits.  Connecting this up to a battery with a length of wire could make this unstable could make the circuit start to sing.  FETS generate heat when they transition conduction state and they require a lot of drive due to capacitance.   It is likely you have enough drive for steady state 8A, my bet is the circuit is oscillating and overheating due to insufficient drive.  Hold an AM radio near it.  I've seen many high current circuits like this and the waveforms are not textbook pretty.

[Nando]

Let me clarify your statement to avoid future miss-conceptions on Power MosFets.

The basic threshold is around 4 volts, this represent certain minimum current traveling through the FET, which for regular MosFets is low.

One needs to observe, study and learn the the curves that show the FET ON behavior, including the saturation voltages at diverse currents.

The MosFet, in this case, is either Fully ON, when the Gate-Source voltage is 12 volts or OFF when the Gate-Source voltage is below 4 volts (the designers, normally, reduce this voltage to almost Zero Volts, and also, the transition of zero volts to 12 volts or 12 volts to zero volts is done fast with a very sharp rise of fall times, many times in the below 100 Nanoseconds, sometimes maybe increased (up to about 2 microseconds), to reduce certain high frequency emissions.

If the Gate-Source voltage is between about 4 volts to about 15 volts and the current drawn is above the saturation current, the MosFet behaves like a variable resistor, about the same way a transistor behaves -- LINEAR MODE --.

I stop here and gone to bed.

Nandio

[ghurd]

"Anything with a 555 in it is junk".  Me too.  That said...

I made a lot of the HP circuits. Way too many. Some with input caps.

I do not think the problems lie in the fet surge currents. Caps did not help me.

Transition conduction state seems unlikely too, because the HP circuit only fires once every 1.1 seconds.  Max.

The HP circuit should not be a waveform, should it?

It's On or Off, right?

I tried a "new" circit today. Again.

"Bang-Bang".  Go figure.

It was firing at <0.1hz to maybe 30hz. No heat anywhere.

G-

[RobC]

I have the same regulator setup. However what I did on mine was to parallel 5 irf540's  

for the output. These mosfets are rated at 33 amps which is ridiculous in my opinion. However in my shunt regulator I was looking for about 25 to 30 amps capacity total and it works great for that as each mosfet only becomes slightly warm to the touch. This unit has been in constant use now for about a year. I would like to rebuild it without a 555

I think it just complicates the circuit. RobC

[Opera House]

Maybe bang bang very second is better for a battery, but I prefer a proportional PWM.  There are any number of switchmode regulator chips that have been around for more than 20 years and are just as available as a 555.  I posted a TL494 shunt regulator made out of an old PC power supply.  With this chip you get an operating supply voltage twice that of 555, a precision voltage reference, and half the number of parts with a single chip.  Keeping the switching frequency under 1KHZ eliminates capacitance driving problems associated high capacitance of multiple gates. By powering the fan with the shunt load with a diode and cap, the fan powers on at a point when the resistors need a little help with heat.

[VinceB]

Hey Nando,

I replaced the burnt FET today and tried the circuit again. Here's what I found out:

When the battery is being charged, and reaches 13.5, the regulator turns the dump load on. The VGS then is around 10 V, maximum. We're far from reaching 12 V. I connected a pull-up resistor to the output pin of the 555, which made no difference at all.

Actually, if you look at the circuit, there's a diode (1N4001) proving supply to the electronics. There is a voltage drop right there of 0.7 V. So even when the battery is fully charged, the supply of both chips is around 12.7 V.

What do you mean by replacing the voltage regulator to a 12 V one ?? The LM723 chip is used as an op-amp (comparator), not a voltage regulator in this application.

Still looking for a solution to raise the Vgs level...

[VinceB]

I also forgot to mention, the FET got warm in about 30 seconds... which probably says a lot. Remember the connected load is only around 8 Amps...

[VinceB]

Quick update:

I modified the circuit a bit and got different results.

I eliminated the "yellow" led, the Vgs increased. I put it in parallel with the load instead.

It increased even further with a pull-up between the supply (after the protection didode) and the output pin of the 555.

Finally, I put the pull-up between the battery + and the output of the 555 and got even  better results.

Here are the numbers:

With a 14V setpoint

Vgs = 11,8 V without pull-up

Vgs = 12,4 V with a pull-up between circuit supply and output of 555

Vgs = 13,0 V with a pull-up between batt. + and ouput of 555

But I also removed the 100 ohms resistor connected between pin 3 of the 555 and Gate.

Does that seem correct?

I've been running it a few minutes and the tranny is not heating up.

Vincent

[VinceB]

Hello Peter,

Why did you add the BC597 and surrounding components?

Why did you change the time constant of the 555 to 4 seconds ? Is a longer time constant desirable for lead-acid batteries?

Thanks!

Vincent

[dinges]

"Why did you add the BC597 and surrounding components?"

As I stated in post nr. 2,

"I used an extra voltage regulator because the NE555 can't stand more than 16V; a lead-acid battery can get dangerously close to 16V when no dumpload is connected and it is being charged."

It's a BC547 btw, not a BC597 (which I have never heard of) but any universal NPN small-signal transistor should work.

I wanted a longer time constant (in reality it's even more than 4 seconds, closer to 6 or 7 seconds) to give a bit more hysteresis in the system (batt + energy source + controller). With a short timeconstant the dumpload could potentially turn on and off constantly with an off-time that might be very short (fractions of a second, possibly). By using a longer time to dumpload the battery (6 seconds), battery voltage gets lower and it thus takes longer for it to reach the dumpload trigger level again; that way the dumpload will not turn on/off so quickly. Some dumploads may not like this rapid on/off cycling. Note that this solution only works when the energy source is relatively small compared to the rest of the system (battery); that short time constant wouldn't help much in the hypothetical case of a 75W panel charging a 7 Ah battery.

[Nando]

The TL494, & Tl594 and other names with the same circuit have been the work horse for many supplies coming from the far east.

One can set the frequency from below 100 Hz to medium high kilohertz.

The TL494 does the job of the LM723 and the 555 many times better.

Using PWM instead of Bang-bang, the regulation is better and more stable, the MosFet does not heat as much and one may be able to use a smaller heat sink.

Also, one may be able to ballast the power source and not the battery, this way the battery receives many less charge/discharge cycles effects, prolonging its life.

Nando

[Nando]

Sorry for the error of the voltage regulator.

This happens when one is trying to answer to more than one place at about the same time and miss read the message that Peter included with the 78L09

The main problem is the way that the MosFet is driven and the heat sink as well as the maximum current the MosFet is capable of accepting, because one thing is dynamic and other thing is the static behavior of the unit.

75 amps is when the MosFet is ON for a limited time according to the saturation RDson times the current and the heat sink attached to it to keep the device temperature below the dangerous point ( keep mine, military worse conditions below 90 degrees C)

Leave the 100 OHms that is in series between the 55 out pin and the Gate of the MosFet

Do not place the resistor from input voltage and the output pin of the 555, in an essence, do not bypass the isolating diode.

Lastly, get a good Heat sink, If the ballast is on for 3 or 4 seconds, then measure the MosFet ON voltage, it should be below 0.1 volts.

Lastly, do place a capacitor (0.4 Uf up across the IC supply pins to stabilize the devices, because in some cases the circuit could oscillate at high frequency depending on the length of the supply leads, including the ballast leads.

Make sure that the ballast leads going to the battery are short and independent from the leads going to the controller, because if the circuitry is common the whole oscillate like a bansheeeeeeeeeeeeeeeeee!

Nando

[RobC]

Personally I would like to try this Tl494 circuit. Anybody want to post a complete schematic? Now that we have been through all this discussion I'm sure others would like to get rid of the Hp design and try something better. RobC  

[Nando]

A better IC would be a UC3823 since it has a totem pole output for fast up/down gate switching pulses.

You can see the data sheet in my user files UC3823.pdf

Nando

[VinceB]

But what would be the desirable operation frequency to send to the FET gate ?

The HP circuit is way to slow I think, creating huge voltage swings at the battery. I'm thinking that if the frequency was higher, the voltage would eventually stabilize near the setpoint, right ?

Once again, thanks guys for that discussion and all the new info provided!

[VinceB]

I would really like to see your TL494 shunt regulator!

[Nando]

For this type of batteries, around 100 Hz to 5 KHZ is quite sufficient and the desired setting will be attained sooner and more accurately with lower battery voltage swings.

Nando

[VinceB]

Hey Nando,

I first used Ra = 1 MegOhm, which gave a rather long time constant. About 2 seconds.

I put Ra = 1 K, and I ended up with a 1,5 KHz switching frequency. The voltage would then stabilize at the exact setpoint. The problem is, the output of the 555 drops to around 8 Volts when using such frequency. So Vgs is then way too low, even with a pull up. Why is that and what can I do then??

Thanks!

[VinceB]

Wow!

I'm having even more erratic behaviour...

I made a mistake in my last post. The switching frequency is about 10 kHz with a 1 K resistor. In that mode, the Vgs dropped, like I said in my previous post, at around 8-9 Volts. I then had the idea to add a 220 uF capacitor between the (+) and Gnd of the circuit. The Vgs went immediately up to 12,5 V... But that created another problem. With the capacitor plugged, the circuit seems to go in an erratic mode. The setpoint voltage becomes lower and the LED flashes in a weird fashion. Removing the cap makes the setpoint go straight back to 14.0 V. What's the problem now?? Oscillation maybe? I'm kinda lost.

Maybe I'd better use the original design with a few IRFZ40.. (

[RP]

At those frequencies you really need to be measuring the gate voltage with an ocsilloscope rather than a meter.  I suspect your still have 12 volts or so on the gate but your meter is reading them lower because of the pulses giving you an average voltage of the square waves.

[Nando]

To really know if you have the right circuit, does the MosFet gets very hot ?>

I do not think so, so the gate pulses are above 12 volts, keeping the MosFet Fully ON.

You need an oscilloscope to see how the MosFet turns ON, low to almost zero volts.

You are reading 8 volts because that is the average of the pulses which maintains the battery voltage stable.

Take a 10k resistor or whereabout from the Drain ( ballast connection) and at least a 0.47 Uf/50 Volts or better or higher value to the other end of the resistor, and the other side of the capacitor to ground ( negative) and read the voltage across the capacitor, which is going to read from battery voltage to zero volts, this depending on the charging current -- the voltmeter should be at least 20,000 OHms/volt if it is a DVM, it will be around 20 megohms.

If you want to see a permanent meter showing the inverse clamp energy, which tells you what excess is being dump, place a milli-ammeter in place of the DVM and adjust the 10 K to give you full scale when the battery is fully charged and no DUMPING is occurring (if using a 1 mILLI-ammeter you may need 1000 OHms/volt then the you need to take 50 ohms from the whole value for the 1 ma meter internal resistance value.

Nando

[VinceB]

I suspect that too.

With an even higher frequency, I've seen Vgs drop to 4 V.. which is ridiculous, because the FET was still in conduction.

[VinceB]

Hey Nando,

The mosfet does not get very hot. After about an hour of operation in a 77 °F room, the case is stable at around 110 °F, with no heat sink.

I plugged a 10K resistor and 1 uF/ 50 V, as suggested in your previous post. The voltage in the capacitor is around 3.8 V. What does this mean?