Mosfet HV regulator : 2 multimeters fried

[Rutherberg]

Hi,

I'm trying to make a pretty basic mosfet regulator but so far I failed miserably.
Have a look at the schematic attached.

My multimeter is between Va and ground, Vdc 600V selected, and when I turn my transformer on, I hear a pretty bad clicking noise (typical noise of a faulty semiconductor or short, you know what I mean) at 100Hz and the DMM is gone.
Same result with a second multimeter.
My multimeters were pretty cheap but have done the work for many years. Its rated CAT II 600V, but I saw the thread on the cheap-ass multimeters shootout... 

But c'mon, its at best 450Vdc so I don't understand why it fails like that.

There is probably a problem with a part in the schematic, but I replaced every semiconductors and the wiring after that and the problem stays the same.
Even without the DMM connected I have this bad clicking noise.

Meanwhile, I ordered a descent Fluke 179  , but I hope it will hold this situation...

Have I done something wrong ? I don't understand what the problem is. Any advice would be appreciated !

Thanks !

[ZeTeX]

Hi,

I'm trying to make a pretty basic mosfet regulator but so far I failed miserably.
Have a look at the schematic attached.

My multimeter is between Va and ground, Vdc 600V selected, and when I turn my transformer on, I hear a pretty bad clicking noise (typical noise of a faulty semiconductor or short, you know what I mean) at 100Hz and the DMM is gone.
Same result with a second multimeter.
My multimeters were pretty cheap but have done the work for many years. Its rated CAT II 600V, but I saw the thread on the cheap-ass multimeters shootout... 

But c'mon, its at best 450Vdc so I don't understand why it fails like that.

There is probably a problem with a part in the schematic, but I replaced every semiconductors and the wiring after that and the problem stays the same.
Even without the DMM connected I have this bad clicking noise.

Meanwhile, I ordered a descent Fluke 179  , but I hope it will hold this situation...

Have I done something wrong ? I don't understand what the problem is. Any advice would be appreciated !

Thanks !

How about measuring the output with a resistor divider (:100) ?

[ovnr]

Your circuit is probably switching hard and generating inductive kickback from something up the line. And of course you use a cheapo DMM that can't deal with the transients.

[Rutherberg]

Ok thanks for the fast replies.

I should receive the Fluke tomorrow, I will try the measurement with a divider.

Without the 10k connected to Va, I don't have the bad clicking noise. So I assume there is transients related to current. Is this what you mean by hard switching ?

Would a 10 ohms resistor (just after the mosfet) help to tame the transients ?
Or a zener between the drain and R2 to suppress the ripple of the mosfet gate ?

I'm puzzled, it should work as it is. Or do I miss something ?

Thanks ! 

[Rutherberg]

D9's polarity is wrong .

Obviously, thanks to point it out. Nevertheless I added it to the schematic for clarity, there is an internal Vgs protection diode in the 2SK2842.

BTW, F179 is not an electronics use DMM. It is an electrician's DMM. It does not have the accuracy and resolution for electronics use.
Get an 87V at least.

179 precision and features are way good enough for my usage and budget.

[Someone]

BTW, F179 is not an electronics use DMM. It is an electrician's DMM. It does not have the accuracy and resolution for electronics use.
Get an 87V at least.

179 precision and features are way good enough for my usage and budget.

Ignore the haters, the Fluke 179 is a very respectable meter and will do you well. I've seen them as standard issue in several engineering workplaces (including for electronics).

[Galaxyrise]

If C2 went to AC instead of GND, I think you'd have a voltage doubler.

[Circlotron]

Circuit could be oscillating. Put a 1k ohm in series with the mosfet gate, right next to the mosfet.

[Zero999]

D9's polarity is wrong

Yes and it's not a proper regulator, just a voltage divider with a common drain amplifier.

To make a regulator a voltage reference such as the TL431 is required.

What's the frequency of the input voltage?

[chris_leyson]

Circuit could be oscillating. Put a 1k ohm in series with the mosfet gate, right next to the mosfet

That was my first thought even before I had read the post, and D9 is the wrong way round.

Even without the DMM connected I have this bad clicking noise.

That's also a sign of high frequency oscillation, the amplitude builds up causing the DC bias to shift until the oscillation stops and then the whole thing repeats.

[ebclr]

Buy a pair of Varistor and put in parallel with the meter probe ( the 2 in series )

https://world.tmall.com/item/522030797114.htm?spm=a312a.7700714.0.0.U6sJes

[ebclr]

1N4007 can't be used if you switch frequency is on the Khz magnitude, Must use a Fast diode

[MagicSmoker]

All kinds of things wrong with the OP's circuit. The usual way of doing this is described nicely at http://electronicdesign.com/power/efficiently-reduce-high-supply-voltages-accurate-low-voltage. Here is the schematic from that page:



Note that while the example is for a >24V input, 3.3V output regulator, the input and output voltage are only limited by the Vds rating (and dissipation, of course). It is not mentioned in the article, but modern MOSFETs optimized for switching tend to perform poorly in linear operation (in fact, they tend to oscillate!); look for MOSFETs expressly designed for linear mode (e.g. IXYS L2 series). Also note that this circuit is not protected from overcurrent/shorts, but neither was yours. It's easy enough to add over current protection with the usual NPN shunting the MOSFET gate to its source when Vbe exceeds 0.6V.

I also agree that the Fluke 179 is a perfectly fine meter for electronics use; it's my general purpose meter of choice, in fact.

[Zero999]

1N4007 can't be used if you switch frequency is on the Khz magnitude, Must use a Fast diode

We don't know what frequency the 300VAC is. I assumed it's 50Hz mains from a transformer because the OP is in Europe but it could be from an inverter.

If it's a much higher frequency, then not only should fast diodes should be used but the 10nF bypass capacitors should be removed.

All kinds of things wrong with the OP's circuit. The usual way of doing this is described nicely at http://electronicdesign.com/power/efficiently-reduce-high-supply-voltages-accurate-low-voltage. Here is the schematic from that page:



Note that while the example is for a >24V input, 3.3V output regulator, the input and output voltage are only limited by the Vds rating (and dissipation, of course). It is not mentioned in the article, but modern MOSFETs optimized for switching tend to perform poorly in linear operation (in fact, they tend to oscillate!); look for MOSFETs expressly designed for linear mode (e.g. IXYS L2 series). Also note that this circuit is not protected from overcurrent/shorts, but neither was yours. It's easy enough to add over current protection with the usual NPN shunting the MOSFET gate to its source when Vbe exceeds 0.6V.

I also agree that the Fluke 179 is a perfectly fine meter for electronics use; it's my general purpose meter of choice, in fact.

That will work but only for output voltages under 15V - the MOSFET's threshold voltage. If a higher output voltage is required, then additional transistors are needed to reduce the voltage across U1 to 15V or below.

[retiredcaps]

My multimeter is between Va and ground, Vdc 600V selected, and when I turn my transformer on, I hear a pretty bad clicking noise (typical noise of a faulty semiconductor or short, you know what I mean) at 100Hz and the DMM is gone.
Same result with a second multimeter.

I can't read a schematic or help with your design, but if you want to try and fix your two dead multimeters, start another thread in the repair section and I can try to help.

I'm travelling this weekend so my responses will not be quick and limited to a few minutes here and there.

[Rutherberg]

Thanks for the replies, the schematic has been modified, D9 is correct now.

Circuit could be oscillating. Put a 1k ohm in series with the mosfet gate, right next to the mosfet.

I'll try that, thanks.

D9's polarity is wrong

Yes and it's not a proper regulator, just a voltage divider with a common drain amplifier.

To make a regulator a voltage reference such as the TL431 is required.

What's the frequency of the input voltage?

The purpose of this circuit is to vary the HV of a cathode biased push-pull guitar amplifier to vary the power output, I don't really need a regulation.
The main frequency is 50Hz.

Note that while the example is for a >24V input, 3.3V output regulator, the input and output voltage are only limited by the Vds rating (and dissipation, of course). It is not mentioned in the article, but modern MOSFETs optimized for switching tend to perform poorly in linear operation (in fact, they tend to oscillate!); look for MOSFETs expressly designed for linear mode (e.g. IXYS L2 series). Also note that this circuit is not protected from overcurrent/shorts, but neither was yours. It's easy enough to add over current protection with the usual NPN shunting the MOSFET gate to its source when Vbe exceeds 0.6V.

I also agree that the Fluke 179 is a perfectly fine meter for electronics use; it's my general purpose meter of choice, in fact.

Ok thanks, I will check for a more appropriate mosfet.
Is it OK to use the zener (D9) as a current protection device instead of the classic NPN ? I mean connect the zener at the gate and after the current sensing resistor ?

Attached, you will find the schematic updated

My multimeter is between Va and ground, Vdc 600V selected, and when I turn my transformer on, I hear a pretty bad clicking noise (typical noise of a faulty semiconductor or short, you know what I mean) at 100Hz and the DMM is gone.
Same result with a second multimeter.

I can't read a schematic or help with your design, but if you want to try and fix your two dead multimeters, start another thread in the repair section and I can try to help.

I'm travelling this weekend so my responses will not be quick and limited to a few minutes here and there.

I appreciate it but those were really cheap and I don't think it worth the repair time 

Thanks !

[MagicSmoker]

...
That will work but only for output voltages under 15V - the MOSFET's threshold voltage. If a higher output voltage is required, then additional transistors are needed to reduce the voltage across U1 to 15V or below.

Good catch. The LM4041 can be directly replaced with an TL431 to extend the input/output differential to ~35V, or it can be replaced with a zener for a fixed output voltage and no real limit on the i/o differential (aside from aformentioned Vds and Pd limits).

Also not included in the EDN circuit is the obligatory 12V-15V zener from gate to source to protect against overvoltage there, but the OP has recognized the need for such so I didn't bother pointing it out before.

[Rutherberg]

So I tried the attached schematic and it works like a charm ! My Fluke too BTW 

I didn't implemented the current limitation circuit for the test, I will add it later.

I changed the mosfet for a beefier one. This is still a fast switching type, I compose with what I have in stock.
Is there any way to further reduce the risk of oscillation in this case ? Except a proper layout ?

About current limitation, is it OK to use the zener (D9) as a current protection device instead of the classic NPN ? I mean connect the zener at the gate and after the current sensing resistor ?

Thanks for your help !

[bktemp]

IRFPExx should be fine up to a few 10W of power dissipation. They are rather old, so have a much larger die compared to more modern MOSFETs.
A source resistor + gate zener diode helps limiting the short circuit current, but unless the current is set to a rather low value (<50mA) the power dissipation will be too high for the MOSFET to survive a short circuit longer than a couple of seconds.

[Zero999]

...
That will work but only for output voltages under 15V - the MOSFET's threshold voltage. If a higher output voltage is required, then additional transistors are needed to reduce the voltage across U1 to 15V or below.

Good catch. The LM4041 can be directly replaced with an TL431 to extend the input/output differential to ~35V, or it can be replaced with a zener for a fixed output voltage and no real limit on the i/o differential (aside from aformentioned Vds and Pd limits).

Also not included in the EDN circuit is the obligatory 12V-15V zener from gate to source to protect against overvoltage there, but the OP has recognized the need for such so I didn't bother pointing it out before.

The TL431 does have a higher voltage rating but it also has a higher minimum drive current.

For higher voltages, another transistor can be added to power the regulator IC, off a potential divider connected to the output.

Attached is an example using a Darlington transistor and the LM431. It will work with a MOSFET and LM4041, with a few modifications.

[Rutherberg]

IRFPExx should be fine up to a few 10W of power dissipation. They are rather old, so have a much larger die compared to more modern MOSFETs.
A source resistor + gate zener diode helps limiting the short circuit current, but unless the current is set to a rather low value (<50mA) the power dissipation will be too high for the MOSFET to survive a short circuit longer than a couple of seconds.

I don't understand, the datasheet announce Pd max = 190W
http://www.vishay.com/docs/91248/91248.pdf

In the worst case, Vgs is about 330V with a quiescent current of 30mA and 120mA peak.
So, Pd quiescent is about 9.9W and Pd peak 39.6W, well below Pd max.
Or am I missing something ?

[bktemp]

IRFPExx should be fine up to a few 10W of power dissipation. They are rather old, so have a much larger die compared to more modern MOSFETs.
A source resistor + gate zener diode helps limiting the short circuit current, but unless the current is set to a rather low value (<50mA) the power dissipation will be too high for the MOSFET to survive a short circuit longer than a couple of seconds.

I don't understand, the datasheet announce Pd max = 190W
http://www.vishay.com/docs/91248/91248.pdf

The specified power dissipation has almost no practical meaning, especially for linear operation at higher voltages:
http://www.digikey.com/en/pdf/m/microsemi-power-products-group/microsemi-make-linear-mode-work

In the worst case, Vgs is about 330V with a quiescent current of 30mA and 120mA peak.
So, Pd quiescent is about 9.9W and Pd peak 39.6W, well below Pd max.
Or am I missing something ?

39.6W should be ok if the junction temperature stays <100°C.
Because there is no DC SOA curve in the datasheet, you can only guess the operating area unless you do some destructive measurements.

[MagicSmoker]

...
I don't understand, the datasheet announce Pd max = 190W.
...
In the worst case, Vgs is about 330V with a quiescent current of 30mA and 120mA peak.
So, Pd quiescent is about 9.9W and Pd peak 39.6W, well below Pd max.
Or am I missing something ?

Ut-oh, you're a young player or a new one, aren't you? To paraphrase an old saying, "there are lies, damn lies and then there are transistor datasheets".

The tell is in the thermal resistance specs: the junction to case resistance is 0.65C/W which results in a 123.5C rise above ambient at 190W all by itself, but then you have to add the rise from the case to heatsink (min. of 0.24C/W) which gets you another 45.6C rise, and then there is the thermal resistance of heatsink to ambient which for a typical convection cooled finned aluminum job is going to be somewhere between 0.5C/W and 2C/W. Meanwhile, the maximum allowed junction temperature is usually 125C or 150C - occasionally 175C - but don't expect very long operational life above 100C.

All in all, that means a more realistic dissipation limit for a TO-247 package is 40-60W, and even that is assuming a very good heatsink (think of forced convection or liquid chill plate).

[Rutherberg]

The specified power dissipation has almost no practical meaning, especially for linear operation at higher voltages:
http://www.digikey.com/en/pdf/m/microsemi-power-products-group/microsemi-make-linear-mode-work

Very interesting, I'll check that.

Ut-oh, you're a young player or a new one, aren't you? To paraphrase an old saying, "there are lies, damn lies and then there are transistor datasheets".

The tell is in the thermal resistance specs: the junction to case resistance is 0.65C/W which results in a 123.5C rise above ambient at 190W all by itself, but then you have to add the rise from the case to heatsink (min. of 0.24C/W) which gets you another 45.6C rise, and then there is the thermal resistance of heatsink to ambient which for a typical convection cooled finned aluminum job is going to be somewhere between 0.5C/W and 2C/W. Meanwhile, the maximum allowed junction temperature is usually 125C or 150C - occasionally 175C - but don't expect very long operational life above 100C.

All in all, that means a more realistic dissipation limit for a TO-247 package is 40-60W, and even that is assuming a very good heatsink (think of forced convection or liquid chill plate).

Exactly, well, kind of...
I have a bachelor degree in electronics but until now I was field service engineer for 7 years. Now I try to evolve in electronics R&D so I certainly have knowledge gaps.
Thanks for the informations !
Any alternative in this case ?

[bktemp]

Another useful application note showing why a mosfet can dissipate much more power at low voltages and why older mosfets are suited better for linear operation:
http://www.infineon.com/dgdl/Infineon+-+Application+Note+-+PowerMOSFETs+-+OptiMOS%E2%84%A2+-+Linear+Mode+Operation+and+SOA+Power+MOSFETs.pdf?fileId=db3a30433e30e4bf013e3646e9381200