Would anyone be willing to give me a couple minutes of help with a circuit?

[Wolfman29]

Best of luck, rhd! I want to see this circuit finished

 

[paul1598419]

rhd, if you are trying to dissipate the heat from the MOSFET it usually depends on the case style you are using. If it is an SMD the heat can be dissipated through the bottom of the case, especially if you leave a patch of copper for it to sit on when you print your circuit board. Even a discrete larger component can dissipate heat through the case into a patch of copper on your circuit board. The main thing to keep in mind is that the patch of copper is not connected in any way to anything it shouldn't be connected to. Most of these MOSFETs have a metal back or bottom of the case which is a good way to heat sink it without using a heat sink per se. I would eliminate the load as a possible problem, though. As long as you arte using the FET as a switch it will be in saturation and you don't need to worry about the threshold gate voltage. The load, however, can cause an initial power surge causing an excess of current to flow through the device at the moment you activate it.

 

[rhd]

I'm still not buying that heat was a factor. Shouldn't heat simply be a factor of Rds and current? So if a MOSFET is capable of handling a particular current, it must be capable of dealing with the heat produced at that current also (assuming proper physical mounting on a PCB).

Asked another way, if this MOSFET can't handle the heat produced by X current, at saturation, in my circuit, how could it EVER handle X current (and thus be rated to do so)?

This package is too small for the manufacturer to have assumed external heatsinking.

 

[paul1598419]

rhd, you may very well be correct that heat dissipation has nothing to do with your problem. I was following the conversation about how to dissipate heat in a MOSFET. The way you explained that the FET self-destructed the moment it was turned on leads me to think that it is a load problem more than anything else. The power or heat dissipated in a MOSFET is the drain current squared multiplied by the Resistance between the source and the drain in conduction, whether that is in the ohmic part of the conduction curve or the saturation part of the conduction curve. In saturation the resistance between the drain and source is very small because the amount of current flowing is at its maximum and can't increase with bias between the gate and source. So, either the FET is under rated for the initial current or the load is such that at the instance of conduction it is seen by the drain as too much load. Does that make sense to you? If not I'll try to state it differently so it does.

 

[rhd]

The power or heat dissipated in a MOSFET is the drain current squared multiplied by the Resistance between the source and the drain in conduction, whether that is in the ohmic part of the conduction curve or the saturation part of the conduction curve.

Exactly my point. Within a given MOSFET, assuming saturation, heat is a function of current, period. So, if a MOSFET is able to handle X current, it must also be able to handle the heat produced by X current, because the amount of heat produced by X current will always be the same. If it couldn't handle that amount of heat, it couldn't be rated to handle the current that produces it.

 

[paul1598419]

rhd, not exactly. There is a thermodynamic part of this equation I left off. But, for the sake of argument, as the power dissipated in the FET increases the RdsON increases as well. You can see from the equation that as RdsON increases and the current remains essentially the same the power dissipated increases and to make a long story short, the FET goes into thermal runaway. This can happen when the temperature reaches ~80 degrees C above the ambient temperature. So, it isn't as cut and dry as the power dissipated is only a function of the drain current. But, if the FET is destroyed instantly, we can probably ignore that part. If the FET lives long enough, you have to take it into consideration.

 

[djQUAN]

The mosfet package is a SOT323 which is VERY small. it won't be able to dissipate any amount of heat.

Did you see my post #29 above? Since the package is really small, it doesn't have much thermal inertia which, with a high inrush current lasting milliseconds or so at startup, can pretty much exceed the device junction temp in an instant without the case getting warm to the touch. I may be wrong but just something else to look at.

 

[rhd]

The mosfet package is a SOT323 which is VERY small. it won't be able to dissipate any amount of heat.

That's contrary to what the datasheet says. Provided the minimum pad sizes are observed (I exceeded them by far), it can dissipate 430mw through the copper.

 

[djQUAN]

Yes, but that is at 25C ambient and the die temp will most probably be at the absolute max junction temp. Try dissipating the full 430mW and put your finger on it and try to feel how hot it is. Whatever the datasheet says, I won't design it that way since it's too hot for my liking.

I'm still leaning on my theory of the driver inrush current blowing the mosfet unless more data is presented.

 

[paul1598419]

rhd, if you don't have access to an oscilloscope, you might be able to get a picture of the current at the time of actuation. I have 3 multimeters left, a Beckman Tech 310, a Triplett 9015, and a Mastech MS8268. The last of these meters has a hold function which you can use on any of the voltage or current settings. If you set it up to measure the current and push the hold function, it will measure and hold the current level at the moment you turn the unit on. That is the only other way I can think of to measure the current at the instant it begins to flow through your MOSFET. If you have a meter like this one, consider doing this.

 

[djQUAN]

Another thing is that there could be enough resistance in the wiring to the FET to cause a voltage drop and
bring it out of saturation. This could be fixed by adding a capacitor near the FET. Good luck, I hope you
get it working.

that will work in the first few microseconds until the cap discharges then the voltage drops again due to wire resistance.

rhd, if you don't have access to an oscilloscope, you might be able to get a picture of the current at the time of actuation. I have 3 multimeters left, a Beckman Tech 310, a Triplett 9015, and a Mastech MS8268. The last of these meters has a hold function which you can use on any of the voltage or current settings. If you set it up to measure the current and push the hold function, it will measure and hold the current level at the moment you turn the unit on. That is the only other way I can think of to measure the current at the instant it begins to flow through your MOSFET. If you have a meter like this one, consider doing this.

A multimeter has an RC filter in the input so it will be much slower than the most basic oscilloscope. If the drop that you're looking for lasts several hundred msecs, then you might be able to see it with the multimeter. Any faster then the meter won't show the peak value and you won't be able to push the hold button fast enough at the precise time anyways

 

[paul1598419]

Yeah. The Mastech has a counter on it that can measure down to 5 microseconds, not as fast as my 100 MHz scope, but fast enough. But, as has been stated, the hold function is limited by ones ability to press hold at the correct instant. It was just a thought I was toying with as a substitute for the scope. Probably not one of my better ideas, though.

 

[Bionic-Badger]

rhd: I'm not really buying the thermal runaway problem either. The current path from drain to source is relatively robust. The gate is where damage usually occurs for assuming no short.

A more likely cause is some kind of effect on the gate: maybe some oscillations or discharge into the gate. Maybe it doesn't like your DC switch, or maybe some noise is causing oscillations. Can you take one of your modules and hack in a resistor between the gate and the switch? Maybe even a cap to smooth things out, and/or a protection diode? Protect that gate!

 

[rhd]

rhd: I'm not really buying the thermal runaway problem either. The current path from drain to source is relatively robust. The gate is where damage usually occurs for assuming no short.

A more likely cause is some kind of effect on the gate: maybe some oscillations or discharge into the gate. Maybe it doesn't like your DC switch, or maybe some noise is causing oscillations. Can you take one of your modules and hack in a resistor between the gate and the switch? Maybe even a cap to smooth things out, and/or a protection diode? Protect that gate!

I will get a bit more creative if the next iteration fails - but I'd like to hold off and test the revised PCB (with resistor between V+ and gate) first to see how it functions, before I start test alternate theories.

I'm not in a rush on this one. The build this is meant for is a long way off, with a lot of other pieces that still need to come together as well. So I have time to iterate and get it perfect.

 

[luxophile]

The problem in your circuit is very hard to diagnose without having your circuit board in front of each of us. Your fragments of schematic all look good as far as I can see. So the best I can offer is a few debugging tips and maybe some anecdotes about pitfalls I have encountered in the past.

First off, don't take anything for granted. Test everything and take notes. For example with the circuit unpowered take an ohm-meter and verify that the transistor's source pin has very low resistance to V-. Similarly verify that the gate pin reads about 1M-ohm to V-, and the drain should have very high resistance to V-. Use the datasheet to tell you where the pins are. This not only verifies some of the circuit board traces; it also verifies that your CAD software has source, gate, and drain connected to the correct pins on the package. (Yes, I've see mis-connected pins and shorted traces before. I don't think that's your problem but it's easy to check these things and rule them out.) Also verify you really are connecting the more positive voltage to V+ and the more negative voltage to V-. (My boss once burned out three power supplies one after the next until he realized he had the leads connected backward. It happens.)

It wouldn't hurt to verify that you are using the transistor you think you are. (It's rare, but I have seen cases where the distributer put the wrong part in a bag and labeled it as the part we ordered.) So check the markings on the transistor if they are still readable. According to the datasheet you should see "DMH" plus a month & year code. If you're as old as I am you'll need a magnifying glass.

Your circuit as originally drawn is just a bit scary. It shows a switch connected directly to the gate pin. (If you don't really have a mechanical switch then you can disregard this paragraph.) Beware that if you really have a mechanical switch it can cause problems. First because it can collect static electricity from your finger and direct it to exactly the worst place it could possibly go. ESD (electrostatic discharge) could cause erratic behavior or complete failure of the transistor. A mechanical switch should never connect directly to a logic IC's input nor to a MOSFET's gate; always provide some form of ESD protection. Secondly a mechanical switch generally has a lot of "bounce". The contacts tend to chatter as the connection is made or broken. Normally that wouldn't be a problem but in conjunction with the 1M resistor it could theoretically cause your MOSFET to remain in the linear region longer and cause heat damage. I suggest using a lower-value resistor in this situation.

Next, never ever rely too heavily on a transistor's current rating. Rated current is derived from the maximum allowed junction temperature and a number of (sometimes highly optimistic and unstated) assumptions about how well the heat travels away to the outside world. Diodes Inc says your NFET junction temperature should not exceed 150 deg C if power is kept below 0.43 Watts and the transistor is soldered down on an FR-4 PCB. They don't state how many square inches the PCB must be and they don't state how much air movement they expect. It might be safe to assume that a small PCB with no other sources of heat is OK. It might not. It might be safe to assume that convection alone in open air will be enough to cool the board. It might not. It certainly would not be safe to assume you could put the board inside an enclosure that restricts air flow.

Also remember that rated current always applies only to purely resistive loads, and for most of us it's rare to be switching a load that has no inductance or capacitance. You should always do at least a cursory check on the characteristics of your load. For example consider the schematic you posted on 3/23 showing a PMOS transistor switching a hypothetical load made up of a couple current regulators and a few capacitors. Assuming you haven't switched it on recently, all the capacitors would be discharged. When you switch the transistor ON, it will briefly see a load that looks like a dead short to ground. The short circuit will disappear quickly as the capacitors charge. But at the instant you switch the transistor ON the only thing limiting current between +8V and GND is the Rds,on resistance of the PMOS. Given its Rds,on rating of 0.050 ohms, you can theoretically expect a brief initial current spike of (8V/0.050 ohm) = 160 Amps. In reality resistances and inductances in traces and in the capacitors will reduce that peak current somewhat. And the RC time constant is short so we're only talking microseconds. Maybe your transistor can withstand such a brief spike. Or maybe you need to select a transistor with bigger thermal mass or beefier bonding wires or lower Rds,on rating.

(You can have similar problem switching an inductive load. But the problems are more likely to appear when you shut the switch OFF rather than when turning it ON.)

Sorry for the long-winded post, but there really are a lot of possible causes of the problem you reported. Hopefully this gives you a little more background to puzzle out what happened and why.

 

[rhd]

luxophile:
Thank you for the advice, and welcome to LPF!

I'm going to take your advice and verify the circuit manually before assuming everything is correct parts-wise. I actually ran into mis-labelled resistors from digikey last order. I caught it, but kicked me back into skeptic mode. I ordered a better DMM that can measure capacitance, and I'm in the market for an affordable meter that can measure sub-ohm resistances (though that's mostly just for verifying current-sense resistors, and isn't relevant here).

I think the gate resistor, as well as the switch to a P-Channel mosfet in my second design, will prove effective. But I'll certainly update the thread once the PCBs arrive and I find out!