MOSFET use in a uC application

[mtechmatt]

Ok, well I am dropping 0.05volts accross the FET wide open 100%, and of course, no heat emitted So its of course the transient.

I have gone ahead and laid the shecmatic using the FET you recommended, if you can give me a drive resistor I am happy to use that, or is there a better way?

Even with a direct short accross the gate resistor, I still get a crappy rise and fall time (and the heat that goes with it) Too big capacitance on the MOSFET for my application maybe?

Matt

[dfnr2]

If your layout is good, you can try leaving out the gate resistor; you can add in a small value if you are having oscillations due to the parasitics.  Keeping the trace short will help reduce inductance.  You need to look at your turn off time as well.  If you include a gate resistor, you can use a PNP to turn off the FET by shunting the gate to ground.  If you are in doubt, you would do well to incorporate a dedicated MOSFET driver close to the MOSFET.  As mentioned above, make sure you choose your FET to be appropriate to the drive levels.  For this little power and this frequency, you should be able to run cool without fancy techniques that get employed in the MHZ+ range.

BTW, this is one of those applications where you want to have the highest BW scope you are able to access, and active probes.

Dave

[georges80]

Ok, well I am dropping 0.05volts accross the FET wide open 100%, and of course, no heat emitted So its of course the transient.

I have gone ahead and laid the shecmatic using the FET you recommended, if you can give me a drive resistor I am happy to use that, or is there a better way?

Even with a direct short accross the gate resistor, I still get a crappy rise and fall time (and the heat that goes with it) Too big capacitance on the MOSFET for my application maybe?

Matt

Ok, you've now got good evidence it's a gate drive issue (slow rise/fall time but good when 100% on). Search for a FET with the lowest gate charge or gate capacitance and it will help speed things up. Your uC pin drive current is what is causing the slow rise/fall time as it tries to charge/discharge the gate of your current FET.

As above, with a clean layout you shouldn't need a gate series resistor or at most a few ohms if you have ringing issues. I'd leave the resistor footprint in the layout and you can then go to 0 ohms if gate signal integrity looks good on a 'decent' scope with appropriate probes and probing.

cheers,
george.

[jmaja]

Can the uC slew rate be controlled? Do you have it at maximum? Have you tried lower frequencies or do you need to have 300 Hz? I had a similar problem with a solar panel PWM controller (50 W, 12 V) I made and it was solved by lowering the frequency. I didn't have the option to change the MOSFET. I started with a few hundred Hz and ended up using 78 Hz.

[dfnr2]

Can the uC slew rate be controlled? Do you have it at maximum? Have you tried lower frequencies or do you need to have 300 Hz? I had a similar problem with a solar panel PWM controller (50 W, 12 V) I made and it was solved by lowering the frequency. I didn't have the option to change the MOSFET. I started with a few hundred Hz and ended up using 78 Hz.

Hmm.  I misread the OP as 300 kHz, but by golly, you're right--it's just 300 Hz.  At such a low frequency, the on/off times for any normal circuit should not add up enough to cause much heating.  Something is not working as you expect.  The scope is your friend.    If you have access to a good scope and active probes, use them.  Also, put a small resistor in series with the output and use it to estimate get output current waveforms.  Try to post some waveforms here.

Dave

[georges80]

If you run through some numbers, I think you'll find that 17us rise and fall time at 300Hz is about 1% of the time being in the linear region of the FET.

With 1A flowing through the solenoid (so about 12 ohms) that means the FET can be 1 - 2 - 3 ohms and still have ~1A flowing. I2R losses will add up.

cheers,
george.

[HackedFridgeMagnet]

Just trying to work this out. Correct me if I misunderstand.

Are you driving a solenoid at 300 hz?
If you are why? And if you are then you want a flyback diode across the coil.

[uwezi]

If you run through some numbers, I think you'll find that 17us rise and fall time at 300Hz is about 1% of the time being in the linear region of the FET.

With 1A flowing through the solenoid (so about 12 ohms) that means the FET can be 1 - 2 - 3 ohms and still have ~1A flowing. I2R losses will add up.

cheers,
george.

At best - depending on the actual value of the threshold voltage for the particular MOSFET in the circuit - according to the datasheet the threshold voltage can lie between 1 V and 3 V!

[HackedFridgeMagnet]

At 2.5volts the FET is well and truly on. At this point is has no more than 15mOhms resistance.

[mtechmatt]

Ok chaps

I will post up some scope images, I have a Rigol DS1052E

In terms of an end game solution I am going with georges MOSFET (IRLML2502) recommendation, which will be fully 'saturated' by 2.5V, with no gate resistor, but a pad for it just in case.

Will order up a sample and dead bug it to see if it works

Matt

PS its driven at 300Hz as that is the stated operating frequency of the solenoid, it is for controlling fluid flow in a valve

[mtechmatt]

At 2.5volts the FET is well and truly on. At this point is has no more than 15mOhms resistance.

Which FET are you reffering to? The current one or the proposed IRLML250 device?

Cheers
Matt

[HackedFridgeMagnet]

I was referring to the initial one,
http://uk.farnell.com/international-rectifier/auirf7805q/mosfet-n-ch-30v-13a-so8/dp/1909590 It seemed ok to me.

I just looked at the Vgs vs Rds on  graph and it showed that the FET was well and truly switch on at 2.5volts and your test confirmed it
re:

Ok, well I am dropping 0.05volts accross the FET wide open 100%, and of course, no heat emitted So its of course the transient.

Assuming you were driving it a 3.3v.

but what about the solenoid, do you have a data sheet on that?

[jmaja]

Why don't you measure the voltage over the MOSFET with your scope? Then you could see how well it works. Maybe you could put a 1 ohm power resistor in place of the solenoid. Then you could easily calculate the current and power as well.

[mtechmatt]

I was referring to the initial one,
http://uk.farnell.com/international-rectifier/auirf7805q/mosfet-n-ch-30v-13a-so8/dp/1909590 It seemed ok to me.

I just looked at the Vgs vs Rds on  graph and it showed that the FET was well and truly switch on at 2.5volts and your test confirmed it
re:

Ok, well I am dropping 0.05volts accross the FET wide open 100%, and of course, no heat emitted So its of course the transient.

Assuming you were driving it a 3.3v.

but what about the solenoid, do you have a data sheet on that?

I see, so the issue must be in the transient states. Maybe better to just use a simple driver instead of changing the fet? Any reccomendations (prefer multichannel as I have 8 of these to drive in my final project), and am I safe to supply the driver (and then ultimately the mosfet gate) the 12-14v supply?

No datasheet for it, just resistance meadures at around 12-15ohms, and that I am running it at 12-14v

Matt

[HackedFridgeMagnet]

I still dont understand why you are driving it at 300hz.

PS its driven at 300Hz as that is the stated operating frequency of the solenoid, it is for controlling fluid flow in a valve

this statement worries me.

That probably means you can switch it on and off at up to 300 hz if you want to. It that what you want? It doesn't mean you drive it with 300hz square wave.

Normally relays and I guess solenoids have either an AC coil or a DC coil. The AC coil is slightly different in that it has a special phasing coil, they are designed to operate at mains frequencies, not 300hz. DC coils are just that, DC.

Its probably DC so just turn on the GPIO for on and no square wave required.

Also if it is a DC coil put a flyback diode across it.

[mtechmatt]

Hi,

No we scoped it in its OEM application and it runs at 300Hz, the flow varies constantly. That is the how the OEM controller runs it.

Cheers
Matt

[HackedFridgeMagnet]

Ok I think I have it now, the duty cycle controls the valve position.

[mtechmatt]

You got it

Really these little IRMLM SOT23's can handle 3Amps??? And be driven by a 3.3 uC?
http://uk.farnell.com/jsp/search/productdetail.jsp?sku=1857300

I just cant see it... maybe thats me being used to clucnky old TO220s!

Matt

[uwezi]

You got it

Really these little IRMLM SOT23's can handle 3Amps??? And be driven by a 3.3 uC?
http://uk.farnell.com/jsp/search/productdetail.jsp?sku=1857300

I just cant see it... maybe thats me being used to clucnky old TO220s!

Matt

It's mostly a matter of R_DS,on for a given gate voltage. By being able to reduce the threshold voltage and obtaining an R_DS,on of 80 mohm the transistor will only have a static dissipation of 240 mW while transporting 3 A. If you then compare the package to a standard 1/4 W resistor it is not so surprising anymore.

But you have to be aware of the fact that the dissipated power can be much higher during the transition between ON and OFF and possibly exceed the safe operating area of the transistor.

[megajocke]

It has been mentioned multiple times already, but there has been no indication of the current state of this matter:

Is there an appropriate freewheeling diode installed across the load?

Without it the transistor will need to dissipate every cycle all the energy stored in the load inductance and then some.

Appropriate decoupling of the power supply is also needed or else the supply wiring inductance will cause high voltage spikes unless switching transistions are made slow on purpose - which is probably a reasonable way to design the circuit at 300 Hz switching frequency.

I just looked at the Vgs vs Rds on  graph and it showed that the FET was well and truly switch on at 2.5volts and your test confirmed it

Yes, but it's not guaranteed. It's the typcial curve. You are not guaranteed more current flow than 250 µA even at 3 V of Vgs. If there is a need to produce more units it has to be redesigned or else transistors need to be screened for Vgs before using.

[uwezi]

I just looked at the Vgs vs Rds on  graph and it showed that the FET was well and truly switch on at 2.5volts and your test confirmed it

Yes, but it's not guaranteed. It's the typcial curve. You are not guaranteed more current flow than 250 µA even at 3 V of Vgs. If there is a need to produce more units it has to be redesigned or else transistors need to be screened for Vgs before using.

Exactly my point from the start. The datasheet only guarantees behavior for V_GS=4.5 V i.e. the transistor is designed for +5 V logic signals. The graph in the datasheet is taken for a typical device, corresponding to a threshold V_GS of about 2-2.5 V. But also according to the datasheet the actual threshold V_GS of the transistors which are produced can vary between 1 V and 3 V.

And at the threshold V_GS the device is only guaranteed to conduct 250 µA.

Remember that the graphs in datasheets almost always only show typical devices. In order to estimate the spread of the manufactured devices you always need to look into the tables in the datasheet as well.

[HackedFridgeMagnet]

Ok I take you point that the FET is designed for 4.5v signals.
But the OPs question was about this particular device.

Ok, well I am dropping 0.05volts accross the FET wide open 100%, and of course, no heat emitted

It seems to me that the FET is well and truly on at 3.3v.

[uwezi]

Ok I take you point that the FET is designed for 4.5v signals.
But the OPs question was about this particular device.

Ok, well I am dropping 0.05volts accross the FET wide open 100%, and of course, no heat emitted

It seems to me that the FET is well and truly on at 3.3v.

Yes, it obviously is - under static conditions. But it is not guaranteed that the same will be true for the next MOSFET of the same type used under the  same conditions in the same circuit.

Also in a PWM design the MOSFET will spend a significant amount of time somewhere between on-state and off-state - the closer the threshold voltage is to the steady-state gate voltage, the worse this will be.

Just on the same topic: you can use a standard MOSFET, specified for a V_GS of 12 V with standard 5 V TTL- or CMOS outputs, since the gate threshold voltage of these devices is normally around 3-4 V. I very often do so for my 1-off designs. But be prepared to find MOSFETs which are still within their specs, but with a too high threshold voltage to be fully on.

[mtechmatt]

Interesting stuff. My design doesent currently have a freewheeling diode but as of tomorrow I will have these new smaller as recommended mosfets in my hands and will also run a FW diode, and report my findings in both cases.

Certainly getting somewhere now

Matt

[mtechmatt]

A little ypdate.. it was flyback...

Added a diode and on the original mosfet now running stone cold!

Thanks for the help everyone, on what was probably a school boy error!

Matt