maximum sink/source current for mosfet driver

[MTong]

Hi all,

I have a doubt: the datasheets for mosfet drivers all mention maximum sink and source current; Do you know if it is internally limited and how?

Thank you

[robotai]

No it's not internally limited, just their guarantee to work below this condition. Once you use it over this limitation, device may break and the result is unexpected.

[solutions]

The result is expected, not unexpected. The limits are to the rest of the specification...it's not just a number.

The currents are typically limited by channel resistance. The limit as a result of the channel resistance is typically on device voltage drop or I*I*R heating, or in really high power drivers, the bond wires can actually act as fuses. These are over the entire temperature range of the part spec.

I think what you are asking is can you run it higher than spec. Yes you can. But there are consequences and if you understand those you should be OK if you take care of those consequences. No manufacturer will guarantee performance, or warranty the part, if you use it outside of spec, though. One lot (batch number) change, and your design may not work - another risk.

It would really help if you told us exactly what you were trying to do and with what specific part, but like everyone else asking for help here, I'm sure that this is a secret project like developing a dilithium crystal oscillator for a warp drive nacelle, and you can't tell us more without invoking the wrath of the Romulan High Council.

[robotai]

Agree with that. The datasheet is what manufacturer design and guarantee. Use device over the specification will still work but definitely increase the risk to break the device.

[solutions]

It does not necessarily "definitely increase the risk to break the device". For instance, a logic level driver will go out of Voh/Vol spec, but it's not necessarily going to release the smoke the factory put into the device if you exceed the current spec.

[MTong]

Thank you guys.
Actually no project here, just a thing me and a colleague of mine were not sure about (we are power elt. students).
How can you exceed the spec if it is limited by channel resistance? One of the datasheets (IR2184?) says that the max current (1.9A in this case) is tested with Vo short circuited, for pulses of 10us.
With exceeding specs do you mean something like using it for longer periods?


Thanks again

[thunderer]

For instance, if the gate capacitance is very high (let's say 6-7nF) and you try to comute very fast, this will result in a lot of power dissipation in the driver. Thus, you will break it because of overheating that our colleagues above stated it is not limited.

Have you seen the application notes and guides? Some examples here:
http://www.ti.com/lit/ml/slup169/slup169.pdf http://ww1.microchip.com/downloads/en/appnotes/00786a.pdf http://ww1.microchip.com/downloads/en/AppNotes/00898a.pdf.

Just type "mosfet driving" in google and read the first pages. This link is good for any beginner: http://robots.freehostia.com/SpeedControl/Mosfets.html.

[MTong]

Those are good read, the first one particularly, thanks a lot.

It's getting clear now.

[hate]

You may pulse the MOS with short gate bursts to achieve overloaded current with the condition that you stay in the allowed maximum power for the device. Another thing to consider is your output voltage when you're sinking or sourcing more current than the allowed limit. You can think the channel resistance like a serial resistance between you supply and your load. Say a MOS has a channel resistance of 1 Ohm (just a simple example, there is no power MOS with a 1 Ohm channel resistance). If you try to source 2A from this device under a 5V power supply, then your load will get 2A but under 3V (5V - 2A*1Ohm). That's another reason not to exceed maximum current limit.

[rinderpest]

I would also recommend this one:
http://www.irf.com/technical-info/designtp/dt04-4.pdf
It's an app note from IR, mainly aimed at driving IGBTs, but the maths and assumptions are very similar.
It goes into some detail about parasitic inductances, resistances etc in the surrounding circuitry.
They have a lot of other useful stuff, obviously it refers to their products but the sums and ideas work for others too.

You'll have fun, its like a country house detective mystery when your FETs start going off like popcorn for no apparent reason.
If you collect app notes and datasheets for FET drivers from all the manufacturers you can find, between them all you'll find the answers even if you are driving FETs discretely.
A fast 'scope helps a lot too (wish I had one), otherwise you have to visualise in your minds eye what you THINK is going on.
It can be done, but it takes concentration and only changing one attribute at once.

[f22kma]

Insufficient gate drive current with high power Mosfets (and low RDSon) can cause the Mosfet to dissipate excessive heat during conduction.

This becomes more of a problem as frequency increases.

A stout gate drive is also important to prevent capacitive coupling to the gate from turning the device on unintentionally.

[FTL]

When driving MOSFETs in a PWM mode, the key is to keep the time they are switching to a very small portion of the overall time.

The gate of a MOSFET is a significant capacitor. Once the gate voltage has been set, it takes almost no current to keep it there. Really just the very tiny bit of leakage current in the gate capacitance. The problem is that since the gate is a capacitor, it takes significant current to quickly charge or discharge the capacitor.

In a switchng mode, the goal is to get the MOSFET from on to off or off to on as quickly as possible. In off mode there is almost zero current flowing though a very high resistance, so there is no significant heating. In on mode, there is significant current flowing through a small resistance (in the ones to tens of milliohms). That causes a bit of heating, hence the maximum continious current rating of the device.

During the transition period, there is current flowing through a significant resistance, hence significant heat.

If your PWM frequency is low (like maybe 1hz), it does not matter if the device spends 1ms switching as it is only switching for 1/500 of the time (remember it switches on and off in one cycle). As long as the time spent during the switching is within the device peak time/current specs you should be OK. I've built devices like that with a slow PWM (driving a heater), and 10ma from a PIC pin was plenty to switch a large MOSFET quickly enough.

The problem is when the PWM frequency is higher. With 500hz frequency, a 1ms switching period would quiuckly burn the device up as it would be spending all of its time as a linear device making heat.

If you want to run your PWM at 1MHz, you need to be able to switch the MOSFET in much less than a microsecond. Probably more like 10-50ns. To charge and discharge the gate capacitance  in that time takes significant current. MOSGET driver IC's are designed for that task.

The goal is to keep the time spent switching a small portion of the time spent either on or off.

[Tekno1]

All the IC manufacturing processes moves towards finer geometries. As a result smaller Ron devices are possible in ever smaller areas (device sizes decreases). One of the major problem with these modern devices is heat dissipation. IC industry tries to mitigate heating problems with newer better package designs. Nevertheless for particularly CMOS devices heating up an IC increases Ron resistance and so decreases current carrying capability of drivers for a given output current, IR drop and efficiency specifications. So data sheets introduce conditions that all the specifications stated are valid. Considering those conditions with IC's intended use model makes all those conditions meaningful. Therefore it is very important to know the intended applications since a specification meaningful for a use model may not be useful for another use model.

Assuming that part is capable to survive for a while (!) over stress conditions, one other issue with over current, heat and voltage conditions is the electromigration issue with tiny (real tiny) metal traces on the IC. Electromigration process over relatively long time fuses the metal connection carrying high currents. This issue presents itself as shorter lifetime of the IC which is subjected to over stress conditions. Nowadays when selecting parts end products lifetime is considered carefully and because of market cost pressures system designers try to get cheapest but long enough lifetime parts. Expensive, longer living and more reliable IC's  are designed more in military and health instruments. Cheapest and short life IC's are designed in consumer products where product lifetimes does not expected to last more than a couple of years.

Of course many circuit design and performance issues becomes evident depends on what is overstressed.
In short it is one of the system designers primary job to pick IC's designed for their application.