The importance (or not) of gate current measurement

[PartialDischarge]

Hi,

Can you give your opinion on why besides the measurement of gate voltage, the measurement of turn-on gate current of Mosfets, GaN, Igbts.. would be important or desirable?
I'm referring to power electronic switching applications of course.
If you think it is not also say it...

Thanks

[Siwastaja]

Is this just "this popped to my mind while sipping morning coffee" thing or are you referring to some design where this is used, or seen discussion on it?

I'm asking because I have never heard about such an idea! Given the complexity, what would be the gain?

You can of course measure gate current on the lab table easily (no point designing it in a product) but given you know gate voltage and have a good approximation of the driver impedance (gate driver transistor R + external Rg + MOSFET parasitic Rg), you also know the gate charging current waveform within maybe some +/- 20% without actually measuring it.

[PartialDischarge]

You can of course measure gate current on the lab table easily

Gate current measurement during development is not that easy measurement, basically there are 3 ways:

1) measure the voltage across Rg, this a floating high speed and low level mesurement, for this you would need a low noise, low capacitance probe that can be floated. Differential probes that have a low Max common mode would be better suited here and they are not cheap. Normal power electronics diff probes are not suited due to the high noise if used to measure 1..2V, and a floating scope is going to add too much common mode capacitance to the gate node

2) Specialized Rogowski coil, PEM offers miniature coils that can be wind up across the pin of a TO-220 or similar packages to measure the gate current. These probes are not cheap and not widely available

3) High frequency current transformer, similar to the Tek P6205 for example. This requires breaking the gate path with extra length of gate cable which is not always available, adding to inductance

[filssavi]

While it is by no means a common measurement to do, it can be, and is done, at least in research...

There are several papers that propose using gate current measurement for all sorts of things:

• Junction temperature measurement
• dynamic current balancing for parallel transistors
• etc...

That said, it is a fairly difficult measurement to pull off properly...

For research purposes, (and if money is not too much of an issue) I would use a optically isolated high voltage differential probe (TeK isovu or lecroy hvfo lines), to measure the voltage across the gate resistor. This way you can get 1 GHz+ bandwidth, which should be enough even with GaN HEMTs. If you are using TO220 or 247 devices, you will probably be fine with the Rogowski coils linked earlier, especially for IGBT (30 MHz banwidth is probably not really enough for SiC).

However be carefull that any additional loading or impedance (especially inductance) you add to the gate circuit can impact (even severely) how the Transistors operate, so you might have some bad surprises when you take away the probe/probe connection predisposition

Also IMHO making this measurement work in a deployed product, as opposed to in a lab with your tongue at the right angle and k$ or gear), with any degree of reliability, would be very difficult bordering on impossible ( I am sure it can be done given a large enough timespan and supply of alchool and curses however., and more crucially custom silicon)

[PartialDischarge]

• Junction temperature measurement

I did not know about this, always thought saturation voltage was the key parameter for junction temperature

optically isolated high voltage differential probe

Just a remark here, these probes are singe ended, not differential. This is what allows them, together with the optical isolation, to have such a good cmrr. The measurements are differential, but the input is single ended

[JayArr]

In 25 years of power supply and inverter repair we have never looked it up, measured it or worried about it. We have been substituting FETs in repair all this time, as long as the voltage and current were as good or better we felt it was OK and it always works. It is noted that each successive generation of FETs has had lower and lower Rdson values so we don't sweat that either. Our spare parts drawers are arranged by voltage first and current second, look up 800V then look for 23A and whatever is in that drawer will work.

We always assumed that there was little to no gate current and as long as the power FET had a good square gate signal of about 10-15V that was all that was required.

[PartialDischarge]

In 25 years of power supply and inverter repair we have never looked it up, measured it or worried about it. We have been substituting FETs in repair all this time, as long as the voltage and current were as good or better we felt it was OK and it always works. It is noted that each successive generation of FETs has had lower and lower Rdson values so we don't sweat that either. Our spare parts drawers are arranged by voltage first and current second, look up 800V then look for 23A and whatever is in that drawer will work.

We always assumed that there was little to no gate current and as long as the power FET had a good square gate signal of about 10-15V that was all that was required.

Clearly repair is different than design or development. Also you’d be surprised at the levels of gate currents, fast currents in time, but not negligible by any means

[JayArr]

If I can substitute at will in power supplies and inverters, without regard to gate current, then my conclusion is that in power switching applications it is not important.

What other components in the circuit do you think might vary based on the gate current of the power FET?

[bdunham7]

Can you give your opinion on why besides the measurement of gate voltage, the measurement of turn-on gate current of Mosfets, GaN, Igbts.. would be important or desirable?
I'm referring to power electronic switching applications of course.
If you think it is not also say it...

The gate voltage and its transient characteristics such as rise time and ringing, are what is important--it is what generates the electric field that operates the device and is also what will destroy the gate if it excessive.  Gate current would only be of interest in analyzing gate voltage characteristics and since you can more easily measure that directly, why would you bother with current?

Another issue I can think of is that the better the gate drive design, the harder it is going to be to insert a current measuring mechanism without changing the characteristics of the circuit.  On a design where you have a TO-220 package MOSFET attached to a heat sink and connected with long wire leads, you could probably measure the current pretty easily.  On a modern, compact, optimized PCB with a gate driver possibly operating at a fairly high frequency, it might be tough to find a spot to measure current and you would likely end up grossly affecting the operation.  Of course Tek, HPAK and LeCroy probably have megadollar probes that I don't know about just for these situations, but I'm not sure they sell any to the power electronics design guys.  The type of measuring you refer to is probably limited to researchers.

[filssavi]

If I can substitute at will in power supplies and inverters, without regard to gate current, then my conclusion is that in power switching applications it is not important.

What other components in the circuit do you think might vary based on the gate current of the power FET?

This is like saying that using the correct diameter wheels and tyres on a car is not important because the car is not scraping the floor and is able to go forward…
I am fairly sure  you could fit carriage wheels to most city cars and they would actually work, horribly of course, but they would still revolve around their axis
With transistor it is the same, just because they are not blowing up this does not  mean they are working well.

Gate current controls how fast the device is switching, too much of it and you have EMI and long term reliability problems, too little of it and your switching losses go up, making the transistor hotter and driving down efficiency.

Of course just as with cars, the mechanic does not need to know why the manufacturer chose the exact diameter for the wheels to change them…

[Siwastaja]

Gate current is important because it affects switching speed (thus EMI) and gate voltage ringing and overshoot, but you can measure both easily and directly by monitoring gate voltage with an oscilloscope, which is something we all do all the time. If it has to be in-circuit (during application), monitoring gate voltage is still much easier than gate current, and IMHO gives more insight than the current measurement.

But maybe there is some oddball application where you gain some important insight by measuring both gate voltage and current, I don't know. In most applications, measuring gate current is definitely totally unimportant, otherwise it would be something that is regularly done. In fact, even just gate voltage (which is much easier to measure) is rarely monitored; if desaturation detection is required, you obviously measure that directly (i.e., Vds). But there are gate drivers that monitor Vgs. Adjusting dead-times automatically is one application; and there voltage matters, not current.

[PartialDischarge]

The gate voltage and its transient characteristics such as rise time and ringing, are what is important--it is what generates the electric field that operates the device and is also what will destroy the gate if it excessive.  Gate current would only be of interest in analyzing gate voltage characteristics and since you can more easily measure that directly, why would you bother with current?

Agreed, but you are missing 50% of the story. Gate charge depends on gate voltage swing, so it may not be specified for your particular application. So the only way to know the gate charge is to integrate the gate current in time, and this is what determines the power. Also the gate current will give you a clear indication of which driver to select. Gate drivers also have maximum specifications that must be derated with temperature, so it may not be clear if the driver will perform well long term under all temperatures and gate resistors.

[PartialDischarge]

but you can measure both easily and directly by monitoring gate voltage with an oscilloscope, which is something we all do all the time.

 In fact, even just gate voltage (which is much easier to measure) is rarely monitored;

You keep mentioning that gate voltage is easy to measure and I disagree, maybe because we are talking about different things, ie different expectations. 3 cases are typically found:

1) High-side mosfet. This is the hardest case with a high and fast common mode that renders differential probes useless due to typical oscillations seen in them. Also in this probes the higher the common mode, the higher the attenuation in differential mode, which creates more noise in low level signals like gate voltages.

2) Mosfet with source current resistor. Many applications feature a Mosfet "slighty" floating over a current sense resistor, similar case as 1) that still creates a common mode that makes a direct single ended probing approach not feasible.

3) Grounded Mosfet. This is the easiest case and still creates distortion when fast switchings and currents are present. This distortion can be enough to render a gate current measurement taken by difference in potential across the gate resistor useless

[Siwastaja]

Well obviously measuring anything high-speed requires the correct gear and know-how to use it. Just probing willy nilly with a x10 probe with a long ground lead mostly gives the same EMI ringing patterns on the scope screen even if you probed your ground plane against the same plane...

But gate shouldn't be that difficult, it's a fairly low-impedance signal (some ohms) with a lot of voltage swing where small fractions of volts are meaningless, you want to know if the gate is at 2V or at 4V at a certain point of time but quite some noise can be accepted in the measurement. Yes, high side has some fast common mode.

Measuring gate voltage is quite easy for the gate driver IC manufacturer with good enough accuracy if they think it gives some benefit to their gate driver operation.

"Kelvin sensing" the source terminal, drain terminal, and switch node is completely possible. The driver IC already has to include the floating driver, the same part can implement floating measurement.

OTOH current shunts are usually chosen to be small enough not to cause serious undesired Vgs negative feedback mechanism, saving the pins and IC design required for Kelvin gate driving. Minimizing the burden voltage is also desirable for efficiency and saving in power dissipation. For example my designs tend to have some 50-100mV max drop under max load and amplifier with some 50x gain. Such low burden voltage can be ignored in gate drive voltage discussion easily.

For really high frequency stuff Kelvin connections are required though to bypass source inductance which otherwise slows down switching.

Again, I'm asking what's the practical point of this discussion, have you seen a design which measures gate current, or have you seen some discussion about it? I.e., what point you are "hiding"?

And are you primarily talking about measurement on the lab table or in-application? Even that isn't clear to me yet, but the difference is huge; for lab, you can just buy decent gear (asuming it's not a shoestring budget hobby thing) and be done with it. In a design, you basically need the IC manufacturers to recognize nont only the brilliance but the real-world benefits of the idea of measuring gate current.

[PartialDischarge]

Again, I'm asking what's the practical point of this discussion, have you seen a design which measures gate current, or have you seen some discussion about it? I.e., what point you are "hiding"?

I'm not hiding anything, I believe gate current is important to choose and confirm the right choice of gate driver, and I want to know if anyone here can add more to the discussion. It is called science.
That you do not have to offer much to this discussion does not mean it is pointless, ie. I do not know everything neither do you.

And are you primarily talking about measurement on the lab table or in-application? Even that isn't clear to me yet, but the difference is huge; for lab, you can just buy decent gear (asuming it's not a shoestring budget hobby thing) and be done with it.

I'm referring to lab-table

[Siwastaja]

Your belief of gate current measurement being important doesn't sound very scientific. All I'm merely saying, I'm sure you have some example, a use case, or a discussion in your mind behind your question; i.e., why you believe gate current measurement is important, so I want you'd share it with us, because I truly have never even considered measuring gate current, it just haven't crossed my mind, so I think I'm going with

If you think it is not also say it...

... but I might be wrong.

In any case, gate current obviously is important but because driver voltage and total gate resistance are well-known and easily adjusted, you can get good approximations of gate current (say +/-20%) without measuring it. You'd need to have a special case where the exact and accurate value (say significantly better than +/-20%) of the current itself, or its integral, would be interesting and any other easier / direct measurement would not suffice as an alternative.

For lab table, problems like fast common mode are best tackled with expensive gear, but that won't be practical for hobbyists or shoestring startups.

[PartialDischarge]

Your belief of gate current measurement being important doesn't sound very scientific.

I was referring to asking others in the field as a way to gather more opinions, that is science to me.

I truly have never even considered measuring gate current, it just haven't crossed my mind,

And I truly appreciate your answer, which has been clear from the start, but that does not preclude this discussion from happening.

[xavier60]

Maybe stating the obvious. The one important thing I look at is the duration of the Gate plateau during turn on and turn off, during typical load, then adjust the Gate resistors if necessary.
Slow voltage transitions before and after the plateau will add a bit to losses.

[IDEngineer]

Still not sure of the original question. Personally, I don't instrument gate driver circuits to know current in real time, but I always characterize gate current (using a Tek current probe) during design to dial in the discrete gate resistor. With all the variables (driver internal resistance, gate resistance, etc.) calculations are at best a rough estimate, so I measure to insure peak current doesn't exceed the driver's spec.

[PartialDischarge]

Maybe stating the obvious. The one important thing I look at is the duration of the Gate plateau during turn on and turn off, during typical load, then adjust the Gate resistors if necessary.
Slow voltage transitions before and after the plateau will add a bit to losses.

True, and changing the gate resistor directly affects the peak gate current, so having this information would allow to confirm if the driver meets specs, and also to calculate gate switching losses

[PartialDischarge]

Still not sure of the original question. Personally, I don't instrument gate driver circuits to know current in real time, but I always characterize gate current (using a Tek current probe) during design to dial in the discrete gate resistor. With all the variables (driver internal resistance, gate resistance, etc.) calculations are at best a rough estimate, so I measure to insure peak current doesn't exceed the driver's spec.

Exactly. I was referring to measuring during development phase. Glad to see that someone does this

[bdunham7]

Agreed, but you are missing 50% of the story. Gate charge depends on gate voltage swing, so it may not be specified for your particular application. So the only way to know the gate charge is to integrate the gate current in time, and this is what determines the power. Also the gate current will give you a clear indication of which driver to select. Gate drivers also have maximum specifications that must be derated with temperature, so it may not be clear if the driver will perform well long term under all temperatures and gate resistors.

Well first, I'm not a design engineer--I mostly fix things.  However, fixing often involves analysis and modification, so I think I do have just enough insight to spot a problem with your reasoning.

In the general case,  you can design something by using the manufacturer specifications and tolerances and leaving an adequate safety margin.  Characterizing an individual part doesn't seem to me to be a requirement for a routine design job, at least for standard, stable circuits.  There may well be situations where you do want to characterize parts--even I do this at times--but the trap is, obviously, that if you rely on your characterization rather than the guaranteed specs, your next batch of components may not work.  A good design should be tolerant of normal variation in component characteristics to the extent possible.  In the specific case of MOSFET gate drive, I think most parts are specified well enough to permit choosing the rest of the circuit components without resorting to doing your own characterization.  If I'm wrong, which I certainly may be, please provide an example.  And I'm certainly not saying that you shouldn't thoroughly test your finished design to make sure it is working as planned, I'm just not convinced that gate current is the way to do it.

You've also mentioned measurement issues.  I frankly don't have any of  your stated issues measuring gate drive voltages within any reasonably necessary accuracy and in the most difficult cases, an isolated oscilloscope that goes for $6K new along with a simple 100X probe with a ground spring does the job better than it needs to.  Measuring gate current would be much, much more difficult for me given that I'm working with an existing PCB.  If I were setting up a custom PCB just for that purpose, I'm sure I could do better.  If measuring current works for you and you do it routinely, could you share your specific methods?

[bdunham7]

Still not sure of the original question. Personally, I don't instrument gate driver circuits to know current in real time, but I always characterize gate current (using a Tek current probe) during design to dial in the discrete gate resistor. With all the variables (driver internal resistance, gate resistance, etc.) calculations are at best a rough estimate, so I measure to insure peak current doesn't exceed the driver's spec.

How do your measured results typically compare to your original estimate?  Are you optimizing the transition time by getting the peak gate current as close to the maximum as possible?  And how many examples do you test to account for component variability?

[dmendesf]

There's some wizardry (now called AI) in setting the gate voltage/current to optimize efficiency:

https://www.powerelectronicsnews.com/using-ai-to-increase-ev-range/

[PartialDischarge]

Found some research papers on the topic:

https://ieeexplore.ieee.org/document/8356960
"Influence of peak gate current and rate of rise of gate current on switching behaviour of SiC MOSFET Naresh K Pilli* and Santosh K Singh"
Emergence of SiC technology brought revolutionary reforms in power electronics area. Ability of SiC MOSFETs to operate at high voltage and current ratings makes it a best replacement for Si IGBT. Mere by replacing the Si IGBT with SiC MOSFET will not give the best results, until special care is taken in driving the SiC MOSFET to extract the positives of SiC material. Gate driver for SiC MOSFET should be quick and hard enough to achieve low switching losses. The paper focuses on modelling the SiC MOSFET, taking peak gate current and rate of rise of gate current into consideration. Switching losses are derived in terms of gate current only.

https://ieeexplore.ieee.org/document/7506090
"Active Current Source IGBT Gate Drive With Closed-Loop di/dt and dv/dt Control"
This paper proposes an active current source gate drive (ACSD) method based on voltage controlled current source(VCCS) feedback control strategy for high-power IGBTs. Unlike the common voltage source gate drive, the proposed ACSD method provides constant drive current to charge and discharge an IGBT. With a large gate drive current, high switching speed and low switching losses can be achieved in a power converter. However, a high-current/voltage overshoot occurs. To solve this problem, a feedback current proportional to the di/dt or dv/dt signal is generated to the IGBT gate. Thus, direct control of the net gate drive current is produced. Then, the current/voltage overshoot is controlled with little sacrifice in switching time


https://ieeexplore.ieee.org/document/7520864
"Gate Sensed Method of Load Current Imbalance Measuring for Paralleling IGBT Devices"
n order to realize load current imbalance measuring for paralleling IGBT devices, the authors propose a novel method for measuring the imbalances from IGBT gate charge. In the study, an analytical model between gate charge and on-state current has been worked out to find the current differences of IGBT among each paralleled sub-circuits. The results from simulation show that the analytical model is effective and on-state static collector current of IGBT can be mapped from gate charge. For dynamic current time delay, it is also can be detected from the gate charge.