Ringing from fast low side driver

[Rubb]

Hi,

I have designed a board that uses IX4427N to scale up a quadrature signal from 3.3V to 24V. Uses a STM32 MCU to generate the quadrature signal.
Have problems with ringing, it polutes both ground and voltage supply.
Frequency of the quadrature signal is < 60 kHz.
Channel 3 and Channel 4 (Blue and Green, are going through another standard N-channel FET to invert the signal, therefore a slower rise time.)
Channel 1 and 2 are directly from IX4427N, however some cables.
The PCB is 4 layers with   Signal - GND - GND - Signal.
I also have an ESD Diode on IC4427N output, don't know if this could cause problems here (ESDA25LY).

When I read about this as this is a gate driver, the output should have gate resistors which I don't have in my design. However, the output is only floating at this point so only stray capacitance in cables then? I have tried solder on 10 and 100 ohm resistors on the outputs but it is at the end of the output with nearly no difference (Or should I put a resistor on the input to IX4427N?).

Looked into snubbers but don't know really where they should be and if I would need both a resistor and capacitor. I am able to solder on extra components to the PCB but as it involves scratching solder mask and such I rather keep the number of extra components low.
I'm thinking of putting capacitor between Vsupply - IX4427N Out AND GND -IX4427N Out, but dont know if this is worth a try and what size roughly I would need.

Regret that I just didn't do an ordinary design with a N-channel mosfet, think it wouldn't have had any problems, but wanted to try a tailored IC but.... something is missing in my design or this was just a super poorly components choice.

Added attachments of some oscilloscope measurements and also the PCB design.

[fourtytwo42]

My first comment is your decoupling sucks!

Unfortunately writing this  reply the site has hidden the picture I was looking at but as I recall you have a single 4.7uF MLCC, this is insufficient. You need at least another 100nF & possible also a 470nF as close as you can get to the pins (Lowest value closest).

Unfortunately your layout also sucks, again as I recall the ground plane is split directly underneath the chip and your present inadequate decoupling is not even directly  connected to the half that contains the ground plane of the chip.

I suggest  you look very carefully at how other people lay out high speed switching circuits, you cannot get away with just throwing them together 

[Rubb]

Thanks for the comment.

That's correct that I only have a 4.7 uF close to the chip. I just put 4.7 uF as I don't really drive anything from the output of this driver (almost no current output as it only simulates an encoder signal that is read by a high impedance input.) Mainly saw the capacitor here as a buffer, but as you say at least a 100 nF close would be nice.

Now when you point it out the supply is really cutting the ground plane here, I see that problem now. Would of course be nice to have a ground plane under the IC.
Layer 2 is however a full ground plane so I thought the GND vias would help with leveling it out and keeping consistent ground.

I'm just a hobbyist so this is my learning I'm literally learning by doing, which includes watching/reading others and then try out som designs myself.
I have done some basic switching of N-channel FET's at these frequencies but never encountered these type of problems.
It all entered when I decided to go for IX4427N.
I mean I even have the ringing overshooting problem at 50 Hz!
Now I'm in fault finding mode to make it better the next time.

The ESD Diode feels like something that could cause problems here as it seems to have a capacitance of around 30 pF?
- I may try to add a 100 nF on the same pad as 4.7 uF or open up the solder mask to the half where the IC GND is and put it there.
- Will try to also remove the ESD Diode to see if the ringing dampens. If it does I guess its better to go without or just try and solder on a 10 ohm resistor before the ESD Diode.

Or I might just be out of possibilities here and no need to try and fix/find the problem.

And yes, I'm a beginner

[tggzzz]

You need to show the ringing in more detail: amplitude, frequency, decay rate etc.

It may be that the ringing is a measurement artefact, so show a photo of your probing technique.

Be aware that many high value MLCC capacitors "lose" 80-90% of their capacitance at their rated voltage. If the manufacturer doesn't specify that, find another manufacturer.

[Jeroen3]

First you need to identify the source of the ringing. Is it ringing in the gate driver circuit or so called switch node ringing on the mosfet drain?
Or measurement of couse, did you measure the with the lead or the springy pin?

What fourtytwo42 correctly mentions is that your gate drive layout is not optimal.
You should recognize the current loop (red) that exists in a gate driver. See screenshot below.
And route accordingly to keep it short and low impedance.
Source: Fundamentals of MOSFET and IGBT Gate Driver Circuits

The diode may not be necessary, and if it is, it's for the mosfet between gate and source. Not the driver.

[Rubb]

The 4.7 uF used is this one: C2012X7R1H475K125AC
https://product.tdk.com/system/files/dam/doc/product/capacitor/ceramic/mlcc/catalog/mlcc_commercial_general_en.pdf
Driver supply voltage is 24 V.

I have to do some measurements again and zooming into the step, forgot to take pictures of these. What I remember from it is that it is super fast ringing, the ringing is just for a few nanoseconds  approximately < 10ns if I remeber correctly, and it is a few periods during that time.
The voltage peak can reach to around 35 V (when it should go to 24 V.) The same goes for the 0V output it can reach almost 10 V, and there is a overhearing between output 1 and output2 which is visible in the oscilloscope pictures as the quadrature signal is phase shifted 90 degrees, it is visible that all edges affect eachother.
I have also measured at the voltage supply which is a standard variable lab supply and I can see the voltage ringing even there, so my guess here is that they are interfereing eachother through the effect to GND and/or Supply.

Even if I go down to say 30 Hz for my input, my understanding was that it still creates the same type of ringing when it comes to peak, decay and frequency.

But I have to do some more measurements looking into these details.

Measurement setup:
During the measurement nothing is really connected, the leads running out of the pictures goes to a board connector where no cable is connected.
The two holes seen are standard 2.54 mm header pins I put there to be able to jack in and analyze the signal.
There I have connected my oscilloscope probes with the "springy pin".
The ground of the probe is connected to a separate ground pin on the board (I guess it becomes a long return path).

I have also connected jumper wires to the 2.54 mm header pins to another inverter board that just inverts the signals and then output to screw terminals, where I have tried to connect a 100 ohm resistor and then measured after the resistor. The resistor dampens the peaks a little but not much.

As I also mentioned earlier, there is really no load connected to the outputs of the IX4427N, basically free floating.
I put the ESD diodes there because the outputs goes to a board connector so I "thought" it would be nice to protect it with an ESD diode. But it was more of an extra layer of protection but they introduce capacitance towards my ground plane, so maybe a good idea to just try without them.


Jeroen3:
"Is it ringing in the gate driver circuit or so called switch node ringing on the mosfet drain?"
Switch node ringing on the mosfet drain, is this the mosfet within the gate driver, because I have no other mosfet in this layout except for the ones inside IX4427N.

Yeah, shouldn't have that long of a route from my driver output back to its own ground. When he has pointed it out it seems so obvious.
But as I don't have a mosfet gate at the output of IX4427N, the return path would be long as it is intended to go through wires to another high impedance input that even maybe on another board with the grounds connected.


I guess one of my mistakes here is that I have used the IX4427N, it feels like it's more targeted at driving another mosfet, I guess I really just wanted a N-channel low side MOSFET with an inverter.
Then I would have been able to tune the gate resistor with the gate capacitance to put it at the correct cut-off frequency for my application?

Is it even possible to have a circuit as IX4427N for the purpose I'm trying here?

[tggzzz]

The 4.7 uF used is this one: C2012X7R1H475K125AC

Do you see a C-vs-Vdc curve or table? If not, nothing can be inferred.

What I remember from it is that it is super fast ringing, the ringing is just for a few nanoseconds  approximately < 10ns if I remeber correctly, and it is a few periods during that time.

That's not "super fast", it is "ultra mega fast" Hint: look just under my moniker on the left

The voltage peak can reach to around 35 V (when it should go to 24 V.)

Key theory: V=L^di/_dt. L is ~1nH/mm of straight wire, more for a loop. Resonance can increase the peak.

Even if I go down to say 30 Hz for my input, my understanding was that it still creates the same type of ringing when it comes to peak, decay and frequency.

Anthropomorphic explanation: the circuit neither "knows" nor "cares" when the next transition might occur.

FFI: https://entertaininghacks.wordpress.com/2018/05/08/digital-signal-integrity-and-bandwidth-signals-risetime-is-important-period-is-irrelevant/

Measurement setup:
During the measurement nothing is really connected, the leads running out of the pictures goes to a board connector where no cable is connected.
The two holes seen are standard 2.54 mm header pins I put there to be able to jack in and analyze the signal.
There I have connected my oscilloscope probes with the "springy pin".
The ground of the probe is connected to a separate ground pin on the board (I guess it becomes a long return path).
....

Show photo. Details matter, and a beginner may not understand which details are important.

[Rubb]

No unfortunately not. No C-vs-Vdc curve, only that it has a 10 % tolerance and is rated for 50 V and I have 24 V.

I understand the moniker It's all in the numbers.

It adds up, it's not unreasonable that I would have 50 cm cables and then the return path on that.

Yeah it doesn't know but it was a little regarding the comment:
"I suggest  you look very carefully at how other people lay out high speed switching circuits, you cannot get away with just throwing them together"
Didn't really see it that way as I couldn't get it working at 30 Hz where I used N-channel + Inverter with no problems like this.
But I guess the comment is towards the IX4427N that favours speed and that's the IC that excites high frequencies?
So the comment was more towards not using IX4427N in a circuit as a beginner.

Will try to set it up again some evening this week in the coming days and take som picture.

EDIT: Apart from looking at the measurement setup, havent I really just picked the wrong component for my task here and that's why I have problems.
This part should drive the gate of a MOSFET where one should keep the trace short to the MOSFET gate and throw in a gate resistor between IX4427N and the MOSFET gate.
In my setup the loop will never be short, as the output of the IX4427N is supposed to go in cables for a distance of say 50 cm to a high impedance input.
The cabling and trace length would give both capacitance and inductance, and the ESD diode doesn't help here either.

I can live with slower rise time as I need say upto 60 kHz 50% duty cycle PWM basically. At 50 kHz the period is 20 us.
I have no need to really drive anything more than overcoming for example the inductance to change the voltage within a step.
If I were to solder on a resistor of ~10 ohm directly on the output of IX4427N before the ESD diode and maybe also a capacitor, which would form a lowpass filter, could that help in dissipating the ringing energy? I will of course loose the switching speed performance but it might be enough to dampen the oscillations enough while still being able to give an acceptable rise time?

IX4427N seems to be a much more complex circuit than I thought. Have done switching power supplies and such before but they usually have component selection, pcb guidelines etc. which indicates the importance of it. Lookong at IX4427N datasheet it looks quite "simple" so I went with the rise time and thruth table., big mistake.

[T3sl4co1l]

Your solution isn't "in here" [zoomed picture], it's out there. Zoom out. Show the board, show how your probes are connected. Show what signals go where and which if any go onto wires/cables/connectors.

This is almost certainly a common-mode phenomenon, i.e. the board's ground potential itself is being pushed around, in reaction to something else (a wire, etc.) being pulled, and this creates a voltage drop across the probe cable (especially the ground clip lead) thus you read a transient even "to ground" (but when is ground not ground, eh?).

Unfortunately writing this  reply the site has hidden the picture I was looking at but as I recall you have a single 4.7uF MLCC, this is insufficient. You need at least another 100nF & possible also a 470nF as close as you can get to the pins (Lowest value closest).

Doesn't matter, and may indeed make things worse.  Graded values and/or capacitor sizes in parallel are an older superstition and don't stand up to critical analysis.  In other words: simulate a typical layout, with body, pad, trace and via inductances modeled, and see what the impedance is like.

Tim

[tggzzz]

This is almost certainly a common-mode phenomenon, i.e. the board's ground potential itself is being pushed around, in reaction to something else (a wire, etc.) being pulled, and this creates a voltage drop across the probe cable (especially the ground clip lead) thus you read a transient even "to ground" (but when is ground not ground, eh?).

"Ground isn't". That's valid not only for PCBs/circuits (e.g. here), but also for large areas of soil/rocks. Consider why, if you are caught out in a lightning storm, you are advised to stand with your feet together. Yup, a nearby lightning strike means ground != 0V to a dangerous lethal degree.

Doesn't matter, and may indeed make things worse.  Graded values and/or capacitor sizes in parallel are an older superstition and don't stand up to critical analysis.  In other words: simulate a typical layout, with body, pad, trace and via inductances modeled, and see what the impedance is like.

Yes, but....

Decades ago graded capacitor values had some validity, due to the very non-ideal properties of old capacitors. Modern components are much less non-ideal (but circuit frequencies have also increased).

Graded capacitor values can also have some use, where there are sharp edges and also large "long term" currents. Not applicable in this case.

Simulation is indeed a great help, but knowledge, experience, and good taste are required so that the important "parasitic" components are included. Difficult for beginners.

[Rubb]

Yeah, I think I understand.
What is the ground where I measure, it might have spots where it differs during these loads in the transients.
But as you are saying how I connect to the ground with the oscilloscope is affecting if I get an antenna loop or not. I have just put in a jumper wire at the power supply (-) on the other side of the board to my probes "claw", so the ground point on the board I measure is a fair distance away from the IX4427N chip.

The board itself is quite large, approx 200 x 200 mm and of course it is alot that could affect. It isn't perfect in anyway but these disturbances I see disappear as soon as I just hold the IX4427N input low. If I turn on IX4427N with no probe connected to the IC4427N output, it is enough to just connect the probe to the power input to the board to start seeing these disturbances in the supply.

But the ESD Diode has really catched my interest, I only put it there as I connect wires to that signaland thought it was a good idea to ESD protect it.
Looking at a gate driver connected to a MOSFET gate we usually need gate resistor as the driver could excite the gate capacitance, reasonable capacitances here are in the range < 100 pF?

If I look at the ESD Diode I have there in the datasheet:
C_line = 50 pF
It has a breakdown voltage of 25-30 V (I at least measure higher than that in the peaks.)
The capacitance is in the order of a MOSFET Gate and on top of that if ringing and overshooting occur the diode will start to short my output to ground which also pollutes my ground, which may spread to the other ESD Diodes nearby which then might start to conduct into other low signals.

So I might be really wrong here but that ESD Diode seems to maybe be a real devil in this setup, so when I have time I will try to just desolder that ESD Diode and see what happens. I will try to take a picture of the setup but I might also try to setup a test circuit (far from optimal will try som header SMD footprints and wires) and see if I can reproduce the behaviour so I can try and experiment a little with different placements of capacitors/resistors and such instead of ruining the solder mask and solder on stuff there.

[T3sl4co1l]

So what connects to those traces, or do they just dead end and do nothing? Or is it super secret sauce that you can never ever reveal, and this whole thing will be forever a mystery to us both?

A nearby ESD diode won't do very much; consider the stray inductance of the trace length between driver and diode, and compare that to the capacitance, and then compare the product to the rise time (tau = pi sqrt(LC)/2), and the ratio to the output impedance (Zo^2 = L/C).

Tim

[tggzzz]

Yeah, I think I understand.
What is the ground where I measure, it might have spots where it differs during these loads in the transients.
But as you are saying how I connect to the ground with the oscilloscope is affecting if I get an antenna loop or not. I have just put in a jumper wire at the power supply (-) on the other side of the board to my probes "claw", so the ground point on the board I measure is a fair distance away from the IX4427N chip.

1mm of wire ~1nH. Add probe tip capacitance, and what do you have?

More complex with coils/loops or ground planes, but the same principles apply.
https://entertaininghacks.wordpress.com/2015/04/23/scope-probe-accessory-improves-signal-fidelity/

For general probing techniques, see the refs at https://entertaininghacks.wordpress.com/library-2/scope-probe-reference-material/

[Rubb]

It is just dead-end now. Just outside of the picture to the right are the pads for the connector outlet. The connector housing is soldered on the board but no connector or cables in it. As I havent even bothered making any connectors yet. But as I said previously these cables are meant to simulate a quadrature signal (encoder) that is connected to a high impedance input, that input is just interested in the order of the signal and edges.
I want the thing I connect it to to believe that it is rotating a certain way.

All the measurements I have taken on these outputs are from the header pins visible in the picture.
With probes directly on the header pins and the ground of the oscilloscope connected to the voltage input of the board, not the IC itself. I have also tried connecting jumper wires from these header pins to an inverter board using 2N7002 and mesured with the probes at that boards output but with the same type of ringing as if no inverter board and just connected directly to the header pins.

So are you saying that the capacitance of 50 pF in the ESD diode connected close (seen in picture) to the IX4427N output shouldnt cause any problems due to the inductance being low?
It wont become the same type of ringing as if I had a Mosfet gate there instead? (Or at least what I have read about when it comes to gate drivers output connected to a gate and the need of for example a gate resistor to dampen). In this case the ESD might even start to conduct to the ground plane or is this transient too fast for the diode to react?

Thanks tggzzz, will look at those links.
But should I see this ringing if I only connect my oscilloscope directly to the power input?

I also just want to thank all of you for the answers, I appreciate it. At least it feels like I know much more now of this than at the first post so it goes in the right direction.

EDIT: tggzzz, looked at the links it cleared it alot with the oscilloscope.
When I have time I'll need to try the probe spring and have as short ground loop as I possibly can.

[Jeroen3]

EDIT: Apart from looking at the measurement setup, havent I really just picked the wrong component for my task here and that's why I have problems.
This part should drive the gate of a MOSFET where one should keep the trace short to the MOSFET gate and throw in a gate resistor between IX4427N and the MOSFET gate.
In my setup the loop will never be short, as the output of the IX4427N is supposed to go in cables for a distance of say 50 cm to a high impedance input.
The cabling and trace length would give both capacitance and inductance, and the ESD diode doesn't help here either.

Yes most definitely the wrong part. It's a gate driver not a line driver.
A gate driver is intended to charge and discharge a gate as fast as possible. And gates are basically capacitors.
Gate drivers barely have any continuous capability. Plus you ask for 20 us, that's not fast 

See if you can find a line driver with your voltages. But you can also use something like an high side switch such as Infineon ITS4141, though these are often much slower.

[T3sl4co1l]

Yes most definitely the wrong part. It's a gate driver not a line driver.
A gate driver is intended to charge and discharge a gate as fast as possible. And gates are basically capacitors.
Gate drivers barely have any continuous capability. Plus you ask for 20 us, that's not fast 

See if you can find a line driver with your voltages. But you can also use something like an high side switch such as Infineon ITS4141, though these are often much slower.

Do you know if there's a current limit, other than thermal?

I would imagine, operating at peak current, unidirectionally, might perhaps run into electromigration issues, but operating at modest currents would solve that.  And then, I'm not aware of any failure mode other than thermal.  For which, running at a couple hundred mA would again suffice.  But I haven't seen such a rating, because, they don't care, it doesn't matter, it's a gate driver, not a... base driver, or load driver; they just don't bother to rate it that way... But that doesn't mean it necessarily can't.

So, just wondering if you've found some useful information you might share.

(And yes, there are types to watch out for, like the boosted (Miller clamp?) types might only work best dynamically, or anything with NPN or NMOS follower isn't going to have as nice of a V_OH saturation characteristic; though the IX4427 in question I believe is a CMOS output, so that wouldn't apply here.)

Regarding edge rate, simply adding a (series) source termination resistor suffices to dampen ringing on a cable; the cable must be shielded, mind, since you're driving a unipolar signal down it.

For slow edge rates, and no DC load, CD4000 logic might well suffice, though none offer a level shift capability from low voltages; you're most likely going to use a open collector/drain output with pull-up to VCC, and then a CD4xxx gate can clean it up.

Might also be tempted to use a discrete circuit, but the half-dozen or so components does get annoying for more than a couple channels.

Tim

[Rubb]

Now there is an update, as I found some time to sit down with this after getting the kid to bed.

@tggzzz thanks for the probing info links it helped alot.
I did some new measurements today and now I only had one probe connected to the scope and I used the ground spring on the probe to measure. I also use the 10X setting at the probe and calibrated the probe with the spring on the calibration input and screw until I had good looking square wave. What a difference!
Took some pictures of the setup but hard to see anything, but I  think the measurements are valid now at least.
I measured with and without the ESD Diodes, no visible difference in the measurements, but now they are removed.

I measured with the probe tip directly on the IX4427N output pin and the ground spring to the decoupling capacitors ground side just below (the ones seen in the picture).
(In the setup picture, you also see that I have two identical IX4427N so I demonstrate the probing on the one below that doesn't have the pin headers but it is easier access with the probe and the spring without shorting things by mistake.)
That gave a nice looking curve with just a small overshoot of almost 5V, no real ringing afterwards. Can't see any difference between the positive and negative transients.

I then took just 2 dupont wires connected to eachother (20 cm each, 40cm in total length), what I had laying around but want to kind of see with "bad" wires, from one of the header pins visible in the PCB layout on one of the outputs. I put the probe tip on the dupont wire end and the ground spring on the same decoupling capacitor as before. Then I start to see larger overshoot and more of a ringing, this time the peak was at 40 V so ~16 V overshoot, attached figures of this as well. 

The middle length picture is from when I only took 1 dupont wire from the header at the IC output to the voltage supply input ~17cm away on the PCB, put the probe tip on the end of the dupont wire and the ground spring on the voltage supply ground input (-). Overshoot and rining again, this time around 10 V overshoot and a ringing.

The quality of the measurements wasn't this good before even when I tried to zoom in, so the ground spring to the probe and measure on different places on the board was a super good tip.
I measured a bunch of other things this way with the spring, the input to the IX4427N had no overshoot at all and it had no real disturbances to it.

It is however, an overshoot as soon as I move away from the IC on the output side and it can get quite large, the oscillations are just a few periods and are rather damped.

Shouldn't this mean that the problem I have is on the output side and it will not become better with larger or longer cables. Is it worth to solder on a say 10 ohm resistor on the IX4427N outputs before the diode footprints? This way it should dissipate some energy during the transient and slow it down? Which in term could dampen the overshoot, as well as the oscillations, at the cost of transient time performance ofcourse, but I have room for it in my time limits.
Could this be damaging to the IX4427N? It is according to the datasheet specified to a peak current of 1.5 A which feels like alot as the output should only be connected to a high impedance input.

[tggzzz]

@tggzzz thanks for the probing info links it helped alot.
I did some new measurements today and now I only had one probe connected to the scope and I used the ground spring on the probe to measure. ...

You're welcome.

I remember learning the consequences of lead length, almost 50 years ago (gulp) on the (non-TTL) clock of a 6800 processor.

[T3sl4co1l]

Gate driver is fine, it's a beefy device.  Peak currents here are comparable to trace impedance.  A wire through air might be closer to a few hundred ohms (i.e. it looks pretty inductive, one or a couple hundred nH if it's the kind of jumpers I'm thinking), and the probe is ballpark 10pF, so Zo = sqrt(L/C) is high, which is approximately the ratio between the voltage step amplitude and ensuing peak current.  High 10s of mA.

Note that your scope is almost certainly not seeing the full picture: a 100MHz bandwidth has a rise time of 3-4ns, exactly what's shown.  The driver might well be producing an edge of 2 or even 1.something ns!  This goes down significantly under load, if it were adjacent to a MOSFET that is.

Tim

[Rubb]

Ok thx Tim.

Do I understand you correct here that it will be in the range of <100 mA current during transients?

There doesnt seem to be a need to change anything to have it connected to a high impedance input?
The intended cables aren't these jumper wires, it will be more suitable cables for the connector.

The receiving end will be connected to the same ground but is there a risk of the receiving end getting way too high voltage overshoot?

[T3sl4co1l]

No idea, you haven't mentioned what connectors or cables, nor what's at the far end.

If the system is linear, and the LC or TL response is dominant without damping, the peak voltage is exactly twice the step amplitude.  Or much more if resonant with repeated edges (square wave).

I did mention shielding earlier. Notice you have an LC circuit (at least around the first resonance; such risetimes propagate as waves through transmission lines (TLs) and this is a 1st/fundamental lumped-equivalent approximation), and that current flows through the signal wire with respect to ground.  That current returns on the ground pin(s) of the connector.  Which will have some voltage drop, and thus the average potential of the cable, or what's at the far end, can differ from local ground: common-mode emissions.

The simple solution for ringing, is to terminate the wave with a series resistor (source termination), shunt (load termination, draws DC current), or both (loses 50% amplitude).  Note this does not reduce the rise time, which as a fast wave, can drop significant voltage (10s, 100s mV?) across ground-return paths, even when shielded cable is used -- i.e., just across a few cm of uncoupled length, like connector/header pins.  A shielded connector is preferred for this reason, with multiple pins or even a continuous shell terminating into the ground plane.

Alternately, a L + (C || (R+C)) filter can be used to slow the wave, which deals with cable reflections (up to some limiting [electrical] length corresponding to cutoff frequency).  For cutoff below a couple MHz, unshielded cable is generally acceptable, but it depends on what and how (for example, USB Low Speed / HID is typically ran over unshielded cable -- mouse and keyboard stuff; note that USB has error correcting functions, making it robust in the presence of noise).

Tim

[Jeroen3]

Yes most definitely the wrong part. It's a gate driver not a line driver.
A gate driver is intended to charge and discharge a gate as fast as possible. And gates are basically capacitors.
Gate drivers barely have any continuous capability. Plus you ask for 20 us, that's not fast 

See if you can find a line driver with your voltages. But you can also use something like an high side switch such as Infineon ITS4141, though these are often much slower.

Do you know if there's a current limit, other than thermal?

I would imagine, operating at peak current, unidirectionally, might perhaps run into electromigration issues, but operating at modest currents would solve that.  And then, I'm not aware of any failure mode other than thermal.  For which, running at a couple hundred mA would again suffice.  But I haven't seen such a rating, because, they don't care, it doesn't matter, it's a gate driver, not a... base driver, or load driver; they just don't bother to rate it that way... But that doesn't mean it necessarily can't.
...

For this specific part they do specify the on-state resistances at 100mA. Which are 8 and 12 ohm. Causing around ~100mW. Not a thermal limit in this package. These resistances do cause voltage drop. But this may not be a problem with 24V.
The 10ns rise time is...

[Rubb]

Thanks for all the help.
Thought I would show what I have tested and landed at trying for now.

Haven't thought of any special cables maybe it will be let's say 0.14 mm2 cables that can be around 50 cm or even a little longer.
These will be connected to an encoder input that reads it as HIGH/LOW. Think HIGH is around ~12 V and LOW should be lower than that.
Input seems to have resistance > 40 kOhm.

As my time demand is rather low compared to the speed of the IX4427N I decided to just try a low pass (RC) filter at the output, in the hopes of removing the higher frequencies that may cause the LC parasitics to resonate.
Took some parts I had laying around (only large 0805 parts), put a 10 ohm resistor in series with the output and then a 22 nF capacitor to ground after it.
That was a mistake, I focused to much on the time constant and realized later that 24 V over a 10 ohm resistor in the transients isn't good
It became hot but I realized it in time as far as I know.

I then tried a 220 ohm resistor and 2.2 nF, which was the lowest capacitor I had laying around that I knew had at least 50 V voltage rating. I think I would rather have had a 220 pF to get it a little faster.

Did some measurements and it looks good for this use case I think (a little slow but hey it is what it is when mistakes are made, I have increased my knowledge at least. )
I used the empty ESD Diode pads to fit the components and I cut the trace where I put the resistor. Is it beautiful? Haha nope but it seems to work.

So how did the measurements look? Well the measurements before the low pass filter is as before with around 5 V overshoot that my oscilloscope can detect.
After the filter there is no overshoot and a typical linear first order system output.
There is no real  difference between measuring towards ground close to the filter or using the double dupont wires and measure the signal towards the voltage supply input ground (-).
The measurement picture is 40 kHz pulse output.