MIC4420YN output oddity

Hello all! I've got a capacitive spot welder that I've been working on intermittently. I had chosen and ordered a few MIC4420YNs for driving the final MOSFETs, and all was right with the world until I noticed that their output pulse widths were much larger than the input they were receiving.

I've got a teensy 4 outputting 3.3v like it should, for a wonderfully precise duration, and the rise/fall times of the MIC4420's are just what they should be.

My circuit is as such:

My complete spot welder has 4 such assemblies, and they all consistently output pulses larger than what I put into them.

Now what's truly baffling is that the deviation becomes narrower the wider my input pulse becomes. At 150ms input width (which is the longest i would EVER want to pulse this device. 1.5 farads of capacitance at 15 volts makes my 2-gauge electrode leads literally jump) the MIC4420 is outputting a pulse width of160ms. At 5ms input width (the very lowest for perhaps welding tin foil, if I'd ever need to), I see output pulses 30-35ms wide.

Now I could likely live with this. 30-35ms is a bit too brief for the battery tab welding that I plan to do. But it's confusing and mildly frustrating, and according to the datasheet for these MIC4420's I'm well within what they should be capable of, and I want this thing to perform as I (attempted) to designed it.

Is this an artifact of their operation? Is it unreasonable expectation on my part? Is my circuit poorly designed? All these questions and more I hope the good people here can answer!

Thank you for your time in reading this.


I've got Diodes 1-6 connected as they are to attempt to protect against any back-EMF from lead inductance. The group of 6 are connected in parallel to each 4 copies of the schematic above.

Capacitors 1-4 across the VCC 10V and GND are shared across all 4 assemblies, in an attempt to reduce voltage drop when a total of 24 amps is (in theory) applied to all 24 PSMN1R1's. I've got a simple banggood-special buck regulator dropping my 14V input to 10V, and hopefully those capacitors are helping out.

If the MOSFETs are connected like your schematic, you don't need gate pulses, the MOSFETs are a short circuit through their body diodes, they will conduct as soon as the electrodes make contact.

Please mark on your schematic, Drain, Gate and Source pins.

Can you post a link to your MOSFETs data/spec please?
Are they P-CH or N-CH?

Are you using 4420 or 4422 as in your schematic?

What is your open circuit spot voltage?

Thanks... Tom.... :smiley: :+1: :coffee: :australia:

Geez, what a silly mistake to make. My excuse is... well nothing. This schematic was tossed together to illustrate my basic setup and I was careless in copying details from my main master schematic. Yes, they should not/are not connected like that. I will edit my schematic.

Will do! I'll edit the schematic along with changes JCA34F pointed out.

They are N-Channel. Here is the datasheet from the Digikey.ca listing that I purchased them from, complete with the CORRECT part number:

Behold, yet another silly mistake! That was indeed supposed to be a 4420 (non-inverting). Datasheet for reference:

Can be anywhere between 5-15 volts, depending on what I feed it from my bench power supply. My main capacitor bank is not shown, but it's 1.5F total rated at 16 volts. I run it at 14V only because the capacitors were expensive and I've scared to harm them, plus exploding these capacitors would probably be unwise.

I've edited my schematic and added a bit of info to my initial post, for clarity. I've added some diodes in that are across all 4 assemblies. I didn't think that they would be an issue, but hey, you never know. Their part number and connections are true to life (be it correct or not)

Thank you both for your questions and pointing out the flaws in my hastily-composed message! I promise that I've had this circuit working "as it should" outside of this pulse width issue. I click the button, my teensy outputs the correct pulse, and sparks happen. But comparing input/output pulses of the MIC4420, some strangeness is afoot.

Hi, @danielwiebe

PLEASE NEVER update/edit previous posts, you have now made ALL the post we have posted redundant and confusing.
If you change something on your project and presentation, PLEASE put it in a new post this will then keep the thread logical and easy to follow.

Tom.... :smiley: :+1: :coffee: :australia:
PS, Thanks for the info and the updates.

Ah, apologies. I'll not do that again!

Thank you for letting me know, and for taking the time to look at my problem!

Salient exception: If you posted code without code tags, please do edit that post to add the code tags to that same code so people do not have to scroll endlessly to get "up to date".

Yeah, that was my rational for editing, but I've got a clearer picture of when to do that now.

Not true. These n-channel devices are connected as low-side switches.

This will be due to interference from the welder affecting the low voltage circuitry I suspect. Welders are about the most electrically noisy environment there is and you need fanatical attention to shielding and EMI rejection in such as environment - 3.3V logic is hyper sensitive as circuitry goes (noise rejection is typically less than 1V).

The MIC4420 switches in tens of nanoseconds, anything on a millisecond timescale is nothing to do with this chip.

One thing to check is that the gate and source connections between the FETs and the 4420 is done properly, thick short wiring / traces, the gate and source signals must run alongside each other tightly in a low inductance fashion, and kept away from the high current path as much as possible (at right angles is good).

The connections to the sources for the high current path and the MIC4420 must be completely separate wiring, only commoned right at the source legs of the FETs. These points at the FET source pins must be the only point the grounds are commoned between the control circuit and the high-current side. Failure to do this results in a ground loop, and a ground loop in a welder will pick up lots of interference (many volts, or even tens of volts) due to the high dI/dt (rate of change of current) flying around.

Why did you use logic-level FETs? That just makes them so much more sensitive to EMI? In a welder you want high voltage parts to reduce the noise/interference sensitivity.

BTW the decoupling on the 10V for the MIC4420 must be ceramic, not electrolytic, and 10µF is plenty. This is ultra-fast decoupling.

And just like that, a perfect example of why I should not have edited my schematic. MarkT, I had indeed displayed them connected in the manner that JCA34F points out. I edited my schematic and thus, confusion! The current state of my first post will be it's last.

I had considered that, but had dismissed it as this is a DC single-shot spot welder, powered by a separate power supply connected directly across a massive capacitor bank. The 5v and 10V SMPS buck converters I use to power the Teensy 4 and MIC4420's respectively are both suspect, but the 5V buck has proper ceramic decoupling in addition to some bulk electrolytics to sustain the Teensy through the massive voltage drop upon trigger. And as I'm getting picture-perfect pulses out of the Teensy consistently, I felt safe in dismissing it.

The 10V buck is still suspect though. I'll add some proper ceramic decoupling and see what happens.

To give my rational as to why they're there in the first place; the bulk electrolytic capacitors across the MIC4420s are more for the 10V buck's sake than anything else. It cannot sustain the instantaneous 24-ish amps that the MOSFET array will pull (for a very brief time), and it's output capacitors are not large enough to handle the instantaneous current.

Well, the traces are a full 1mm wide in 2oz copper, so no problems there. As for location, wellllll.... Not so good. The MOSFETs border a 5/16th inch (sorry, Canadian, we work in both!) hole to which the negative electrode bolts to. The MOSFETs' sources/drains bridge 2 large planes that they surround, one of which goes to the bolt hole, and the other to the capacitor bank's negative terminal. Lots of current flowing through that area. The MIC4420 output traces go parallel to this area, about 20mm away.

I'll admit, I'd rather this particular fix be a last resort. The board(s) was expensive and very time consuming to put together!

If it helps, I got the same behavior with the MOSFETS installed, and without but with a 10ohm load to ground.

Heh, they were fairly cheap, their SOA at the desired instantaneous was acceptable when ganged, and honestly I didn't pay much attention to whether they were logic level or not, just their Rds ON when at 10V. 1.3 mOhm is quite low, especially when 24 are ganged together!

I will do this tomorrow! I have an exam first thing so I must review, and all my good parts (well, not salvaged parts) are in my lab toolbox at school!

Another thing that I would like to try is adding a 10k Ohm pull-down resistor to the output of one of the MIC4420s, without the MOSFETs connected with a 10 Ohm load, to see if that helps.

Thank you very much for your input and your time!

I have performed the following changes, in order:

Removed all FETs and am using a single 10 ohm gate resistor to ground, to help keep the circuit in question isolated. No change to circuit behavior.

Removed all 4 4420's and replaced only 1 with an unused, fresh spare (yay for sockets). Pulse width at 150ms remains 9-11ms longer than input, but the discrepancy at lower input widths does not increase. It remains at 9-11ms longer than input, down to 5ms. Improvement?

Added a 10 kohm pull-down resistor to the single 4420's output network, with no further change to circuit behavior (Expected, I suppose. the 10 ohm gate resistor is doing more for pull-down than anything ever could)

Added a 0.1 uF ceramic decoupling capacitor directly across the single 4420's 10V and ground. No further change to circuit behavior.

I realize I failed to address one of MarkT's comments:

That might be a problem. While yes, all the grounds DO go to the same place electrically, they certainly don't JOIN at the same place physically. Currently the 4420's all route through the 10V buck converter which is non-isolated and a straight through to ground. Still, not ideal at all. I will try to shorten those long traces with a small jumper to a FET source pin hole (which is as direct as I could make it to the capacitor ground plane, it literally encircles the pin) and see what happens. One moment please!

Sadly no further change after jumping straight to the ground plane about 15mm away from the 4420's ground pin.

I'll have to file a proper fix to that under "last resort", as with the 4420's output trace locations.

Again, thank you all for your responses, and I hope that a solution presents itself before we all lose patience!

This topic was automatically closed 180 days after the last reply. New replies are no longer allowed.