Core saturation with overcurrent protection?

[sapien]

Hi all!

Recently I've build a buck converter, which I suspect might be saturating the core (the IC is blowing up). It however has overcurrent (peak) protection.

Is it possible this inductor is saturating during sudden turn-on of a load?
https://lcsc.com/product-detail/Power-Inductors_COILMX-MTP2920-150MC_C2858867.html
The core saturates at about 18A, and the overcurrent peak detection is set to 100mV, with a 7.5mOhm shunt => shouldn't exceed 13.3A

Does such an overcurrent protection react fast enough to avoid core saturation?

We're talking about a 13A load (electronic load), in theory divided across 2 inductors, since the LTC3892 is working in dual phase mode. Ofcourse during turn-on transient the current will probably exceed the 6.5A per inductor easily.

My second question is; can such a core saturation in this configuration destroy the IC?

LTC3892 datasheet: https://www.analog.com/media/en/technical-documentation/data-sheets/38921fc.pdf

[fourtytwo42]

Recently I've build a buck converter, which I suspect might be saturating the core (the IC is blowing up).

I take it you mean the control IC however you do not mention the condition of the MOSFETS ?

Also you have not posted your layout, is this a pcb or breadboard prototype ?

Unfortunately the data sheet does not mention Tpd from the sense input's to the driver turn-off, also you do not mention how much testing you have performed on the circuit so maybe the MOSFETS you have chosen have a high gate capacitance and need some assistance to turn off faster.

If you increase RS you should be able to test the overcurrent operation at a non-destructive current and find out why it does not seems to be doing it's job!

[sapien]

Thanks for your reply!

I take it you mean the control IC however you do not mention the condition of the MOSFETS ?

What exactly do you mean with the MOSFET condition?
The MOSFET's never die, only the IC. And this is after literally 20 or something blown IC's (and never another blown component).

Also you have not posted your layout, is this a pcb or breadboard prototype ?

I've added the layout of the board. It's a 4 layer board. It's my first ever real design of a buck-converter, so any feedback is greatly appreciated!

If you increase RS you should be able to test the overcurrent operation at a non-destructive current and find out why it does not seems to be doing it's job!

I've tested the overcurrent protection quite extensively: it does (somewhat)work. Making RS higher lowered the max current accordingly (exactly as expected), or simply lowering the threshold voltage works as well. However, the proper behavior of graceful limiting has always only been achieved with slowly lowering the 'constant R' value on my electronic load. Whenever I would drop the CR either too much, or too quickly (stepping from 10.5 to 0.5 Ohm, or slowly decreasing to something like 0.3Ohm, both with 12V at output) it seemed the current limiter didn't gracefully cope. The IC would either blow up, or the output voltage 'slowly' oscillate all over the place (instead of a bit more jittery and lowered output voltage).
Might this be due to core saturation?

[Terry Bites]

You may find this interesting. e2e.ti.com/blogs_/b/powerhouse/posts/waveform-audit-inductor-core-saturation-saturation-current

[sapien]

Thanks for thinking along! The inductor I'm using certainly seems to have quite a hard saturation. The temperature rise isn't noteworthy. 20 to 25 Celsius increase max., and all this temp rise is 90% due to it's own heating, so should be accounted for in the datasheet inductance, right?

It also still leaves the two questions; does the overcurrent protection react fast enough to prevent saturation?
And can saturation lead to only destroying the control IC, not any other component?

[fourtytwo42]

It also still leaves the two questions; does the overcurrent protection react fast enough to prevent saturation?

That is what I hoped you would find out with lower current non-destructive testing, using a scope to measure the time between the current exceeding the limit and the FET drive output turning off (as it's not specified in the data sheet).

I also suggest you make yourself a saturation tester to find out if the inductor meets it's spec and also how hard the saturation is.
The attached red trace is an ETD?? core, vertical 4A/cm so onset of saturation is 12A.

Ignoring the inductor for a moment I just noticed there appears to be something seriously wrong with your schematic! What is the purpose of R12 ?
The decoupling capacitor C19 is clearly inadequate, why did you not follow the data sheet schematic closely, was there a reason ?

[sapien]

I initially followed the board layout of the DC2493A.
https://www.analog.com/en/design-center/evaluation-hardware-and-software/evaluation-boards-kits/dc2493a.html#eb-documentation
It shows a low pass filter at the input of Vin, with values corresponding to R12 and C19 in my schematic.
Right now I've even increased R12 to 5Ohm, and C19 to 4.7uF, which provided me with somewhat better results (survived a few more transients before breaking the IC).
Might this be harmfull? As you're saying 'seriously wrong'...
What might be a suitable decoupling cap? It's only feeding the internal LDO's, so we're talking around 200mA max, right?

I will try out building a saturation tester. I've also bought an 10uH 30A saturation replacement to test out.

[fourtytwo42]

Right now I've even increased R12 to 5Ohm, and C19 to 4.7uF, which provided me with somewhat better results (survived a few more transients before breaking the IC).
Might this be harmfull? As you're saying 'seriously wrong'...
What might be a suitable decoupling cap? It's only feeding the internal LDO's, so we're talking around 200mA max, right?

My feeling was the other way around that transients on various signal lines into the chip particularly current sense & fet drive would couple into vcc and if that was a high ac impedance the voltage would rise and destroy the chip. I confess the fet drivers have separate decouplers however for such a complex chip 100nf alone on vcc is inviting trouble. It seems very strange to me that you are able to destroy control chips without apparently affecting the mosfets, after all the normal failure mode of such chips is a gate to drain short in a mosfet. Have you absolutely checked all the mosfet gates are high impedance with the control chip removed ?
The other possibility is the layout, I must confess I cannot understand your plots but maybe thats just me, for example the top layer seems to have copper pour over everything obliterating for example the inductor pads that I cannot see. Also maybe the ESR of the particular capacitors you have selected or if ceramic there actual capacitance at the working voltage might be a problem. What is the operating frequency & have you checked it's correct ?
Good luck

[Andy Watson]

I think if saturation was the problem it would be the mosfets that would be suffering.

Have you seen this thread - in particular reply #4 and #16.

https://www.eevblog.com/forum/projects/gate-driver-keeps-blowing-up/

[sapien]

Thank you both for your replies!

The MOSFET gates are high impedance with the IC removed.
It is my first time designing such a board, so my layout has probably maaaany improvements. If you see any you'd like to point out, feel free!
The controller works at 150kHz, with a 180 degree phase shift between the 2 sides, therefore generating a 300kHz at Cin and Cout. I have measured this to be actually the case. I wouldn't know how to test the ceramic caps at their used voltage and frequency, but the voltage ripple is pretty much what I expect, so should be fine then right?

The other possibility is the layout, I must confess I cannot understand your plots but maybe thats just me, for example the top layer seems to have copper pour over everything obliterating for example the inductor pads that I cannot see.

Most of the power electronics stuff, i.e. the inductor, is on the 'bottomlayer'. Is it the convention to put large components on the 'top'-labeled layer? 
I guess the picture could be way more cleared up, disabling the view of some layers/masks when taking a picture. I'll upload a better one tomorrow!

A 100nF cap might not do the trick, but my 4.7uF does?

I think if saturation was the problem it would be the mosfets that would be suffering.

Have you seen this thread - in particular reply #4 and #16.

Do I not understand how this forum works, or is there no #16? Thanks for pointing out though! I also thought the FETs should suffer, so nice to see my electronics understanding somewhat confirmed. I will try to get a good look at the gates of the MOSFETs, and check whether they for example go below 0V. I'm wondering though, is the load for the gatedriver increasing as the load increases? As far as I understand it doesn't at all. All in all doesn't seem to be my problem, but I'll certainly look into it.

[Andy Watson]

or is there no #16?

Did I say #16? Apparently I did - oops - sorry! I meant #8 (on the 16th Novemeber). But it's a short thread - wouldn't hurt to read the whole lot.

[DavidAlfa]

The MOSFET gates are high impedance with the IC removed.

Bad idea anyways. Always add a high value pulldown resistor to ensure nothing randomly charges it up, 100K should be ok...

[sapien]

Thanks for the tips and clarification!
Ones again; the FETs blowing up isn't my problem.
Is it at all possible (high frequency) noise and oscillations managing to get through the low pass filter at Vin of the IC are breaking it? I'm using a cheap, quite noisy power supply, which I measured to have quite a bit of noise and some weird oscillations. They're being excited at around 20kHz (the oscillations are way higher freq). Noise is up to 400mV peak-to-peak. Not all does make it through the low pass filter, but certainly some does.

[Siwastaja]

Well, if Isat = 18A and your calculated overcurrent detection is at 13A, there is ample margin. Of course you may have calculated wrong, or maybe made a mistake wiring up the current sense to the controller, or something like that.

Actually measure the current on oscillosscope, that will tell. Set the trigger near 20A and see if it ever triggers. Well of course it will trigger when the MOSFETs fail, but then you can see what happened instantly before that failure - is it current ramping up linearly and then this ramp-up speed accelerating until seriously exceeding MOSFET limits?

EDIT: Read more carefully and the IC is failing. IC usually fails too when MOSFET fails due to gate shorting, but if your MOSFETs survive and only the IC fails, then the problem is very likely something else than saturation. Use oscillosscope, probe the supply voltage first.

[fourtytwo42]

I'm using a cheap, quite noisy power supply, which I measured to have quite a bit of noise and some weird oscillations. They're being excited at around 20kHz (the oscillations are way higher freq). Noise is up to 400mV peak-to-peak. Not all does make it through the low pass filter, but certainly some does.

An unreliable test bench doesn't help debugging so you need to replace that psu pronto!
However the ripple you state should be impossible given all the capacitors in your POWER input filter, so either your measurement technique is poor or your capacitors are poor OR that so called psu should be shredded immediately!

I assume we are talking psu related noise and not your loads switching ripple ??

[sapien]

However the ripple you state should be impossible given all the capacitors in your POWER input filter, so either your measurement technique is poor or your capacitors are poor OR that so called psu should be shredded immediately!

I assume we are talking psu related noise and not your loads switching ripple ??

Thanks for your comment! We're talking psu related noise. This is my measurement across one of the input ceramic caps. It's the same I measure the source directly, when it's detached. In both instances, the load has its display on, but is NOT powered on. Is there any way of how my scope might measure, or some common mode noise which might lead to this? The high frequency oscillations are around 2.4MHz, and occur with a regularity of around 20kHz.
I've even measured after the low pass filter of the Vin pin of the IC, and it's identical! How can this be?
Diving into the capacitors I used, the MLCC I use at Cin have an 5% ESR rating at 1kHz. This seemed, and still seems typical (not the best, but certainly not the worst, right?) to me.

https://datasheet.lcsc.com/lcsc/2110121030_Kyocera-AVX-CM316X7S475K100AT_C2900750.pdf

Ofcourse there might be some weird ESL effect at way higher frequencies at play, but that's just guessing, since it isn't mentioned in the datasheet.

Is my source really this crappy? And are the MLCC's I'm using somewhat proper for the job? Or did I overlook something?

EDIT: Read more carefully and the IC is failing. IC usually fails too when MOSFET fails due to gate shorting, but if your MOSFETs survive and only the IC fails, then the problem is very likely something else than saturation. Use oscillosscope, probe the supply voltage first.

What do you mean by probing the supply voltage first? What should I look for?

[Siwastaja]

Well, take that tiny steel spring which came with the oscillosscope probe (or make a new one from a component leg) and probe directly between the GND and actually any other pin, one by one. None of the pins should exceed maximum ratings. You can look at Vin, INTVcc, DRVcc for starters.

The long ground lead with the alligator clip picks up too much interference which is not actually there at the pins.

[sapien]

Thanks for that tip! I made myself such a steel spring gnd probe, since it didn't come with my (cheapo) oscillosope.
I'll report my progress tomorrow.

It seems most pins on the IC only can handle -0.3V. So it might just be that some oscillations somewhere, due to bad board layout, cause it to dip below that.

[sapien]

After a day of testing, for now it seems the problem could be related to oscillations at the top gate, and it going too far below 0V.
I've added an 5Ohm resistor in series with the gate, and so far the IC hasn't broken down on me.
However, instead of keeping on trying, and only after a year of doing so being able to say 'it definitely works!', I would like to know if this is logical to be a/the point of failure (I'm going to be attaching a next version of this board to some expensive equipment, which I really don't want to break because of this board failing).

Sidenote about my cheap scope:
I'm really running into the limitations of my cheap scope. It shows a big voltage dip (measured at the gate since it's the only point I could way more easily access), which would certainly destroy the IC if true. However I've measured it with a 10x probe, to minimize capacitance and not noticeably change the gate capacitance. Whenever I zoom in more to get a better graph (both voltage and time) my scope gos crazy. Zooming in time, the voltage will NOT actually be crossing the threshold when being triggered. Zooming in voltage will or sort of have the GND reference break, making the voltage swing all over the place.

My question:
The demo-board has top FETs with a gate resistance of 1.1 Ohm. Mine are 0.68. Is it logical for such a small difference to let the IC break? There's also no optional components on the demo-board, making it possible to increase the gate resistance in case of for example a noisier source. So is it logical such a 5Ohm resisitor does the trick reliably?

[MrAl]

Hi all!

Recently I've build a buck converter, which I suspect might be saturating the core (the IC is blowing up). It however has overcurrent (peak) protection.

Is it possible this inductor is saturating during sudden turn-on of a load?
https://lcsc.com/product-detail/Power-Inductors_COILMX-MTP2920-150MC_C2858867.html
The core saturates at about 18A, and the overcurrent peak detection is set to 100mV, with a 7.5mOhm shunt => shouldn't exceed 13.3A

Does such an overcurrent protection react fast enough to avoid core saturation?

We're talking about a 13A load (electronic load), in theory divided across 2 inductors, since the LTC3892 is working in dual phase mode. Ofcourse during turn-on transient the current will probably exceed the 6.5A per inductor easily.

My second question is; can such a core saturation in this configuration destroy the IC?

LTC3892 datasheet: https://www.analog.com/media/en/technical-documentation/data-sheets/38921fc.pdf

The first is a hard question to answer you'd have to take some scope pics.  Usually an inductor will not suddenly saturate, it takes a buildup of current over some time period.  If the control circuit is faster than the buildup then the control circuit protects the system, but if it is too slow to respond then it may not be able to protect anything.  To find out for sure you have to scope it out.

The second question depends on the interconnections and what the control IC is responsible for.
If the control IC has the driver transistor on it, then it can definitely be destroyed with an over current condition.
If the control IC has an external driver transistor, then it depends on how the external driver transistor blows out.  If the drive transistor blows completely open then the IC may not be affected unless it gets an over voltage.  If the drive transistor blows such that say the collector base shorts out then the IC internal driver transistor could easily blow out too.  It could be like a chain reaction.  This actually happens with circuits that have main driver transistors and drive transistors that drive those transistors.  The lower current transistors blow out too.  One simple explanation for this is that when the main higher current transistors blow out the circuit topology changes suddenly to something totally different than what the design had called for, so you never know if that new topology is subject to new current paths or higher voltage potential nodes that can blow out some other parts in addition to the main transistors.  Because of this there could be other parts that blow out too.
It's even a little more complicated than that unfortunately.  If the main transistors blow out it could damage the lower current transistors and you'd never know it.  Replacing the main transistors would appear to fix the problem when really the reliability of the entire device could be very compromised.  That's why it is best to replace the lower current drive transistors in addition to the main transistors, and if there is an associated control IC it may be a good idea to replace that too.  The damage occurs to the silicon dies so you cant see it and cant measure it unless you get 'lucky' and the transistor blows out completely so you can measure it.

[sapien]

Thanks to everyone helping out!
Although I might run into new problems in the future, the biggest problem as it turns out was the gate of the top MOSFET causing the LTC3892 IC to blow up. Either by pulling the voltage too low for the rating of the IC, or by having such a low gate resistance the current spike was blowing up the IC (although the first option seems way more likely!). This was solved by adding (for now 5 Ohm) resistor in each path to the gate.