Can someone please help me find why the LM2596 is blowing up on my PCB ?

Thread Starter

nishantnidaria

Joined Sep 27, 2021
34
The problem with that TP0610T MOSFET is its Rds(on) is 10Ohms, not milliohm, ohms!!!

View attachment 249008

You need something like a Vishay Si2393DS in a SOT23- 3 case which can handle about 700mW on a 25 x 25mm copper area, with Rds(on) of 22mOhm. Ideally a larger case form would be better and more flexible on board area.
Vishay Si2393DS in a SOT23- 3 seems like a good option! Thanks! but I am using NCE3407AY ( P-Channel Enhancement Mode Power MOSFET ). Is it not good ?1632838487283.png
 

Attachments

Irving

Joined Jan 30, 2016
3,181
Vishay Si2393DS in a SOT23- 3 seems like a good option! Thanks! but I am using NCE3407AY ( P-Channel Enhancement Mode Power MOSFET ). Is it not good ?View attachment 249011
Its OK, The Vishay is the best I could find in a SOT23 case - its a little better than your NCE3407, but is still marginal.... You really need a bigger case, TO252 or similar and an Rds(on) <15mOhm. Your inrush current is 11A for about 1mS to charge up those big capacitors, though run current is only around 3.5A peak and averages 1A. An ONSemi FDD4141 for instance has Rds(on) of 10mOhm and will lose only 94mW.
 

Thread Starter

nishantnidaria

Joined Sep 27, 2021
34
Its OK, The Vishay is the best I could find in a SOT23 case - its a little better than your NCE3407, but is still marginal.... You really need a bigger case, TO252 or similar and an Rds(on) <15mOhm. Your inrush current is 11A for about 1mS to charge up those big capacitors, though run current is only around 3.5A peak and averages 1A.
Everything you said is correct, i do need a bigger case. I will change that in my design.

Do you think the reason intrinsic body diode of the P MOS is shorted because of over/under voltage at the gate? However the datasheet tells us the limit +-20V.
 

Thread Starter

nishantnidaria

Joined Sep 27, 2021
34
No, I think it just overheated... with Rds(ON) of 10ohm its junction temperature would have exceeded 150C in a few mS and turned it into a low resistance silicon bar. I'm just surprised it didn't pop or smoke...
But my RDS(ON) is <52mΩ @ VGS=-10V, the name in the schematic is not the same as the one i actually have on the PCB, i am using NCE3407AY.

I am not using TP0610T MOSFET
 

Attachments

BobaMosfet

Joined Jul 1, 2009
2,053
Thank you for your comment, My input voltage is between 12 to 13.5 V, I have two output voltages one dc-dc each, 5V and 8V, the 5V one is blowing up on reverse polarity. Desired current is 250mA to 500-600mA.
Look at the figures you've told me, and then consider how much heat you're forcing that regulator to dissipate.

12-13.5V is forcing the regulator to dissipate 7-8.5V * 250_to_600mA = 5.1W (max)- you're overheating your regulator because you didn't know how to determine (based on the data sheet) the maximum current you can get out of it at those parameters. At a minimum, at a huge heatsink to the regulator.

Here is an example of an LM7805 ('cause I have that calc handy) showing you what it could output if you tried the same parameters (with a TO-220-3 heatsink):

1632843156072.png

You couldn't expect the regulator to output more than 322mA and remain at 38.8C.
 

Thread Starter

nishantnidaria

Joined Sep 27, 2021
34
Look at the figures you've told me, and then consider how much heat you're forcing that regulator to dissipate.

12-13.5V is forcing the regulator to dissipate 7-8.5V * 250_to_600mA = 5.1W (max)- you're overheating your regulator because you didn't know how to determine (based on the data sheet) the maximum current you can get out of it at those parameters. At a minimum, at a huge heatsink to the regulator.

Here is an example of an LM7805 ('cause I have that calc handy) showing you what it could output if you tried the same parameters (with a TO-220-3 heatsink):

View attachment 249020

You couldn't expect the regulator to output more than 322mA and remain at 38.8C.
But i have absolutely no problems of any kind during normal operation, no heating whatsoever. However i would mention that the 250-500mA current consumption is not constant, (it's 200mA -250mA (from raspberry pi 2) for maybe 35 seconds then 500mA (from a camera) for maybe 10 seconds then again 250mA for 5 seconds, then again 500mA for 5 seconds and then 0.2mA consumption for 5 minutes) this cycle repeats every 5 minute.

Problem is only during reverse polarity because Mosfet's internal diode shots as soon as i start the circuit (in normal condition--> right polarity.).
 

Irving

Joined Jan 30, 2016
3,181
But my RDS(ON) is <52mΩ @ VGS=-10V, the name in the schematic is not the same as the one i actually have on the PCB, i am using NCE3407AY.

I am not using TP0610T MOSFET
Oh...

Do me a favour - in future:
  • make sure your schematic reflects reality;
  • disclose everything in your first post as to what you did and what's happened
  • be very clear and specific about what you are asking,...

Otherwise people who want to help will very quickly get fed up...

To answer your question... its hard to say...
If you look at the thermal impedance chart Fig 14 and specifically the single pulse line (the bottom one) then for a 1mS pulse you have a normalized thermal impedance, Ti(1mS) of around 0.07. The thermal resistance Tr(ja), junction to ambient, for SOT23 is around 100degC/W on 1" x 1" copper but here you have virtually no copper except the gate is tied to the ground plane. So the actual thermal impedance is probably two or three times that. But lets say 200degC/W. And the power dissipated is 11A through 52mOhm = 11 * 11 * 0.52 = 6.3W. So your junction temperature for that single pulse at an ambient of 30C will be at least:

Tj = Ta + Tr(ja) * Ti(1mS) * Power = 30 + 200 * 0.07 *6.3 = 118C​
It wouldn't surprise me if it was 20 or 30C higher...

And since the only real path for that heat is through the gate pin to the ground plane its more than likely you had a hot spot there, melted the silicon and punctured the gate isolation, or at least weakened it to the point that any nasty voltage spikes could subsequently puncture it.

By the way, what is your 12v source?
 

Thread Starter

nishantnidaria

Joined Sep 27, 2021
34
Oh...

Do me a favour - in future:
  • make sure your schematic reflects reality;
  • disclose everything in your first post as to what you did and what's happened
  • be very clear and specific about what you are asking,...

Otherwise people who want to help will very quickly get fed up...

To answer your question... its hard to say...
If you look at the thermal impedance chart Fig 14 and specifically the single pulse line (the bottom one) then for a 1mS pulse you have a normalized thermal impedance, Ti(1mS) of around 0.07. The thermal resistance Tr(ja), junction to ambient, for SOT23 is around 100degC/W on 1" x 1" copper but here you have virtually no copper except the gate is tied to the ground plane. So the actual thermal impedance is probably two or three times that. But lets say 200degC/W. And the power dissipated is 11A through 52mOhm = 11 * 11 * 0.52 = 6.3W. So your junction temperature for that single pulse at an ambient of 30C will be at least:

Tj = Ta + Tr(ja) * Ti(1mS) * Power = 30 + 200 * 0.07 *6.3 = 118C​
It wouldn't surprise me if it was 20 or 30C higher...

And since the only real path for that heat is through the gate pin to the ground plane its more than likely you had a hot spot there, melted the silicon and punctured the gate isolation, or at least weakened it to the point that any nasty voltage spikes could subsequently puncture it.

By the way, what is your 12v source?
Really sorry about the lack of information and creating confusions, it is my very first post in any forum, for sure next time i will take all the things you said under consideration before posting.

It's a great and detailed analysis! Helps me understand a lot of things a lot better! Thank you so much for that!

My source changes, sometimes it comes from 220V through a 12V transformer and sometimes a 12V 12AH Lithium Ion battery.
 

BobaMosfet

Joined Jul 1, 2009
2,053
You cannot apply the power calculations of a liner regulator to a switcher.
True- but the lesson I'm trying to impress is the same- you need to understand your thermal junction temp even with a switcher. Switchers are designed to work within an envelope of parameters.

While I am not familiar with the switcher the TP is referencing, I can provide an example using the ubiquitous MC34063A:

1632851082337.png

That's pushing 7W throught the IC- again- engineering is a discipline where you must learn and understand what you are doing- how things actually work. Not just throw it together and hope for the best. Is the TP aware of the maximums for their regulator, and where is power, derating, and thermal curves are based on his necessary figures?
 
Last edited:

Irving

Joined Jan 30, 2016
3,181
Notwithstanding the reverse polarity issue... I think there are issues with the SMPS... here are the results of a simulation I ran:

1632844679091.png1632846918458.png
The specified 470uH inductor is running at double its peak current rating at the high output current, and close to it at the low end.

The specified output capacitor has an ESR of 0.45ohm and a ripple current rating of 450mA, its running out of spec over much of the output range. A good capacitor would have an ESR <0.25Ohm in that value range.

Was there a reason for using the LM2596? There are better chips that have a smaller footprints, better regulation, higher efficiency, etc, etc

I'll deal with layout issues in another post...
 
Irving points out a common issue with COTS chokes - they are all horribly over specified, i.e. if you want 2A out buy a 4A choke - else it will overheat, core losses for high delta-I ( = delta B ) are also very high in these cheaper parts - hence keeping the current ripple low is a good idea.
We have 10's of 1000's of units in the field with the 52kHz variant of this part - and no field failures - to date. We do buy the good quality certified parts and we do protect against reverse input.
 
p.s. in the analysis provided by Irving the Fsw should be 150kHz - which is what the part switches at ( always ) so the current ripple for 12V to 5V would be 28mA - so the simulation is quite wrong unfortunately. This is a function of the library part supplied by the vendor - and these are often quite flawed.
 

Irving

Joined Jan 30, 2016
3,181
p.s. in the analysis provided by Irving the Fsw should be 150kHz - which is what the part switches at ( always ) so the current ripple for 12V to 5V would be 28mA - so the simulation is quite wrong unfortunately. This is a function of the library part supplied by the vendor - and these are often quite flawed.
This wasn't a manufacturer's model but one developed by, I believe, eetech00 (I think that's how he spells it) and as I understood it was reckoned to be a good emulation, but maybe you have a better one?
 
Yes we have better models - but we don't share - any way - pencil & paper is all that is needed here V/L = di/dt

12-5 = 7 volts applied, D = 0.42 approx, thus dt = 2.8uS L = 470uH, therefore di = +/- 20.8mA or 15mA rms if you prefer.

for 47uH simply mult x 10 = +/- 200mA
 

Thread Starter

nishantnidaria

Joined Sep 27, 2021
34
I think i would go with this Mosfet: HSU0115 If my Vcc is between 11.5-13.5 V, with this particular p channel mosfet do i still need a zener diode for Vgs and current limiting resistor at the gate ? Thanks!

Or is it too big for the purpose of reverse polarity protection? My device's current consumption is no more than 600mA

P.S. Vgs limit is +-20V
 
Last edited:

Irving

Joined Jan 30, 2016
3,181
I think i would go with this Mosfet: HSU0115 If my Vcc is between 11.5-13.5 V, with this particular p channel mosfet do i still need a zener diode for Vgs and current limiting resistor at the gate ? Thanks!

Or is it too big for the purpose of reverse polarity protection? My device's current consumption is no more than 600mA

P.S. Vgs limit is +-20V
No you don't need zener/resistor if your input is <20v and clean, i.e no spikes. If you're not sure then put a 15v zener in there.

Your average consumption is around 600mA but peak is around 3A running and start-up in-rush is around 11A. That device, in theory, is going to be more resilient/robust than the little SOT-23 devices you were using...

Don't you have parts from mainstream suppliers like Infineon, STM, etc.?
 

Thread Starter

nishantnidaria

Joined Sep 27, 2021
34
No you don't need zener/resistor if your input is <20v and clean, i.e no spikes. If you're not sure then put a 15v zener in there.

Your average consumption is around 600mA but peak is around 3A running and start-up in-rush is around 11A. That device, in theory, is going to be more resilient/robust than the little SOT-23 devices you were using...

Don't you have parts from mainstream suppliers like Infineon, STM, etc.?
Unfortunately i don't but i found one from Diodes Incorporated "DMPH6050SK3Q-13" it's a TO-252 package size. Do you think it is better than HSU0115?
 
Last edited:
Top