Exploding IC in buck converter

[Teun]

Hi guys,

I have an interesting problem.
A client of ours has sent back several buck converters we supply. Never had much trouble with them espescially for this client.

It is used to charge a mobile device with an onboard receipt printer. These are used on busses for public transport.

Over time most of the PCB's have blown IC's. All on the side of the input pin.(see picture) TVS and varistors are used to handle spikes on the input. The converter should be able to deliver 9V/3A (IC rated 6A output) on the output and from what i can see with the printer activated it uses 28W of power max. The current limit/OCP of the converter is set to 3.5A.

Input range is +/-   12-32V DC. (Busses can be around 30VDC)

Can anyone give me an idea of why the input would be blown?

[Ian.M]

Automotive DC supply transients are *NASTY* especially on older or poorly maintained vehicles.  Probably you haven't got enough headroom between the TVS and varistor clamping voltage during a high energy transient spike and the abs.max. input voltage of the IC.   Post a schematic for your input protection circuit, + links to datasheets for the IC, TVS and varistor.

[Hast]

Hi

I have had a similar problem.
I did a buck converter for a industrial system with "industrial 24V".
At first everything worked fine and then we started to get returns where the input was blown similar to yours.

After some measuring we found that the "industrial 24V" was a strange sinusoidal signal from 0-48V. The buck converter I used had a max of around 32 V in or something like that.
The customer had switched out the 24V supply to a cheaper one.

So measure the input voltage with an oscilloscope and not a multimeter is my advice.

BR

[max_torque]

Load Dump for 24v nominal systems is speced at 212 volts!   What LD and RBP (Reverse Battery Protection) do you have in place in this unit!

[T3sl4co1l]

Hmm, that's exactly where the bondwire ought to be, too.  That thing ate a lot of amperes, very quickly.

Yeah, 30V, useless.  What knucklehead designed this thing?  To what standards was it tested?

Automotive parts start at 50V and go up from there, and that's mainly for 12V systems with modest (70A) alternators.  You might want an 80V rating here for example, just so there's enough headroom to handle the transients how ever you can (probably by disconnecting the load?).

Tim

[Teun]

Can't post the controller datasheet. But here are the input filter and varistor /TSV.

https://www.bourns.com/docs/Product-Datasheets/SMAJ.pdf
https://www.sfi.com.tw/index.php/en/2012-05-14-07-49-39/ml-c

[Ice-Tea]

Kinda difficult to comment without posting details about the controller?

[Teun]

Kinda difficult to comment without posting details about the controller?

I know and I'm sorry for that.
It has a -0.3 to 40V absolute maximum input rating. So I would think it has enough headroom.

Also once again... we use them in many of our products and they work fine. It seems this problem only occures with this type of load or supply.

[Ice-Tea]

The rating of D2 seems a bit strange to me? With a 12V nominal bus (yeah, yeah, automotive blah blah) it has a max clamping voltage of 48V? By that time, your 40V DC/DC controller is long gone..

[Ian.M]

Well its fairly obvious what's wrong (and Ice-Tea has also picked up on it). 

The nearest match in your linked  datasheet to the TVS in your schematic (D2, 36.8V breakdown) is the SMAJ30CA.  Its max. clamping voltage passing a 8.3A peak pulse >10us is 48.4V, well above the 40V abs. max input voltage you claim for your controller.  There just isn't enough headroom between the max normal operating voltage from the vehicle DC bus and your controller's abs.max. input voltage rating for *ANY* TVS diode to be effective.

Load dump isn't a myth although its fairly rare, but nearly all vehicles still on the road have central load dump protection, so the 24V system unclamped 212V peak load dump will only be seen if someone is running a poorly maintained vintage truck or similar with no ECU.  Without knowing the exact specs of the specific vehicle's central load dump protection, you cant know if it clamps under or over your controller's 40V abs.max input voltage rating.

[Siwastaja]

36.8V nominal breakdown TVS & 40V abs. max spec - not gonna work.

The actual clamping voltage, when current flows through the TVS, is much higher than the nominal value.

Add unit variation on the top.

[Teun]

Ok thanks! Trickery is that I need a reverse voltage of at least 28V for the TVS. So that's just not possible?

[Ice-Tea]

The SMAJ26CA is at least closer: 28.9 breakdown (satisfies your 28V minimum) and 42.1V max clamping voltage. Still beats your 40V max, but better at least and in a lot more cases it will not surpass the 40V max (when the current isn't too high).

Still, not a good way to do things. You simply need a higher rated DC/DC controller.

[Siwastaja]

The problem with TVS diodes is, they are not precision devices. The clamping voltage depends on current and unit variation. This is why the downstream components need to be rated higher. If your input is x volts, and you pick nominal clamping voltage of, say, 1.25*x to ensure it's never unintentionally clamping, then the worst case clamping is at least 1.5*x or maybe even closer to 2*x.

The problem in automotive systems is, the voltage varies quite a lot even in normal conditions. Assuming this is a 24V automotive system, you don't want to clamp at below 30V, otherwise you'll be dissipating power in the clamp unnecessarily. (BTW, you should have a thermally coupled polyfuse to protect against long-lasting slight overvoltage.) Now, if you pick the closest minimum guaranteed clamping voltage (with acceptable leakage figure) from 30V, the nominal clamping voltage will end up somewhere between 35-40V, and the worst-case, under load clamping will be somewhere around 45-55V. There is no easy way around this; I'd strongly consider replacing the DC/DC controller IC to something that can handle at least 60V absolute maximum.

The SMAJ26CA is at least closer: 28.9 breakdown (satisfies your 28V minimum)

For an even remotely robust system, you need to look the "working peak reverse voltage" spec, which is only 26V; it will likely conduct and leak excessively in a 24V automotive system when the engine is running, causing another set of problems (excessive heating, possibly only in some units, which may only become a problem after many hours of runtime...) . SMBJ30CA would be the lowest acceptable from that series, but then again, we are back to 48.4V max clamping voltage, way over the abs. max. rating of the IC. I think there is no way out by just replacing the TVS with another.

[T3sl4co1l]

Put another way, this design process is backwards.  It should be:
1. Start with nominal input voltage.
2. That selects TVS nominal voltage.  TVS peak voltage is typically 33-50% above that.
3. That selects converter max voltage.
4. If adequate ratings are not available, consider higher order methods, like a switching or linear prereg to deliver ride-thru, or a switch to disable load during transient.

If the TVS needs to handle massive energy, enough that a TVS (as such) isn't economical: use a MOV instead, which will be selected in the same manner.  Note the MOV will have a peak voltage typically 100-200% above nominal, putting much more pressure on the converter rating.

Tim

[magic]

I wonder if it would be feasible to cascode the converter with a depletion FET whose gate is pulled to the input by a resistor and clamped to ground by a low power TVS/zener? I mean, if you really can't find a converter with suitable ratings...

[Teun]

Well it's not that I don't want to use other converters. My boss keeps coming up with cheap ones he finds online and he expects me to to just use them

But al least it gives me another reason to try and convince him otherwise

Thanks for the help so far!

[Gyro]

Do you have any resistance in series with the supply. I know you'll drop the efficiency a bit but, for example, a 1 Ohm 5-10W wirewound resistor (higher resistance if possible) in series with the input (after the fuse but before any reservoir capacitance) will significantly improve any clamping measures by limiting the peak transient current. It's also the sort of thing that could be retrofitted outside the unit (or clamped to the inside of the case if metal).

This, in combination with the lower suggested TVS voltage rating, might get you past the problem.

[Siwastaja]

Well it's not that I don't want to use other converters. My boss keeps coming up with cheap ones he finds online and he expects me to to just use them

Well someone needs to do the design math, this is engineering and you can't just slap random parts together. It's not rocket science. If you are the responsible designer, do the math, show it, and just say "no" if that is the truth. It should be the end of discussion.

[Alti]

Well it's not that I don't want to use other converters. My boss keeps coming up with cheap ones he finds online and he expects me to to just use them

If you are the responsible designer, do the math, show it, and just say "no" if that is the truth. It should be the end of discussion.

If your boss's decision was to use component X then use component X.

Such decision frees you from doing the math necessary for component selection process and speeds up work. The disobedience and wasting your time for formally proving your boss makes poor decisions is not your job, you are an ee and not a babysitter.

[Siwastaja]

Alti, I usually detect sarcasm quite well but now I have to ask, are you serious or joking?

[thm_w]

For an even remotely robust system, you need to look the "working peak reverse voltage" spec, which is only 26V; it will likely conduct and leak excessively in a 24V automotive system when the engine is running, causing another set of problems (excessive heating, possibly only in some units, which may only become a problem after many hours of runtime...) . SMBJ30CA would be the lowest acceptable from that series, but then again, we are back to 48.4V max clamping voltage, way over the abs. max. rating of the IC. I think there is no way out by just replacing the TVS with another.

I agree that it is not an ideal design. But it looks like the maximum continuous automotive voltage tested in this certification is 28.2Vdc, does that sound right? Everything else is ms bursts.
http://www.emcprima.com/FileUPLoad/DownLoadFile/635792103688970689.PDF
http://read.pudn.com/downloads226/ebook/1062022/ISO16750-1.pdf

[T3sl4co1l]

True enough, but I would add these two caveats:
1. Add your formal complaint that you disagree with the choice, for such-and-such reasons;
2. Record evidence of the decision process (conversations -- insist on e-mailed discussion; emails, design documents, datasheets, and approval by the boss guy).

A faulty converter for an aftermarket automotive gizmo doesn't seem to be a very serious concern, but this is well worth keeping in mind anywhere the stakes are higher.

Tim

[mzzj]

For an even remotely robust system, you need to look the "working peak reverse voltage" spec, which is only 26V; it will likely conduct and leak excessively in a 24V automotive system when the engine is running, causing another set of problems (excessive heating, possibly only in some units, which may only become a problem after many hours of runtime...) . SMBJ30CA would be the lowest acceptable from that series, but then again, we are back to 48.4V max clamping voltage, way over the abs. max. rating of the IC. I think there is no way out by just replacing the TVS with another.

I agree that it is not an ideal design. But it looks like the maximum continuous automotive voltage tested in this certification is 28.2Vdc, does that sound right? Everything else is ms bursts.
http://www.emcprima.com/FileUPLoad/DownLoadFile/635792103688970689.PDF
http://read.pudn.com/downloads226/ebook/1062022/ISO16750-1.pdf

Real-life is definitely worse than that. Maybe you misinterpreted the standard or its not very good stardard to go with at all.
On a cold day I remember measuring 15.8v from VW alternator.  That would be 31.6v for 24v nominal system.

This is a graph from Bosch (Bosch Automotive Electrics and Automotive Electronics book)


I wouldn't design anything for below 32v. 

[magic]

Alti, I usually detect sarcasm quite well but now I have to ask, are you serious or joking?

We are in a Poe's law singularity. True believers are so crazy and trolls are willing to go so far in their exploits that there no longer is any practical difference.