Switching regulator is failing from voltage spike, how to handle?

[Corporate666]

Hi guys,

I am using an Alpha Omega AOZ1021 switching buck regulator to make a 7.5V supply from a 12V source. This drives a series of LED's and has a max current draw of about 1.35A at the output.

The problem I am having is the regulators are dying on me. I connect the power supply, and instantly, the regulator is dead. I have experienced this problem before, and after talking to the manufacturer, it was determined that a very short high voltage spike when connecting the circuit to my power supply was killing the regulator chip.

I am pretty sure the same is happening, but last time I solved it by adding a TVS to the circuit. This time I have done the same, and it still happens.

The AOZ1021 has a max Vin of 18V. I am using a TVS with a breakdown voltage of 16V. The TVS is bidirectional and installed between Vin of the switcher and ground.

Does anyone have any ideas? I am not sure why voltage spikes would still be causing failure of the regulator, unless the TVS wasn't fast enough to catch it. But then if that's the case, I am not sure what to do about it... I thought about caps, or a reverse biased zener... but then that's what a TVS essentially is.

Any ideas or advice?

 

[kabiru]

Well, since you are using buck converter i expect the output voltage to be more stable,and what if the current you are consuming is higher than that of the regulator, the regulator can also burn.

 

[dick_freebird]

How do you know it's a voltage spike? Simple conjecture?
Secondhand conjecture?

Maybe it's inrush current and you need to work on your
soft start. Especially if it doesn't happen with no output
load & no excess filter caps connected to the converter.

Maybe you need to hook a 'scope to the input and
see what the real deal is as you connect the converter.
Rather than believing what a vendor tells you, who really
just wants to avoid taking a field return and doing FA
on it. Or maybe two probes, on either side of a 0.1 ohm
sense resistor, at VIN.

For automotive, you want way more than 18V abs max
rating.

 

[mtwieg]

A high voltage spike on the input doesn't sound likely, unless you've completely neglected to use bypass capacitors on the input. Does it happen if you lower the supply voltage, connect it, then ramp the supply slowly back up to 12V?

 

[Corporate666]

Thanks for all the replies. First off, obviously I am not an engineer by education, so I am learning as I go, but very happy to learn, even (and especially) from my mistakes

As for how I know - well, I was having some issues with another boards last year with Alpha Omega switcher regulator failures, so I contacted them and they did a failure analysis on a few of the regulators, and said the internal FETs had failed. My boards worked fine for their engineer, so we focused on what was different and zoomed in on the power supply. Putting a scope on the input showed a voltage spike. Adding a TVS to the board solved the problem.

This time around, a TVS was already on the board. I put a scope on the input to the switcher and I see a 22 volt spike that lasts around 20us when the power supply is connected. The regulator has a Vin Max of 18V, so I assumed this was what was causing the problem. What I am unsure about is that I have a TVS with a 17.5V breakdown voltage that is connected between Vin and ground right where power enters the board. So I am not sure why I would be seeing a 22V spike in the first place. It's a bidirectional TVS. I checked the obvious, it starts to conduct at 17.45V in one direction, 17.6V in the other.

I realize that a TVS alone is not sufficient protection for connecting this circuit up for automotive use, but I am stuck as to why the regulators are failing in testing, and if it's the voltage spike, I am not sure why I am even seeing that spike with the TVS there.

I haven't put a resistor in line with the power input and checked the inrush current, I will do that next. The engineer at AO Semi said this particular switcher has internal soft start, and that inrush current should not be an issue. I have also adjusted the code in my microcontroller to delay when the LED's are turned on, and the failure seems to happen whether there is any load on the switcher or not.

 

[Orson Cart]

If you apply a step impulse to any LC ckt, you can get a ring of up to 2x the input volts if the ckt is lightly damped, this may be what is happening here (if there is some non zero L in the connecting leads etc, and you have caps on the input to your buck regulator), or your power source may be doing a wee overshoot when it reacts to the switch on.
Either way you need an active clamp to stop the over-shoot, a 5W, 16V zener across your bypass caps should be fast enough and of low enough dynamic impedance to clamp this overshoot, making the bypass caps bigger will lower the rate of rise of the overshoot and allow the zener more time to work.
Hope this is of some help, Regards, Orson Cart.

 

[sreepss]

Use a simple buck convertr like LM2576T, In ur design , which type of inductor is used in the output??? check its current rate.

 

[enjunear]

You might investigate putting a small-valued series resistor right at the power input to the board, before the TVS. When the voltage on the supply is a 22V, I'm expecting that it has a lot of charge capacity, so it can hold that >18V level for a significant period of time (20us, in your case). The TVS is simply a diode (well, two) that go into avalanche breakdown, shunting the excess charge to ground. When it is conducting, it will be pulling a lot of current from the supply to discharge the excess energy, and thus reduce the voltage. A small resistor will take that large instantaneous current and turn it into a moderate instantaneous voltage drop across the resistor (single ohms, maybe smaller). You'd have to play with the resistor size and value, such that it helps reduce the peak input voltage seen at the regulator during connection of power, and small enough that it doesn't dissipate a significant amount of power when your system is operating normally.

I had a similar setup with a start-up regulator on a 28V line that could see transients of 70V, and the part was only rated to 55V input, abs max.

 

[mtwieg]

Thanks for all the replies. First off, obviously I am not an engineer by education, so I am learning as I go, but very happy to learn, even (and especially) from my mistakes

As for how I know - well, I was having some issues with another boards last year with Alpha Omega switcher regulator failures, so I contacted them and they did a failure analysis on a few of the regulators, and said the internal FETs had failed. My boards worked fine for their engineer, so we focused on what was different and zoomed in on the power supply. Putting a scope on the input showed a voltage spike. Adding a TVS to the board solved the problem.

This time around, a TVS was already on the board. I put a scope on the input to the switcher and I see a 22 volt spike that lasts around 20us when the power supply is connected. The regulator has a Vin Max of 18V, so I assumed this was what was causing the problem. What I am unsure about is that I have a TVS with a 17.5V breakdown voltage that is connected between Vin and ground right where power enters the board. So I am not sure why I would be seeing a 22V spike in the first place. It's a bidirectional TVS. I checked the obvious, it starts to conduct at 17.45V in one direction, 17.6V in the other.

I realize that a TVS alone is not sufficient protection for connecting this circuit up for automotive use, but I am stuck as to why the regulators are failing in testing, and if it's the voltage spike, I am not sure why I am even seeing that spike with the TVS there.

I haven't put a resistor in line with the power input and checked the inrush current, I will do that next. The engineer at AO Semi said this particular switcher has internal soft start, and that inrush current should not be an issue. I have also adjusted the code in my microcontroller to delay when the LED's are turned on, and the failure seems to happen whether there is any load on the switcher or not.

So wait, are you saying that you think the spike is killing the control IC itself (through its supply voltage) or just the FET? If it's through the supply voltage, then solution is easy; just supply the chip itself through maybe a 10 ohm resistor and then a good bypass cap. That will damp any spikes without creating losses in the actual power path.

If it's just the FET that is dying, then you really need to dampen the transient or clamp it better. Keep in mind that not all TVS diodes are equal. Some won't to clamp well until they're well above their reverse standoff voltage. For example, some TVS devices are specified for protection against lightning, and may clamp hard at 25V even though they are rated at 18V. Try several devices from different manufacturers and you will probably find some that work.

 

[Orson Cart]

10 ohms is a bit large given your power requirement, pretty hard to beat a 15/16V 5W zener for speed or clamping ability for this app (can put a suitably large diode in inverse series if you don't want the zener to conduct in its diode direction).
Regards, Orson Cart.

 

[mtwieg]

10 ohms is a bit large given your power requirement
Regards, Orson Cart.

You misread my post, I meant putting the resistor in the path leading to the supply pin of the buck IC, not the entire supply input. That would filter out transients seen by the supply pin without causing losses in the converter.

 

[Orson Cart]

Surely the chip you mention - AOZ1021 - has only one Vcc connection? not a seperate one for power and another for the control part, or are you considering another IC entirely?
Regards, Orson Cart.

 

[mtwieg]

Surely the chip you mention - AOZ1021 - has only one Vcc connection? not a seperate one for power and another for the control part, or are you considering another IC entirely?
Regards, Orson Cart.

Oh wow your right, I wasn't aware the the chip supply and the FET input were shared by one pin. I've never seen that in a buck IC before. If that's the case, I would ditch the chip entirely; that's just really, really poor design.

 

[Orson Cart]

Well, quite few low power IC's are like it, just means a good clamp needed at the Vcc pin.
Regards, Orson Cart.

 

[Corporate666]

So I have done some more testing, and have been mulling this over quite a bit, and doing some searching online for solutions.

It seems there are many solutions out there, and some (expensive) IC's which perform this function.

It also seems there are simple solutions using transistors or FET's to disconnect the load from the supply above a clamp voltage.

Looking at the following:

**broken link removed**

I was thinking of implementing the 2nd diagram (just over halfway down the page). This uses a simple NPN transistor and a zener w/resistor to set a voltage at which the base on the NPN will be low enough to turn off the transistor.

I realize it's not as efficient as a FET solution, but it is cheaper and (more importantly) uses a lot less board real estate. The maximum current draw will be 1.5 amps in my design, but normal draw will be 0.5A about 99% of the time. A power NPN transistor would have to dissipate 0.9W with a forward drop of 0.6V at the full 1.5A of current. A DPAK diode rated for 3-4W and adequately sinked seems like it would be more than adequate.

Any comments on this solution?

I was also going to leave the TVS in the equation as well as move to an alternate version of the AOZ regulator which has a max Vin of 27.


Only one thing I am unsure of... in the diagram on the Maxim website, it appears they have a zener on the base setting the clamp voltage... and another on the output of the protection circuit. I am unsure what this second diode is for, other than simple reverse hookup protection, but I am unsure why it would be on the emitter side of that NPN. Maybe I am missing something obvious?

Thanks again for all of the help, it has been very enlightening.

 

[Orson Cart]

it appears they have a zener on the base setting the clamp voltage... and another on the output of the protection circuit. I am unsure what this second diode is for, other than simple reverse hookup protection, but I am unsure why it would be on the emitter side of that NPN. Maybe I am missing something obvious?

It is a schottky diode to stop reverse current in the xtor.

I see you are looking at a pre-regulator here, a switcher with a higher Vin max may may be a better long term solution.

Regards, Orson Cart.

 

[Corporate666]

Ahh, that makes sense on why the diode is there. Thank you for your replies!

As for the circuit, when I was looking into the issue, it seems that for automotive use, the circuit must be able to handle voltages of 80V or sometimes higher, for up to several hundred milliseconds. There is also the case of a prolonger overvoltage (a few seconds). The Maxim link I posted shows various tests that automotive circuits must withstand. It seems that many people/companies recommend something which will disconnect the load from the power when a voltage threshold is exceeded.

I can certainly look at switchers which are "automotive rated" and claim to be able to withstand automotive transients, but the specs for automotive still seem to call for resistance to sustained overvoltages, and even a switcher which claims to be able to handle transients seems like it would still need additional protection from sustained overvoltage.

Thoughts?

 

[Orson Cart]

LM5085 75V Constant On-Time PFET Buck Switching Controller., www.national.com/ds/LM/LM5085.pdf
Regards, Orson Cart.

 

[Corporate666]

Thanks for the information on that regulator, Orson. I was looking only at regulators which had a built-in FET switch, and I realized when I saw your post I was ignoring a whole group of suitable products. I think I will switch the design over to the LM5085 and see how I do with it.

Thank you for all your help.

Cheers!

 

[FvM]

I think the basic point is to design a higher overvoltage margin. If you use TVS diodes, you have to look for the clamp rather than the breakdown voltage. A TVS that safely limits the voltage to 18V will have e.g. 13V breakdown voltage. Although there may be differences between types, you can expect a similar ratio for all TVS and zener diodes.