Problems with TPS715xx regulator?

Hello Folks,

Did anyone have TPS715xx regulators mysteriously blow out? I tested the heck out of the prototype circuits here and all did well but back at a client several TPS71550 (5V version) blew, burned out with a sizzle.

We have a 5V zener up front to knock off a few input volts but even without it these regulators should work fine up to the spec'd 24V. Somehow they don't. Load current is very low, between zero and a few mA. Unfortunately these are LDO which I don't like but I had no choice in this case because we really needed the sub-3uA quiescent.

--
Regards, Joerg 

http://www.analogconsultants.com
Reply to
Joerg
Loading thread data ...

Joerg,

You may be exceed> Hello Folks,

Reply to
Dave

Any input transients taking them out? Serious ground bounce?

Reply to
Jeff L

Hello Dave,

Under normal loads it doesn't. Unfortunately there seems to be no thermal shutdown feature so if it were to go into oscillation it could overload. That's what I am going to check when I arrive there. The caps and all are correct and I verified stability, but who knows. In the past I have found some undocumented pathology modes in (other) LDOs. For example, one went beserk if the source resistance exceeded a certain value. When discussing this with the app engineer and poring over the chip's schematic together I could hear him mumble something like "oh s..t".

--
Regards, Joerg

http://www.analogconsultants.com
Reply to
Joerg

Hello Jeff,

That's one reason why I'll have to fly out. The circuit has a full common ground plane so no worries there. There is a cap at the input and they told me that they used a top of the line bench supply in one case, a battery in the other. It's quite puzzling, at least from the distance.

--
Regards, Joerg

http://www.analogconsultants.com
Reply to
Joerg

I've saw some pretty decent PS's put out a damaging spike. One HP PS that we had (late 90's vintage, digital meters) had a errata bulletin on HP's website on how to modify the PS to avoid a large spike under certain conditions during turn on (ours had the newer serial numbers, so was OK)! One common way to make this happen on some supplies was to turn the supply off, and then fairly quickly back on. Turning it off allowed some internal voltages to drop, while others decayed much more slowly. Turing it back on brought voltages in the circuitry back up in a non-normal sequence, and if fed the wrong way into a error amp, ect, bad things could happen.

That's interesting. Is there anything else connected to the battery? Is anything else connected to the device? was it blown out before the battery was connected?

Were the units working, and then just die *during* use, or do they just one day fail to power up?

Is there any significant capacitance on the output side of the regulator?

78xx style regulators would blow up if current from those caps forced current to the input. A reverse biased diode going from the input to output prevents this problem. This could easily happen if the power wires become shorted somehow when the unit is powered or while being disconnected.

Also, a decent sized TVS slapped on the input side can prevent a lot of problems.

Customer not telling you something? Why doesn't this work when I connect this to the 48V battery bank, or why did this stop working in the car (few hundred V load dumps possible)!?

As for the thermal possibility mention in the other post, there are special crayons available that leave marks that melt or change color once a certain temperature is reached. You may want to mark the regulators with one to see if they are reaching unacceptable temperature levels for some reason.

Reply to
Jeff L

Hello Jeff,

Thanks for the hint. Interesting. Didn't they test the heck out of these before releasing them as a product? HP is supposed to be like Mercedes Benz. My main lab supplies are quite mundane. Russian made, simple, supposedly certified to work from Siberian climate conditions to tropical, potted transformers and all that. I can't read the certs that came with them. Although I bought them from a lab gear dealer in Germany these docs were in Russian. Anyhow, they don't create spikes. But the client's supplies might.

Nothing else connected and they all worked fine here in my lab even though I gave them a good work-out. The batteries are small, like the ones used in dog training collars.

They died pretty much right after turn on, with an audible sizzle. They ran for quite some time in my lab (all of them) at max rated Vbat.

There is about 1.2uF on the output and the data sheet states there should be at least 0.47uF. Very low ESR so lots of margin there. The input cap is also twice the recommended value and low ESR (all ceramic caps). The data sheet explicitly says "any capacitor >0.47uF properly stabilizes this loop". Now I just hope that is correct. The problem with modern chips is they do not release the full schematic of the innards anymore, only a coarse block diagram. So you can't retrace and vet their design.

Tricky in this case. It's one of those designs full of 0402 parts that has to fit into a thimble. Even a mosquito wouldn't be able to find a comfy spot in there.

Nah, long-term client and we are always fully honest with each other. It's meant to be operated on a 24V battery, that's all they need from it.

That's the first thing I am going to do. Thermocouple measurements. Basically it'll be enough to touch the chip. It should not have any noticeable temperature above ambient. BTW, when I ran it here I had the AC turned off because I get sinus problems from it. So it was about 95F yet no thermal issues at all.

--
Regards, Joerg

http://www.analogconsultants.com
Reply to
Joerg

one

Is this device protected from reverse polarity? Is it possible to hook them up backwards?

What about ESD - is someone handling them, and getting a little static into the input? It does not take much sometimes - especially when the device does not draw significant current at times. If it fails when I zap an input / output with a barbeque lighter, I'm usually not happy with the design!

regulator?

output

become

Including all bypass caps on the board in parallel with the output cap?

Actually looking at the data sheet says this has a MOSFET pass device so there will be inherently a diode included, and it mentions this in the "regulator protection" section.

Apparently if you go too low on the ESR, some LDO regulators will oscillate!

It's a big PITA, and with all the new parts released these days, with ever shorter time to market, with all the newbie's (and sometimes seasoned people) that make mistakes in the designs, mistakes get made that make it to market. It's hard to know what's good and bad sometimes.

Tiny - What about a TVS stuck on the power cord / connector?

(few

24V is pushing the limits for those regulators. You said that there is a 5V zener in series - but at near zero current, the diodes leakage might be enough to charge up the input cap to the supply voltage. Fresh batteries might put out a few extra volts, but I highly doubt that the 24V is that critical, as there will be headroom for tolerance in the semi process, likely fairly substantial.

special

certain

see

Reply to
Jeff L

Hello Jeff,

They do have the diodes but reverse polarity isn't really an issue in this case since the battery remains connected.

They work with ESD protection just as I do in my lab. It would take quite a charge to drive a 0.1uF above 24V.

Yes, a 1uF ceramic right next to the regulator plus a couple 0.1uF bypass caps in parallel.

Yep.

This one supposedly doesn't. Others have a rather mushy ESR "range curve" in the data sheet. Those give me the goose pimples since you typically can't cajole them into stating a guaranteed range. At least not in writing.

Well, like that rather hokey reset circuit I came across in an expensive piece of lab equipment. I just shook my head.

Besides board and battery there is pretty much zero space anywhere.

Well, it says 24V and the batteries are really not much above that. The zener is also there to lop off a little dissipation. I made sure that the reverse current at 2V is well under the Iq of the regulator and that

2V drop is more than we need.
--
Regards, Joerg

http://www.analogconsultants.com
Reply to
Joerg

Very strange!

Oscillation or strange startup behavior in the circuit somewhere else would be about the only things I can think of, as everything else seems to be ok. Is it possible that something during startup could draw significant current (possibly caused by the regulator oscillating), causing the regulator to go in current limit mode, and get stuck there from the resulting brown out, or causing the rare possibly of an IC latching?

Also note that unless you have a thermocouple that is made with very fine gauge wire, significant heat will be drawn from the thermocouple leads them selves, thus cooling the thermocouple due to the thermal resistance between it and the device, and cooling the device it's self. I've seen temperatures drop 20 to 30 deg C with 30 gauge wire on much larger devices. The best place to measure a device this small is to solder the smallest thermocouple you can find/make to the ground pin of the IC.

Also, were these units that are failing the ones that you tested in the lab, since if not they may have parts that were not quite up to spec, poor assembly, ect?

Reply to
Jeff L

Hello Jeff,

Indeed.

That could be the case because there is logic on board that becomes rather unhappy when Vo would rise above 6-7V. But it should not oscillate.

All we'd need to see is whether the temperature rises because in this circuit it should remain almost at ambient.

It's the ones, I tested every single board of this prototype run. All there was between here and the destination is a Fedex flight.

--
Regards, Joerg

http://www.analogconsultants.com
Reply to
Joerg

Is it possible that your customer has longer leads to the battery (more inductance)? I had a situation once where the supply lead inductance in conjunction with the very low ESR of the ceramic input cap caused a large voltage overshoot due to turn-on inrush into the cap. I added a small-value series resistance to cure the problem. Also, I have seen some recommendations to use capacitors with higher ESR so as to dampen things.

Good luck.

John

Reply to
John - KD5YI

Hello John,

It was the power supply, just as Jeff had hinted above. Couldn't really believe it because it was a top of the line Lambda. It did not like load changes to the tune of 30dB and reacted with pretty nasty regulation spikes.

--
Regards, Joerg

http://www.analogconsultants.com
Reply to
Joerg

This is something that also can pose a serious threat, but generally needs a large very low ESR capacitor at the end of the cable. I've read an app note a year or two ago about this, mostly about the huge ceramic filter caps that are on the inputs of laptops. This could cause a large transient upon power up. The cure was to use a standard low ESR electrolytic, not the super low ESR ceramic, or add a small series resistance to the ceramic cap.

spikes.

Interesting (I read your other post about this). Shows you can't even trust top notch equipment vendors! I bet the owners / users of the power supply were not too impressed.

Anyway I'm glad you got the problem sorted out.

Jeff

Reply to
Jeff L

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.