Voltage regulator instability

[suputnic]

I have attached a LM1086CT-3.3 voltage regulator to a 12V 500mA
unregulated wallwart. It is attached very near the wallwart. Then there
is a few metres of coiled up twin core wire after the regulator. The
load initially drags about 500mA then settles at about 20mA. I did not
bother to add any caps as the load is quite voltage tolerant (it used
to be batteries). But this one output about 8V one time I tried it.
Lucky the device (gas water heater) could handle it. Every other time
it output 3.3V. Actually it drops to about 3.1 volts briefly with the
initial 500mA but that's OK. I was counting on any instability due to
the lack of caps to be in the order of a few tenths of a volt. Is this
wild instability normal? If it was a precision device it would be toast.

 

[Johnny Boy]

I have attached a LM1086CT-3.3 voltage regulator to a 12V 500mA
unregulated wallwart. It is attached very near the wallwart. Then there
is a few metres of coiled up twin core wire after the regulator. The
load initially drags about 500mA then settles at about 20mA. I did not
bother to add any caps as the load is quite voltage tolerant (it used
to be batteries). But this one output about 8V one time I tried it.
Lucky the device (gas water heater) could handle it. Every other time
it output 3.3V. Actually it drops to about 3.1 volts briefly with the
initial 500mA but that's OK. I was counting on any instability due to
the lack of caps to be in the order of a few tenths of a volt. Is this
wild instability normal? If it was a precision device it would be toast.

Many regulators need the recommended capacitors, especially on the output,
to prevent oscillation. A couple of years ago, I set up a 5V regulator and
had an output approaching 10V, due to oscillations, purely from forgetting a
10uF cap on the output. I would suggest that you add capacitors to the input
and output as suggested in the datasheet and see if you still have problems.
.... Johnny

 

[John Popelish]

I have attached a LM1086CT-3.3 voltage regulator to a 12V 500mA
unregulated wallwart. It is attached very near the wallwart. Then there
is a few metres of coiled up twin core wire after the regulator. The
load initially drags about 500mA then settles at about 20mA. I did not
bother to add any caps as the load is quite voltage tolerant (it used
to be batteries). But this one output about 8V one time I tried it.
Lucky the device (gas water heater) could handle it. Every other time
it output 3.3V. Actually it drops to about 3.1 volts briefly with the
initial 500mA but that's OK. I was counting on any instability due to
the lack of caps to be in the order of a few tenths of a volt. Is this
wild instability normal? If it was a precision device it would be toast.

Low drop out type regulators, like this are inherently unstable,
unless they are compensated with a correct combination of output
capacitance and resistance, and with input capacitance. I would place
at least the recommended 10 uF capacitor at the input and a 10 uF in
series with .1 to 1 ohm in series at the output, to improve the
stability.

 

[Eeyore]

I have attached a LM1086CT-3.3 voltage regulator to a 12V 500mA
unregulated wallwart. It is attached very near the wallwart. Then there
is a few metres of coiled up twin core wire after the regulator. The
load initially drags about 500mA then settles at about 20mA. I did not
bother to add any caps as the load is quite voltage tolerant (it used
to be batteries). But this one output about 8V one time I tried it.
Lucky the device (gas water heater) could handle it. Every other time
it output 3.3V. Actually it drops to about 3.1 volts briefly with the
initial 500mA but that's OK. I was counting on any instability due to
the lack of caps to be in the order of a few tenths of a volt. Is this
wild instability normal? If it was a precision device it would be toast.

Add the caps !

Graham

 

[colin]

I have attached a LM1086CT-3.3 voltage regulator to a 12V 500mA
unregulated wallwart. It is attached very near the wallwart. Then there
is a few metres of coiled up twin core wire after the regulator. The
load initially drags about 500mA then settles at about 20mA. I did not
bother to add any caps as the load is quite voltage tolerant (it used
to be batteries). But this one output about 8V one time I tried it.
Lucky the device (gas water heater) could handle it. Every other time
it output 3.3V. Actually it drops to about 3.1 volts briefly with the
initial 500mA but that's OK. I was counting on any instability due to
the lack of caps to be in the order of a few tenths of a volt. Is this
wild instability normal? If it was a precision device it would be toast.

IT NEEDS THOSE CAPS MAN!

 

[Michael Black]

IT NEEDS THOSE CAPS MAN!

I learned that over thirty years ago, when I got my first three
terminal regulator. What a great oscillator it made. I've never
used one since without capacitors.

Michael

 

[Leon]

I learned that over thirty years ago, when I got my first three
terminal regulator. What a great oscillator it made. I've never
used one since without capacitors.

When 5 1/4" floppy disk drives first became available I built my own
regulated 5V and 12V supply for one, to interface to my TRS-80. A
friend of mine copied my circuit but left off the capacitors and
wondered why it didn't work. As soon as I stuck a DVM probe on the 12V
output it started working, it had killed the oscillation. I added
capacitors across the regulator pins, which fixed it properly.

Leon

 

[default]

I have attached a LM1086CT-3.3 voltage regulator to a 12V 500mA
unregulated wallwart. It is attached very near the wallwart. Then there
is a few metres of coiled up twin core wire after the regulator. The
load initially drags about 500mA then settles at about 20mA. I did not
bother to add any caps as the load is quite voltage tolerant (it used
to be batteries). But this one output about 8V one time I tried it.
Lucky the device (gas water heater) could handle it. Every other time
it output 3.3V. Actually it drops to about 3.1 volts briefly with the
initial 500mA but that's OK. I was counting on any instability due to
the lack of caps to be in the order of a few tenths of a volt. Is this
wild instability normal? If it was a precision device it would be toast.

Not sure what you are getting at. Are you concerned it got up to 8
volts, or drops to 3.1?

Twin core wire? Like that cheesy 30 gauge stuff that comes with wall
warts? Four meters total?

Have you looked at it with a scope? Not all DC output wall warts
contain filter capacitors.

8 volts out seems impossible and 3.1 entirely likely. I'd guess the 8
volts is a mistake in measuring and 3.1 with 1/2 amp normal if
measured at the load.

 

[Rich Grise]

When 5 1/4" floppy disk drives first became available I built my own
regulated 5V and 12V supply for one, to interface to my TRS-80. A
friend of mine copied my circuit but left off the capacitors and
wondered why it didn't work. As soon as I stuck a DVM probe on the 12V
output it started working, it had killed the oscillation. I added
capacitors across the regulator pins, which fixed it properly.

Some decades ago, I stumbled on a 9" B&W monitor for about ten bucks
at some surplus store. Built a nifty TV typewriter using it, which
was intensely rewarding.

But the monitor needed 12V at about 1.5A. So, I took a 12V, 2A
transformer, diodes, and caps, and used a 7812 plus that PNP pass
transistor trick.

I must have capacitated it right (ceramics right up almost against
the body of the regulator - actually, wrapped around its leads just
below that discontinuity), because the pic was as clean and crisp
as the frequency response of the video chain allowed. ;-)

Cheers!
Rich

 

[Sjouke Burry]

I have attached a LM1086CT-3.3 voltage regulator to a 12V 500mA
unregulated wallwart. It is attached very near the wallwart. Then there
is a few metres of coiled up twin core wire after the regulator. The
load initially drags about 500mA then settles at about 20mA. I did not
bother to add any caps as the load is quite voltage tolerant (it used
to be batteries). But this one output about 8V one time I tried it.
Lucky the device (gas water heater) could handle it. Every other time
it output 3.3V. Actually it drops to about 3.1 volts briefly with the
initial 500mA but that's OK. I was counting on any instability due to
the lack of caps to be in the order of a few tenths of a volt. Is this
wild instability normal? If it was a precision device it would be toast.

Did you really want to built an oscillator?

 

[suputnic]

Thanks for all that information. I just checked today, the voltage is
definitely 8v. I have also used the same make and model of adapter on
another unit, with the same regulator. Today I found out it has cooked
the regulator, and destroyed the water heater by putting 14V or so into
it. The wallwart / adapter is 12V 500mA nominal. It measures close to
16V open. Does the high open circuit reading definitively mean it is an
unregulated adapter? This is what a regulated adapter would do, ie cook
the regulator right? Or could a regulator with no caps added fail so
badly when added to an unregulated adapter?

 

[John Popelish]

Thanks for all that information. I just checked today, the voltage is
definitely 8v. I have also used the same make and model of adapter on
another unit, with the same regulator. Today I found out it has cooked
the regulator, and destroyed the water heater by putting 14V or so into
it. The wallwart / adapter is 12V 500mA nominal. It measures close to
16V open. Does the high open circuit reading definitively mean it is an
unregulated adapter?

Yes. It also means that any linear 3.3 volt regulator has an awful
lot of voltage to burn up and will produce a lot of heat doing it.

This is what a regulated adapter would do, ie cook
the regulator right? Or could a regulator with no caps added fail so
badly when added to an unregulated adapter?

The lack of regulated adapter is not the problem. An unstable
regulator combined with about 10 volts drop is the problem. You
should look for a 5 or 6 volt adapter for your next attempt.

 

[Johnny Boy]

Thanks for all that information. I just checked today, the voltage is
definitely 8v. I have also used the same make and model of adapter on
another unit, with the same regulator. Today I found out it has cooked
the regulator, and destroyed the water heater by putting 14V or so into
it. The wallwart / adapter is 12V 500mA nominal. It measures close to
16V open. Does the high open circuit reading definitively mean it is an
unregulated adapter? This is what a regulated adapter would do, ie cook
the regulator right? Or could a regulator with no caps added fail so
badly when added to an unregulated adapter?

If you have an oscilloscope, you'd find that the 8V is actually an
oscillating output that averages 8V. It's probably swinging from about 3V to
13V and averaging 8V. The recommended caps would have stopped this and kept
the output within spec. As John Popelish mentioned, low-dropout regulators
are very sensitive in this area. A 78xx series reg. is a bit more tolerant,
but I don't think that they come in a 3.3V version.
By the way, are you sure that the original voltage was only 3.3V? That
sounds a mite low.
Using an unregulated adaptor, (transformer w/ rectifier), the ideal
input to this regulator is about 1.5V above the output voltage. (That's
during the input troughs, not peak, average, RMS etc.) Anything else is
burnt off as heat.
Also, if as mentioned by an earlier poster, some adaptors don't have a
filter cap, the troughs would approach 0V. No regulators will work under
these conditions.

.... Johnny

 

[[email protected]]

I have attached a LM1086CT-3.3 voltage regulator to a 12V 500mA
unregulated wallwart. It is attached very near the wallwart. Then there
is a few metres of coiled up twin core wire after the regulator. The
load initially drags about 500mA then settles at about 20mA. I did not
bother to add any caps as the load is quite voltage tolerant (it used
to be batteries). But this one output about 8V one time I tried it.
Lucky the device (gas water heater) could handle it. Every other time
it output 3.3V. Actually it drops to about 3.1 volts briefly with the
initial 500mA but that's OK. I was counting on any instability due to
the lack of caps to be in the order of a few tenths of a volt. Is this
wild instability normal? If it was a precision device it would be toast.

This part is particulary forgiving of a lack of load capacitors, mostly
due to the PMOS pass device, which is an easier design to stabilize.
Also, the internal reference is double buffered in the sense that the
reference itself runs off of another internal voltage regulator.
Though the datasheet indicates it needs a 10uF output cap, even 0.47uF
will work.

http://datasheets.maxim-ic.com/en/ds/MAX1658-MAX1659.pdf

If the 500ma is temporary, exceeding the 350ma limit shouldn't be a
problem.

TI also makes LDOs with PMOS pass devices.

 

[Johnny Boy]

This part is particulary forgiving of a lack of load capacitors, mostly
due to the PMOS pass device, which is an easier design to stabilize.
Also, the internal reference is double buffered in the sense that the
reference itself runs off of another internal voltage regulator.
Though the datasheet indicates it needs a 10uF output cap, even 0.47uF
will work.

http://datasheets.maxim-ic.com/en/ds/MAX1658-MAX1659.pdf

If the 500ma is temporary, exceeding the 350ma limit shouldn't be a
problem.

TI also makes LDOs with PMOS pass devices.

Yeah, 0.47uF might work under some conditions, especially if an input
cap is used. Generally, if the datasheet says 10uF, then use it!
Believe me, the designers of this chip know what they're talking about.
Beware of posters that suggest anything outside the ratings, as it's likely
to cause premature failure, as you've discovered and can destroy more
important circuit elements, as you've also discovered.
Sorry to step on your toes, miso, but that's how it is.
.... Johnny

 

[[email protected]]

Yeah, 0.47uF might work under some conditions, especially if an input
cap is used. Generally, if the datasheet says 10uF, then use it!
Believe me, the designers of this chip know what they're talking about.
Beware of posters that suggest anything outside the ratings, as it's likely
to cause premature failure, as you've discovered and can destroy more
important circuit elements, as you've also discovered.
Sorry to step on your toes, miso, but that's how it is.
... Johnny

Actually, I designed that part. The idiot in applications didn't
believe it was stable with a small output capacitor, even though my
simulations and actual silicon proved otherwise. Besides the fancy
double regulated reference, the part has replica parasitic devices to
give it better real life line rejection. The only negative to the part
is high reference noise, but they wanted low Q current, and low noise
is hard to do with low power.

 

[Johnny Boy]

Actually, I designed that part. The idiot in applications didn't
believe it was stable with a small output capacitor, even though my
simulations and actual silicon proved otherwise. Besides the fancy
double regulated reference, the part has replica parasitic devices to
give it better real life line rejection. The only negative to the part
is high reference noise, but they wanted low Q current, and low noise
is hard to do with low power.

In that case, I'll shut up. My apologies. I'm glad that I wasn't too
rude. My suggestions would normally be relevant, or, at least, I'd like to
think so). Sometimes, we all need a wake-up call. As I said earlier, "the
designers of this chip know what they're talking about". (Talk about putting
your foot in it).
What do you think would have caused the out-of-spec output and eventual
failure of the part? I suspect that there was no capacitor on the input AND
that the O.C. input of 14V was actually higher, and that without smoothing,
it might have exceeded the input voltage spec of the device.
Again, thanks for the (very polite) correction. You've now got me even
more interested in this problem.

.... Johnny (Steve)

 

[[email protected]]

In that case, I'll shut up. My apologies. I'm glad that I wasn't too
rude. My suggestions would normally be relevant, or, at least, I'd like to
think so). Sometimes, we all need a wake-up call. As I said earlier, "the
designers of this chip know what they're talking about". (Talk about putting
your foot in it).
What do you think would have caused the out-of-spec output and eventual
failure of the part? I suspect that there was no capacitor on the input AND
that the O.C. input of 14V was actually higher, and that without smoothing,
it might have exceeded the input voltage spec of the device.
Again, thanks for the (very polite) correction. You've now got me even
more interested in this problem.

... Johnny (Steve)

Actually, you should always obey the datasheet, well, unless you know
better. This was a case where nothing else in the company was as
stable, so dammit, Apps won't skimp on that output cap. The max1658 was
targeted to replace a similar Maxim LDO that used a PNP output device.
That part was notorious for bad line rejection and stability problems,
so the new part was supposed to be user friendly. The component count
is a bit low as there are many intentional parasitic diodes in the
design to balance out active devices in an attempt to get very good
line regulation.

I never did a pure bipolar chip, so I can't comment on the failure
mechanism. My guess would be the output device was no longer in the
SOA.

It would be interesting to know which type of LDO (bipolar or the PMOS
pass) is more reliable. For my own board level designs, I always buy
the PMOS LDOs, generally TI since they are well stocked. Bipolar output
devices can get funky as you approach saturation (i.e. approach the
drop out limit).

The "anti-saturation" electronics in a bipolar design (at least mine)
start to draw base current from the output BJT as they approach
dropout. That is, the circuit senses the saturation level and tries to
cut back the base drive. I never got a warm and fuzzy feeling about
such a design. It seems like you have yet another loop to stabilize.
The PMOS pass device design is much simpler. The only drawback is the
output capacitance of the power fet and the load capacitance form a
capacitive divider, essentially determining the ultimate line rejection
at high frequency. It is predictable, but not as good as bipolar
designs.

 

[suputnic]

Yes. It also means that any linear 3.3 volt regulator has an awful
lot of voltage to burn up and will produce a lot of heat doing it.

It is only drawing a 20mA current steady state, so I thought heat would
not be a problem.

The lack of regulated adapter is not the problem. An unstable
regulator combined with about 10 volts drop is the problem. You
should look for a 5 or 6 volt adapter for your next attempt.

I don't have any adapters of that size. Besides the spec sheet says the
regulator handles up to 30 or so Volts.

 

[suputnic]

I added 2 x 33uf 16V caps to the regulator putting out 8V, and tested
it. It was working OK putting out the right voltage 3.3V with an
oscillation with the initial transient current. I will redo and test
the other adapter, the "heater killer" soon. Thanks for all the input,
the parts I could understand were interesting!

PS to the person who thought the required voltage seemed low,
originally the gas water heater took 2 X 1.5V batteries which power a
12000V spark ignitor.