Searching \ for '[EE]: op amp latch-up problem' in subject line. ()
Make payments with PayPal - it's fast, free and secure! Help us get a faster server
FAQ page: techref.massmind.org/techref/index.htm?key=amp+latch+problem
Search entire site for: 'op amp latch-up problem'.

Exact match. Not showing close matches.
PICList Thread
'[EE]: op amp latch-up problem'
2006\01\25@004532 by Rob Robson

flavicon
face
I apologize in advance if I've missed something that should be obvious, but this one's making me crazy.  I have an LM1972 uPOT on +/-5V rails feeding a TL072 on +/-12V rails wired as a unity gain buffer as per the LM1972 datasheet.  I've used this circuit for years with a linear power supply without any problems, but when the circuit was ported to a system with a (noisier) switching power supply, the TL072 would latch up after only a few hours of operation .  First problem: I had put the TL072 on +/-5V rails without consulting its spec sheet (too low, as it turns out), so I substituted the more forgiving LF353.  The latch-up condition would still occur after several hours of operation.  I then replaced the 353 and raised the rails to +/-12V as they had been in the functioning legacy system.  No dice: the latch-up still occurs.  Besides power supply noise, the only other difference between the functioning legacy circuit and the faulty new one is the impedance they're each loo!
king into: 2.7k for the former; 22k for the latter.  Does anyone have any ideas about what might be causing the op amp to latch up?

Many thanks,
RR      

2006\01\25@064702 by Gerhard Fiedler

picon face
Rob Robson wrote:

> Besides power supply noise, the only other difference betwee n the
> functioning legacy circuit and the faulty new one is the impedance
> they're each looking into: 2.7k for the former; 22k for the latter. Does
> anyone have any ideas about what might be causing the op amp to latch
> up?

Not really. But possibly you're one step further if you check whether your
circuit works with a 3k resistor from the output(s) to ground (so simulate
the earlier impedance).

What did you change to cause the impedance change?

Gerhard

2006\01\25@082206 by Mark Scoville

flavicon
face
>  king into: 2.7k for the former; 22k for the latter.  Does anyone
> have any ideas about what might be causing the op amp to latch up?

Nope... But, if you suspect the switcher why don't you temporarily use a
linear supply to power the *new* board and see if the latchup problem goes
away. Maybe the problem is somehow layout related?

Gerhards idea of simulating the lower impedance by adding a 3K seems like a
good thing to try as well.

-- Mark



2006\01\25@094139 by Spehro Pefhany

picon face
At 09:44 PM 1/24/2006 -0800, you wrote:
>I apologize in advance if I've missed something that should be obvious,
>but this one's making me crazy.  I have an LM1972 uPOT on +/-5V rails
>feeding a TL072 on +/-12V rails wired as a unity gain buffer as per the
>LM1972 datasheet.  I've used this circuit for years with a linear power
>supply without any problems, but when the circuit was ported to a system
>with a (noisier) switching power supply, the TL072 would latch up after
>only a few hours of operation .  First problem: I had put the TL072 on
>+/-5V rails without consulting its spec sheet (too low, as it turns out),
>so I substituted the more forgiving LF353.  The latch-up condition would
>still occur after several hours of operation.  I then replaced the 353 and
>raised the rails to +/-12V as they had been in the functioning legacy
>system.  No dice: the latch-up still occurs.  Besides power supply noise,
>the only other difference between the functioning legacy circuit and the
>faulty new one is the impedance they're each loo!
>  king into: 2.7k for the former; 22k for the latter.  Does anyone have
> any ideas about what might be causing the op amp to latch up?


Are you sure the op-amp itself is the problem, and not, say, the LM1972?

Usually latch-up problems are caused by exceeding the power supply range or
input voltage range of a part. This causes parasitic devices to trigger, or
puts the part in a mode what is unusual and may self-latch because of
internal or external feedback paths. So, we'd tend to be less suspicious
of a device that is running off the highest supply voltage unless the
input(s) can go outside of the normal range of operation (for example, the
common mode range on an op-amp). If the latter is happening, like banging
your head against the wall, the solution is simply not to do that (eg.
clamp the voltages into the inputs so they stay within permissible limits).

Perhaps you can get a full set of measurements in the failure state and
record the voltages at all relevant nodes (perhaps including some that
don't initially appear to be relevant) and put them on a schematic.
I suspect the real cause will become abundantly clear at that point.

BTW, the TI TL072 advertises that its operation is "latch up free".

Is  it possible that your switching supply is so horrible that it is
exceeding the absolute maximum input voltage or supply voltage ratings
under some conditions? Or maybe the particular system you're checking it
on has a problem such as an intermittent in the feedback loop that causes
the SMPS regulator to go "balls to the wall" intermittently.

Best regards,

Spehro Pefhany --"it's the network..."            "The Journey is the reward"
spam_OUTspeffTakeThisOuTspaminterlog.com             Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog  Info for designers:  http://www.speff.com
->> Inexpensive test equipment & parts http://search.ebay.com/_W0QQsassZspeff


2006\01\25@101534 by Bob Axtell

face picon face
Spehro's got it.

I can count on one finger the number of latchup's I've ever had- and it
was caused
by a faulty ground in a PCB layout.

You will need to rent a FAST scope, fast enough to catch those noise
spikes and
locate 'em. The TL072 is a very rugged chip; to kill a TL072 it will
need to be FAST
and HIGH (at least 5V above the VCC rail). That's what those rental
companies
are for- special tests like this. Use 'em.

Hint: if your design has a switching supply, and you are not using a
ground plane
(PCB of 4 layers or more, with a layer reserved for GND) that's the
reason. Switchers
operate in the Mhz+ region, and if the ground is poor, RF noise will
swamp the other
circuits.

--Bob

Spehro Pefhany wrote:

{Quote hidden}

--
Note: To protect our network,
attachments must be sent to
attachspamKILLspamengineer.cotse.net .
1-520-850-1673 USA/Canada
http://beam.to/azengineer

2006\01\25@120707 by Rob Robson

flavicon
face
Thanks very much for the suggestions.  The load impedance change was caused
by a change in resistor values around the 5532 summing amp that follows the
latching-up LF353.  I've now reduced those values by a factor of 10 and I'm
retesting.  It took about 22 hours for the fault to appear each time, so
it's too early to tell whether this change will fix it.

In researching this problem, I came across an article that claims that the
TL072 is particularly susceptible to latch-up, although I've used them for
years without problem.  I'll provide the link if I can find it again.

This is a 2-layer board with an external switcher, but the bottom layer is
virtually all ground.  Should that suffice?

What would be the simplest way to protect the op amp input from spikes?
Zener to ground?

Thanks again,
RR

{Original Message removed}

2006\01\25@125028 by Bob Axtell

face picon face
Rob Robson wrote:

>Thanks very much for the suggestions.  The load impedance change was caused
>by a change in resistor values around the 5532 summing amp that follows the
>latching-up LF353.  I've now reduced those values by a factor of 10 and I'm
>retesting.  It took about 22 hours for the fault to appear each time, so
>it's too early to tell whether this change will fix it.
>
>In researching this problem, I came across an article that claims that the
>TL072 is particularly susceptible to latch-up, although I've used them for
>years without problem.  I'll provide the link if I can find it again.
>  
>
I'd like to see it. Wow, TL072 is in hundreds of designs, never a failure.

>This is a 2-layer board with an external switcher, but the bottom layer is
>virtually all ground.  Should that suffice?
>  
>
That should be enough.

>What would be the simplest way to protect the op amp input from spikes?
>Zener to ground?
>  
>
Try current-limiting with resistors, if you can, if it won't screw up
the design. 100 ohms
might do the trick.

Zeners are no good for transient suppression. Use Transorbers or other
TVS types.
The good news is that there are TVS in 0603 packages, and they work great!

--Bob

>Thanks again,
>RR
>
>{Original Message removed}

2006\01\25@141735 by Dwayne Reid

flavicon
face
At 10:44 PM 1/24/2006, Rob Robson wrote:
>I have an LM1972 uPOT on +/-5V rails feeding a TL072 on +/-12V rails
>wired as a unity gain buffer as per the LM1972 datasheet.  I've used
>this circuit for years with a linear power supply without any
>problems, but when the circuit was ported to a system with a
>(noisier) switching power supply, the TL072 would latch up after
>only a few hours of operation .  First problem: I had put the TL072
>on +/-5V rails without consulting its spec sheet (too low, as it
>turns out), so I substituted the more forgiving LF353.  The latch-up
>condition would still occur after several hours of operation.

Which latch-up condition do you refer to?  There are two: SCR type
latch-up where the chip draws significant current *or*
phase-inversion latch-up where the output of an op-amp swings hard to
one rail because one or both inputs went out of the common-mode voltage range.

I'm assuming a phase-inversion type latch-up but it would be useful
to have that confirmed.

Like Bob, I've been using the TL07x family for a couple of decades,
in at least a hundred different products (probably way more than 100)
without incident.

The TL07x family does include the positive supply rail within the
common-mode voltage range.  Can you try adding an external negative
supply such that the negative rail is at least 3V lower than the
lowest expected signal?  I'd suggest starting off at -8V or
-9V.  Leave the + supply rail at 5V.  Then let the thing run and see
if the problem occurs.

Are the TL072 inputs exposed directly to the outside world?  What
signal range are you expecting?

dwayne

--
Dwayne Reid   <.....dwaynerKILLspamspam.....planet.eon.net>
Trinity Electronics Systems Ltd    Edmonton, AB, CANADA
(780) 489-3199 voice          (780) 487-6397 fax

Celebrating 21 years of Engineering Innovation (1984 - 2005)
 .-.   .-.   .-.   .-.   .-.   .-.   .-.   .-.   .-.   .-
    `-'   `-'   `-'   `-'   `-'   `-'   `-'   `-'   `-'
Do NOT send unsolicited commercial email to this email address.
This message neither grants consent to receive unsolicited
commercial email nor is intended to solicit commercial email.

2006\01\25@151537 by Rob Robson

flavicon
face
The op amp non-inverting input is coupled to the output of an LM1972 digital
attenuator via a 1uf electrolytic.  This part of the circuit is identical to
one that has been in use flawlessly for years.  The inverting input is tied
directly to the output, which feeds a 5532 summing amp via a 0.1uF ceramic
cap.  This part is different from the old circuit in that the
now-problematic op amp also fed a 5532 buffer amp with a Z in of 2.7k via a
10uF electrolytic.  I realize my LF cut-off through the summing amp is
higher now due to the smaller coupling cap, but this is desirable in this
application.  All traces are short.  The input of the digital pot (also via
a 1uF electrolytic) does go off-board to another board within the unit, but
this was also the case in the working legacy system.  Would an input current
limiting resistor really make a difference on a JFET input op-amp with its
characteristically high Z-in?  Would this unity-gain buffer be more stable
if there was a nonzero resistance in the feedback loop?

I'll try to increase only the negative rail as you've suggested and see what
happens.  The time it takes for the fault to appear (approx 22 hours) is
slowing down the troubleshooting process, but I have several units to play
with.

RR


{Original Message removed}

2006\01\25@153357 by Bob Axtell

face picon face
Rob Robson wrote:

>The op amp non-inverting input is coupled to the output of an LM1972 digital
>attenuator via a 1uf electrolytic.  This part of the circuit is identical to
>one that has been in use flawlessly for years.  The inverting input is tied
>directly to the output, which feeds a 5532 summing amp via a 0.1uF ceramic
>cap.
>
<respectful snip>

>The time it takes for the fault to appear (approx 22 hours) is
>slowing down the troubleshooting process, but I have several units to play
>with.
>
>  
>
Hmmm, 22 hrs for the fault to appear? Methinks this is a big, BIG clue....
I'm suspicious of something wrong externally, like a piece of equipment
that kicks on
when the office gets cold, or an arc in a phase-correction box down the
street, or
maybe cellphone interference...?

"Curiouser and curiouser", said Alice...

--Bob

--
Note: To protect our network,
attachments must be sent to
EraseMEattachspam_OUTspamTakeThisOuTengineer.cotse.net .
1-520-850-1673 USA/Canada
http://beam.to/azengineer

2006\01\25@155907 by Rob Robson

flavicon
face
I'd thought about that, but the odd thing is that it doesn't seem to matter
what time of day the test is started - the op amp just seems to take a long
time to latch up.  I don't think it's thermal: the lid is off, the
components are cool.  Curiouser indeed.  BTW, it does seem to be the
"phase-inversion" type of latch-up, resulting in significant power supply
draw.

RR

{Quote hidden}

> --

2006\01\25@162718 by Ed Browne

flavicon
face
It sounds like you have no dc path to supply bias currents, which would
explain the long time constants:

"The op amp non-inverting input is coupled to the output of an LM1972
digital
attenuator via a 1uf electrolytic." Do you mean it's ac coupled?

"The inverting input is tied directly to the output, which feeds a 5532
summing amp via a 0.1uF ceramic cap.  This part is different from the old
circuit in that the now-problematic op amp also fed a 5532 buffer amp with a
Z in of 2.7k via a 10uF electrolytic. "  Do you mean that the output is also
AC coupled?

If so, then neither the inverting or noninverting inputs has a dc bias path.

2006\01\25@163936 by David Minkler

flavicon
face
Rob,

What's keeping the non-inverting input (of the TL072) from drifting off
into a rail?  I didn't hear anything in your description that sounded
like a DC path.

Dave

Rob Robson wrote:

{Quote hidden}

>{Original Message removed}

2006\01\25@172047 by Rob Robson

flavicon
face
This is the kind of embarrassing "light-bulb moment" I suspected might come
out of this posting.  How I could've overlooked the need for bias is beyond
me.  I'll try it, but the suggestion is already gonging with the ring of
truth.

Sheepish thanks,
RR

>
> What's keeping the non-inverting input (of the TL072) from drifting off
> into a rail?  I didn't hear anything in your description that sounded
> like a DC path.
>
> Dave
>


2006\01\25@173924 by Dwayne Reid

flavicon
face
At 02:44 PM 1/25/2006, David Minkler wrote:
>Rob,
>
>What's keeping the non-inverting input (of the TL072) from drifting off
>into a rail?  I didn't hear anything in your description that sounded
>like a DC path.
>
>Dave
>
>Rob Robson wrote:
>
> >The op amp non-inverting input is coupled to the output of an
> LM1972 digital
> >attenuator via a 1uf electrolytic.  This part of the circuit is
> identical to
> >one that has been in use flawlessly for years.  The inverting input is tied
> >directly to the output, which feeds a 5532 summing amp via a 0.1uF ceramic
> >cap.

I think that Dave nailed it.

Try adding a 1M0 resistor from the non-inverting input (+) input of
the opamp to ground.

dwayne

--
Dwayne Reid   <@spam@dwaynerKILLspamspamplanet.eon.net>
Trinity Electronics Systems Ltd    Edmonton, AB, CANADA
(780) 489-3199 voice          (780) 487-6397 fax

Celebrating 21 years of Engineering Innovation (1984 - 2005)
 .-.   .-.   .-.   .-.   .-.   .-.   .-.   .-.   .-.   .-
    `-'   `-'   `-'   `-'   `-'   `-'   `-'   `-'   `-'
Do NOT send unsolicited commercial email to this email address.
This message neither grants consent to receive unsolicited
commercial email nor is intended to solicit commercial email.

2006\01\26@074639 by Gerhard Fiedler

picon face
Dwayne Reid wrote:

>> What's keeping the non-inverting input (of the TL072) from drifting off
>> into a rail?  I didn't hear anything in your description that sounded
>> like a DC path.
>
> I think that Dave nailed it.

Rob said that this circuit worked for a long time. If I understood this
correctly, it didn't have a DC path either. Or maybe the leakage of the 10u
cap (in the original working circuit, but not in the new circuit) provided
a DC path?

Gerhard

2006\01\26@111140 by Rob Robson

flavicon
face
You're quite right, Gerhard.  The difference turns out to be is that the old
circuit was only used intermittently, while the new application is 24/7.  I
guess the problem was always there, but the symptoms never had time to
appear.  The first unit with Dwayne's fix installed has now been running 14
hours without a latch-up, and I suspect that the problem will not reappear.

BTW, here's the link that mentions that the TL072's relatively narrow CM
range can responsible for latch-up problems.  Because it's more convenient
for me to use rails of +/-5V in this application, I'm now using the LF353.
http://www.dself.dsl.pipex.com/ampins/webbop/072.htm

Thanks to all,
RR

>
> Rob said that this circuit worked for a long time. If I understood this
> correctly, it didn't have a DC path either. Or maybe the leakage of the
> 10u
> cap (in the original working circuit, but not in the new circuit) provided
> a DC path?
>
> Gerhard
>
> --

More... (looser matching)
- Last day of these posts
- In 2006 , 2007 only
- Today
- New search...