op-amp locks up on powerup

[JLavoie]

Some of you might remember my little mixer project here http://www.groupdiy.com/index.php?topic=42359.0

I'm having a wierd problem that I'm sure is a rookie mistake that there's a simple fix for.
on power up sometimes (expecially if there is a signal present) the op-amp for the mic pre rails low and locks up.
it can be broken free by a burst of signal, but other than that it's stuck for good.

any ideas?
I've attached a picture since the picture links from my original thread don't seem to work anymore.
the circuit is pretty much the same as it was except R9 & R19 are now 200k (to boost the mic gain to 50dB, and R25 is now 10k

Any help is appreciated
thanks!

Jason

 

[JLavoie]

as a bit of an update, it seems that even shorting the input (during powerup) causes it to lock up. so not so much signal as just something loading the input on powerup does it.
also, when this happens the other opamp on the same chip is still working fine.. so it's definitely not a power supply rail issue.

it's got me stumped

 

[pstamler]

Not sure where the latch-up is coming from, but I see a couple of other problems.

First off, this thing is going to be way noisy. R11 and R12 will be in series with the microphone's own source resistance, so the impedance on each leg will be about 1069 ohms. That will generate noise - Johnson noise and current noise -- way excess of what the microphone at 150 ohms will generate. That'll give you an equivalent input noise of about -117.4dBu. For reference, your average cheap mixer has about -127dBu, and the microphone itself has about -130.8dBu. So this preamp will add about 13dB of noise to the mic's own noise. Also, the common-mode rejection will depend on the match between the four resistors comprising the differential circuit. If these aren't carefully selected, rejection will suffer.

Also, if you're planning to use phantom power, the input capacitors C9 and C10 are facing the wrong way.

Also, the 22k input resistors on the second stage (R6 and R7) will load down the preceding pots, playing hob with their tapers.

Peace,
Paul

 

[PRR]

While locked-up, MEASURE voltages around the stuck amp, an un-stuck amp, also Vref.

Since there's no fault in the plan matching the symptom, the problem is in the build. Input caps backward (or Phantom power present) is a possibility.

 

[JLavoie]

Hmm. you may be onto something..
an un-stuck amp hovers around Vref (3.6V) on all 3 pins. the + input is a bit lower than the rest (about 100mv less)
on a stuck amp ALL pins are low (~0.8V) including the + input
while locked, Vref is still holding at 3.6, so that means the + input is being held low inside the chip. is that even possible?

so something happens on power up that causes the opamp to fail internally. but only one opamp on the chip.. the only difference between the two opamps power-wise is that the reference input of each is pulled to Vref with a different value.

Jason

 

[Walrus]

I'm not good at circuit analysis, but the Vref circuit means your ref voltage is 250mS behind your rail voltage. Hence the IC's +ve input will be at 0v for 250mS, making it's output 0v too  (I think).  Try changing R23/24 to 2.2kOhm as this will lower the time taken for Vref to 25mS.

Warning: This could be complete rubbish.  ;D

 

[JohnRoberts]

+1 to PRR's advice... With opamps the pin voltages will usually lead you to finding the problem.

From looking at the schematic,,, the + input of the mic pre opamp should be sitting at v/2 (+6V?), even with no opamp plugged in... If the + pin voltage is not v/2, figure out why? (Bad solder connection, leaky input cap, connected to some voltage? , whatever).

The opamp output will be pushing or pulling the - input trying to get the - input pin close to the + pin.

Simple opamp math... if the + input is more than several mV higher than the - input the output will be pegged at + rail. If the + input is lower than the - input the output will be pegged low. For the opamp output to be at some mid supply voltage the inputs will be close to each other.

=====
If vref is only 3.6v and + supply is 12V that is wrong figure out why.
If + input is 100mV different from - input that is wrong... figure out why,

Here is where Ohms and Kirkoff laws are useful for troubleshooting...

Measure the voltage on both ends of R19. Then using ohms law, you can figure the current flowing in R19. since everything else connected to the + input is cap coupled, if you measure current flowing through R19 it has to be going somewhere. There is always a possibility that some caps are leaking. measure for any voltage across R13 to identify current flowing in that path. For cap leakage, there generally needs to be voltage across the cap before there is leakage.

Mesure everything and the data will be your friend.  Note: a low impedance meter, or capacitance associated with the probe, could alter the circuit while probing, so sometimes measurements can be a little obscure.

It's always something.

JR

PS.. the symptom sounds like open solder joint on R19 if all parts are good and I am forced to guess, but with simple voltmeter we don't have to guess...

 

[JLavoie]

Thanks for the help.

the reason for the low Vref is that V+ is not 12V. I have a 100 ohm resistor on the power input that helped reduce noise greatly. it didn't have any adverse effects so I left it that way even though it reduced V+ by a few volts.

when stuck, there is 2.8V across R19, which is not a lot of current, and I would go looking for what's sinking it, but it really does seem like the current is going into the chip.
if we had a leaky cap would it not be like that all the time?
and if it were a bad solder connection I wouldn't be seeing it on multiple units (some hand built, some wave soldered).

I've quadruple checked the board layout vs the schematic, and the boards passed an electrical test.
the only thing I haven't been able to verify are the values of the 100pf caps, but even a wrong value there shouldn't cause these symptoms.

the - input is higher than the + input simply because the output is at the rail and can't go any lower.

if it wasn't for the fact that it does power up properly some of the time, I would be hunting for a short somewhere, but it just doesn't add up.
I guess I might try removing R13 and see if it behaves.

Jason

 

[JohnRoberts]

Thanks for the help.

the reason for the low Vref is that V+ is not 12V. I have a 100 ohm resistor on the power input that helped reduce noise greatly. it didn't have any adverse effects so I left it that way even though it reduced V+ by a few volts.

when stuck, there is 2.8V across R19, which is not a lot of current, and I would go looking for what's sinking it, but it really does seem like the current is going into the chip.
if we had a leaky cap would it not be like that all the time?
and if it were a bad solder connection I wouldn't be seeing it on multiple units (some hand built, some wave soldered).

Except there should be 0V across R19...  That 50+ uAmps is going somewhere.

If it is going through R13 that would be 50 mV.

When working properly the opamp input would not be sinking that much current.

Is the opamp + supply pin getting good solid voltage supply?

I've quadruple checked the board layout vs the schematic, and the boards passed an electrical test.
the only thing I haven't been able to verify are the values of the 100pf caps, but even a wrong value there shouldn't cause these symptoms.

the - input is higher than the + input simply because the output is at the rail and can't go any lower.

if it wasn't for the fact that it does power up properly some of the time, I would be hunting for a short somewhere, but it just doesn't add up.
I guess I might try removing R13 and see if it behaves.

Jason

I would suggest making more measurements first...  Kirkoff teaches us the currents going into a node must equal the currents leaving a node. R19 is the path for 50 uA in... you can tell if any is flowing out through r13 with a simple voltmeter measurement.  If it isn't leaving via r13 the only choices left are a bad c22, or opamp fault. Either a bad opamp, or missing PS rail voltage to the opamp, so it's input is acting funny.

JR

 

[abbey road d enfer]

Thanks for the help.

the reason for the low Vref is that V+ is not 12V. I have a 100 ohm resistor on the power input that helped reduce noise greatly. it didn't have any adverse effects so I left it that way even though it reduced V+ by a few volts.

when stuck, there is 2.8V across R19, which is not a lot of current, and I would go looking for what's sinking it, but it really does seem like the current is going into the chip.
if we had a leaky cap would it not be like that all the time?
and if it were a bad solder connection I wouldn't be seeing it on multiple units (some hand built, some wave soldered).

I've quadruple checked the board layout vs the schematic, and the boards passed an electrical test.
the only thing I haven't been able to verify are the values of the 100pf caps, but even a wrong value there shouldn't cause these symptoms.

the - input is higher than the + input simply because the output is at the rail and can't go any lower.

if it wasn't for the fact that it does power up properly some of the time, I would be hunting for a short somewhere, but it just doesn't add up.
I guess I might try removing R13 and see if it behaves.

Jason

Have you tried replacing the 4562 with a 5532 or even a TL072?

 

[ricardo]

Firstly, I don't see any decoupling on your '12V' rail.  You need several 50u and these should be situated near each OPA.[1]

Secondly, several OPAs have nasty locking-up and signal reversal behaviour when signal levels exceed the rails.  TL07x and the NS equivalent are the most notorious.  Even with normal +/- 15V supplies, on turn on, TL07x latch momentarily.  NE5532 dun have this problem.

I have no experience with LM4562 but if replacing them with NE5532 solves the problem but TL072 doesn't, they have this problem.

But the "12V" decoupling is a must.  Try using a 0-6-12V battery supply and see if the problem goes away.

[1] Another issue is you need to keep supply (12V decoupling caps) and "signal / feedback/Ref decoupling" earths separate but that's beyond the scope of this discussion.

 

[JLavoie]

Except there should be 0V across R19...  That 50+ uAmps is going somewhere.

If it is going through R13 that would be 50 mV.

When working properly the opamp input would not be sinking that much current.

R13 measures 0.000V across it. current is going into the chip..

Either a bad opamp, or missing PS rail voltage to the opamp, so it's input is acting funny.

the other half of the chip is working fine, PS rail voltage is within spec.
checked input voltages at powerup with a storage scope and nothing exceeds V+

Jason

 

[JLavoie]

Firstly, I don't see any decoupling on your '12V' rail.  You need several 50u and these should be situated near each OPA.[1]

oops. that's an old version of the schematic, with some stuff left out for simplicity..
the incoming 12V actually goes through a diode, then a 100 ohm resistor (may lower this) then there are two 10u and 100u decoupling caps. plus .1uf caps from Gnd to Vref and Vref to +V (which isn't 12V anymore as explained above because of the resistor)

the chip in question happens to be the closest to the supply filter caps, and is literally on the opposite side of the PCB. traces less than 1/4" from filter caps to the chip.

Secondly, several OPAs have nasty locking-up and signal reversal behaviour when signal levels exceed the rails.  TL07x and the NS equivalent are the most notorious.  Even with normal +/- 15V supplies, on turn on, TL07x latch momentarily.  NE5532 dun have this problem.

I have no experience with LM4562 but if replacing them with NE5532 solves the problem but TL072 doesn't, they have this problem.

Hmm. NE5532 may be borderline for my low operating voltage, but I'll have to try it if I can't find a way to keep the 4562 happy.

But the "12V" decoupling is a must.  Try using a 0-6-12V battery supply and see if the problem goes away.

Sounds like a worthwhile test. not quite sure how I could turn on both batteries simultaneously enough to keep the chip happy though.

All great ideas, and food for thought. still no magic bullet though. somehow this chip is locking up seemingly without the input pins going out of bounds..

Jason

 

[JohnRoberts]

Except there should be 0V across R19...  That 50+ uAmps is going somewhere.

If it is going through R13 that would be 50 mV.

When working properly the opamp input would not be sinking that much current.

R13 measures 0.000V across it. current is going into the chip..

OK, back to earlier questions.. what are the opamp input pin voltages? you tell me Power + and - are good, and other half of opamp is OK...

there may be some obscure latchup mode, but first lets prove the input voltages are in invalid states.  It still could be something else.

Either a bad opamp, or missing PS rail voltage to the opamp, so it's input is acting funny.

the other half of the chip is working fine, PS rail voltage is within spec.
checked input voltages at powerup with a storage scope and nothing exceeds V+

Jason

If opamp input pins don't agree with output state, the opamp is bad, or at least misbehaving.

What are output pin, - input pin, and + input pin voltages?

It's always something...

JR

 

[ricardo]

oops. that's an old version of the schematic, with some stuff left out for simplicity..
the incoming 12V actually goes through a diode, then a 100 ohm resistor (may lower this) then there are two 10u and 100u decoupling caps. plus .1uf caps from Gnd to Vref and Vref to +V (which isn't 12V anymore as explained above because of the resistor)

Can you post an exact circuit?  Including the 100R etc.

 

[JLavoie]

What are output pin, - input pin, and + input pin voltages?

output pin is 0.826V, - input is 0.960V, and + input is 0.934V

Keep in mind that 0V is the -ve rail

Jason

 

[JLavoie]

UPDATE:

if I reduce R19 to 50k (a value I previously had in there in a prototype run) it seems to be stable..

so now I guess the new questions are:
does that help explain the problem?
does that mean that if I leave it that way it'll be stable forever? or have I just made it less likely to fail?
and more importantly, in the short term, can I adjust R9, R12, and R13 to still give me ~50db of gain with this new value of R19 without messing with the input impedance too much?

Jason

 

[JohnRoberts]

What are output pin, - input pin, and + input pin voltages?

output pin is 0.826V, - input is 0.960V, and + input is 0.934V

Keep in mind that 0V is the -ve rail

Jason

OK, those voltages are actually valid... the - input is 25 mV higher than + input so telling the output to peg full down.

So question remains where is the current from R19 going?  A faulty c22 (not likely but worth inspecting).

If the IC is in a socket I would pull it to confirm the drop across R19 goes away.

From the data sheet they specify input CM range from (V+)-2 to (V-)+2, with typical closer to 1V from either rail... So that input condition is not in their valid CM range. Some opamps get stupid when inputs are driven outside their nominal operating range, but I'm not ready to ASSume that is the case here yet... In fact the absolute input range is given as (v-)-.7 to (V+)+.7, so a full diode drop below ground and above + rail... So we're OK there.

The conundrum here is that the opamp pin voltages actually look like the opamp may be OK, but it wouldn't be normal "and" sucking 50 uA into it's + input. So the mystery is where is that current going? Is it really there?

I would repeat the measurements, perhaps make more.  With power off, confirm that you measure 50k from opamp + input to Vref, and some larger than 50k resistance from + input to ground (60k?).  If the 50k resistor is open circuit, or some alternate path to ground is messing with the input bias that could be the issue. .

Old school troubleshooting is to add another resistor (say a few k ) from + input to vref or maybe to + supply to see what happens.

If you don't find a smoking gun with measurements, pull the opamp, and confirm the R19 biases up to Vref with no current draw.

I have never worked with 4562, but that behavior sounds too bizarre to suspect it right away.

Quasi random behavior like that can be caused by open circuits (bad solder)  too...

JR

 

[JohnRoberts]

UPDATE:

if I reduce R19 to 50k (a value I previously had in there in a prototype run) it seems to be stable..

I thought it already was 50k

so now I guess the new questions are:
does that help explain the problem?

If there was a bad solder connection, and the former resistor wasn't really making a good circuit, the opamp could drift up or down...  With old school opamps they would always drift one way or the other, but new modern opamps cancel input bias current, so the direction of what is left is random.

does that mean that if I leave it that way it'll be stable forever? or have I just made it less likely to fail?
and more importantly, in the short term, can I adjust R9, R12, and R13 to still give me ~50db of gain with this new value of R19 without messing with the input impedance too much?

Jason

R 19 needs to be same value as R9

The value of the resistor there should not be making the difference you see... I suspect the former resistor was not making a good connection, or very unlikely a bad (open circuit) resistor.

JR

 

[JLavoie]

I thought it already was 50k

no, the schematic I posted was old, and I noted in the first post that R9 and R19 had been changed to 200k

The value of the resistor there should not be making the difference you see... I suspect the former resistor was not making a good connection, or very unlikely a bad (open circuit) resistor.

but we've been able to reproduce the problem on a hand-bult prototype that I soldered myself, AND several units pulled at random from a batch that was wave soldered..

will try your suggested measurements above and see if there's any explanation.

Jason