op-amp locks up on powerup

[JohnRoberts]

Kirkoff and Ohms law are our friend.

With a simple VOM we can troubleshoot most issues.  But be careful, we can sometimes get hung up on an assumption or sloppy measurement, that leads us down a dead end street... Measure twice, measure multiple ways...

The opamp input pin voltages, look like that opamp is not the bad actor,, I suspect the input bias string is intermittent open circuit.. but that is just my guess from a distance.

FWIW the normal input currents for that opamp are nano amps, (9 zeros) so very small currents.

JR

 

[ricardo]

Can you post the EXACT circuit that misbehaves and put in as many measured voltages as possible.

These should include

Vcc
Vref

pins of misbehaving OPA
pins of working OPAs

You've already mentioned one 'cure'.  Making R19 = 50k.

Was the old circuit with R9/19 = 50k OK?
Was the old circuit w/o the 100R OK?

 

[JohnRoberts]

Can you post the EXACT circuit that misbehaves and put in as many measured voltages as possible.

These should include

Vcc
Vref

pins of misbehaving OPA
pins of working OPAs

You've already mentioned one 'cure'.  Making R19 = 50k.

Was the old circuit with R9/19 = 50k OK?
Was the old circuit w/o the 100R OK?

In case you haven't been following this exchange these questions have been essentially asked and answered.

Not only were power pins OK, but the device is dual and other half working OK.

R19 changed from 200k to 50k, but my suspicion is this is a red herring and the value is not the actual fault vector.

opamp input/output pin measurements reveal that the opamp is in a valid state for those pin voltages.

I suspect the unaccounted for tens of microamps presumably pulled through R19 and into the opamp + input are really not there and a erroneous measurement artifact.  Replacing the R with a different value might have worked just as well as replacing it with the same value, if solder connections or copper voids are present. But I haven't seen a copper void in decades. 

opamps are not rocket science... and that one appears to be behaving in agreement with it's input pin voltages... , it just isn't getting the right input voltage to its  + input pin,,,

of course I could be wrong...  nah

JR

 

[ricardo]

Duu.uuh!  Sorry Guru JR.  My small brain sometimes needs to be shown the (whole) big picture to see stuff.

Here we have

- intermittent (?) problem (I don't think its solder or components cos Jason tells us he can replicate the problem on other samples)
- latches when 'something' connected to mike amp.  Consistently?
- low Vcc
- single supply with ref coming up slowly at a indeterminate rate cos we don't know exact position & values of PSU caps.
- i/p pins out of 'working' CMR when latchup occurs

An accurate circuit with voltages marked is useful to the IQ challenged like me  :-[

From my experience with TL07x and the NS equivalents, this looks like similar latching behaviour.  Their i/p pins also start taking current when latched.  (FET i/p while LM4562 appears BJT)

There has been one 'cure' R19 -> 50k
I'm trying to establish if the original with R9/19 = 50k, 12V w/o 100R was OK

The rate of rise of Ref, Vcc & the i/p pins is a factor in this latching behaviour on TL07x.

I (very) humbly second your guru directives on finding out where currents are going.

Thought it might be worthwhile to check another possible cause.  But my small brain needs to see a clearer picture to attempt this

 

[JLavoie]

There has been one 'cure' R19 -> 50k
I'm trying to establish if the original with R9/19 = 50k, 12V w/o 100R was OK

the original started out with I believe 72k, and no sign of problems. then I sent out a proto run that are in the field working fine where we bumped it to 100k
before the production run I added a bit more gain by going to 200k and then noticed the problem. (somehow the modified prototype never acted up on the bench, and a gain change didn't seem like it needed super thorough testing before sending it off)
after noticing the problem, I dropped back to 100k and found that at this level it works fine as long as the input isn't connected (which might explain why tests and field units have been fine)
I had to reduce it to below 100k to make it stable on power-up both with and without a load on the input..
these are typically left on permanently, so the problem wouldn't show up unless there was a power bump while a mic was plugged in.

The rate of rise of Ref, Vcc & the i/p pins is a factor in this latching behaviour on TL07x.

is quicker rise better or worse?
this circuit has a very slow rise due to the filtering R/C on the input, but in one test I did short the 100R and it didn't seem to help.

Jason

 

[Chris_V]

As ricardo is asking, please post the right schematic.
I was wondering why you put different values for R9/19 and now you are saying that they are the same ...

 

[ricardo]

I dropped back to 100k and found that at this level it works fine as long as the input isn't connected (which might explain why tests and field units have been fine)

Do you mean with R19 = 200k, it sometimes latches even with no source connected?

How do you short the input? XLR p2 to p3 or both to p1?  What source are you using?

As this R19 -> 50k seems to be the only thing that 'works' we need to be clear about what actually happens with the various R19s.

is quicker rise better or worse?

Generally faster is better but the order in which each pin rises is also important.

Try w/o the 100R and removing C15 or making it eg 100n.  If it still latches with R19 = 200k, I'm going to bow out as I need much more info to pontificate at long distance.

And you need R9 = R19 too.  R19 has only a small effect on gain.  R9 has a big effect.

 

[JLavoie]

Do you mean with R19 = 200k, it sometimes latches even with no source connected?

yes. the majority of the time.

How do you short the input? XLR p2 to p3 or both to p1?  What source are you using?

I have a short cable with all three wires twisted together.
other sources that cause it to fail (with R19 at 100k) are an NTI minirator signal generator (whether it's on or off) and a plain SM58

I'm hesitant to post the entire schematic as that would just make it too easy for my competitiors, but tonight I'll post views of the power section and updated values for this preamp.

Jason

 

[JohnRoberts]

I am still not enthusiastic about the latch up scenario, but IF it is latching up intermittently, AND  the value of R19 is the marginal factor, one proof to confirm that fault mode is to make R19 even larger to force it to latch up every time and confirm that sensitivity to R19 value.

I have already suggested other things I would look at also, like network voltages without the opamp in place, but that's like work so how about another set of opamp input and ouput pin voltage measurements when it isn't latched up and working? Does it properly bias up to Vref, or is it biased at some other intermediate voltage? just not pegged all the way down. With the cap coupled output there wouldn't be a very noticeable audible difference if the bias is off slightly.

The opamp input bias current is nominally 10 nA so the drop across R19 should be only mV.

If XLR inputs are shorted to ground, a very leaky input electrolytic cap could also be contributing to the mis-bias, but there should be some evidence of that current through R13, while it could be a very small voltage.

My apologies for being argumentative, but the opamp latching up with the + input sinking current, still seems unlikely to me based on the evidence.  I still suspect some external factor (passive components/connections). Of course it has to be something.

JR

PS: I'm not a guru, I've just been in these kind of situations and fooled by appearances before.

 

[PRR]

> the opamp latching up with the + input sinking current, still seems unlikely

Yeah, but it jams near Vbe. Possibly it does the right thing until it runs out of headroom (actually bottom-room). I'm picturing a PNP input with a current-mirror under. When V c-e gets down to 0.1V (maybe 0.1V above a 0.6V mirror b-e drop), the input device starves for collector-suction, Beta falls way-way off, to unity. 50uA is a very reasonable emitter current.

So we have an amp marketed for bipolar supply being used single-supply. I am the first to say "it doesn't matter" "the chip does not know where ground is!"; but sometimes it does. Also a chip designed for 34V use being worked far lower. Yes, the specs do cover quite low voltage, but still with a bipolar assumption. The "error" is tens of mV... it appears to be "barely latched". Maybe it has just slipped off the corner of the chip's design envelope.

 

[JohnRoberts]

I have even more questions than answers about this, but will stop blathering, until we hear more feedback.

It's always something.

JR

 

[ricardo]

the opamp latching up with the + input sinking current, still seems unlikely

The latching mechanism for TL07x & LF35x is a hidden thyristor incorporating the substrate.  If you look carefully (eg with a storage scope) at how a simple TL07x OPA circuit powers up, you'll see it latch to each power rail in turn for a significant period of time.  By twidding with PSU rise times, the actual amp configuration & input levels you can increase or decrease the latch time.  This is a problem with some LF filter circuits on overload when the latching can be a second or two.  NE553x stabilise almost instantly.

QUAD switched to TLC??? OPAs from the TL07x (used in the 33 & 44 preamps) for their later models for this reason.

There are other mechanisms too.  One 200W Golden Pinnae Amp, if powered up without a load would latch to one rail.  The cure was to short the output.  Caused by badly thought out cascode input stage.

Jason, you said R19 -> 200k was to increase gain.  But R19 has only a small effect on gain.  R9 has a big effect.  What happens with R9=R19=200k?

 

[JohnRoberts]

Jason, you said R19 -> 200k was to increase gain.  But R19 has only a small effect on gain.  R9 has a big effect.  What happens with R9=R19=200k?

I am still resisting arguing about the ghost in the machine without more data.

But I can answer the easy stuff.. as I mentioned before R9 needs to be the same value as R19. So hopefully they already are the same value.

U3 is a simple differential amp, so for
R12=R13 = 1K,
R9 =R19 =  gain/1k  or 200k for 200x gain

To optimize rejection of CM noise they need to be the same value. .

JR

 

[tubologic]

This is a quite simple circuit and from all your previous tests it appears that we can safely exclude the circuit topology,a few passive parts,power supply decoupling,poor soldering, etc... What remains is the OPAMP itself , nothing else. Did you tried to substitute it with another type or make to check if the problem persists ? Don't simply assume that because the other half of the dual Opamp works the I.C can't be faulty. Is your  LM4562 from a reputable source (N.S official distributor) ? The electronic parts market is currently plagued with counterfeit/rebranded and/or sub-standard parts (factory rejects).
I personally wasted considerable time in debugging electronic circuits to discover at last the problem was due to a defective (or sub-standard) NEW part, not the circuit design. If anything else fails I would not disregard this possibility and try to order a few  LM4562's from another source.   

 

[JLavoie]

Wow, lots to get caught up on.. sorry I wasn't able to get back to you guys. I was reading along but have been away from the office and unable to do any testing on the units..

From JohnRoberts:

opamp input and ouput pin voltage measurements when it isn't latched up and working? Does it properly bias up to Vref, or is it biased at some other intermediate voltage? just not pegged all the way down.

when it isn't latched, everything looks good. the + pin floats at Vref (which is half of V+ as it should be) and there is no sign of any significant current through R19

If XLR inputs are shorted to ground, a very leaky input electrolytic cap could also be contributing to the mis-bias, but there should be some evidence of that current through R13, while it could be a very small voltage.

right, but if that is the culprit then that doesn't explain the latch-up when the input isn't shorted (which happens with higher values of R19)

My apologies for being argumentative, but the opamp latching up with the + input sinking current, still seems unlikely to me based on the evidence.  I still suspect some external factor (passive components/connections). Of course it has to be something.

no problem.. I've been arguing with myself, so you're just one more voice
and you're right. the latch-up is indeed possible, but according to National it should only happen under conditions that we're not seeing.

From Ricardo:

Jason, you said R19 -> 200k was to increase gain.  But R19 has only a small effect on gain.  R9 has a big effect.  What happens with R9=R19=200k?

R9 and R19 were increased in sync. but in testing, it seemed that I only had to reduce R19 in order to make the problem go away. naturally I lowered R9 later.
R9=R19=200k is where we noticed the problem and started the diagnosis.

From turbologic:

This is a quite simple circuit and from all your previous tests it appears that we can safely exclude the circuit topology,a few passive parts,power supply decoupling,poor soldering, etc... What remains is the OPAMP itself , nothing else. Did you tried to substitute it with another type or make to check if the problem persists ? Don't simply assume that because the other half of the dual Opamp works the I.C can' be faulty. Is your  LM4562 from a reputable source (N.S official distributor) ? The electronic parts market is currently plagued with counterfeit/rebranded and/or sub-standard parts (factory rejects).
I personally wasted considerable time in debugging electronic circuits to discover at last the problem was due to a defective (or sub-standard) NEW part, not the circuit design. If anything else fails I would not disregard this possibility and try to order a few  LM4562's from another source. 

as mentioned, we're seeing this on production units as well as a prototype that I pulled out.
our parts came from Digikey (who I'm now noticing is not listed on National's distributor list, but would assume they would be unlikely to be selling counterfeit parts)
I believe there is a chance the builder also ordered from Digikey.


so I guess I'm looking for suggestions for a test op-amp that has similar specs to the LM4562 (similar noise specs, and supply down to+/- 3v?)
at the same time I'll order some more LM4562's from a different source just to be sure.

Jason

 

[JohnRoberts]

I agree an opamp substitution seems like an easier path.  At the minimum it would prove that the 4562 is the bad actor.

I am not inclined to suspect a counterfeit 4562 from Digikey. They move a lot of product. But I am still not completely on board with the latch up mechanism.

Perhaps first drop in any dual you have handy (TL072, 5532) just to confirm that the latch-up fault goes away, and is specific to the 4562.

I do not follow the modern opamp offerings so don't have a recommendation.

=======
OK, I do have an academic interest in understanding what is going on.

If the latch up does not need for the inputs to be grounded or terminated, I might be inclined to build up a stripped down circuit on the bench that I can make latch every time... Figure out how to make it latch in isolation, and that gives us insight into how to make it stop.

---------


If your circuit works with the value of R19/R9 reduced to 50k, you can get that gain back by dropping the resistance of R6 a like ratio. You may see some difference in the 100k pot taper but it will still deliver the full endpoint gain and should maintain decent S/N.

JR

 

[JLavoie]

I have some DIP 4562's that I just might try on the breadboard..

for the few units that I've needed in the field, we reduced R11/R12 to get half of the gain back. the change in input impedance didn't seem to have much effect.
also, don't forget that we changed the pots to 10k after that first schematic, so log curve shouldn't be affected if we decide to change the whole gain structure around.

I'd like to avoid that though as we would have to change the gain of our link input, and due to a few other additions to the circuit that's about a dozen component changes just to make everything else happy with the gain change.
I can do it, but would rather find out what's going on obviously.

Jason

 

[JohnRoberts]

I am starting to suspect there are more than values different about the schematic I am looking at...

If R11 and R12 are the 1k input resistors (R12 and R13 on my schematic), you can drop them some, but will run into diminishing returns quickly, and dropping that value too low will load down the microphone output at some point. I wouldn't advise much below 750 ohms to maintain a bridging termination to 150 ohm source impedance.

The R6 I mentioned is in the wiper of the mic input level pot, and should not interact with the link input gain (at least on the schematic I am looking at.) If R6 is feeding an inverting input to the output opamp, only it's gain will be changed (increased) by dropping the resistance value.

JR

 

[JLavoie]

My apologies.. I had meant to post updated schematics earlier and thought I did, but obviously the're nowhere to be found..

here are the updated values for the mic pre:

 

[JLavoie]

and here is the power section.. +12V comes in, then a diode, 100ohm resistor, then filtering, and a divider to create Vref