Problem with simple OpAmp Circuit, need help

[Jacques St-Pierre]

Hello,



I have been searching for a bug with a simple analog circuit for weeks
without results, so now I am trying to get help from you people.



The final design is obviously more complex than that, but I have been able
to isolate the problem to that particular circuit, build around a single
LMC6282 chip. In fact, that section is part of an amplifier/filter/A/D
circuit for a sound recording system.



The audio signal is feed to Vin, a 2.5Vdc reference from the A/D chip is
feed to Vref and the circuit Vout is feed to the A/D input. The circuit is
use to add a 2Vdc offset to the audio signal that is couple with the 1uF
capacitor. The circuit is supply via the same +5Vdc used for the digital
section, but that +5Vdc is cleaned using a T filter using a couple of 1mH
coils & 10uF capacitors (not on the schematic) and the 10 Ohms & 1uF filter
on the supply of the LMC6282 Opamp.



That circuit work fine for years until now. Until we had been force to use
larger compact flash memory card, in our case 4gb and more. We suspect that
those newer cards used more energy on the +5Vdc supply when they are in
write mode and that put glitch on the supply rail. We have not been able to
prove this, but the card speck sheets confirm that they use a lot more
current than our older card. Nothing can be detected on the supply using our
current instrumentations.



In turn, those supposed supply glitches appear amplified at the output of
the circuit (Vout) and are digitalize by the A/D. They arrear has single
random value sample in the data stream. By example, when all samples should
be +2Vdc, we have isolated sample at 1 or 3Vdc, during the write data phase
to the compact flash card. We have been able to prove the presence of those
glitches at Vout, but since they are very occasional, they are quite hard to
catch on our scope, they are easily spotted on the recorded digital audio
stream.



We try to power the circuit via an different supply, the problem disappear,
so it's clear it come from the supply, but I have no idea why that
particular design is so sensitive to glitches on the supply rail.



Any ideas?



Here an ASCII representation of the circuit.



+5Vdc Supply

|

\

Vref 2.5Vdc / 10

| \

\ |

/ 20K --------- *-----|

\ | | |\ | --- 1uF

| | |\ | -----\/\---------*---|-\ | ---

| |-|-\ | | 1.0K | | \| |

\ 50K | \ | | | | \ GND

/<------|+ /---*-----* | | \--------*----- Vout

\ | / | | | / |

| |/LMC6482 | 100K 825 | | /LMC6482 |

| -----\/\--*-\/\------|+ /| |

\ 95K | | | / | |

/ | | | |

\ Vin ---||----- | GND |

| 1uF | |

| -------------\/\---

GND 5K





Thanks



Jacques

 

[Jacques St-Pierre]

Is it a digital scope?

Yes this is, we have been able to catch the glitch at Vout a few times, but
not on the supply or Vref.

The circuit looks reasonably proof from short glitches on the supply.
Since you can't actually see what's going on, how do you know that the
problem is in the circuit you show and not in the ADC?

When we disconnect the Vout from the input of the A/D and connect the A/D
input on its Vref, the data stream is constant with no glitch at all.
If we put a large capacity at Vout & GND, the glicth goes away, but it can
not be use since it attenuate the audio signal too.

I'd work with your software engineers to make a test that pounds writes
into the flash card (to make the problem show more often). If you can't
figure out how to trigger the scope to make it reliably show the problem,
see if the software guys can't toggle a pin on the processor when they see
an outlier. Then trigger your scope off that pin.

We did work on software to find a way to find the source of the problem.
We know it's link to the compact flash.

What we can not figure out, is a way to stop the problem, since we can
identified where the glitch came in the circuit.

Bye
Jacques

 

[Jacques St-Pierre]

When you say you power the circuit from a separate supply are you
including the ADC and Vref? My guess is the noise would becoming
through the reference either through Vref or Ground. Try putting a cap
to ground at the non inverting input on the first opamp.

Yes when we use a separate supply for our analog +5Vdc supply, it's include
the analog section of the A/D too.
We try to put a 10uF Tantalum capacitor on Vref, the positive input and
output of the first Opamp with no impact on the problem.
We try to raise the T filter capacitors from 47uF to 100uF on the analog
+5Vdc with no result.
We try to raise the 10uF capacitor to 100uF on the power pin of the Opamp
with no result.
We change the 100K to 1K with no result.
The only thing that appear to stop the problem, is to put a large capacity
on Vout, but it can not be use since it destroy the audio signal.

We try to catch the glitch on the Vref and power pin of the Opamp, but we
did not catch it yet with our digital scope.

Bye
Jacques