DC offset issue with this Amplifier

Jump to Latest
#21
mtx489 said:
Anyway the feedback node follow the same offset variation.
Click to expand...
C11 is doing its work properly, the offset originates from the first stage.
There are two current sources (Q12-Q13: pnp's and Q14-Q15: npn's) in use, but these are not coupled.
Any fluctuation caused by temperature will yield in slightly different currents, which are summed at the feedback injection point.
A new setpoint is the result and this will go on forever.
It's better to have them coupled. One 'master' and one 'slave', but that will 'break' your symmetric design.
 
#22
I doubt temperature or mismatch in Q-(12,13,14,15) can cause as much offset and drift as is observed. A mismatch of 2X in the current sources that these transistors create should result in an offset on the order of about 80mV. This 2X mismatch represents a very large current error assumption which is not realistic. I just used it to demonstrate the sensitivity to current mismatches in general.

All designs can be "improved", but I don't see a design error that would explain the actual offsets being observed. This leads me to believe an intermittent component or connection is at the root cause of this large and variable offset.
 
#23
Potentiallyincorrect said:
I doubt temperature or mismatch in Q-(12,13,14,15) can cause as much offset and drift as is observed. A mismatch of 2X in the current sources that these transistors create should result in an offset on the order of about 80mV. This 2X mismatch represents a very large current error assumption which is not realistic. I just used it to demonstrate the sensitivity to current mismatches in general.

All designs can be "improved", but I don't see a design error that would explain the actual offsets being observed. This leads me to believe an intermittent component or connection is at the root cause of this large and variable offset.
Click to expand...
Both Q3-Q10 and Q2-Q9 are (almost) unity gain, but the first stage is still susceptible to all kinds of (natural) drift.
The resulting offset at the feedback injection node is also amplified by Q10-Q1 and Q9-Q11 and tighten together again with Q6. Then the follower stage and back to the same fi-node. There is no stable reference point ('ground') here, so the closed loop amplifer part cannot do anything else then to follow / amplify the offset from the first stage. The bases of Q10 and Q9 are not related to each other, they just wander around depending on Q12-Q13 and Q14-Q15, and to their liking.
 
#24
Hello
just an update: after the last action with hot air, at soldering temperature, the amplifier works good.
No more offset drift and variations since two days.
I used hot air in the current source zone, so probably there is a bad solder joint or something else in that zone.
I hope is not a coincidence.
 
#25
The second PCB I mounted is quite better from the beginning, there is a little offset variations at the cold start but after some minutes it tends to stabilize.
Maybe the design itself is not so stable, not as a VFA.
Do you see something I can do, without change the PCB layout, to help improve stability at the start ?
Lowering the first stage current? Increase the feedback resistor ?
 
You must log in or register to reply here.
Top Bottom