We have a customer EEG application using the INA121U as a front end with a gain of 50 and bipolar +/-15V supplies (derived from a dc-dc converter switching at 800kHz). When the EEG inputs are open, a dc offset can build up over several hours. When this offset exceeds +/- 300mV across the inputs, the INA output saturates and the device is observed to lock into oscillation at approx. 100 hz. This oscillation sustains itself until the dc offset at the input is discharged by connecting a subject.
The original circuit had an “ac coupling” configuration similar to Fig 9 of the INA121U datasheet, using feedback to the REF pin. We originally thought this might be causing the oscillation. However even with this bypassed and REF set steady at GND, the oscillation still occurs.
This is a legacy design in use for 5+ years. The changes I have noticed in the schematics are:
- early designs were dc-coupled to the following gain stage (hence the Fig 9 “ac-coupling” setup. later designs were ac-coupled (10uF) but the “ac-coupling” setup was still retained.
- early designs had 1uF supply decoupling. This was reduced to .33uF with 0.01uf in parallel, then reduced again to 0.1uF.
- all designs incorporate a Right Leg Drive circuit using a signal tap from a center-tapped gain resistor
Our questions:
- any ideas on the possible source of this oscillation and how to squelch it?
- any comments on whether there may be conditions where the oscillation would not be eliminated by connecting a subject?
If needed, more detailed circuit info and results can be supplied off-line to a TI applications engineer.
Thanks in advance,
Jim