Comparator oscillation in meter circuit - pollutes audio
Posted: Fri Sep 01, 2017 3:43 am
I'm using a rather standard metering circuit with full wave op-amp rectifier that feeds a string of LM339 comparators. Comparators outputs are connected to the string of LEDs that are fed from the constant current source. All well and good. I had this circuit working on the dual rails supply with no problems at all.
However, in this device I have a single supply, so I'm using an op-amp rail-splitter and I'm deriving virtual ground reference from there. Now, here's what happens. When the LED just starts to turn on (that is, input V at the comparator = Vref), I see a high order harmonics apearing on the FFT analyzer. Once the LED is fully on, these harmonics disappear.
I hunted this problem down to LM339 oscillation between on and off and this somehow couples into audio path. The only solution that I found is to put a cap in the NFB loop of each comparator. Here're the quesions:
1. Why does this happen? Is it because of split rails? Didn't have any issues with regular dual supplies.
2. Is there any more elegant solution than slapping 16 additional caps on the LM339 (I've got 4 chips on the board).
Thanks for any comments!
However, in this device I have a single supply, so I'm using an op-amp rail-splitter and I'm deriving virtual ground reference from there. Now, here's what happens. When the LED just starts to turn on (that is, input V at the comparator = Vref), I see a high order harmonics apearing on the FFT analyzer. Once the LED is fully on, these harmonics disappear.
I hunted this problem down to LM339 oscillation between on and off and this somehow couples into audio path. The only solution that I found is to put a cap in the NFB loop of each comparator. Here're the quesions:
1. Why does this happen? Is it because of split rails? Didn't have any issues with regular dual supplies.
2. Is there any more elegant solution than slapping 16 additional caps on the LM339 (I've got 4 chips on the board).
Thanks for any comments!