In the following two circuits, I am using a TLE2426 as a low impedance virtual ground to bias multiple single supply opamps/diff amps. At the output is an LRC filter, but I'm not sure whether to connect the filter capacitor to GND or virtual ground. My gut says that shunting the unwanted signal frequencies to GND directly is better and would keep them off of the virtual ground line, but I'm not sure if this is correct, or if there's really any difference. Performance-wise I can't distinguish a difference. Which of the two options is better?

Option 1: enter image description here

Option 2: enter image description here


I believe option 2 is the better one.

Reason being transients being driven out on U2B have a shorter signal path to the ground of the device. If you go with option 1, the transient current need to make their way though the virtual rail driver to ground instead. Obviously a longer path through active devices.

I am however a little surprised there are no capacitors on your virtual ground connecting to the other rails. Various examples on line show some pretty large caps there/

  • \$\begingroup\$ That confirms what I was thinking. Re the virtual ground caps: I've actually been told that numerous times in response to questions I've posted where I've used a TLE2426 in the schematic, and every time I go back and review the datasheet, and there is no mention anywhere that large caps (or any caps for that matter) are recommended to be used at the output. I have yet to have any obvious issues as a result of not using them, but I'll have to dig into this more. \$\endgroup\$ – User7251 Nov 15 '17 at 6:55
  • \$\begingroup\$ The graph on page 17 of the TLE2426 says you may well need some cap on the output if you don't want the thing bursting into oscillation when it has to source current. On past, painful experience it is not wrong. You want to keep out of the region marked unstable at all times. \$\endgroup\$ – Dan Mills Nov 15 '17 at 13:52
  • \$\begingroup\$ @User7251 yes I looked at the datasheet too, and expected to see recommendations there. That's why I said I was surprised. However, I think that is just an error on the writers part. It makes sense to have one or better two. Also, Dan's experience indicates it is more crucial. If you were using the 8 pin part with the noise reduction cap it might be different. \$\endgroup\$ – Trevor_G Nov 15 '17 at 14:19
  • 1
    \$\begingroup\$ @Trevor, the version I'm using is actually the 8-pin NR version, with 100nF and 1uF caps to ground on that pin. In the data tables they commonly used 1uF and I think I tested this a while back and found that that combination gave me the best noise reduction. But the schematic on page 3 suggests that the NR pin is at the input rather than output, suggesting that the NR caps aren't a substitute for the load capacitance you and Dan have suggested \$\endgroup\$ – User7251 Nov 16 '17 at 0:55
  • 1
    \$\begingroup\$ They are not a substitute as they provide filtering to the voltage divider not capacitive load to the output driver (Which is weird, but seem to be what that part likes). \$\endgroup\$ – Dan Mills Nov 16 '17 at 11:53

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.