Better Vref ripple reduction using active filter

[composition4]

I was working on something today for which I wanted to use a buffered 4.5V vref, and went to my normal voltage divider + capacitor into an op amp voltage follower (top example).

Given that I'm using an op amp anyway, I though why not use an active filter, and came up with the bottom example using a Sallen-Key low pass. 33x better ripple filtering according to the simulation at the cost of using an extra 10uF capacitor and another resistor.

Are there any negatives to using the bottom example? I feel like I might be missing something.





Jonathan

[ElectricDruid]

What's R5 for? I haven't seen that before on a Sallen Key.

Oh, I get it. It's the other half of the voltage divider. Nice.

[composition4]

Yeah I figured I could cheat and put it in there.

So I breadboarded it as a comparison in a high gain circuit... seems to work but it increases the propensity of the circuit to oscillate with the gain way up. I'm not sure why.

[diffeq]

Are there any negatives to using the bottom example? I feel like I might be missing something.

Settling time. How long does it take to climb to 4.5v after power supply is enabled?

Since it is a TL072, you could use two 10Megaohm Rs and 220uF cap, no extra components. Sim shows 100mv ripple reduced to 46uV, how's that for ripple reduction? Only downside is that it takes 30 minutes to get there. 

[diffeq]

Yeah I figured I could cheat and put it in there.

So I breadboarded it as a comparison in a high gain circuit... seems to work but it increases the propensity of the circuit to oscillate with the gain way up. I'm not sure why.

Try 470R in series the opamp output. Or move gain duties to a separate opamp.

[composition4]

Are there any negatives to using the bottom example? I feel like I might be missing something.

Settling time. How long does it take to climb to 4.5v after power supply is enabled?

Since it is a TL072, you could use two 10Megaohm Rs and 220uF cap, no extra components. Sim shows 100mv ripple reduced to 46uV, how's that for ripple reduction? Only downside is that it takes 30 minutes to get there. 

Thanks for the input... I calculate both examples as I gave them as less that a quarter of a second, nothing significant?

Try 470R on the opamp output. Or move gain duties to a separate opamp.

I don't know why extra resistance on the output of the opamp would make a difference, it's not driving a capacitive load. Also the reason I'm buffering is for as low impedance vref as possible, kind of defeats the purpose.

[antonis]

I think diffeq means to place 470R resistor inside the NFB loop (right after op-amp output..)

[composition4]

Still not sure how this would help, given that it's not driving a capacitive load. I think some more experimentation is needed in future. For now I'll just go with the usual top way, with a big 100uF cap instead... works well for my purpose

[BJF]

Hi there,

To the point


With an OP amp buffered divider what is connected to midpoint sees the output impedance of the OP amp and this works most of the time because the output presents a capacitance equivalent to 1Farad to either supply line. But capacitances have a tendency of diminishing at high frequency.......
High gain circuits a problem is often feedback via floating ground and even a buffered midpoint may need something like 100nF ( 0,1uF)  connected at output of OP amp to either supply line. This can be a solution if there is high frequency oscillation.

If supply line is noisy then try first making as fine DC as you can by filtering supply line such as regulating .

Back in the late 1980s I had a  digital delay and and an extreme gain distortion both powered from same Mascot supply. Just connecting the DC inlets caused something high pitched to leak from delay to distortion and the solution was to make an extension box with a 7809 regulator and to set the Mascot for 12V and the ripple from the Mascot at 1mV could be reduced about 60dB. The extreme distortion box had a gain of 80dB nd resulted in 3dB dynamics at output so it anyway needed a noise gate at output.

If midpoint is only filtered by a capacitor even at circuit gains of 46dB this capacitor may need to be 100uF to prevent feedback via midpoint if there are several stages in series connected to the mid supply.

With an oscilloscope you can monitor the squealing at midpoint and trace the feedback path.
Squealing in high gain circuits is an oscillation and by nature then there is positive feedback to trigger oscillation and that can occur via mid supply point.

In regards to feedback via power supplies and restricting that to 9V circuits OP amps have usually very high PowerSupplyRejectionRatio
but transistor circuits not so much. In particular transistor buffers have almost none, so it is a good idea to filter seperately any such stages.
Otherwise oscillation can be triggered via supply line. Likewise any oscillators connected can leak through supply line and should be given there own mid point if one is needed to ensure minimal leakage.

It is a good idea to consider midpoint more a source of bias and then create separate midpoints for sensitive circuits and locally.

In summary it is a good idea to makes as fine DC as possible. Use of OP amp buffered midpoint is a very good idea if current draw from mid supply is asymmetric and then OP amp output can force stable voltage and in the process create a virtual ground.

Ground at a distance from ground pole should be viewed as a  DC potential and hopefully a point of no AC;)

Building high gain on breadboard opens a whole set of possible leakages and microphonics and makes life interesting

At your service
Bjorn Juhl
BJF Electronics
.