Feedback cap in + fb to create neg. cap due to Miller effect

With a differential amplifier, can we tie capacitors (from output to input) in a positive feedback manner to get negative capacitance due to the Miller effect? I heard it can be used in practice to negate the loading of the gate-to-drain capacitance at the input.

If this is the case, what prevents us from going so far to increasing this feedback cap to negating all the capacitance loading the input (parasitics, gate-to-source capacitance of input transistor, etc...) as well?

I know we could run into instability problems if it goes too far, but there must be other terrible things that prevent frequent use of this technique in practice. It looks too good to be true.

Any help on this would be great. Thanks a lot.

Re: Feedback cap in + fb to create neg. cap due to Miller ef

I think I understand what you are saying, but the amount of capacitance required would depend on the gain in order to just cancel and not cause oscillation. I have used positive capacitive feedback to remove the gate capacitance of fets, the capacitance of a photodiode, and the stray capacitance of compensation networks but it relies on having a controllable, predictable gain. I have always added a unity gain buffer and used that to generate the positive feedback.

The easiest way to see if your idea works is to try it (i.e. simulate it).

Keith.

Added after 5 hours 3 minutes:

What I was describing is slightly different to what you are suggesting. My ideas (which I know work) are "bootstrapping" the opposite end of a device to cancel out capacitance. What you are describing puts it directly on the device input.

I put together a simple circuit to prove it will work and have attached it. The green trace is with no additional coupling capacitance (C1&C2). The red is with 12fF and the blue with 20fF. So, yes the bandwidth can be increased by this technique, but I don't know how you could control the value well enough for it to be practical. The value required will depend on the gain which depends on quite a few things (e.g. temperature & supply voltage).

You could probably do it if you had a low gain or unity gain front end.

Keith.

Re: Feedback cap in + fb to create neg. cap due to Miller ef

Thanks for the helpful reply and the attached simulation.
Sounds like bootstrapping is what is more commonly used! (as you mentioned, the gain should be well-controlled). Although, it seems like bootstrapping would only help "kill" those capacitances tied from input to output since those caps would be Miller multiplied by 0. To "kill" other caps (like parasitic caps coupling to ground), we would need a controlled gain of greater than 1? Any clarifications would be great. Thanks.

Bootstrapping is often used to increase input impedance. e.g. increase the effective gate resistor on a JFET amplifier - an example is here **broken link removed** at the top of the second page.

I have used them in transimpedance amplifiers for pulling up the opposite end of the photodiode - anode to amplifier, cathode to positive supply through a resistor and then pull the cathode with a buffered version of the input. It reduces the effective photodiode capacitance considerably. While this isn't exactly what I was doing http://cds.linear.com/docs/Design Note/dn399f.pdf it shows the principle.

They are all examples of unity gain bootstrapping. Miller cancellation like you suggest sounds interesting and may be practical if the gain stage had a moderate and reasonably well controlled gain. For example, if the stage had a gain of 10 and a known input capacitance of say 10pF, then coupling through 1pF should cancel out the input capacitance. This is somewhat different to normal bootstrapping but operating on similar principles.

Keith.

Re: Feedback cap in + fb to create neg. cap due to Miller ef

Miller effect cancellation, or also known as capacitance neutralization is a well known technique. It is well suited in integrated fully differential circuits. In the attachment above, the extra capacitances, cross-coupled between the output and input can be done by connecting MOS transistors with half the size of the input transistors - drain, source shorted together forming one terminal of the cap, and gate is the other. This way you achieve matching between the Cdg of the input transistors and the neutralization caps. Accurate gain is not important. The cancellation is not complete, meaning you don't completely cancel Cgd, but instead of it being Miller amplified, the residual seen at the input is 2Cgd - if memory serves me well. Check Gray and Meyer.

I couldn't find it in Gray & Meyer but I see what you mean. I was forgetting about the fact that the feedback capacitance is subjected to the same gain as the Miller capacitance so the same gain is applied to both.

Thanks.

Keith.

Re: Feedback cap in + fb to create neg. cap due to Miller ef

4th edition, p.849

Re: Feedback cap in + fb to create neg. cap due to Miller ef

sutapanaki said:

4th edition, p.849

Click to expand...

Mine is 3rd edition (and fewer pages). Do you have a section heading in case it is still in mine?

Thanks.

Keith

Re: Feedback cap in + fb to create neg. cap due to Miller ef

I guess it might not be in the 3rd edition, since in the 4th it is in ch.12 for fully differential operational amplifiers and the section is called Neutralization. But I'm sure you can find the 4th edition somewhere online.

Re: Feedback cap in + fb to create neg. cap due to Miller ef

sutapanaki said:

I guess it might not be in the 3rd edition, since in the 4th it is in ch.12 for fully differential operational amplifiers and the section is called Neutralization. But I'm sure you can find the 4th edition somewhere online.

Click to expand...

No, it's not in the third edition. No matter, the concept is easiest enough to understand & simulate.

Thanks.

Keith.