Rail to Rail MOSFET Buffer with Error Correction

[jxdking]

Cordell uses Zobel's on Gates of output MOSFET (47p and 47ohm.) Would that help?

MOSFET has hundreds pf input capacitance already. I tried with 100pf. It does nothing.

 

[stigigemla]

I think 100pF for C5 and C10 will be enough with a little capacitor at the feedback. I think the bandwith can be lifted to 4 or 5MHz without problems.
I would simulate with 10 KHz square wave large signal from the generator and an oscilloscope to check if the slew rate limits the performance.

 

[Mark Tillotson]

MOSFET has hundreds pf input capacitance already. I tried with 100pf. It does nothing.

A Zobel network is heavily damped, the MOSFET internal capacitance is not so well damped and can resonate with wiring inductances in the VHF region.

 

[m0rten]

Hi @m0rten, how is your board doing? My board has arrived. To my surprise, the Error Correction loop is not fully stable. Removing C311/C312 (#25) fixed oscillation (about 50mVpp @30MHz). That means the oscillation is caused by Error Correction alone. Yea..I can reduce the Error Correction feedback ratio, but it is not longer holding the advantage to other solution such as CFP.

I expect the boards to arrive tomorrow.

 

[jxdking]

I expect the boards to arrive tomorrow.

If you still want to buid it, here is the mod required to make it work. I have tested on my board.
Based on my simulation, the fix below provide at least 20dB local NFB, 10dB+ less THD than that without Error Correction. Sadly, I don't have audio analyzer to prove that.

Alternatively, you could run it without C6/C7 installed. In this way, it will run without any Error Correction. The performance without EC based on the simulation is on par with the Blameless EF2.