Flipped Voltage Followers (FVF), any hidden draw backs????

[jgk2004]

Hello all,

I am considering use these lost and kinda forgotten FVFs in a design since when reading all the publications they still seem amazing for SC filters. When building them up, they seem to be fine typical but I have not looked yet over MC and Corner cases yet. I know.... I need more time.....
So when looking at using this type of FVF which is reported to be the best, the only drawback I can see is increased cap loading at input, more current which is in the flip, and maintaining stablity due to loading of inputs(sizing for 1/f) on the flip node and then needing area to compensate above the Ib's. Is that what it costs to make these? Is there any other hidden pain which I am missing........

Kind Regards,
JGK

 

[sutapanaki]

Obviously you will be using this buffer perhaps with some global feedback. So, you will have that external feedback plus the local feedbacks of the FVF. You will have to make sure the local feedback has loop gain working to higher frequencies (maybe 10x) compared to the loop gain of the external feedback.

 

[jgk2004]

Thanks Sutapanaki,

Super good point, thus sizing up for 1/f, since I don't want to chop, might be an issue since I will slow down the FVF and have to compensate them, thus this might be where I might need more current in the flipped branch. I will report back after sign off for anyone else interested in these cool FVFs

Kind Regards,
JGK

 

[mirror_pole]

Hello guys,

Why is it important that the loop gain of the FVF works to higher frequencies then the global feedback?
Is it because otherwise the FVF will slow down the circuit? But why it has to be even 10x higher frequency.
Sorry im pretty unexperienced but would like to understand why this is important.

 

[jgk2004]

Hi all,

So after looking into it more, if the FVF loop isn't fast it will not increase the current fast enough for the class AB expanding. It totally makes sense. So if you want to have super fast slewing, your FVF loop needs to be in the GHz range, so easily like 100X my fs., This is easy since you can design them small at the cost of current mirror matching. You just have to be very aware of that loop. I just designed the input stages for reasonable 1/f so like 50um/300nm, and this killed the FVF bandwidth down to 300MHz and I needed to compensate the FVF for PM, thus here you have to increase current and do some sizing tricks to get the GBW up to >1GHz with good PM which I need for my slewing requirement.

So are there any other hidden draw backs?

JGK

 

[sutapanaki]

Hello guys,

Why is it important that the loop gain of the FVF works to higher frequencies then the global feedback?
Is it because otherwise the FVF will slow down the circuit? But why it has to be even 10x higher frequency.
Sorry im pretty unexperienced but would like to understand why this is important.

You have few loops that interact with each other. They may not interact always in a constructive way. When a loop starts to die, because of its frequency limitations it introduces additional phase shifts that can cause troubles in the main loop. So, it is a general practice to move one of the loops out of the way. Since the external loop is the one that we aim at to define our functional performance, that means we move the internal loops out of the way. Usual way for that is to make them fast enough so they kind of become transparent for the operation. 10x is not a magic number, just a first cut guess. Usually for engineers 10x often means cheesy infinity. If internal loop is 10x faster, its phase shift is just starting to be noticed by the main loop at 10x lower frequency wrt the cut-off frequency of that internal loop and the hope is that by that frequency the main loop has already crossed 0dB