Hi,

I know of one solution, however the idea is PATENTED so i don't think you can use it. It applies to fully differential MOSFET-C filters where you cross couple the MOSFETs. So you have two control voltages, Vc1 and Vc2. Vin+ is connected to two MOSFETs, one with Vc1 gate voltage and the other with Vc2 gate voltage. The one with Vc1 control voltage is connected to the negative opamp terminal and the one with Vc2 control voltage is connected to the positive terminal (and vice versa for Vin-). Like i said, its patented, but maybe you can come up with something along those lines......and then patent that too

Bear in mind you need the opamp with feedback otherwise you won't get the AC ground at the MOSFET source terminal.

Also, i have a feeling that this idea won't really remove any odd ordered non-linearity...not sure about that one though.

cheers,
Aaron

Hi,

Is your application switched or continuous-time? In the former case, I would use a bootstrapped switch where V(g,s) is constant regardless of input signal. If it is continuous-time and you have relatively small signal swing and speed, then the same can be realized using simple level shifting circuitry.

For good bootstrapping circuits, look at the JSSC paper by Dessouky & Kaiser. I think it was from Mar. 2001.

If your signals are continuous-time and fast, then you may consider if it is possible to use 2 sets of overlapping switch branches using bootstrapping. That would work too. Of course, the easiest is using linear resistors available in a process.

Otherwise, check out the book by Johns & Martin (chapter on Gm-C filters). They have lots of ideas over there about linearizing Gm stages, which can all be applied with some modifications. The idea that Aaron is talking about seems suspiciously like a variant of Barrie Gilbert's multi-tanh principle, where the different voltages mimic the effect of an offset and increase the linear range of the switch. If so, then it is doubtful how strong that patent is. If not, then excuse me. I would have to look at it.

Regards,
Vivek

Independend how do you count a genius, outstanding patent or simply a variation from a basic math insight applied to a circuit, feed to the patent office and let the court decide what was before (chicken or egg)...

I remember one of my first MOS books

http://www.abebooks.de/products/isbn/9783519000730/Gad-Horst/Feldeffektelektroni...

where a MOS used as a single ended resistor is linearized by adding 1/2 of the resistor across voltage is added to the gate control voltage.

The normal expression for Ids is:

Ids=u*Cox*(W/L)*(Vgs-Vth-Vds/2)*Vds

The question is what is the genius of what to do with Vgs so that the expression is linear in Vds?

I would advise that the patent is trivial if the key idea is a math manipulation.

Or simply put the solution "Ids linear on Vds" into a symbolic math engine. Copy the resulting equation and built it by basic (non-genius) construction rules. I am shure that many circuit patents are based by math rules and manipulations but then are put under the the magic umbrella of a circuit genius.

is IP the next (third bubble in this century) ...