Charging a supercap from a boost converter - help!

[Pete_UK]

I'm having a hard time trying to find any examples of charging a supercap from a boost converter architecture. I need to be able to charge a few farads of supercap to 20V, from a battery power source which is nominally 6V, so I was looking at using a boost converter running in constant-current, with some means of then shutting off when the required voltage is reached. In this application one side of the supercap is permanently wired to ground, so this will necessitate high-side current sensing. The charging current is relatively low, at around 100mA.  I've had a play around in LTspice with some regular LT boost converters, and unfortunately it all starts to get a bit messy right away, eg. the high-side sensing requirement makes it difficult to derive the feedback, which then initially massively over-shoots due to the boost topology etc. If anyone has done something similar, or has any advice on parts or better suited topologies (eg. buck-boost, or SEPIC?) then that would be great to hear about. Thanks!

[SuzyC]

Do post a schematic. From what you describe, you've seemed to found an elegant but over complicated solution to your design problem.

Post your charging power source voltage its current capacity.

What are your requirements about the footprint and also efficiency and current accuracy you need for  boosted V CC charging?

Do you have room for a small heat sink for a power transistor that would be the constant current source?

There could be an easy solution, a quick and dirty solution.

[langwadt]

Since a boost has an inductor and diode directly from input to output it can't limit the current when the output is less than the input

you'd need something like a sepic, but then something similar to this current limit for a buck might work

[SuzyC]

A boost converter raises the voltage to the proper level. A single transistor configured as a constant current generator then sets the charging current.

[lastguy]

your idea is too far for such design.

[Pete_UK]

Thanks all for the replies, and thanks langwadt for the schematic idea (looks quite useful). I agree that it looks like Buck-Boost or SEPIC is probably the best way to tackle the problem. I'll try out some more ideas in LTspice & report back later how it works out..