From: email@example.com (Ken Smith)
Subject: Re: mitigating charge injection in switched capacitor systems...
Date: Sat, 14 Dec 2002 18:01:20 +0000 (UTC)
Organization: a2i network
References: <_moK9.2019$X56.firstname.lastname@example.org> <_YBK9.58$SQ.email@example.com>
NNTP-Posting-Date: Sat, 14 Dec 2002 18:01:20 +0000 (UTC)
X-Newsreader: trn 4.0-test76 (Apr 2, 2001)
Originator: firstname.lastname@example.org (Ken Smith)
In article <_YBK9.58$SQ.email@example.com>,
Kevin Aylward wrote:
>Ken Smith wrote:
>> In article <_moK9.2019$X56.firstname.lastname@example.org>,
>> Kevin Aylward wrote:
>>> Ken Smith wrote:
>>>> Yes. what happens is: a charge is capacitively coupled in from the
>>>> gate to the source when the gate moves. The channel then becomes a
>>>> very high resistance and traps this charge as the source of the FET.
>>>> If you are making your own gate driver for the fet, you can control
>>>> the shape of the switching wave form to reduce this effect quite a
>>> Vary dubious if the load is also a capacitor. It wont matter what the
>>> ramp time of the control voltage is. Its a capacitor divider.
>Yes it is. Have you actually tried simulating this in Spice?
No I designed, built, and tested one. It worked just fine.
>Hint: I have:-)
>> until the FET turns off it is an RC circuit with the resistance
>> being the channel resistance of the FET.
>If you switch faster than the RC time constant, then the voltage accross
>the resistor will peak at the switch voltage, so the resistor can not
What resistor are you talking about? There was no resistor in my post.
>> If the gate moves slowly,
>> most of the charge goes to the driven electode and not onto the load
>Ahmmm.... sort off, on *rare* occasions, depends on just how *slow* you
>switch. Sure if you switch *very* *very* slow such that i=dv/dt * R
>switch is small, then that can limit the feedthough voltage.
So you now admit I was right.
>this effect is all but *useless* in sample and hold or switched
>capacitor filter applications because you usually have to switch much
>faster than any RC time constant.
"usually" means "sometimes you don't" . In my situation, I had lots of
time for the FET to switch compared to how fast of a FET I was using.
>Its actually a common misconception that you can slow down edges to
>minimise feedthrough. It can only work when the system response required
>is slow, e.g. an audio switch can be slowed to ms ramps.
A "common misconception" that works just fine for audio, interesting.
email@example.com forging knowledge