[sdiy] Quad s&h math verification

[harrybissell]

Just for the record... you might be better off with a smaller
cap even if the math isn't as good. 1uF is a hell of a big cap
to hang on a 4066 switch.  It can take a lot of current when you
try to slew it.

You might even find that when that happens you can suck some
supply current into the switch. Check the fine print on the
data sheet, there is some disclaimer there.

A 1uF cap is likely to have quite a bit of leakage and dielectric
absorption.   You will probably not get a good dielectric in this
size cap.

A .1uF to .01uF is more typical of the values that have been used
(successfully) in the past.  In a .01uF you can get polystyrene
caps, which are about as good as they come for this application.

H^) harry

Karl Ekdahl wrote:
> 
> Hi list, i'm just trying to figure out wheter the Quad
> S&H i designed 'n built just works out of sheer luck
> or if i'm actually learning stuff (reading The Art of
> Electronics all the time), therefore i'd like to
> verify my maths...
> 
> Basic s&h's built around a 4066 as FET switch, a 1uF
> cap and TL074's. Input is fed from a TL072 follower.
> Schematic:
> http://www.sdiy.org/knas/ars2/schematics/quadsh.gif
> 
> Max voltage swing is 0 - 5v, maximum tolerable droop
> is 1mV (yes it's a lot). Worst case leakage from 4066
> + input-bias from TL074 is 200nA + 200pA = 2002nA, i
> used the formula dt = (C * dV)/I to calculate droop
> time for 1mV like dt = (0.000001*0.001)/0.0000002002
> ~= 0.005 seconds.
> 
> For the charge time i assumed 0-5v charge @ 20mA from
> the TL072, the 4066 having a ON resistance of 350ohms.
> Using U/I=R on 5/0.02=250 i get total resistance of
> 600ohms thus U/R=I -> 5/600=0.00833A supply current.
> Again using dt = (C * dV)/I i get dt =
> (0.000001*5)/0.0083333 ~= 0.0006 seconds charge time.
> 
> So i need to refresh every s&h every 5mS and the
> refresh period has to be at least 0.6mS long.
> 
> phew... so, am i totally out biking (i used that
> before i read it Karl!) or am i correct? Any case, it
> works just fine, can't hear any VCO "flutter" or other
> inconsistencies.
> 
> thanks
> 
> Karl