[sdiy] sample and (infinite) hold

[Magnus Danielson]

From: "RMC" <RMC at richardcraven.plus.com>
Subject: Re: [sdiy] sample and (infinite) hold
Date: Wed, 5 Jan 2005 23:04:08 -0000
Message-ID: <088701c4f37a$dc9b2010$0201a8c0 at cruncher>

> >> Has anyone built a sample and infinite hold and can shed some light
> >> in my direction?
> 
> polystyrene caps are the right sort and indeed, construction methods play a 
> large part in minimising leakage currents.
> 
> Take a look at how some oscilloscope circuits are constructed:
> 
> 1) mill a slot around the critical components so that the slot cuts right 
> through the board
> 
> 2) No copper on the opposite face
> 
> 3) Use "lands" made from PTFE - Tektronix used small PTFE "studs" which push 
> into the FR4 fibreglass board. The stud contains a solder well, into which 
> you solder the components.
> 
> 4) Build the circuit in a "skeleton" style using such lands, so the critical 
> components are hanging in the air.

Indeed.

The PTFE stud is good, since it lends itself to a more stable physical design
than the airborn soldering. You can use guard-rings. If you mount a capacitor
either to ground or to the output of the op-amp (when sitting in the negative
feedback path) then the other end of the capacitor will be very high-impedance
sensitive where as the ground/output end will not be. What you now do is
placing a guard-ring around the full trace (both sides) having the same
potential as the high-impedance trace. This potential should be low-impedance,
either from direct earthing or driver by a separate op-amp.

Another technique is to compensate for droop, such as expamplified in the
autonulling amplifier found in the Art of Electronics, chapter 7 on Precision
circuits and low-noise techniques. The amplifier works such that it S/H the
output voltage at the current time and then subtract the sampled voltage from
all those comming afterwards. To be usefull it needs a looooong timeconstant of
the S/H. The S/H gate is made out of transistors and resistors, the capacitor
sits in the feedback path over an OPA111B and an additional op-amp inverts the
output of the buffer op-amp, divides it down though a multiturn trimpot and
then feed back to the high-impedance node through a 500 MOhm resistor. The
trimpot is trimmed to minimize droop.

A thing to care about regardless is cleanness. Not only when mounting, but
also encapsulating it so that dust, dirt, fumes etc. don't deposit on
components and boards to create a leakage path. Also, after mounting the board
should it be cleaned so that soldering residues is cleaned off. Isopropanyl
based cleaners should be used, but be carefull to be in a well vented space
when doing it! When talking encapsulement, I mean both sides of the board,
preferably the full board should be encapsuled in some semi-hermetical package.

How far you need to go depends on what you need to do.

> People like bob Pease at Natsemi wrote datasheets on this kind of 
> technique - worth looking for  on their website.

It's also good info available on his website.

Cheers,
Magnus