# How reset an integrator at 1 Mhz?

Discussion in 'Electronic Design' started by enzino, Sep 24, 2004.

1. ### enzinoGuest

Hallo
I'm new here
I'm designing a special integrator with an operational amplifier LM6172in in
inverting configuration.
Rin = 1Kohm
On the feedback loop there is a 220pF capacitor.
The non-inverting input is connected to ground.
The problem is that I need a reset every 1us, and the reset operation should
bo no longer than 0.2us.
I've tried with some analog switch, but these were too slow !
I've tried with the 2n3819 fet, but the control signal pass to the rest of
the integrator circuit, vanishing the results.
How can I design this circuit?
Which the correct component to perform the reset operation?

Ideally the reset operation may be realized with a variable resistor
connected to the operational output and the inverting input, in parallel
with the 220pF capacitor.
During the integration, the variable resistor should be set at extremely
high values (for example 1Mohm).
During the reset operation, the variable resistor should be set at low
values (for example 10 ohm).
Which component is able to perform this task at a rate of 1 Mhz?

Some idea?

2. ### John LarkinGuest

(being tired of hacking embedded code, here's a chance to take a break
and solder something. Actually, I've been wondering for some time if
gaasfets can act as bidirectional analog switches.)

This seems to work: connect a small GaasFet, FSU02 maybe: source to
the summing point, drain to opamp output. Drive the gate through a 1K
resistor maybe. Drive the gate negative to integrate, ground to reset.

You'll have to drive the gate more negative than opamp max neg swing
minus Vp. For the FSU02, with the opamp output swinging +-5 max, that
would be -7.5 maybe. If you can arrange for the opamp to swing
positive only when it integrates, gate drive could be, say, -2.5 to
integrate, 0 to reset. This might leak a few uA off and be maybe 10
ohms on. It will be very fast and charge injection will be pretty low
for such a low Ron.

Works just like a jfet! Feedback appreciated.

John

3. ### JoergGuest

Hi Enzino,
Welcome to the club!
You could send that kind of pulse through a toriod transformer to get
rid of unwanted loops etc.
Check out the SD5400 series. Used to be made by Siliconix/Vishay but now
it seems it went to Calmos:

http://www.calmostech.com/SD5400.pdf

The non-SMT version would be the SD5000, but watch out when ordering
since there is a diode with the same name.

These are chips that cost just a few Dollars but contain four nicely
matched FETs. You can make differential switch gates with these and with
a transformer circuit you can achieve a charge injection close to zero.
Charge injection is what causes the control signal to penetrate into
your signal path. It is caused by the gate capacitance of the FET.
Basically what a differential approach does is inject the opposite
charge into the path at the same time the switch acts.

Regards, Joerg

4. ### enzinoGuest

Thanks John
I'll try and report some feedback

5. ### enzinoGuest

Thanks Joerg
Very usefull...

6. ### Winfield HillGuest

enzino wrote...
You need a fast low-capacitance MOSFET with low ON resistance.

One good choice is a part from the old Siliconix (now Vishay)
SD210 family, with Cdx = 1.1pF and Ron = 45 ohms. The SD211
has a useful gate-protection zener. Future Electronics stocks
the SD210DE, but requires a minimum 50-piece purchase. These
parts are also made by Calogic and Linear Integrated Systems,
who may have better inventory sources.

You will find that modern apparently-superior cmos IC switches
have very high capacitance, compared to the sd210 family.

7. ### JoergGuest

Hi Winfield,
That's the dream of every high speed designer. Something like 10Ohms and
a fraction of a pF. Maybe some day it comes true.

BTW, one other solution for a switch without charge injection would be
to use (or rather abuse) a double-balanced mixer such as those from Mini
Circuits. But it would add a few Dollars. I did that on a sample and
hold a long time ago although I was greedy and rolled my own. I think
Mini Circuits made dedicated "balanced diode switches" in those days,
I believe that one also went to Calmos, like the SD5400 did..

Regards, Joerg