Comparator with FET hysteresis

[cupoftea]

Hi,
Woudl you agree the attached FET hysteresis has some advantages over using a feedback resistor. For example, this will be less noise susceptible?
LTspice and jpeg attached

 

[dick_freebird]

Switched hysteresis can give you high accuracy on one of the two thresholds and the other, not worse than R/R feedback. I've used this style in RS-422 Rx front ends where hysteresis band is part of input sensitivity spec window.

One small discovery is that wiring the FET so its cgs / cgd aid rather than oppose switching helps prop delay more than you might expect. Source follower switch better than common source switch. Just have to find the right spot.

 

[KlausST]

Hi,

why do you think it is less susceptible to noise?

Klaus

 

[cupoftea]

...thanks, because you dont have to have that series resistor to develop some voltage across..(the hysteresis voltage) ...so the impedance "seen" at the comparator pin is less...so less noise susceptible.

 

[dick_freebird]

Actually higher input network impedance would tend to
filter HF noise (other than kickback) more, than lower.
But what noise you might really encounter, who knows?

 

[KlausST]

Hi,

you dont have to have that series resistor

I don´t see it . In both cases you need at least two resitors to develop a hysteresis.

In your case the "input" is developed by R1, R2, R7
and the hysteresis by R6

Klaus

 

[cupoftea]

Thanks,

Actually higher input network impedance would tend to
filter HF noise (other than kickback) more, than lower.

So would adding a cap, say 1nF, to the comparator pin, make it more noise susceptible?

I think with the fet method, it facilitates adding a cap to the comp'tor pin.
With the feedback R method, the RC forms a time constant......with the fet method, the fet just discharges the cap quickly, as if its not really there, so this is good for tripping wihtout additional delay.

Also, its easier to work out trip points with the " fet method".

 

[dick_freebird]

Again this relates to the noise coupling / sources.

If your output likes to hammer the local ground, adding
a capacitor to that ground (especially asymmetrically,
IN+ vs IN-) will couple ground impulses back to the
inputs, and now you must reject that by HF CMRR
(if balanced) or just eat it if unbalanced pin-C-Gnd.
Else you may get chatter despite an attempt at
hysteresis (a 5V output edge vs a few hundred mV
of hysteresis, comes down to the C*R high pass
corner as regards effectiveness of hysteresis).

 

[KlausST]

Hi,

So would adding a cap, say 1nF, to the comparator pin, make it more noise susceptible?

a capacitor and a resistor creates a filter. If you want to reduce noise, then you need an LPF.
An LPF lets low frequencies pass but high frequencies become attenuated.

But I don´t understand the idea.
There is a 7ns fast comparator, thus I guess you are interested in fast signals.
So it makes no sense to use a capacitor / LPF.

Or maybe it´s just a lack of information. We just see the schematic ..
So we don´t know what´s important and what´s not.

Klaus