Comp. Prop Delay: Significance & How to Choose High Speed Comp

[likephysics]

How is comparator prop delay significant?
It's just delay, isn't it. Input occurs at say, t secs, output will occur at t+x secs.
I am looking for a high speed comparator, I don't understand why manufacturer's classify speed based on propagation delay. Isn't rise time the factor that decides speed -
tr= 0.33/freq?

 

[berkeman]

How is comparator prop delay significant?
It's just delay, isn't it. Input occurs at say, t secs, output will occur at t+x secs.
I am looking for a high speed comparator, I don't understand why manufacturer's classify speed based on propagation delay. Isn't rise time the factor that decides speed -
tr= 0.33/freq?

The equation you list is not applicable to the rise and fall times of a gate output. The equation relates the spectral width of a square wave's spectra to its rise and fall times.

The prop time of the comparator is important for several reasons, but one is that you need the positive feedback hysteresis to get back to the input quickly enough to prevent oscillations. That's one of the most important parts of high-speed comparator circuit design and PCB layout -- preventing oscillations. The high-speed comparator that you have chosen probably has some layout guidelines to help with the oscillation issues.

 

[likephysics]

The equation you list is not applicable to the rise and fall times of a gate output. The equation relates the spectral width of a square wave's spectra to its rise and fall times.

The prop time of the comparator is important for several reasons, but one is that you need the positive feedback hysteresis to get back to the input quickly enough to prevent oscillations. That's one of the most important parts of high-speed comparator circuit design and PCB layout -- preventing oscillations. The high-speed comparator that you have chosen probably has some layout guidelines to help with the oscillation issues.

aha! Got it. So if I don't use +ve feedback, prop delay shouldn't matter?
Anyway, I was just testing LM311 this morning on the breadboard. Worked great upto 1MHz, but then the output was high all the time. I thought, the rise time of the input was too fast for LM311.
Now I get it. The feedback is taking too long. Any equations that relate prop delay with +ve feedback?

 

[berkeman]

aha! Got it. So if I don't use +ve feedback, prop delay shouldn't matter?

I suppose it won't matter, because the comparator will be singing (oscillating)

Anyway, I was just testing LM311 this morning on the breadboard. Worked great upto 1MHz, but then the output was high all the time. I thought, the rise time of the input was too fast for LM311.
Now I get it. The feedback is taking too long. Any equations that relate prop delay with +ve feedback?

I don't know that it's the feedback not being fast enough that's keeping the output from responding. It may just be internal delays that can't keep up with the fast input switching.