0
\$\begingroup\$

enter image description hereenter image description hereI wanted to make an comparator with cascade of amplifiers of equal gain, output impedance etc. We know that we can approximate it as single pole system , thus a resistor and a capacitor in the output will do the approximation. I was able to calculate the gain for first stage i.e $$-g_m * R_o * \frac{t }{R_o C_o} $$

This is when t<< time constant. Now as stages increases , I want to derive the speed and delay. Can someone help me or share materials that contain this derivation?

\$\endgroup\$
9
  • 1
    \$\begingroup\$ What is your definition of speed and delay. Usually, we talk about phase delay for a given frequency. For that, you need to express the gain as a transfer function which depends on frequency and not tie it to the time variable t. By the way how did you calculate the expression for the gain given in the question? Does it take into account the effect of the input impedance of the next stage? I see only \$R_o\$ in the expression. Does that include the effect of loading due to the next stage? \$\endgroup\$
    – AJN
    Commented May 30, 2021 at 6:23
  • \$\begingroup\$ Can you please go through the image ? its a snippet from a text book. I wanted to derive that equation. I got stuck when they just convoluted to get the output for stage two. \$\endgroup\$ Commented May 30, 2021 at 6:33
  • 1
    \$\begingroup\$ "I was able to calculate the gain for first stage". This sentence is probably wrong. The expression written below this sentence seems to be the time response for step input \$t^0 u(t)\$ for first stage and corresponding inputs for later stages \$k_n t^n u(t)\$. You use the word convolution; but it is not mentioned in the snippet. The previous page of the snippet probably has that part. Did you understand the derivation of the impulse response and the calculation of amplifier response that would have been given in the previous page? \$\endgroup\$
    – AJN
    Commented May 30, 2021 at 6:43
  • 1
    \$\begingroup\$ What is the form of the impulse response ? \$u(t)\$ or \$e^{-t/\tau}\$ ? Whatever it is, convolve the input (probably step input; you have not specified) with the impulse response to get output of second stage. Then take that result and convolve with the impulse response of the second stage to get its output. Then take its output and convolve with impulse response of third stage and then ... \$\endgroup\$
    – AJN
    Commented May 30, 2021 at 7:12
  • 1
    \$\begingroup\$ It would interesting to compare this work, which optimises the gain of a comparator with time as a constraint using identical sections, with that of Oliver Collins 'Design of low jitter hard limiters', which optimises the timing jitter performance of a comparator using a cascade of sections with different gains and time constants. \$\endgroup\$
    – Neil_UK
    Commented May 30, 2021 at 9:38

1 Answer 1

1
\$\begingroup\$

Assume that \$V_i = 1\$. It doesn't change the essence of the problem since a linear time invariant system is assumed. Just multiply the final answer with \$V_i\$.

The step response is approximated as \$k\cdot t\$ in equation (6.2), where \$k\$ is some constant (\$\frac{g_m}{C_0}\$). So the corresponding (approximate) impulse response is \$k\ u(t)\$, the unit step function. So, you need to convolve the output of previous stage with this impulse response to get output of next stage.

Input of first stage is unit step. Output of first stage is \$k\ t\$ equation (6.2).

So, the input of second stage is \$k\ t\$. When convolved with \$k u(t)\$ we get the output as $$ \int_0^t k \cdot u(\tau) \times k\cdot (t-\tau)\ d\tau = k^2 t^2 - k^2 t^2/2 = k^2 t^2/2 $$

This is the input of the third stage. When convolved again with \$k u(t)\$ we get the output as (checked via Wolfram Alpha) $$ \int_0^t k \cdot u(\tau) \times k^2 \cdot (t-\tau)^2/2\ d\tau = k^3 t^3/(2\cdot 3) $$

Continuing, the nth stage output is \$k^n t^n / n!\$.

This is what is given in (6.4).

note

If you are familiar with Laplace transform, the difficult to perform convolutions simplify to multiplication. Try that route if convolutions are difficult for you.

\$\endgroup\$
1
  • \$\begingroup\$ Thanks for the answer @AJN \$\endgroup\$ Commented May 30, 2021 at 9:08

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.