How to generate an impulse to test a system's impulse response

[alkhafaaf12]

what is the hardware we use to generate an impulse ?thanks

 

[Aussie Susan]

In theory an impulse has zero width and infinite height and so it not physcially realisable.
If a 'very narrow' pulse will do then you might manage it but that WILL a non-ideal spectrum and so not give you the exact measurments you want.
Depending on what you are trying to do, you might be able to use the relationships between a system's overall transfer function and its impulse and step responses. That way you can probably create a step input, measure the result and then use that to determien the transfer function.
Susan

 

[alkhafaaf12]

Hi Susan , thank for that, the whole idea of step input is to include "all" frequencies and be hence by applying this to the system we can able to determine the system's responce, the step function used in the function generator and spectrum analyser ( looks like DC On and Off to me ), its like switching a power supply On and OFF , which means the pulse it self is a pure DC on and off, how can the step in this case be useful if its a DC on and off , or am I missing something here? thanks.

 

[KlausST]

Hi,

An impulse containing all frequencies is called "dirac impulse".

which means the pulse it self is a pure DC on and off,

"Pulse" - and especially a "dirac impuls" - versus "DC" is the absolute opposite.

DC ON and DC OFF just means the pulse has a flat top and rather high rise and fall rate.
But here also a "flat top" is contradicting to a dirac impulse. A dirac impulse has no flat top.

As Susan already wrote: an ideal pulse is impossible.
Thus (liky one should do with all designs) you need to define the borders = specifications, requirements.

With this information you can design the "non ideal dirac like" pulse for your application.

But currently I see zero technical data, like voltage, current frequency range....so detailed assitance is impossible.

Klaus

 

[wwfeldman]

what is the hardware we use to generate an impulse ?thanks

what do you want to test?
how an impulse or a step is done depends on power levels, voltages, currents, etc
to get a reasonble result, one needs to build an appropriate gadget

 

[crutschow]

what is the hardware we use to generate an impulse ?

Depends upon what you are testing.
A pulse generator may work to provide a short pulse that will adequately emulate an "impulse" for your test purposes, if the pulse width is much shorter than the high frequency response of your system.

 

[kaz1]

Hi Susan , thank for that, the whole idea of step input is to include "all" frequencies and be hence by applying this to the system we can able to determine the system's responce, the step function used in the function generator and spectrum analyser ( looks like DC On and Off to me ), its like switching a power supply On and OFF , which means the pulse it self is a pure DC on and off, how can the step in this case be useful if its a DC on and off , or am I missing something here? thanks.

impulse input is not very practical for actual hardware testing (I don't see how step input can help). I suggest using frequency sweep (chirp) as input covering a useful range of frequencies then checking output by applying fft on both input and output and directly comparing.

 

[Aussie Susan]

If all you are doing (and we are all guessing here until you tells us) is to measure how a circuit responds over the designed input frequency range then use a spectrum analyser with a swept frequency input. They are (relatively) cheap.
I also think that you need to read up on Fourier transforms, and the various relationships that exist around transfer functions.
Susan

 

[dick_freebird]

A single step is the integral of an impulse applied,
you can gather about the same information from
either but a step is easier to create with basic
hardware - you only need one edge, not two
plus an infinitesimal dwell. Just have to massage
the data (in, and out) differently. A step function
capacitively coupled to a low-Z input will apply
an impulse current and create an impulse-ish
voltage on the other side, but this may not be
measurement friendly (Heisenberg didn't like
'scope probes much).

If you follow EPC they put out some marketing
chum about very narrow pulses for LIDAR etc.
made with their GaN FETs. The timescales shown
are sufficiently impulse-y for things like baseband
analog and sub-GHz RF.

 

[crutschow]

As an example, below is the LTspice simulation of a quasi-impulse (single pulse) applied to a 400Hz, 3-pole Butterworth, low-pass active filter.

You can check if the pulse width is short enough for an accurate test, by changing the width and seeing if it significantly affects the shape of the response.
The simulation shows the results of a 2V, 50µs pulse (red trace), a 10V, 10µs pulse (yellow trace), and a 20V 5µs pulse (green trace).
[The amplitude is inversely changed with the width to keep identical impulse energies (the volt-second product)].
Since the three outputs are essentially the same, this indicates the pulse-widths are short enough to emulate an impulse for this circuit with good accuracy.

 

[FvM]

The question has been posted in Digital Signal Processing forum, respectively we can't know which kind of hardware the OP imagines. In high speed analog circuit design, a step signal is an appropriate means of wide band excitation, e.g. for characterization of OPs or TDR measurement of transmission lines. Rise time varies depending on the application, can be down to ps with tunnel diode generators.