EMRFD Message Archive 4008

Message Date From Subject
4008 2010-01-18 13:27:25 aa7hq@comcast.net impulse response, graceful or ???
rick, i have some questions about impulse response functions at radio frequencies.

one, would you or some other member of this group please send me one or too examples?
my simple-minded googling did not find any such.

two, if you or some other member of this group have in fact measured the impulse response of some rf device, would you please tell me how you did it. it seems
formidable to me.

de ed, aa7hq

[Non-text portions of this message have been removed]
4009 2010-01-18 14:25:52 pchamilt1126@aol.... Re: impulse response, graceful or ???

I am not Rick (KK7B ? he is my first Elmer and one of my instructors ) but I am both an electrical engineer and a grad student in EE, so I will take a swing at helping you. One cannot get an impulse generator because an impulse's energy is confined to infinitely small time. Even step generators like the ones in TDRs have a specified rise time, they are not ideal steps. As such one cannot measure the impulse response of an RF device because one cannot get an impulse generator.

An impulse is the time derivative of a step and the impulse response is the time derivative of the step response. EE's use the step response and the impulse response to characterize the frequency response of a circuit because it is mathematically convenient [ sorta :-) ] to do so.

The classic teaching example is the response of a low-pass RC filter. If you will write me back direct I will find such an example (to be sure I do not miss-lead you ) and lay it out in the best detail I can manage. You need differential and integral calculus to appreciate such exercises so some serious skull sweat will be required.


Paul Hamilton, KE7UAE

-----Original Message-----
4013 2010-01-18 19:11:25 Gary Re: impulse response, graceful or ???
> two, if you or some other member of this group have in fact measured the impulse response of some rf device, would you please tell me how you did it. it seems
> formidable to me.

I currently work in ultrawideband systems where the time response is extremely important. We use two methods to measure impulse response. First, we use a vector network analyzer (VNA), which measures the complete transfer function of the device under test, including magnitude and phase. By definition, that is the impulse response in the frequency domain. You can do an inverse Fourier transform on that data, and it will produce the equivalent time-domain waveform. Modern commercial VNAs include this as a built-in feature. Look up impulse response on Wikipedia.

The other way we do it is directly in the time domain. We use very fast impulse generators, for instance from Picosecond Pulse Labs. These instruments typically generate a step with a very fast edge, then differentiate it with a circuit known as an impulse forming network, or IFN, which is basically a series capacitor. The resulting Gaussian impulse has a known amplitude and pulse width and after passing through the device under test, the output can be observed on a fast oscilloscope. Some math is required to deconvolute the properties of the driving impulse if you want really accurate results.

We do both of these measurements routinely up to 26 GHz, but it's not cheap nor is it trivial to do... For HF applications, the ham-grade VNAs such as the N2PK VNA or the DG8SAQ VNWA both have software that will do the inverse FFT. One could also make a fast step generator (74AC-series logic can give you a 5 V step with a 300 ps risetime), then design an IFN to provide an impulse of the desired width.

(In my experience, I think a simple step response is probably the most useful time-domain measurement when characterizing filters, AGC loops, and other radio-related circuits.)

Gary, WB9JPS