How to understand analog signal distortion as a funtion of varying group delay?

[bcro]

My question really is about as general as the title indicates, but here's the specific context:

In the lab we often need to digitize spectroscopy signals, which consist of a string of more or less Gaussian pulses. The shape and spacing of these pulses are what we measure (we often fit curves to them in the computer). I'm trying to design some anti-aliasing filters, and it would be very convenient to use the LTC 1564. I'm attaching a snippet of the spec sheet that shows the change in group delay over frequency:



I'm realizing I just don't have enough experience with such things to look at that plot and decide if I should worry about distorting my pulse shapes. Lets say I'm conservative, and keep my signals down around 1/2 the stated cutoff frequency. Does a variation of 25 microSeconds over that range translate into a measurable difference in the pulse coming out of the filter? How does one go about answering such questions?

Thanks in advance for your help. Have a great day.

--Brad

 

[crutschow]

My (limited) understanding is that if you have a constant group delay within the frequency of interest, than the waveshape is preserved. Since the group delay is fairly constant up to 1/2 the cutoff frequency in the graph, then I would expect the pulse shape to not be significantly changed. But I don't know how to quantify that. Simulation would be the easiest way to do that but there doesn't seem to be a Spice model for the LTC 1564.

 

[bcro]

Yeh, I've been thinking about it more, and I guess without simulation there's really not much more you can extrapolate that's not already in some of the other input/response type plots on the spec sheet already. I guess its one of those things that just takes some more experience to get a feel for. All in good time.

Well, thanks again, crutschow, for jumping in there with me. I'm having to shelve this project for a few days, but I'm sure I'll be back again with some other question before long.

Have a great day.

 

[chuckey]

In TV engineering a raised cosine pulse is often used for this very purpose. It is viewed on a CRO with a special graticule with 2 and 4% limits engraved. You could try and google "2T pulse", or "raised cosine pulse". I am not sure how close it is (or indeed your pulse) to a true Gaussian pulse.
Frank