Vibroseis seismic data processing

[jefry123]

Hi everyone,

I am going to use the SEG Wiki open data (https://wiki.seg.org/wiki/Open_data) to learn more the processing. There is a 2D land data with Vibrator as the source named as "Poland 2D Vibroseis Line 001" (https://wiki.seg.org/wiki/2D_Vibroseis_Line_001). So, I read about vibroseis data processing and its difference with explosive source. There exists a sweep generated by vibrator and the uncorrelated trace (which is the field data, raw shots) is actually the convolution of this sweep with earth reflectivity. Then, this uncorrelated trace should be cross-correlated with the sweep to have the correlated trace, and this correlated trace goes to be processed, right?

My first question is that since the sweep length has a long length, what would be the length of field data (uncorrelated trace)? Shouldn't it be sweep length+listen time?

My second question is that how can we realize if the raw data is the correlated one or the uncorrelated one? The data I mention, Poland 2D Vibroseis Line 001, says that the source is vibroseis and the characteristics of the sweep is also given, but, the length of the data is 3 seconds. Should I do the cross-correlation of this data with the synthetic sweep generated by the given characteristics?! or it's already done?! The following is a picture of the first shot which I taught may be useful:

[D_Reynolds]

Hi Jefry123

This data is correlated vibroseis and most likely zero phase. You are correct to realize that uncorrelated vibroseis data should be much longer = listen time + sweep length.

Because typical processing will at some point apply a deconvolution step, this assumes input to Decon will be minimum phase. With the sweep trace there is a way to use that to compute a filter operator to convert your zero phase data to minimum phase before doing decon.

Good luck with your processing!

Dave Reynolds

[GuyM]

Looks to me like the data is already correlated; its usual to get the data with the correlation already applied.
The record would be a lot longer if it was uncorrelated - you'd have the target record length plus the sweep length.

[jefry123]

So; somehow the length of the record could be a criteria to know if it is correlated or not.

And, why the sweep characteristics are given? My first taught was that it may be used to find the Klauder wavelet obtained from autocorrelation of the sweep and then be used to find the minimum phase operator to be applied to the data.

[GuyM]

Well if you wanted to collect data to a maximum depth of 3 seconds with a 6 second sweep, you'd need to record for a total of 9 seconds.
It's pretty usual to give all of the acquisition parameters even of the data has been deconvolved already..