Hi, Thanks for the responses. I have been using the frequency grid for both LUT and RT calculations. This would be f_grid, right? I am using the same Q structure to create the table as to invert it.
I set Q.OUTPUT_FILE_FORMAT and Q.INPUT_FILE_FORMAT both to 'binary' but it did not change the error. After rounding the values in f_grid to 10kHz, there is no error anymore. If there are differences smaller than 10kHz in the f_grid values, then the error returns. This new f_grid is fine for the inversions I am doing anyway (the old f_grid was found by some optimization technique, I think). I still don't know what exactly causes the error, or whether it is specific to my system setup, but it would appear to be rounding or a check of some sort. I should try it on another system to see. Also, I find that, with the 'binary' option for both the Qpack input and output, the inversions (using a LUT, anyway) are about twice as fast. Which is great. Cheers, Niall Quoting Stefan Buehler <[email protected]>: > Hi, > >> do you use the same freq grid for both the LUT calculation and the >> RT calculation? if so, see patrick&stefan. if not, it might be an >> interpolation issue. i don't remember what is the status on freq >> interpolations when using LUT (whether allowed, forbidden, or >> allowed by setting some flag...) - stefan, do you remember? > > The latter. It is allowed, for example to allow wind retrievals from > the Doppler shift. But the default is to be strict. I think this is > documented in built-in documentation and/or user guide. > > /Stefan _______________________________________________ qpack mailing list [email protected] https://www.sat.ltu.se/mailman/listinfo/qpack
