Jonas, Patrick, all,

I think this is fixed as of the latest commit.  There is one thing I am
unsure about --- there is an absolute sign that I am a bit confused about
because I have to normalize to -1/c --- but I think this is working now.

The qpack test seems somewhat reasonable at least.

//Richard

2016-11-05 2:52 GMT+09:00 Patrick Eriksson <[email protected]>:

> Hi,
>
> Jonas: Yes, I confirm that retrieved winds in qpack2_wind3d_demo are
> coming out as zero. I have run this demo several times, but failed to
> notice this. So this error has probably been around for quite some time!
> Sorry.
>
> Richard: I have confirmed that spectra with zero and non-zero winds
> differ, but that wind jacobians seem to be zero. The cfile produced by
> qpack attached. Can you have a look? Have I missed to adjust qpack in any
> way?
>
> No more time today to look at this. I will try to look at this further
> tomorrow. Just wanted to give a quick response.
>
> Bye,
>
> Patrick
>
>
>
> On 2016-11-04 11:08, Richard Larsson wrote:
>
>> All,
>>
>> This might be my fault.  I have no idea how this particular qpack-case
>> works but there was an overhaul in the design of the Jacobians in ARTS
>> last year that made their calculation less forced and more theoretical
>> (i.e., the jacobian before was from partial perturbations but it is now
>> from direct derivation as much as possible).  For the wind field this
>> means that the Jacobian is calculated based on the derivative of the
>> transmission with respect to frequency before being scaled back to wind
>> field parameters.
>>
>> There could be edge-cases that I have miss-implemented.  It sounds like
>> this at the least.
>>
>> Could you explain your test case to me so that I can look into fixing
>> the problem?  What is the atmospheric geometry and what is the a priori
>> wind field?  What kind of results do you expect?
>>
>> //Richard
>>
>> 2016-11-04 18:01 GMT+09:00 Jonas Hagen <[email protected]
>> <mailto:[email protected]>>:
>>
>>     Dear all
>>
>>     I'm currently porting my wind retrievals from ARTS/atmlab v2.2 to
>>     v2.3 (developement version). But somehow the wind retrievals just
>>     return zeros (wind_{u|v}_x and wind_{u|v}_A in L2 are exactly 0).
>>     This is also the case for the qpack2_wind3d_demo.m. Can anyone
>>     confirm this? I could need some help with the debugging. There are
>>     no error messages or warnings.
>>
>>     Best regards,
>>     Jonas
>>
>>     --
>>     Jonas Hagen
>>     Institute of Applied Physics
>>     University of Bern
>>     Sidlerstr. 5, CH-3012 Bern, Switzerland
>>
>>     _______________________________________________
>>     qpack.mi mailing list
>>     [email protected] <mailto:[email protected]>
>>     https://mailman.rrz.uni-hamburg.de/mailman/listinfo/qpack.mi
>>     <https://mailman.rrz.uni-hamburg.de/mailman/listinfo/qpack.mi>
>>
>>
>>
>>
>> _______________________________________________
>> qpack.mi mailing list
>> [email protected]
>> https://mailman.rrz.uni-hamburg.de/mailman/listinfo/qpack.mi
>>
>>
_______________________________________________
qpack.mi mailing list
[email protected]
https://mailman.rrz.uni-hamburg.de/mailman/listinfo/qpack.mi

Reply via email to