From AstroWiki

DR: CaveatsWhenPlottingTheGroupDelay

When plotting the groupdelay ( d=getdetlay(tag+'.groupdelay.fits') ) on top of an image of the Fourier transformed spectra (see Diagnosis of fringe tracking), one has to be careful with both the time / frame axis and the delay axis. For the time axis the main caveat is that the time of the fringe track is not NDIT * DIT, but given by the difference between the times (e.g. in t=oirgetdata(tag+'.fringes.fits') & t = t.time ) of the last and first entry. For a track of 8000 frames this is roughly 170s (where NDIT*DIT is only 144s for the default DIT of 0.018s). To find out about the delay scale, one can make use of the header keywords OPD0, OPD1 and OPD2. These keywords hide in the IMAGING_DATA table of any .groupdelay.fits file and you can read them by doing

OPD0=1.e6*double(midigetkeyword('OPD0',tag+'.groupdelay.fits',extna='IMAGING_DATA'))

They give the scale and offset of the image in delay space:

• OPD0 is the OPD of the middle pixel (in micron)
• OPD1 is the difference between the middle and the upper border (in micron)
• OPD2 is the scale (in micron/px), also called 'OPD stepping'. It is mostly negative which means that you should either mirror your image horizontally (IDL: reverse) or plot the delay axis such that values increase towards the lower border of the image