There is a possible problem in the flux calibration of S06 science observations, traceable to memory effects left over from reference scans. The intended science objectives may then be compromised in portions of the spectrum.
A user entering data for a S06 or S07 SW wavelength range scan has to enter three wavelengths - the start wavelength, end wavelength and a reference wavelength - and two fluxes - the flux at the reference wavelength and the maximum flux expected over the wavelength range. If the signal at the reference wavelength is considerably higher than that in the rest of the wavelength region memory effects may be introduced by the reference scans that remain in subsequent scanning operations.
An affected S06 observation, a deep scan concentrated in band 2, is shown in
figures 2.1 to 2.3. Figure 2.1 is a plot of
the wavelength seen by detector 13 during part of the observation. At
approximately 2100, 2700, 3300, 3900 and 4500 s reference scans are taken. The
user selected a wavelength of 7.7 
for the reference scan. Between 2100
and 3300 s the grating is scanning down in wavelength (up in LVDT), interrupted
by a reference scan, and between 3300 and 4500 s it is scanning up, again
interrupted by a reference scan. We therefore have four scans and five
reference scans.
Figure 2.2 shows the SPD output for the same time period. The
reference scans have a higher signal than the astronomical source for all but
the reference scan at 3300 s. In particular, the signal at 7.7 (2100 and
4500 s) is factor 
8 higher than at 12 (in 
V/s).
The net result of this is that the reference scans leave memory effects in all
up and down scan fragments, mainly because the signal at the reference
wavelength is too close to peak signal in the 7-12 region. The shape of
the SPD after the first reference scan (2100 - 2500 s) is markedly different
from that just before the last (4100 - 4500 s) even though they cover the same
wavelength. This memory effect is purely because of the first reference scan.
Lesser effects can be seen around the other reference scans.
The effect on the AAR, shown in figure 2.3, is just as marked. The
difference between the up and down scans can clearly be seen between 11 and 12
.
Of the four scans, three are too badly affected to be used in their entirety. Only the third scan is useable, but this too may have memory of the previous flux history (scans 1 & 2 and preceeding reference scans).
Figure 2.1: Wavelengths as a function of time
Figure 2.2: SPD data from detector 13 for the same time interval
Figure 2.3: AAR from the above. The dots come from the up-scans, the
crosses from the down-scans. The cluster of points around 7.7 microns come
from the reference scans
Observations which may be affected are limited to band 2 or 4 observations with AOT S06, and only if the reference scan signal is several times the average signal.
Band 1 and 3 are not be affected (as they do not suffer from memory effects) and AOTs S01 and S02 do not use reference scans.
AOT S07 observations also use reference scans, as these are needed to tie the SW dark measurements to the corresponding SW grating scans done in parallel with the Fabry-Pérot operations. Possible effects are limited however, since there are no AOT S06 type discontinuities, shown in Fig 2.1.
The solution adopted by the SOC is to delete reference scans from all observations with S06. This change will take place from mid-June onwards.