Fringes pattern
in spectra

Page written by B. Mauclaire

Page updated  2010-05-24


Many spectrograph users observed in their spectra fringe patterns. In low resolution such a pattern looks like waves in star's continuum which can be interpreted as small emission lines. In high resolution spectra the pattern limits resolution for small detail analysis. These patterns are caused by the {spectrograph-ccd camera} system since they are visible in flat-fields.

In order to know whether these patterns come from the spectrograph or from the camera, and who is responsible of what, we gather several amateur observations to study as many as possible {spectrograph-ccd} combinaisons. We attempt at drawing preliminary conclusions.

Finally, we review techniques proposed by some amateurs that may help, in somes cases, to remove fringes.

Flat-field obtained for several combinaison {spectrograph-camera}:

1. Working with low resolution spectrographs (eg. 150 g/mm gratting):

Here are some flat-fields obtained with spectrographs up to R=2000.

Configuration description Flat-field picture
KAF1602E(Audine)+LHIRES3 150 g/mm
B. Mauclaire
KAF1603ME(ST8)+LHIRES3 150 g/mm
B. Mauclaire
P. Lailly

Images's top comes from 100 µ slit and bottom comes from 50 µ slit.

Links to related pages:

2. Working with medium resolution spectrographs (eg. 1200 g/mm gratting):

Here are some flat-fields obtained with spectrographs up to R=10000.

Configuration description Flat-field picture
KAF1603ME(ST8)+LHIRES3 1200 g/mm
P. Lailly
kaf1603me 1200
KAF1600(Audine)+LHIRES3 1200 g/mm
V. Desnoux
K640+LHIRES3 1200 g/mm
L. Schanne

Links to related pages:

3. Working with high resolution spectrographs (eg. 2400 g/mm gratting):

Here are some flat-fields obtained with spectrographs up to R=20000.

Configuration description Flat-field picture
KAF1603ME(ST8)+LHIRES3 2400 g/mm
P. Lailly
ST8 2400
20D+LHIRES3 2400 g/mm
P. Schlatter
DSI2pro+LHIRES3 2400 g/mm
P. Schlatter
ATK1HS2+LHIRES3 2400 g/mm
P. Schlatter
Canon 350D+LHIRES3 2400 g/mm
C. Buil
350d 2400
KAF1603ME(ST8)+LHIRES3 2400g/mm
B. Mauclaire
KAF1603ME 2400

Links to related pages:

4. More tests are required

We would be very grateful if spectrograph providers could help their customers to understand what happens. We would like to make available a methodic study involving:

Possible explanations:

1. Pnenomena related to the camera:

Peter Schlatter has carried out a thorough study about the relationship between ccd sensor's protection glass and ripple's width. As this glass has parallel faces, it seems to produce interferences at visible wavelengths and with normal incidence light, with fringe patterns depending on the glass's thickness and on the grating used in the spectrograph.

In the other hand, some people seem not to get such fringe patterns. This could come from their ccd sensor's glass that may have a specific thickness. So we now need to know every cover glass's thickness, especially for Canon 350D and KAF400/1600 ccd sensor, in order to confirm or not the above-mentioned relationship.

2. Phenomena related to spectrographs:

It has been shown that the tuning of LHIRES3's main mirror has an effect on fringes' intensity.

Possible solutions:

1. Spectrograph's setup:

As C. Buil showed in OHP 2007, fringe patterns depend on LHIRES3's main mirror setup and may desapear when neon lines are as close to vertical as possible. But this has consequences: more important geometrics deformations and perhaps less light concentrated on ccd.

J. Ribeiro tilted his main mirror and fringes appear with less stretch.
O. Thizy suggests to tilt a bit the grating position.

2. Software's solution:

C. Buil and V. Desnoux have described a method in OHP 2007's spectro star party (bottom of the page) and OHP 2009 to deal with ripple patterns by extracting the "high frequency" component of the flat-field.

  1. Take a flat-field during the same night as you acquire other spectra:
  2. Do a blur filter or Gaussian filter on flat-field in order to get only the low frequency variations:
  3. Extract the high frequency intensity's variations by dividing the original flat-field by the low frequency flat-field:
  4. Divide all spectra by the low frequency flat-field;
  5. Do a wavelength calibration of the high frequency flat-field in order to determine the horizontal offset to apply on it;
  6. Divide all spectra by this new high frequency flat-field.

As B. Mauclaire and V. Desnoux showed in their tests, this method can be applied only if:


Today the origin of fringe patterns in spectra is not clearly understood: more tests are required.
Fringes pattern are visible in flat-field but also in stellar spectra. Note that fringes have its pseudo-period and intensity growing toward spectrum's blue side. So, dividing the sepctrum by such flat-field may sometime remove fringes and sometime increase fringes' intensity.
There is no universal method to easily remove ripples in any material setup.