Context:

For this example, I compare EW's measurements of zet Tau computed by authors of Ernst Pollmann's survey group with the results obtained using SpcAudace and MIDAS.

Data used for this study:

We can see that measurements of JD-2450000 54506, 54509 and 54512 are rather different. Let's do some computations.

Comparative computations:

We can see that measurements of JD-2450000 54506, 54509 and 54512 are different. Here follow EW computations between 6535 A and 6595 A:

JD-240000	Date (Y-M-D)	Pollmann's spectra (Vspec)	Pollmann's spectra (SpcAudace)	Pollmann's spectra (MIDAS)	Mauclaire's spectra (SpcAudace)	Mauclaire's spectra (MIDAS)	Mauclaire's spectra (Vspec)
54506	2008-02-09	-14.3	-14.33±0.20 (SNR=388)	-14.34	-13.47±1.07 (SNR=73)	-13.75	-13.5
54509	2008-02-12	-15.6	-15.64±0.26 (SNR=490)	-15.54	-14.75±1.38 (SNR=93)	-14.72	-14.6
54512	2008-02-15	-16.0	******	******	-15.50±1.87 (SNR=54)	-15.59	-15.6

Fits files can be download by selecting blue EW's values.

1. MIDAS's results:

Here are MIDAS's screenshots:

bmauclaire : 2008-02-09 (continuum n°1)	bmauclaire : 2008-02-09 (continuum n°2)
bmauclaire : 2008-02-12	bmauclaire : 2008-02-15
epollmann : 2008-02-09	epollmann : 2008-02-12

It appears that when selecting point' limits with mouse on the graph, EW's value is sensitive to the intensity selected at one wavelength. I used to select the two points that makes the continuum's line formed going through star's continuum mean value especially if continuum is noisy.

2. SpcAudace's results:

On the other hand, the automatic algo in SpcAudace (spc_autoew command) does:

1. spectrum normalization by continuum extraction (as shown in left picture bellow);
2. detection of integration wavelength's limits by finging when intensity comes closed to 1 (as shown in right picture bellow);
3. integration by trapeze method and dividing by mean continuum's intensity.

Examples of spc_autoew two ways usage:

> spc_autoew zeta_tau_20080215 6535.0 6595 # Le continuum vaut 29.3647846959 # Les limites non nulles sont 3 ; 1468 ... # # Date: 2008-02-15T23:06:07.77 # JD: 2454512.4625899307 # EW(60.0=6535.0-6595.0)=-15.5 A. # Sigma(EW)=1.87 A. # SNR=54.

> spc_autoew zeta_tau_20080215 6563 # Le continuum vaut 29.3647846959 # Les limites non nulles sont 3 ; 1468 ... # Le continuum entre 6605-6671 vaut 0.999699 # L'aire sans le continuum vaut 131.883706 # # Date: 2008-02-15T23:06:07.77 # JD: 2454512.4625899307 # EW(38.13=6545.1-6583.23)=-15.38 A. # Sigma(EW)=1.3 A. # SNR=54.

If the spectrum is allready normalized:

> spc_ew zetatau_20080212 6535 6595

# Le continuum vaut 0.99789038961
# Les limites non nulles sont 0 ; 767
...
#
# Date: 2008-02-12T17:44:32
# JD: 2454509.2392592593
# EW(60.0=6535.0-6595.0)=-15.64 A.
# Sigma(EW)=0.26 A.
# SNR=388.

The following function may allow you to normalize spectra:

> spc_autonorma zeta_tau_20080215

# Le continuum vaut 29.3647846959
# Les limites non nulles sont 3 ; 1468
...
# Fin de la division : 70 divisions par 0.
# Division des 2 profils sauvée sous zeta_tau_20080215_div.fit
# Profil normalisé sauvé sous zeta_tau_20080215_norm
# zeta_tau_20080215_norm

There are no human actions or mouse selections that may introduce errors. It's only pure computation. All limits are found with mathematical cirteria.

Conclusions:

As in any group surveys, measurements at one date may have dispersion and this avoid systematic error's propagation and bias. However we have to control what are softwares' limits in such computations.

If we base our comparaison on MIDAS's results, as it is used by professionnal astronomers, SpcAudace and Vspec (with a continuum well estimated by user) computation's results fit for use as they are closed to MIDAS's one.