5 Three-telescope software architecture and control
5.1 System Overall Architecture
6 Three-telescope acquisition and stabilization results
The cophasing of two telescopes was presented in our Zero OPD Breadboard Design Report [4] as well as our preliminary ideas on the cophasing (active control of the phase) of 3 telescopes in particular the control strategy of the couple of interferometers of the 3 interferometers in serie or in parallel (e.g. Fig. 2 of [4]). Indeed if the OPD (Optical Path Delay) between the telescope 1 and the telescope 2 and the OPD between 2 and 3 are equal to zero, then the OPD between 1 and 3 is necessarily null. The third interferometer, between telescopes 1 and 3, allows to verify this last point.
This simple servo-control scheme is made possible by the form of the fringes’ pattern. For the zero OPD (and for each extrema) the signal given by the synchronous detection at simple frequency is null while the signal given by the synchronous detection at double frequency is a minimum or a maximum (see Fig. 1).
Fig. 1 - Fringes’ signal and signals from the two lock-in amplifiers.
With a large spectral source (like the Sun) we have only a few fringes and a distinguishing central peak. Yet, the fringes’ signal is symmetric only if the different beams pass through the same type of glass and the same thickness (see Fig. 2). For our current set-up, this effect is not particularly important, but for the final realization (with molecular adherence) it can be a major source of error, when several interferometric couples are considered, if tolerances on the setup are not tight enough.
Fig. 2 - Influence of an unequal differential thickness (∆e) of glass.
Fig. 3 – Early setup of the 2-telescope breadboard used in laboratory and on the sky in 1997.
Fig. 4 - Specific 2-beam setup used for the low flux study with 6 beamsplitters and 9 reflections (total transmission of only 0.3 %).
Although the fringe signal was not perfect (low contrast, bad servo control, electronic noise) we made measurements of phase control with two telescopes:
|
Optical
density |
Flux
(nW) |
Visual
Magnitude |
Vpp
SD (V) |
s
fluctuations (mV) |
Stability
l/
(pp@ 6s) |
|
None |
0.48 |
2.2 |
2.6 |
1.2 |
361 |
|
0.3 |
0.24 |
3.0 |
3.4 |
2.6 |
207 |
|
0.6 |
0.12 |
3.7 |
3.4 |
3.8 |
149 |
|
1.0 |
0.048 |
4.8 |
1.3 |
6.6 |
50.5 |
|
1.3 |
0.024 |
5.5 |
3.5 |
38.5 |
15.2 |
|
1.6 |
0.012 |
6.3 |
1.6 |
26.1 |
10.2 |
|
2.0 |
0.005 |
7.3 |
0.8 |
23.1 |
5.8 |
These are indicative, as reported in the OPD Breadboarding Activities handouts presented at Progress Meeting II [5]. In fact, if we account for the low contrast (28%) and improper collimation used at that time, 2 to 3 magnitudes could be gained and a magnitude 7 star could probably be used to achieve a 3s stability of l/100.
Following several years of studies of UV imaging interferometers and demonstrations of the cophasing of 2 telescopes, we have developed the actual laboratory breadboard of a 3-telescopes interferometer (see Figs. 5 & 6).
Fig. 5 – Conceptual layout of the three-telescope breadboard.
|
|
|
|
|
|
|
|
|
Fig. 6 - Demonstration breadboard of the 3-telescope cophased interferometer SOLARNET.
The control software, "InterfLab" (standing for Interferometry Laboratory software) was designed to:
This document shortly presents the architecture chosen to implement the system and describes its main components.
To maximize the responsiveness of the servo-controlling machine, the system was distributed over multiple machines. The Experimentation Server realizes interferometers cophasing task. It also exposes the acquired data and sends events and log messages to a unique client. This client can be of two types: a final user client or a dispatching and archiving server named the Data server. While the first resulting configuration was designed for an Intranet single user usage, the second configuration was imagined to interface the experimentation to the Internet and therefore, to be accessible to multiple users. Those users can enter the virtual InterfLab Laboratory as simple visitors or, if they have the required privilege, as a controlling experimentator.
Fig. 7 - The single user intranet architecture
Fig. 8 - The multiple users Internet architecture
The experimentation server is the core component of the system; its purpose is related to the cophasing process and should not be dependent on the execution environment. To respond to this last requirement, the component was written in pure C++, STL (the C++ Standard Template Library available on every good implementation of the language), and a personal library named the XTL (eXtension to Template Library) easily portable to any specific platform. Consequently, the current component – fully functional and debugged – can be directly reused on a configuration different than the laboratory configuration, for instance, for a spatial version.
To interface this platform free component to the laboratory needs, a COM layer was added. Therefore the experimentation component is easily insertable in any COM compatible application. The best demonstration of this possibility is the final user client mentioned previously. This single user interface to the experimentation was implemented within Excel in VBA (Visual Basic for Application) language. Just imagine: to start a new experimentation on any machine of the laboratory domain, you just have to open Excel, select the InterfLab document template, configure the cophaser and nested interferometers, click on Start command of the InterfLab custom toolbar and results fill appropriate worksheets of your Excel workbook!
Fig. 9 – The single user configuration
In the same way, the data server is a COM application, which play the role of a bridge between the experimentation server and the Internet: it assures the asynchronous communication with Java applets of multiple clients, those applets embedded into laboratory site web pages.
Fig. 10 - The multiple users configuration
As described previously, system main component is the experimentation server. In its implementation, a class abstracts each part of the experimentation table; therefore we can encounter a synchronous detection class, a plate class, an analog_input class, etc. But the principal classes of abstraction are the cophaser and its nested interferometers:
Lets focus on them.
The cophaser does not correspond to a physical entity. It is the software abstraction of the interferometers cluster to servo-control. It sequences interferometers runs and centralizes interferometers notifications and events. It is the only one operational public interface of the experimentation.
Two modes of sequencing are possible: the serial running mode and the parallel running mode. In the serial mode, each interferometer is processed the one after the other. The whole process is then long but it allows the maximum hardware reusability. In the parallel mode, the interferometers cluster is splitted into two branches and each branch is processed simultaneously. It is two times faster but requires more hardware.
The role of the interferometer class is to detect fringes region, to detect the central fringe and to servo-control. All of those operations are performed on only one specific interferometer. Three running mode was defined for an interferometer run; each mode is an extension of the previous one.
a) The “simply scan” mode. In this mode, the interferometer run algorithm is:
Fig. 11 - The "simply scan" mode algorithm
b) The “find fringes region then stop” mode. In this mode, the interferometer run algorithm is:
Fig. 12 - The "find fringes region the stop" mode algorithm
c) The “find fringes region, the central fringe and servo-control” mode. In this mode, the interferometer run algorithm extend the previous one as follow:
The quality of the experimentation server component results rests on the quality of the background data acquisition process. We explain here how each measurement is done.
For each signal channel, the analog signal is digitized in 12bit samples:
è
0x13, 0x24, 0x27, 0x28, 0x2b, 0x12, 0x04 …
This flow of data is segmented into windows of n samples. ‘n’ is fixed by the user for each step of interferometers run under the XMeasurementWidth parameters (for instance, ScanningMeasurementWidth):
0x13, 0x24, 0x27, 0x28, 0x2b, 0x12, 0x04 è 0x24, 0x27, 0x28
Windows are not overlapping; therefore the next window of the example is 0x2b, 0x12, 0x04. On each window, signal value and sigma is calculated. The resulting couple {dc, sigma} is what is named the elementary measurement.
Depending on the plate move speed, the system can perform more than one elementary measurement for each position. The resulting number of elementary measurement is named the weight of the measurement.
A measurement is also time stamped. The final structure of a measurement is then: {dc, sigma, weight, time stamp}.
Our 3-telescope cophasing breadboard is now operational in laboratory. On the next figure, are shown (on the excel spreadsheet) the signal of the synchronous detection and double frequency synchronous detection during the acquisition and phase control.
Fig. 13 – Control interface of the 3-telescope breadboard of SOLARNET. Fringes are obtained with a white-light source (Xenon) at low flux : an aperture of only 5 µm is placed at the focus of a 2 m parabolic mirror (200 mm in diameter) and a selection is made using a 60 mm aperture for each telescope. Therefore, there is less than 9 % of the flux for each beam at the entrance pupils (transmission less than 6 %) and the contrast is lower than 20 % due to the optical aberration and the field of view.
For the two interferometers that control the 3-telescope interferometer we have two curves (cf. Fig. 11). The lower ones shows the signal issued of the synchronous detection at simple frequency (brown curve DS1) and the one above, the synchronous detection at double frequency (orange curve DS2), and this during the Principal Delay Line move. The phase between these two signals is well equal to π/4 as the mathematical simulation showed it. The signals’ form is not perfect, because of the average of multiple measures and the elementary move of the delay line (0.1 µm) which is not infinitely small.
The upper curve shows this same signal during the search of the central fringe (null OPD) with a phase control. Then the signal of the synchronous detection at simple frequency is null while the signal of the synchronous detection at double frequency grows until the null OPD is reached.
Stabilization performances are excellent and will be reported in a subsequent paper. Yet, we can show us the following result, about l/20 at 3s, but we had a lot of noise of s = 2.4 mV outside the fringes when signal during the active phase control had a residual noise of s = 3.2 mV
Fig. 14 – During phase control we measure for each interferometer the signal of single (blue curve) and double frequency synchronous detection (red curve). We trace the Fourier Transform for each signal and we can see some peaks that correspond of electrical noises (50, 100 et 150 Hz), vibrating mode of the optical table (10 Hz ?). The 88Hz peak is generated by the signal acquisition (for an unknown reason), it is not a true noise. The quality of the phase control is very good (>250 at 6s for a contrast of only 12 % and a flux of 0.25nW on the detector).
Fig. 15 – Comparaison of the pictures obtained for a simulated star (hole of 5 µm on the focus of a 1500 mm parabolic mirror) when the three beams are cophased (left picture) and when the beams are not cophased (right picture). We could recognize at the left image the PSF of SOLARNET telescope.
During the beginning of 2000, the whole experiment will move, again, to the “Grand Sidérostat de Foucault” (cf. Fig. 14) at Meudon Observatory for a demonstration directly on the Sun (and the major planets) of the cophasing of the 3 telescopes. This time, fine pointing is implemented (3 active mirrors) and the interferometric table is placed inside the building behind a large optical window in order to carry a complete evaluation of the cophasing performances.
