Measurements of electron cyclotron emission (ECE) spectra on tokamaks have shown that electron temperature profiles can be derived from harmonics for which the plasma is optically thick .
The large variation of vacuum magnetic field with the distance from the torus axis provides a relationship between spectral frequency and spatial position via the cyclotron resonance condition; this relationship is univocal provided that there is no harmonics overlap. For FTU the optically thick second harmonic in the extraordinary polarization is overlap free over two thirds of the plasma diameter and can be conveniently used to infer temperature profiles.
An ECE diagnostic system has been put in operation on FTU during 1990 and has been extensively used since then. The main parts of the system are:
The performance of ECE diagnostics critically depends on the characteristics of the light collection and transport system which connects the spectrometers to the tokamak plasma. In fact the width of the field of view must be compatible with the required spatial resolution, while the signal to noise ratio (which determines the sensitivity of the polychromator and the quality of the absolute calibration) depends on the étendue of the light collection system and on the losses in the light transport system.
The requirements on the light collection system for FTU were quite stringent: the antenna pattern half width at half maximum was required to be 2.5 cm; a large light grasp was required by the need to perform an absolute calibration of the diagnostic with existing standard sources; furthermore a large bandwidth (more than two octaves) was imposed to fit the range of toroidal fields at which FTU can operate.
These requirements could not be fulfilled using conventional techniques and a novel light collection system was developed  employing a combination of guided wave propagation and quasi-optical techniques. The front end is the aperture of a square-section (3.8x3.8 cm2) light pipe equipped with a pair of focusing mirrors having a focal length equal to the distance from the plasma center.
The other aperture of the light pipe is close to the crystal quartz vacuum window. A quasi-optical system placed ouside the tokamak vacuum determines the spatial resolution and selects the extraordinary polarization. A second oversized waveguide connects the quasi-optical section with the spectrometers, which are located beyond the shielding wall. The first mirror of the quasi-optical system can be turned in order to substitute the vacuum section with a replica placed in air which is used for calibration.
In Fourier transform spectrometry the spectrum of electromagnetic radiation is obtained by Fourier transforming the interferogram produced by a two-beam interferometer with variable path difference.
The resolving power is proportional to the maximum path difference, while time resolution depends on the time taken to effect the path-difference scan.
A fast scanning FTS has been developed in our laboratory  since the time resolution attainable with commercially available instruments based on vibrating mirrors (15 ms) was not satisfactory. A novel scanning system based on the rotation of a helicoidal reflector was employed which allowed to scan linearly the optical path difference from 0 to 4 cm in 5 ms. Such scanning system turned out highly reliable and did not require any maintenance over six years of routine operation.
The data acquisition system for the FTS is based on the ADC Le Croy model. The detector output is a sequence of interferograms; it is sampled at a 100 KHz rate and stored in the ADC 512 Kbytes internal memory during the tokamak shot; the rotating reflector gives a trigger that allows to determine the starting address of each interferogram.
A complex software package has been developed for data elaboration and display. The basic functions are:
- Selection of each interferogram in the recorded sequence.
- Computation of the discrete Fourier transform and application of the calibration function to obtain the calibrated radiation temperature spectrum.
- Spectrum to profile conversion using the vacuum magnetic field (including ripple) at zero order and taking into account plasma fields at first order. The time evolution of measured electron temperature profiles in the first 250 ms of FTU shot #7525 is shown.
More complex operations can also be performed including: reduction of signal carried noise by averaging several spectra; display of temperature time evolution at a given position; display of full spectra to evaluate the suprathermal electron content. A supratermal ECE spectrum measured on a discharge with Lower Hybrid Current Drive and a reference thermal spectrum are shown.
A 12-channel diffraction grating polychromator (GPC) has been built  to measure the plasma temperature at 12 different positions with a time resolution of 30 s. Two parabolic mirrors has been used in a Czerny-Turner mounting to reduce the astigmatic effects due to the off-axis reflection. The GPC lay-out is shown
The optical components are described in the following table together with their relevant sizes.
|Input lens (L)||-||-||120||400||-||-|
|entrance slit (S1)||20||30||-||-||-||-|
|Output mirror (M2)||400||150||-||800||-||-|
|Exit slits (S2)||17||30||-||-||-||-|
The beam coming out from the input waveguide is focused onto the entrance slit (S) of the polychromator by a TPX lens (L). Close to the lens focus, where the beam size become small enough, there is a set of four gratings (LPF), with relatively small dimensions, that act as spectral low-pass filter . The transmission of this filter set is close to one for spectral wavelengths greater than the cut-off and for lower wavelength fall down to the noise level with a slope of 600 db/dec. The cut-off wavelength is chosen according to the channels range of wavelengths to reject the higher diffraction order of the grating. Since this filter works with linearly polarized radiation, a grid polariser (P) has been inserted after the lens to reject eventual spurious polarization.
The geometry of the spectrometer has been chosen in order to have a the resolving power of about 50 and a spectral range extending at least over 2/3 of the ECE second harmonic.
Three diffraction gratings and three adapted low-pass filters are mounted on moveable structures to allow a remote exchange when the magnetic field applied to FTU is changed from 4 to 8 Teslas. These movements are highly reliable, the angular position of the diffraction grating and the filtering group are kept within 0.5 millirad. Furthermore the diffraction gratings can rotate around a vertical axis in order to change the angle of incidence and to permit a fine tuning of the polychromator channels. The expected efficiency of the diffraction grating is greater of 0.8 for all channels for an incidence angle of -7.5 degree, however the measured efficiency falls down to 0.5 for low frequency channels (greater diffraction angles).
The exit light-pipes that carry the radiation to the cryogenic InSb detectors are made by a 17x30 mm2 rectangular waveguide, a rectangular/circular transition and a 10 mm diameter circular waveguide. Three mitre bends are necessary to adapt the rectangular waveguide direction to the detector position. The losses inside the whole output ligth-pipe has been estimated to be about 50% .
Detectors are high sensitivity fast InSb crystals. They are merged in liquid helium inside two 60 litres cryostats (six detectors each) with a hold time better than 40 days.
The polychromator has been calibrated in frequency, by means of a slow scan Fourier transform interferometer. The calibration source (a high pressure mercury lamp) and the Fourier transform interferometer has been placed in a break of the input waveguide coming from the tokamak, close to the polychromator and have been aligned with the waveguide itself. For a fixed angle of incidence, a complete scan of the interferometer has been done and an interferogram for each channel registered. This procedure has been repeated for a set of incidence angles and for all gratings. To reach a significant signal-to-noise ratio, integration was performed over more than 8 s; consequently a complete scan lasted more than 2 hours. The spectra have then been compared with those produced substituting the diffraction grating with a mirror, in order to get the grating efficiency as well as the channel wavelengths and bandwidth. The polychromator is routinely working since Autumn 1990, plasma temperature oscillations as sawtooth and MHD activities have been studied with this diagnostic. Here is an example of sawtooth oscillations as appear on the twelve expanded traces.
In thes next figure the temperature quench following a deuterium pellet injection is shown. The drop of temperature at each channel (marked with an arrow) is induced by the passage of the pellet at the corresponding radius.
 A. E. Costley, Proc. Course and Workshop 'Diagnostics for Fusion Reactor Conditions', EUR 8531-1 EN, 1, 129, Varenna (1982).
 P. Buratti, O. Tudisco, M. Zerbini, Infrared Phys. 34, 533 (1993)
 P. Buratti, M. Zerbini, Rev. Sci. Instrum. 66, 4208 (1995)
 O. Tudisco, F. Berton, P. Buratti, E. Grilli, S. Mantovani, Rev. Sci. Instrum. 67, 3108 (1996)
 O.Tudisco, Infrared Phys. 33, 373 (1992)