JET far-infrared interferometer/polarimeter diagnostic system—40 years of lessons learned

Originally designed for 5 years of plasma operations, the JET far infrared interferometer/polarimeter diagnostic system was still operating at full capability nearly 40 years later in ITER-relevant conditions (e.g. metal wall, tungsten divertor) for multiple D–T campaigns, albeit with significantly lower neutron fluences. The original design had to be adapted substantially over the years due to machine changes, leading to reduced signal and access to mirrors etc, and the diagnostics still worked due to the excellent dynamic range of the detectors. This paper will discuss invaluable lessons learned from designing, operating, optimising and enhancing such a complex system and how these can be used for developing a new class of laser-based diagnostics for next-generation reactor-grade machines.


Introduction
First-generation burning plasma devices such as ITER, STEP or DEMO will operate in a very challenging environment for diagnostic systems that can only be found in some existing fusion devices around the world such as JET [1].Some of these conditions include high ambient temperatures, strong electrodynamic forces due to high magnetic fields, long pulse lengths and uninterrupted periods of operation, and, most notably, very low or zero access to some parts of the diagnostics. 1See the author list of 'JET machine operations in tritium & D-T' by D B King et al to be published in Nuclear Fusion Special Issue: Overview and Summary Papers from the 29th Fusion Energy Conf. (London, UK, 16-21 October 2023).* Author to whom any correspondence should be addressed.
Original content from this work may be used under the terms of the Creative Commons Attribution 4.0 licence.Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Additionally, the presence of tritium and of D-T neutrons and the resulting activation constrain the diagnostic design and operation.With that in mind, one must develop a system with enough redundancy and robustness to survive reactor-relevant plasma conditions for many years.
The JET far-infrared (FIR) interferometer/polarimeter was one of the first systems developed for JET back in the 1980s, and it operated until the last day of plasma operation of JET.
The original design changed substantially over the years, for example introduction of the divertor significantly reduced access and the number of laser beams and D-T readiness required double vacuum windows that greatly reduced the level of the laser signal through the plasma to 5%, but the diagnostics still worked due to the excellent dynamic range of the detectors.
In JET, complete alignment of the FIR system was required only once every decade, like the timescale for fusion reactor maintenance shutdown.The FIR system operated as a nearly fully automated hybrid interferometer and polarimeter system, with state-of-the-art electronics for phase counting, improved redundancy in both optical hardware (multiple lasers) and data acquisition and control, and real-time integration of measurements with the JET plant (active plasma control and additional heating system interlocks).This was done in conjunction with other diagnostic systems and magnetic reconstruction codes that were developed and enhanced at JET over the years.
One notable enhancement was the integration and use of polarimetry for real-time plasma density control and machine protection in a fusion plant for the first time.This is being replicated in most of the current FIR developments.Also, during the latest D-T campaign on JET, the FIR system was the single point failure device for density control, replacing the original backup system based on visible spectroscopy that could not be used since the introduction of the metal ITERlike wall in 2011.At JET, a fault of the single point failure device meant stopping JET operations to restore device functionalities.The FIR system operated with nearly 100% reliability 16 h aday, 5 days a week.This paper will discuss invaluable lessons learned from designing, operating, optimising and enhancing such a complex system and how these can be used to develop a new class of laser-based diagnostics for next-generation reactor-grade machines.

Latest diagnostic capabilities
The JET FIR diagnostic [2] (see figure 1), at the end of JET plasma operation on 18 December 2023, operated as a hybrid Mach-Zehnder interferometer/polarimeter with diffraction grating wheels used as beam modulators [3] probing the plasma with four vertical channels and four lateral channels (see figure 2).
The instrument used FIR or terahertz lasers, as these frequencies (2.5 THz and 1.5 THz corresponding to 195 µm and 118.8 µm in wavelength terms) are far from the plasma frequencies, which are typically in the gigahertz range; the plasma is thus a transparent medium for these types of lasers.Therefore, by analysing the change in the optical properties of a laser beam passing through the magnetically confined plasma one can obtain important information on electron density (interferometry) as well as information on the magnetic structure using internal measurements (polarimetry).If a sufficiently large number of probing channels are used, one can also derive a profile.
Both diagnostic systems operated in real-time, 16 h aday, for all JET plasma operation days and are part of the essential diagnostics (sometime referred to as 'basic control') without which JET could not operate.

Lasers
The lasers used by the JET FIR system were as follows: for 195 µm, two continuous wave (CW) deuterated cyanide (DCN) FIR lasers (high voltage driven at 4 kV and 1.2 A) using a mix of CD 4 , N 2 and He gases with a typical power of 200 mW; for 118.8 µm, three CW methanol lasers optically pumped from a 10 µm CO 2 laser (high voltage driven at 16 kV and 40 mA) using a laser mix of CO 2 :He.These methanol lasers were commercial products from Edinburgh Instruments Ltd that were modified by us to improve their performance to nearly steady-state.
The 118.8 µm lasers were used for interferometry compensation and there were two sets: the original on PL4/FIR295, generating around 100 mW FIR power, and the second set (originally installed on the UKAEA COMPASS machine) being a PL6/FIR395 twin cavity with power in the range 250-350 mW each.
Only one DCN and one compensation laser were used at a time, the secondary ones were for redundancy reasons.The DCN laser was based on 1971 design from CEA Cadarache, France and has been deployed on several machines (JET, Tore Supra, FTU, ASDEX-Upgrade, START).
The latest implementation of the interferometer provided line-integrated electron plasma density with a time resolution of 1 ms in real-time but also had the ability to perform offline measurements with a 10 µs time resolution.This proved to be a very valuable addition for magnetohydrodynamics (MHD) studies [5,6] as well as for characterising fast events such as edge localised modes (ELMs), pellet injection and, more recently, shattered pellet injection (SPI) for disruption mitigation experiments [7,8].The polarimeter diagnostics were implemented later (in 1988) to measure the Faraday rotation angle (FAR) and line-integrated density from Cotton-Mouton angles with 1 ms time resolution [9,10].
For alignment, originally only an 18 mW 632 nm red He-Ne gas was used.Later, in 2008, a 10 mW 532 nm green diodepumped solid state visible laser was added due to the poor reflectivity of the original red laser beam that made alignment impossible for some lateral channels.
All FIR and alignment lasers had a very low full beam divergence of about 0.1 mrad.

Mechanics
The diagnostic components spanned three areas within the JET main building called J1, as depicted in the schematics in figure 3.

Diagnostic room (J1D)
This area hosted many of the JET diagnostics, and in the case of the FIR system it is where all lasers, large granite optical tables (approx.11 m × 6 m), ancillary equipment, data acquisition and control as well as detectors were sealed (see figure 4).

Basement (J1B)
The JET basement (below the torus hall) contained a lot of equipment typically linked with JET ventilation and vessel pumping but also many areas dedicated to diagnostics.
With respect to the FIR diagnostics, the basement area only contained optomechanics and optics used for beam transfer to/from the diagnostic area to the JET machine area, which was all enclosed in a large duct of approximately 25 m × 1.5 m × 1.5 m (see figure 5).More specifically, this duct contained the optics for the four input and ten return laser beams.

Torus hall (J1T)
The JET FIR diagnostics was the largest element inside the JET torus hall (followed by neutral beam injection boxes) and consisted of a ∼15 m diagnostic tower, as shown in figure 6, weighting 70 tons.It was installed on a crane carriage (to be moved out ∼7 m if access to the octant 7 median port windows area was required) with a repositioning accuracy of about 100 µm.Inside the tower there were the optics for splitting the beams in the eight probing channels and two reference channels as well as the recombination interference plates (Z-cut quartz) that generated the beat signals.
Due to its long optical path of 80 m and strong transient magnetic fields, to minimise vibrations below 10 µm as required by the interferometer, all the optics were overengineered, driven by pneumatic motors and made of nonmagnetic metals (aluminium, brass) with fixings of glass resin.An example of such a mirror assembly is displayed in figure 7. The duct and tower enclosure panels were made of a sandwich of two plates containing a honeycomb structure of a phenolic-based resin to increase mechanical stability and minimise weight.

Detection and data acquisition and control (DAQ)
The system had several parallel systems and architectures installed over the years as per figure 8.
The detection system contained three cryostats each having up to seven high-quality cryogenic detectors.They were indium antimonide (InSb) bolometers operating at temperature of 4 K with a noise optical power of less than 10 −12 W Hz −1/2 and were made by QMC Instruments [11].Each detector was coupled with a pre-amplifier set (with fixed gain −40 dB, −60 dB and −80 dB) using bias currents of the order of µA.Each cryostat had a double-wall construction and contained two chambers, an outer one for liquid nitrogen and an inner one for liquid helium.Before cryogens could be used the cryostat vacuum interspace was pumped to high vacuum level of 10 −6 mbar.
The system was split into two branches.One set was sent, via a CODAS-developed unity gain amplifier, to a UKAEA proprietary fast transient recorder [12] that could acquire data up to 2 MHz and was used for fast interferometry measurements.The second set connected to a set of analogue filters for laser modulation frequencies.The original set contained a set of filters for 100 kHz and 5 kHz corresponding to DCN 195 µm and methanol 118.8 µm laser wavelengths, respectively, and (later) another system with 23 kHz for the second colour (118.8 µm laser) on vertical channels.These signals were then amplified and sent to two DAQs based on CAMAC (1984), C40 (1995) and PowerPC (2001) [13] architectures.The amplifier gain was controlled by a standard PC operating the Microsoft MSDOS operating system until 2018, when it was upgraded to an industrial PC based on Beckhoff technologies [14].
In 2012, a new field-programmable gate array (FPGA)based system was developed in collaboration with our colleagues from CEA Cadarache.This contained two sets of EUROcards for filtering raw data, zero-crossing phase counting and built-in fringe-jump correction based on FPGA and Linux technologies [15][16][17].
Interferometers are instruments that provide historydependent measurements, meaning that at any measurement time point one records only the phase variations between the probe beam that passes the plasma and the reference outside the plasma in a specific time interval and this value is added to the total measurement.Phase variations are measured in fringes that can be easily converted into lineintegrated electron density measurements (e.g. for the 195 µm laser wavelength one fringe corresponds to 360 • phase variations that translate to a line integrated density vaue of 1.143 × 10 19 m −2 ).However, if the signal is lost for some reason (e.g.beam refraction, absorption) phase measurements are not possible, causing so-called 'fringe-jumps'.
Most of the laser controls and environmental controls were based on Beckhoff and ADAM technology [18,19].The DAQ was remotely controlled via Solaris UNIX systems for both monitoring and setting up the parameters.

Physics data available
The system routinely produced measurements of lineintegrated electron density (LID), Faraday rotation angle (FAR) and Cotton-Mouton phase shift angle as part of the real-time systems used for basic density control or machine protection (e.g.additional interlocks for heating systems on JET).
The physics data produced were saved in what was called a JET pulse processed file (PPF) and was manually validated for interferometry, with automatic validation for polarimetry.
For advanced studies, it was possible to generate PPFs for several channels of interferometry with a time resolution up to 10 µs.This was very useful for fast transient data events such as disruptions, pellets or ELMs.Raw data from the fast interferometer were already integrated in a dedicated code for spectrogram analysis, providing important qualitative information on the MHD modes using internal plasma measurements.
Over the years, several codes were developed using various channels of interferometry to reconstruct a density profile and to compare and benchmark other diagnostics (e.g.high-resolution Thomson scattering diagnostics and other microwave diagnostics).The polarimetry data were integrated into various magnetic reconstruction codes such as Equinox [13] and EFIT++ [20].The diagnostic capabilities are listed in table 1 and were retained for the entire lifetime of the diagnostics.

The original 1984/85 version
The original diagnostic was designed for a much simpler JET machine with Ohmic plasmas (up to 7 MA plasma current), no divertor, no additional heating systems, no H-mode plasmas and no pellet injector.
The DCN laser technologies at the time (see figure 9) were still in the prototype [21] period and the lasers were operating at maximum power capabilities (up to 400 mW), which caused regular tube breakages, leading to higher running costs.
The original setup had six vertical channels (see figure 10) and vacuum windows had a single window made of Z-cut quartz (to minimise birefringence effects).
The power losses were substantial and were found to be mainly due to water absorption (moisture in the air) of the FIR radiation through the long optical path of 80 m.A Perspex enclosure was added to the lab optical tables (see figure 4) and focusing optics telescopes under the detectors, and all were fed with dry air (dew point −50 • C).The duct enclosures in the basement were sealed with bitumen tape to try to minimise the ingress of moisture with a dry air supply again from the basement.
To get better profile information on the electron density from the interferometer, a lateral system with four channels was originally envisaged (see figure 11) with two new sets of vacuum windows (one for the input beam and one for the reflected beam) and a set of in-vessel reflectors for three channels.
The first implementation of the lateral system had three in-vessel retroreflectors with a beam profiling set of mirrors before and after the vessel and one additional edge channel with main reflectors on the JET mechanical structure.Due to vibration of the vacuum vessel, a compensating interferometer was required to discriminate between phase changes of the laser beam due to plasma and in-vessel vibration.The original laser was a 3.3 µm He-Ne laser as this wavelength was compatible with the beam splitters and window materials.

Alignment setup
A complicated telescope viewing system was designed to allow observation of the beam position along the beamlines directly from diagnostic labs.The implementation had a set of lights that could be enabled one set at a time for each torus hall mirror to evaluate the position of the visible alignment beam with respect to the side illuminated edges of the mirror.The rationale behind this was that there was an expectation that the JET torus hall would be so radioactive after the D-T experiments that people would not be allowed inside the torus hall for tasks such as alignment.However, due to losses of the visible beam from, for example, Kapton foils around the system, and mirror flatness quality for FIR not being designed for visible laser beams, this never actually worked.Also, the original 3.3 µm He-Ne laser was found not to be suitable due to large displacement of the in-vessel mirrors of 10 mm and large phase change perturbation due to air turbulence in front of heated vacuum windows (around 150 • C).
Visible alignment was very hard to achieve due to bad ergonomics and physical access for human beings (see figure 12) and the fact that some captive optics were sealed inside enclosures.This was valid for the entire life of JET.
Due to all these factors, four out of ten channels were not operational for the first 4 years, nearly half of the expected operational lifetime of JET at the time.
In 1987, another type of laser to replace the original 3.3 µm one, produced by Edinburgh Instruments, was installed.This consisted of a 40 W CO 2 CW wave laser called PL4 that optically pumped a methanol cavity (FIR295) for a FIR power of about 100 mW at 118.8 µm wavelength.This wavelength was partially compatible with the transmissive optics designed for DCN operation and made the system operational within parameters deemed acceptable at the time.

1987 polarimetry
In 1987, polarimetry capabilities were added with the installation of eight half-wave plate rotator assemblies in the torus hall in front of the vacuum windows.All these rotators were pneumatically driven (see figure 13) but were very slow (3 s per step), which was not practical for regular calibration.
In the detection section eight wire grid analysers were installed to separate the two orthogonal polarisations of the incoming FIR beam plus a new cryostat with seven detectors dedicated for polarimetry.The wire-grid analysers were manually adjusted and there was a lot of prototyping required to adapt these to the detection section to fit the existing optomechanics and allow redirection of the two orthogonal polarisations to the detectors.
Calibration was based on ideal mathematical formulae [9] and ignored mutual interaction of the various effects (e.g.birefringence of transmissive optics, interaction of the Faraday rotation angle with the Cotton-Mouton angle, and so on).
Another important aspect was that only polarimetry physics data were available and the calibration for one channel took hours to perform.There were some weaknesses in the design that caused loss of the optical axis of the half-wave plate after a few months of operation, requiring manual reset in the holder and a lot of logistics (e.g.moving the tower out, raising scaffolding and so on).This was corrected later with the new design (see the following sections).

Divertor introduction, 1991
The introduction of a divertor had a great impact on the interferometer as, due to the divertor coils, two vertical channels were blocked and one was nearly tangential to the plasma, reducing its usefulness for physics measurements.For example, during all high-performance plasmas (H-mode), the beam was outside the plasma during flat-top.
The beam apertures were reduced from 130 mm to 12-60 mm for the vertical channels, causing some vignetting as the original window dimensions assumed a beam waist (1/e) at the window of about 35 mm.

In vessel mirror redesign, 1992
The original mirrors (see figure 14) suffered severe displacement (up to 10 mm) during the plasma shots, so the JET central column needed reinforcement.The second set was affected by arcing/carbon deposition due to the very close proximity to the plasma, therefore reducing reflectivity and the beam profile and these were redesigned to be hidden behind a limiter as well as being recessed by a few centimetres below the front of the first wall tiles (see figure 15).They were aligned in-vessel, then secured with bolts to the first layer of the wall.The associated locking nuts were then welded in place.These were used successfully from 1992 to the last day of operation in December 2023.As part of the JET decommissioning programme it is planned to retrieve them for analysis.

DCN laser controls and DAQ upgrades
In 1995, there were upgrades in preparation for the first highpower D-T campaign (DTE1).The manual flow control for the HeNe and CD 4 gases was replaced with mass flow controllers and associated valves controlled by proportional integral derivative (PID) controllers linked with the high-voltage power supply.UKAEA developed unique CD 4 gas introduction control modules with built-in timers and functionalities for remote control.
Also, new analogue filters and amplifiers for 100 kHz and 5 kHz modulation frequencies were added, as well as an automatic gain controller running on MSDOS PC technology.

Deuterium-tritium experiments 1 (DTE1), 1997
In preparation for DTE1, the diagnostic hall, the basement and torus hall, as per the requirement for a tritium safety barrier, were separated and isolated with negative pressures of about 100 Pa.In the case of the FIR diagnostics, a secondary tritium barrier was required, and this translated to the installation of additional pellicle windows at interfaces between areas and in replacing vacuum windows with double windows.All these measures reduced the FIR signal level by an additional 90%, causing losses of laser power in the range of Only the fact that the detectors had an exceptionally large dynamic range allowed the system to still be operational to the end, but the signal-to-noise -ratio (SNR) was reduced from 40 000 to about 100-400.
The FIR system was one of the few diagnostics that was fully operational during the DTE1 campaign, and was used for plasma density control as the primary diagnostic.Apart from upgrades in previous years, no special requirements were defined specifically for D-T operation.

Real-time DAQ and polarimetry upgrade calibration, 2002
A programme to upgrade the real-time system using Performance Optimization with Enhanced RISC -Performance Computing (PPC) technology as well as new polarimeter controls started in 2001 and was completed in 2002.This was part of a large programme to develop a realtime current control called EQUINOX [13].
Part of the polarimeter upgrade related to controls as follows: the half-wave plate and wire grid mechanisms were upgraded using high-torque stepper motors (see figures 16 and 17) as well as an automatic in-pulse calibration mechanism that took only 10 s (7 s for actual rotation and 3 s for damping to neutral-position) [9].This allowed automatic FAR calibration at the start of each pulse as well as the integration of raw data with EQUINOX.However, the system relied on calibration data from the previous pulse running from a FORTRAN offline code.
However, no real-time data quality tracking was implemented, and therefore the polarimetry system was not usable for basic control but only for advanced control [22].To understand the difference between basic and advanced control, basic control is for controlling a simple parameter such as plasma density while advanced control is more to control the spatial profile shape of plasma density that relies on a combination of multiple measurements so as to map individual values on a spatial profile density information profile.

Cotton-Mouton addition, 2004
By altering the initial polarisation angle of laser beam polarisation with respect to the toroidal field direction from 0 • to 45 • , Cotton-Mouton measurements were made possible.These measurements were very important for control since they provided absolute information on the line-integrated electron density.This was demonstrated as feasible on a small scale [9] and later a more comprehensive statistical analysis for a large group of JET pulses was performed [10,23].

Fast data acquisition for interferometry, 2007
A relatively modest upgrade in terms of hardware cost greatly expanded the diagnostic capability by recording the raw unfiltered data from the detector using UKAEA's own fast transient recorder electronics [12] on three vertical channels (and later for the lateral channels).This was tried earlier when the fast magnetic system was originally developed but due to the capacitance of long cables the noise level was very high.
The upgrade enabled high-contrast spectrograms of a core channel of the interferometer, making it possible to observe core-localised toroidal Alfvén eigenmodes [24], which are instabilities driven by energetic particles, as well as the ability to provide line-integrated measurements with 10 µs time resolution by analysing laser intensity variations that are smaller by a factor of 10 −3 than the main carrier frequency.The new measurements allowed a fuller description of fast events such as ELMs or pellet injection with 150-300 points versus 3-5 points previously [5].This system was replicated at the CEA Tore Supra (now WEST) tokamak in Cadarache, France.

CEA FPGA prototype, 2008
Following the ITER-like wall programme we investigated a way to upgrade the DAQ to ensure FIR diagnostic operation for another decade.Our approach, together with our colleagues from CEA Cadarache, was to use the more modern FPGA technology.The first step was to test the viability of this technology by installing a full-rack FPGA prototype to record the data during a plasma experimental campaign [15].This proved to be 100% reliable and allowed for the first time an on-the-fly fringe-jump correction algorithm and direct integration with JET's real-time control network.

Real-time EP2 programme upgrade, 2012
With the decommissioning of the UKAEA COMPASS machine in 2008, the FIR laser became available.This was based on an Edinburgh Instruments PL6 laser feeding a twin cavity providing completely independent laser beams.We used this laser to add redundancy to the existing compensation laser by adding a second wavelength (sometime described as second colour) laser measurement.The power level of the new laser compared with the original compensation laser was three times more on each laser cavity, including much better power stability.A new DAQ system was developed using commercial off-the-shelf (COTS) component and FPGA technology with CEA Cadarache as mentioned in section 2. A novelty for this implementation is that we used a full engineering approach for direct integration with machine control and a unique design suitable for both Tore Supra and JET, including the firmware.To select the right machine, one needed only to change position of an on-board switch (0 for CEA mode, 1 for JET mode) [15][16][17].This data acquisition hardware proved to be 100% reliable from the outset.
Since 2014, the new system, in fact, became the primary system used for basic density control, extending the lifetime of the interferometer for another decade as the old DAQ was failing increasingly regularly.

PPC real-time polarimetry upgrade, 2014
In 2014, we completed and validated a major upgrade of the real-time software including an innovative calibration using complex amplitude ratio [25] and a comprehensive mechanism for self-validation polarimetry physics measurements every millisecond during a plasma pulse.This was challenging, as the PPC VME software implementation did not have mathematical libraries to compute basic operations with complex numbers so we had to build and test these libraries even for elementary operations such as addition or simple trigonometric functions.
Due to these upgrades, JET was the first machine in which polarimetry was used unattended for basic control [25].The impact of this upgrade was so relevant that any future interferometer in the fusion world will probably have polarimetry capability for backup measurements of electron density (due to its lower resolution than interferometry).

DCN laser automation and optics upgrade, 2016
As part of preparations for the second high-power D-T campaign (DTE2), all the DCN local controls were upgraded from manual to fully automatic using hybrid industrial personal computer(IPC) technology from Beckhoff [14,18].The rationale for using this technology was as follows: due to the criticality of the diagnostic we chose technology that had proven successful for 4 years by then (since 2010) on the JET plant essential monitoring module system; secondly, this was fully supported by the JET information technology team.This system was upgraded further in 2019, when automatic gain control of the old analogue electronics was implemented.This reduced the time needed for the daily start-up/shutdown operation by 90 min as most of the time sequences no longer required the presence of personnel in the lab to go to the next step (e.g.waiting 5 min for the vacuum level to go down to the expected value before the gas was activated).

New integrated control DAQ for deuterium-tritium experiments (DTE2), 2017
In preparation for DTE2, we transferred most of the local control and environmental monitoring (e.g.laser status and power, voltages, humidity levels etc) to remote.
An integrated software package was developed with the scope to unify all the interferometer/polarimeter systems on a simple single page control panel mimic using only a fourcolour coded system.This simplified tracking errors by exception only and proved to be easy to monitor even for nonexperts.This proved essential during the coronavirus disease (COVID) pandemic when, together with the Beckhoff upgrade, we were able to manage the JET FIR operation during DTE2 with an extremely limited number of staff and under very stringent health-related controls (e.g. the 2 m rule between individuals when working in the same area).

Post-DTE2 DCN modular laser tube upgrade
The DCN modular laser tube upgrade was done as part of a resilience programme started in early 2016 to ensure JET remained in operation until 2024.Since the time that the upgrade was approved the JET plans changed significantly for various reasons, including delays to the DTE2 schedule and COVID.Nevertheless, the upgrade proved useful in JET's final years of experiments.
The project started at a time when the sole manufacturer closed its business, leaving no supply chain available; this led to the team developing, together with our CEA colleagues, a modular laser tube using mostly COTS components that doubled the laser power, increased laser mode performance and halved maintenance time.This new design was tested for the last 2 years of JET's operations on both DCN lasers during deuterium-tritium experiments (DTE3) with no faults.

Diagnostic operation
The FIR diagnostics were required whenever JET was in operation, and any work on the diagnostics, including regular maintenance and daily operation, depended strictly on JET shutdown/intervention operation planning.

Daily operation
The system was the only diagnostic that was still fully supported with shift personnel at the end of JET.This was mostly to ensure the availability of the right personnel for emergency recovery of diagnostic capabilities during JET operations.Some of the daily operational activities were covered by  JET shift technicians and the diagnostic coordinator, and some by responsible officers or their deputy.
The daily activities were as follows: • start-up/shutdown of lasers • refilling of detector cryostats with liquid nitrogen (twice a day) • tuning of lasers when required (a few times a day) • CO 2 bottle change (every 1.5 days of laser operation) • interaction with JET control room staff • provision of support to the scientific team with respect to measurements at any time during operation shifts • emergency recovery actions (ideally between pulses within a 20-min timeframe that was the typical waiting time between two consecutive pulses).

Weekly operation
There were several tasks to be performed on a weekly basis, typically by the responsible officer and deputy, and they were as follows.
• refilling of cryostats with liquid helium • checking the CO 2 laser gas stock • maintenance of the DCN laser when required (typically 2-3 h).

Monthly operations
These mostly concerned logistics and planning, as follows: • DCN laser maintenance • monthly roster preparation • ordering CO 2 gas bottles • stock checks of pare parts.

JET shutdown operations
During shutdowns and interventions there were several activities to be covered.The key ones are listed in the next paragraphs.Note that not all these were necessary during every shutdown or intervention, depending on various factors such as access to areas of diagnostics, how long since the last time the task or activity was performed or observations during operations.All the activities related to the diagnostics were recorded in the diagnostic commissioning checklist document.

4.4.1.
Record the status of the diagnostics.This could be done at any time during shutdown, but when access to torus hall was allowed a complete checklist had to be completed.The checklist contained the following parameters: • laser power before shutdown at source • environmental control parameters (e.g.humidity) • detector settings, preamplifier setup, voltages • gain level the last operation day • visible/FIR alignment position for the probe beam at specific points (e.g.torus hall floor input/floor windows, return floor windows in the diagnostic hall, position of FIR/visible laser beam at the recombination beam splitters inside the tower).

Diagnostic tower operations.
Due to its very close proximity to the machine, the diagnostic tower had to be moved from its location for access to octant 7 to allow vacuum leak checks and inspections and commissioning of other diagnostics.This operation was always strictly controlled by the responsible officer and was done via the machine operation group document, requiring a team of up to six observers due to the proximity to the machine (5 mm only at a few points).This operation could be a problem when reinstating the tower following a shutdown when new installations could impede full repositioning.This happened, for example, prior to DTE2 when the pipework attached to the newly installed tritium injection module no.7 was in direct line of contact with the top of the tower for about 20 cm, requiring modification to the pipework.The speed of the move is typically 1 mm s −1 but could be up to 10 mm s −1 .With the 'faster' speed there is a stopping distance, due to mass inertia, of about 25 cm.In 2010 that was deemed too dangerous as there are no brakes on the tower wheel mechanisms, so the procedures were changed.It is important to note that whenever the tower was moved from its operating position there was a risk to the alignment.

Vacuum cleaning of the vacuum windows.
Due to the carbon-wall, debris was always generated inside the torus.This affected operation of the interferometer as carbon dust made its way to the bottom of vacuum windows blocking the laser beams, therefore necessitating cleaning at any opportunity during shutdowns.This operation was always done by the remote handling team to clear up the debris that was typically inside channels 2 and 3.This was done in-vessel using a very long pipe with a plastic tip attached to the remote handling arm (see figure 18).Underneath the window there was always an isolator enclosure, certified by the health physics department, to capture any debris in case of catastrophic damage to the vacuum windows during this procedure.• laser maintenance by the manufacturer (CO 2 laser) • software/firmware upgrades • power supply checks.

Diagnostic commissioning
All diagnostics on JET had to follow strict facilities commissioning procedures that needed to be approved by the nominated safety authority, various specialist safety risk specialists and discussed at the weekly JET coordination meeting.Any modification of procedures, hardware and software of basic control/essential systems had also to be discussed and approved by the JET Machine Protection Working Group.At the end of any diagnostic commissioning, the responsible officer and group leader had to sign a readiness for operation form before enabling use of that diagnostic within JET facilities.JET plasma operation could not go ahead until readiness for operation of the FIR interferometer system were approved.The documents for commissioning the interferometer and polarimeter were very large, comprehensive and complex, with the following key points:

Lessons learned
During 40 years of operation the diagnostics went through many changes, but some original design choices proved to be relevant in ensuring operation for such a long time.In this section, only a subset of lessons learned are shared-these were either discovered by the authors or passed down by our predecessors.As a general comment, one must consider the entire functionalities of the design phase as having the same importance: measurement requirements, maintenance requirements and long-term operation.

Lesson learned #1 (Mechanics and over-engineering)
The mechanical parts of the system were over-engineered to keep vibration levels below 10 µm and also due to the very low data acquisition capabilities.This proved important in the long term, as full alignment was required every 10 years with checks and small optimisations during each shutdown.The original design proved impossible to align with the remote system that was originally designed.
Over-engineering in mechanics and optomechanics, together with oversized mounts, allowed long-term optical stability.

Lesson learned #2 (Small issues matter)
For many years, a good signal level was very difficult to achieve following an alignment.There were many causes, but most were simple and relatively easy to fix.Here are a few examples: (a) The Kapton/Mylar absorption level was very high for visible and methanol lasers: this was fixed by replacement with 90 µm thick TPX.(b) Lack of proper sealing of large optical enclosures (e.g.duct 25 m × 1.5 m× 1.5 m, tower 14 m × 7 m × 3 m) made it impossible to reach the required level of dryness (of the order of dew point −50 • C), thus delaying the time for which FIR beam detection became possible (only after 1 week); this was easily fixed by adding 5 mm Neoprene seals on most of the enclosures' panels that we could access.(c) There was a 30% loss of methanol laser power due to incorrect design of the half-wave plate placed in front of the laser.Replacing it with a similar one that was just 5 µm thicker improved transmission to 99%.(d) Poorly maintained DCN laser pumps and seals/gaskets around the system caused laser glass tubes to break often (up to six times per year in 2004, down to none in 4 years by the end of JET).Replacing the seals every year and servicing as advised by the manufacturer fixed this problem.(e) A catastrophic failure of the vacuum pump caused oil to contaminate and damage all the mass flow controllers.Replacing these oil-based pumps with dry pumps eliminated this potential issue at source.
Typically, big problems/faults were caused by many small issues that accumulated.

Lessons learned #3 (Calibration)
The performance of the original polarimeter design, especially the calibration, was not acceptable.Upgrading to motorised optomechanics and real-time estimation of physics measurements yielded a very powerful diagnostic that was elevated to the level of basic control.The upgrades to the real-time system to actively check data between pulses and during plasma pulses ensured diagnostic resilience for a decade.
Consider how you will calibrate the instrument to get the proper data for the actual physics application.Identify sources of error that would affect the final deliverables.Software development has the same importance and cost, if not more, than hardware development.

Lesson learned #4 (Redundancy)
During the lifetime of the instrument, we increased redundancy as follows: we had multiple redundancy for laser availability, in detection and on measurements for density control and so we ensured that JET operation was not stopped on even a single day due to the FIR system not being operational.
Increased redundancy provided seamless availability and increased machine performance.

Lesson learned #5 (System requirements
The interferometer was not designed for H-mode plasmas, and with 30 MW of neutral beam injection heating it suffered signal loss during plasmas (e.g.high density gradients due to a hollow density profile at ramping).However, second colour lasers (smaller FIR wavelength, smaller refraction and smaller losses) plus the LID measurements via Cotton-Mouton from polarimetry were used routinely as a backup system and seamlessly allowed correct plasma control during highpower, high-performance pulses that ensured the success of JET.Modifications done for DTE1 requirements caused a loss of laser power reaching the detector by a factor of 99.9%,However, the high dynamic range of the detectors (down to pW level) allowed the system to operate during DTE1 and for another few decades.
Have the most reliable and powerful sources (lasers) and detectors from day one.Involvement of diagnosticians in mechanical design/modifications of the vacuum vessel is essential.

Lesson learned #6 ('Smart' upgrades)
The first major upgrade of lasers, optics and data acquisition systems in nearly two decades of the JET FIR system lasers added further lessons: Consider safety at all steps in the design.
Collaboration is key when developing new diagnostics.Build in safety measures in the design from day one.Non-biased feedback from external experts is extremely valuable.
Use upgrades to add extra redundancy.Use COTS to substantially reduce the operating cost.

Lesson learned #7 (Integration)
Over 40 years the system had many small modifications, ad hoc fixes, patching and software upgrades, but these sometimes proved counterproductive for the entire system.
Ask expert colleagues in various fields for suggestions/ideas and to help and review.Software integration/development is often not properly evaluated and beta-tested when upgrading.Reserve lots of time for beta-testing the new/upgraded software functionalities.

Lesson learned #8 (Reactor-grade conditions)
This lesson was learned during preparation for DTE2 in 2017 and it was based on the following assumptions: • no access to the torus hall after the first D-T pulse • components must be radiation hardened • access to the diagnostics area would be very limited.
Several actions were implemented, and are listed below.

Torus hall
• check and secure ALL accessible wiring, optics and optomechanics • perform final optical alignment optimisation of the diagnostics • ensure the space around vacuum windows is clear of obstructions.

Diagnostics area.
• extensive refurbishment of lasers and ancillary equipment • as much of the monitoring as possible should be made available remotely • revise operation/maintenance procedures.
Preparation for failure is critical in situations with noaccess areas.Allow plenty of time for planning and implementing deep refurbishments.

Lesson learned #9 (Long-term operation)
Another aspect that is not often considered is that design choices that can have a negative impact on long-term system performance, such as cost and frequency of equipment servicing and limited access, so that this becomes very rare and very expensive, system flexibility for upgrades in the medium term (e.g.every 5 years), system degradation in the long term (e.g.in-vessel mirror degradation or slow loss of alignment).
Consider staged development for build/operation and chose equipment more carefully for cheaper long-term running/maintenance costs.

Lesson learned #10 (Maintenance)
Regular maintenance is rarely covered in design reports.There are logistical considerations to be taken into account, such as lead time for parts, the number of spare parts on site, time to recover after a fault, that can sometimes be substantial (e.g.vacuum failure of a single JET FIR detector cryostat would take a minimum of 1 week to recommission, assuming nothing else is wrong), tasks that require large teams of people (e.g.full alignment of the system requires large scaffoldings and 10 working days for a team of four people).
Spare parts for most components were kept ready and available and various emergency action procedures were developed to restore diagnostic capabilities as fast as possible (less than 20 min, which was the time between two consecutive JET pulses).The lessons to be learnt here are:

Challenge your procedures to improve maintenance. Standardise common tech between systems (COTS).
Preventative maintenance is key to smooth operation.

Lesson learned #11 (People)
There is a single aspect that is rarely considered during the early design phase of any diagnostics project, in particular those that will be operating over decades, and yet has important implications: human resources capacity to maintain and operate the system.Having only one responsible officer is a serious single-point of failure and business risk for projects such as JET with a weekly operating cost that was of the order of millions of pounds.

Figure 1 .
Figure 1.Artistic view based on real computer-aided design of the JET machine section and JET FIR C-frame tower system in the torus hall on the carriage (right side).

Figure 2 .
Figure 2. Overview of JET FIR channels (blue colour) through the plasma (indicated in green) and with magnetic flux surfaces indicated by red curves.Reproduced from [4].Published by IOP Publishing Ltd on behalf of the IAEA.All rights reserved.CC BY 4.0.

Figure 3 .
Figure 3. Schematic of the JET FIR interferometer areas and laser beam structure with relative approximate dimensions in metres.

Figure 4 .
Figure 4. JET FIR diagnostic lab area as per the 1985 implementation showing DCN no. 2 laser (front) and granite optical tables with Perspex covers (middle).

Figure 5 .
Figure 5. Section of the FIR basement duct in 1984, containing four returning FIR beams enclosed in 80 mm Pyrex waveguides and two of the focusing mirror parts of the beam profiling telescopes.

Figure 6 .
Figure 6.JET FIR diagnostic tower inside the JET torus hall moved back from its working position as per the 1985 installation.Reproduced with permission from JET.

Figure 7 .
Figure 7. JET FIR mirror optomechanics assembly used in the Jet torus hall diagnostic tower.

Figure 8 .
Figure 8. Overview of the JET FIR diagnostic DAQ.

Table 1 .Figure 9 .
Figure 9. Overview of the FIR terahertz/FIR lasers platform used for the JET interferometer, ToreSupra, FT-U and ASDEX.

Figure 10 .
Figure 10.JET FIR diagnostics vertical distribution channels as per 1986 (from 1991 only channels 2, 3, 4 and 5 were available).Reproduced with permission from JET.

Figure 11 .
Figure 11.JET FIR diagnostics lateral distribution channels as per 1986 (three channels having in-vessel reflectors).Reproduced with permission from JET.

Figure 12 .
Figure12.Details of the lower boom of the JET FIR diagnostic tower with mirror assemblies inside a very narrow enclosure 50 cm wide and 1.5 m depth that was difficult to access even with the cover removed (a white shoe visible on the bottom left side is for reference).

Figure 15 .
Figure 15.In-vessel mirror for JET FIR diagnostic (installed in 1992, photographed in 2014 during an in-vessel photographic survey).

Figure 16 .
Figure 16.JET FIR polarimetry half-wave plate assembly installed in front on the JET vacuum windows on the diagnostic tower.

Figure 17 .
Figure 17.JET FIR polarimetry motorised wire-grid analysers (Ni wires 10 µm diameter, 25 µm interspace centre-to-centre) assemblies installed in front on the JET FIR detector inside the diagnostic lab.

Figure 18 .
Figure 18.Lower vacuum windows for JET FIR diagnostics corresponding to channels 2 and 3 before and after cleaning.

( 1 )( 4 )
Prerequisite/inspections before commissioning • vacuum-vessel windows and in-vessel mirrors (via scaffolding/remote handling) • pellicle windows • wire grid status for polarimetry • half-wave plates • penetration pellicle windows (part of the tritium safety barrier) • cabling and insulation • pressure systems • power to main cubicles • high-voltage laser power supplies • local control supplies.(2) Lasers • power of lasers at various settings (gas pressures, voltages etc) • perform maintenance if required.(3) Alignment • visible alignment for vertical and lateral systems • matching FIR lasers with visibility in the diagnostic hall • FIR optimisation on vertical and lateral channels.Data acquisition control electronics and software