In plasma reflectometry, an electromagnetic wave propagates through the plasma until it reaches a plasma cutoff layer at which the refractive index drops to zero, at which point the wave reflects back out of the plasma. Higher frequency waves reflect from higher electron density cutoff layers deeper in the plasma. In swept-FM reflectometry, the frequency of the probing wave is swept across a large portion of the plasma density profile. Mixing a portion of the outgoing wave with the reflected wave generates an intermediate frequency (IF) frequency that increases linearly with the time-of-flight of the reflected wave – digitizing the IF signal allows the phase delay or time-of-flight measurement at each frequency to be calculated and then inverted to yield the corresponding electron density profile.
In the swept-FM approach commonly employed on tokamak plasmas, a broadband microwave source is swept in frequency across a considerable span of the characteristic frequency profile with the received signal mixed with a portion of the outgoing signal. Assuming a linear frequency sweep, the intermediate frequency (IF) of the downconverted signal is proportional to the time difference between the outgoing and received signals. In the case of an extremely short sweep time, this yields a very high IF frequency which in turn translates to the need for a very high digitization rate (can exceed 1 GHz on some fusion plasma devices). The presence of large density fluctuations, however, create spurious phase changes that are difficult to separate out from the multifringe phase changes formed by the reflectometer. To avoid the deleterious effects of plasma turbulence, the measurements must be completed in a time that is short compared to the time constants of the plasma motion such that the plasma is essentially “frozen” in time. In the case of most tokamak plasmas, sweep times of the order of 10 µsec or less are often required. Even at this high rate, the effects of plasma turbulence are sufficiently strong that 10 or more successive 10 µsec sweeps are often averaged together. This represents a big problem in the case of spheromak plasmas, for example, whose pulse duration typically ranges from 2 to 5 msec.
Ultrashort pulse reflectometry (USPR) transmits extremely short duration (~few nsec) chirped waveforms that illuminate and reflect from the target plasma. The reflected waveforms are passed through a multi-element filter bank, with Schottky diode detectors (< 400 ps risetime) to convert each filtered wavepacket into ~1 ns FWHM pulses. Low-cost time-of-flight electronics then measure the time delay of each frequency component with high spatial (~few mm) resolution, eliminating the need for high speed digitizers and the fast processors needed to analyze the large amounts of data collected by the digitizers.
The first USPR systems were employed in the 1990’s on the CCT and TEXT-U tokamaks in the US, followed by a system on the GAMMA-10 mirror machine in Japan in the 2000’s. The most ambitious of the early implementations was a system that UC Davis installed on the Sustained Spheromak Experiment (SSPX) in Livermore, CA, which employed a 3-chirp transmitter. To boost the output at the higher IF frequencies (up to 18 GHz), a portion of the low frequency (LF) chirp is frequency doubled to form middle frequency (MF) and high frequency (HF) chirps. Two SP2T switches select the appropriate waveform, cycling all three frequency chirps for full coverage. The transmitter chirps are directed to one of three millimeter-wave mixers for upconversion and propagation into the plasma. The reflected signals are downconverted, amplified, and gated to remove spurious reflections. The signals are then split into 8 different filter bands with time-of-flight electronics measuring the time delay at each frequency. A third order polynomial fit is performed on the filtered time delay data, and the fitted time delays are Abel-inverted to generate electron density profiles.
UC Davis has now been funded by ARPA-E to develop a USPR system for compact, magnetically-confined fusion-energy concept devices. The new USPR system takes advantage of advances in active frequency multiplier and high speed electronics technology to realize a compact, highly adaptable, high resolution diagnostic instrument that can be readily transported from device to device with minimal setup required. Here, the passive upconverting mixers employed on SSPX in the USPR transmitter are being replaced by active multiplier chains (AMCs), and connectorized Schottky diode detectors and commercial time-of-flight (TOF) electronics replaced by new custom printed circuits employing high performance integrated circuits (ICs). Lastly, the stand-alone digitizers employed on SSPX will be replaced by field programmable gate array (FPGA) based controllers, which can not only record TOF data but also process and analyze the data including the ability to internally generate the desired density profiles without the need to install diagnostic software into the data acquisition system of the target plasma.
The accuracy of the electron density profiles generated by pulsed techniques such as USPR has been limited primarily by 3 distinct factors: (1) the number of TOF data points or frequency separation between data points, (2) the total time required to acquire these data points, (3) by a limited dynamic range (in terms of the post-detector signal amplitude) over which the TOF measurements are made. The SSPX system required 9 µsec to acquire one set of 3×8=24 TOF measurements from the 3 mixers that spanned 33 to 75 GHz. The limited dynamic range of the previous detection system resulted in the need to average the lowest 4 out of 9 TOF measurements at each frequency to minimize the effects of plasma turbulence on the reconstructed density profile, such that each density profile reflected measurements taken over a 9×9 = 81 µsec window.
The new approach will acquire ~60 TOF measurements from 4 mixers with an increased span of 22 to 75 GHz, and do so in only 4 µsec due to the faster response times of the new time-of-flight electronics. By effectively doubling the dynamic range of each time delay measurement, we anticipate that the time acquisition window required for each profile to be reduced to perhaps as low as 3×4 = 12 µsec. Hence a >6-fold increase in the time resolution combined with a nearly 4-fold increase in the number of data points with a corresponding improvement in the spatial resolution of the reconstructed density profile.