GigaFRoST: the gigabit fast readout system for tomography

Abstract

Owing to recent developments in CMOS technology, it is now possible to exploit tomographic microscopy at third-generation synchrotron facilities with unprecedented speeds. Despite this rapid technical progress, one crucial limitation for the investigation of realistic dynamic systems has remained: a generally short total acquisition time at high frame rates due to the limited internal memory of available detectors. To address and solve this shortcoming, a new detection and readout system, coined GigaFRoST, has been developed based on a commercial CMOS sensor, acquiring and streaming data continuously at 7.7 GB s^-1 directly to a dedicated backend server. This architecture allows for dynamic data pre-processing as well as data reduction, an increasingly indispensable step considering the vast amounts of data acquired in typical fast tomographic experiments at synchrotron beamlines (up to several tens of TByte per day of raw data).

Keywords: 4D imaging; X-ray tomographic microscopy; evolution of dynamic systems; fast readout system; high-speed camera.

Figure 1

Photographs of the GigaFRoST camera’s custom-made data and control boards attached to the image sensor head. All covers of the housing are removed.

Figure 2

Sketch of the GigaFRoST camera architecture. The pco.Dimax headboard (blue) is connected to two custom-built data boards and a control board (green). Data connections for the image stream are shown in red, control connections in blue.

Figure 3

Network topology for the GigaFRoST data acquisition system.

Figure 4

An example showing the differences between signal trains for the (a) physical and (b) virtual enable scheme operation. The external enable signal (blue) is identical for the two modes of operation. The camera triggers (red) are provided by the internal auto trigger mode (see section on trigger modes). Only frames with an active frame store bit (green) will be added to the data ring buffer. Note the shift in frame timing between the physical and the virtual enable scheme operation. Even though the second and third gate have the same length, the number of accepted trigger pulses is different (three and two pulses, respectively) in the virtual enable scheme due to the timing jitter of the gating signal with respect to the fixed frequency trigger signal. Also shown are the frames selected for preview (black), assuming a preview strategy displaying every fourth frame acquired by the camera.

Figure 5

Illustration of the four possible different modes using the start or end of the enable signal to trigger the acquisition of predefined frame series, both for the (a) physical enable and (b) virtual enable schemes. The fixed number of frames to be taken has been set to N = 4. Note that the frame store signal is not shown in (a) since it is identical to the exposure trigger signal. The exposure trigger in this example is assumed to be produced internally by the camera at constant frequency (auto trigger). Due to the different timing of the trigger pulses between the physical and virtual enable schemes, the exact frame timing is different for both cases.

Figure 6

Panel (a) shows a radiographic projection of a gold Siemens star using a 20× objective coupled to a 20 µm LuAG:Ce scintillator. The numbers indicate the thickness of individual gold lines at the given location. In panels (b) and (c) we compare two tomographic reconstructions of a polymer foam acquired with the GigaFRoST and pco.Dimax, respectively. Acquisition parameters were: pixel size = 3 µm, exposure time per projection = 0.5 ms, number of projections = 500, size of the reconstructed volume = 1008 × 1008 pixels, X-ray energy = 20 keV.

Figure 7

Volume rendering of a liquid foam flowing through a constriction. The acquisition rate was set to five tomographic scans per second to resolve the flow in time and track the individual cells for a period of 27 s. In the volume rendering, each color represents a time frame of the flowing foam: t ₀ (yellow), t ₀ + 2 s (green), t ₀ + 4 s (blue). In (a) the overview image of the constriction is shown, in (b) the side view shows how individual bubbles progress in the vertical direction from the top towards the bottom of the flow cell (constriction) and in (c) the view from the bottom demonstrates the radial displacement of the bubbles towards the walls below the constriction. See also movies 1 and 2 of the supporting information.

Figure 8

Three tomographic cross sections of an AlSi₈Mg₄ + 0.5 wt% TiH₂ foam captured in situ during the foaming process. The images are tomographic cross sections. The entire foaming process is recorded at 20 Hz. The time step between the selected snapshots is 72 s, while the full four-dimensional volume represents the state of the system every 50 ms.