The configuration of the camera system is shown in Fig. 1. We adopted a rotating waveplate driven by a stepper motor for the polarization modulator. The incident light goes through the rotating waveplate and an “analyzer” (a linear polarizer) and reaches the H2RG detector. As described before, in polarimetry the synchronization between the polarization modulation and image acquisition is crucially important.
We employed a MACIE card for the interface between a “SIDECAR” (System for Image Digitization, Enhancement, Control And Retrieval; Chen et al. 2014) application-specific integrated circuit (ASIC) focal plane electronics (which outputs A/D-converted signals) and a personal computer (PC). Typically a “SAM” (SIDECAR Acquisition Module) card by Teledyne (Blank et al. 2012) is used for the interface; however, it is difficult to synchronize the polarization modulation and image acquisition using a SAM card. Nevertheless, the MACIE card has a function to send and receive timing signals. The synchronization with the polarization modulator becomes possible using these signals. New assembly codes (firmwares) developed by Markury Scientific have been installed into the SIDECAR to realize the synchronizing operation.
The synchronization is managed by a timing controller. The timing controller functions are implemented by a Complex Programmable Logic Device (CPLD), which receives the “row clock” (line sync signal) from the MACIE card and provides motor drive pulses to the polarization modulator. It also produces trigger signals to start the image readouts synchronizing with the rotation of the waveplate and sends the signals to the MACIE card.
Figure 2 shows the timing relations among various signals and data readouts in detail. Using the 32-channel readout in fast readout mode, a H2RG and a SIDECAR run under the row clock of \( \sim \;70~{\text{kHz}} \). One cycle of the row clock corresponds to the processing time of each row of the detector. Specifically, a H2RG/SIDECAR processes approximately 70,000 rows in a second in fast readout mode. The row clock is an internal signal produced by the SIDECAR; therefore, it can be used as the base clock to control the overall timing of the system.
When the MACIE card receives an external trigger signal, it starts the data readout of a frame by the SIDECAR. To drain out the photoelectrons, a reset of the detector is implemented at the row clock after the subsequent one of the readout at each row. An exposure starts after the reset signal and ends at the readout. The exposure needs to be longer than the time needed to read the data of all the rows, i.e., 2048 row clocks (and a margin) for a full frame. As the dead time between a “read” and a “reset” is only two row clocks, it is virtually negligible (the duty ratio, the ratio of the exposure time to the exposure interval, is \({>}~99\%\)). Hence, this is a rather effective system, where most of the photons coming into the detector are used to produce images.
Nevertheless, such a “reset–read” operation has some shortcomings. The noise level in the “reset–read” operation is generally higher than that in the “reset–read–read” operation (correlated double sampling, CDS), and therefore, the CDS is commonly used in night-time astronomical observations. In fast readout mode, the CDS also lowers the noise level to a certain extent; however, it is not very effective (Blank et al. 2012). Nevertheless, the CDS lowers the frame rate and increases the dead time. Another shortcoming is the residual image or persistency due to the residual photoelectrons, which remain after the reset operation (e.g., Mosby et al. 2016), and which are also among the sources of noise on images. It is known that a single reset operation cannot completely drain out the photoelectrons accumulated during the exposure. Repeated reset operations are effective to reduce the residual electrons, and the CDS is also useful. However, these operations result in a reduction of the frame rate. Therefore, there is a trade-off relation between the noise and the frame rate. The effectiveness of the introduction of the CDS and/or multiple resets will be analyzed in future studies.
The polarization can be fully expressed by four Stokes parameters (I, Q, U, and V), and to determine them, typically 16 images are taken during one rotation of the waveplate of a polarization modulator. In our case, one rotation of the waveplate is completed by 400 drive pulses for the stepper motor, and thus, the frame time to take an image corresponds to 25 drive pulses. In the case of a full-frame image (2048 rows), we typically set the frame time to be 2400 row clocks including a margin. It means that the exposure time starting at a “reset” and ending at a “read” (see Fig. 2) is 2398 (\(= 2400-2\) clocks of dead time) row clocks. The timing controller divides 2400 row clocks to 25 drive pulses in this case.
The rotating waveplate unit is equipped with an origin sensor. The waveplate hits the origin sensor once in every rotation, and the timing controller receives the origin signal. Then, the timing controller starts to send out the external trigger signals to start the image readout by the H2RG/SIDECAR every 25 drive pulses. The actual number of drive pulses between two origin signals, which needs to be 400, is constantly monitored to detect any step-outs of the motor. In this manner, the external trigger signals are issued at the same phase angles in every rotation, and thus realizing the synchronization. The data acquired by the MACIE card are transferred to the PC. The PC receives the origin signal via the timing controller, and uses it as the trigger to start to record the transferred data.
As mentioned above, we typically set the number of clocks to be 2400 for a full-frame image. The number of row clocks of 2400 per frame corresponds to a frame rate of 29 fps and a waveplate rotation of 1.8 rps. The new assembly codes for the SIDECAR enable vertical windowing. By decreasing the width of the window from 2048, the frame rate increases. For instance, a 1024-row window (1200 row clocks per frame) enables 59 fps with a rotation of 3.7 rps, and a 512-row window (600 row clocks per frame) enables 117 fps with 7.3 rps.
The drive pulse frequency for all rotation speeds mentioned above is higher than the maximum pull-in pulse rate of the stepper motor. Therefore, the motor starts to rotate with low-frequency pulses produced by the timing controller, and then it is accelerated to a target speed such as 1.8 rps for the full-frame readout. After reaching the target speed, the drive clock is gradually switched to that produced from the row clock. Thus, the synchronization between the waveplate rotation and the row clock is established. In this way, using the H2RG, we realized polarimetry with high frame rates such as 29–117 fps.
As shown in Fig. 2, the exposure of each row starts progressively. Such an operation corresponds to the rolling shutter, and each row undergoes different polarization modulations. To demonstrate the synchronization and the progressive readout, Fig. 3 shows measurement results of completely polarized artificial light. The incident polarized light (linear polarizations of Stokes \(Q/I=1\) and \(U/I=1\)) traveled through the rotating waveplate and the analyzer, and the light intensity reaching the detector varied sinusoidally due to the polarization modulation. We set the window of 512 rows (\(2048\times 512\) pixels), and the waveplate completed one rotation with 9600 row clocks (600 row clocks \(\times \) 16 images). The x-axis in Fig. 3 represents one rotation of the waveplate. At row clock 1, row 1 of the first image was read, and at row clock 512, row 512 of the first image was read. At row clocks 601–1112, the second image was read and at row clocks 9001–9512, the last (16th) image was read. In Fig. 3, the measured light intensities averaged every row are plotted at the row clock corresponding to their readout timings. Each segment of the curves (one of which is marked by an orange circle in Fig. 3) is comprised of 511 data points (row 1 contains dummy data) in an image, and as a whole, the measurement results for both Stokes inputs Q/I and U/I show sinusoidal variations as expected. In Fig. 3, the measurement results of 16 continuous rotations are overplotted. All data points of 16 rotations look like to fall on a single curve. The fluctuation of the measured signal during the 16 rotations is about \(4\times 10^{-4}\), even including the possible brightness fluctuation of the light source. This result indicates that the quality of the synchronization of the data readout and the polarization modulation is very high.