2.1.

Photosensor

The photosensor of the CPS is identical to the rest of the pixels in the MAGIC-II camera. They are PMTs from Hamamatsu (super-bialkali R10408), each with 25.4-mm diameter, a hemispherical photocathode, and 6 dynodes.¹ Relative to the PMTs of the original MAGIC-I camera (Electron Tubes ET9116B) used for the old CPS, the new PMTs have an enhanced quantum efficiency in the blue wavelength ($\sim 32\%$ at 330 nm⁶). The PMTs are grouped in clusters of seven to form a modular unit to facilitate installation and maintenance. Each single cluster weighs $\sim 1\mathrm{kg}$, has a length of 50 cm, and a width of 9 cm, with a distance of 3 cm between the pixel centers.

Every PMT is connected to several electronic boards as seen in Fig. 2. The first three closest to the PMT come from the manufacturer of the PMT and are responsible for generating high voltage and feeding it to the PMT dynodes. The high voltage is generated by a Cockroft–Walton voltage multiplier, which is controlled by the slow-control board of the cluster.

Fig. 2

The PMT of the CPS assembled with its high voltage generator, a preamplifier, and a VCSEL driver. These electronic boards control and process the sensor’s response into optical output signal ready to transmit via multimode optical fiber.

2.2.

PMT Preamplifier

The PMT preamplifier and DC sensing circuitry are placed in the fourth board, as shown in Fig. 2. To achieve a bandwidth spanning from 1 Hz to several kHz in the DC branch so that the fast signal (ms regime) can be detected, several passive components in this preamplification board were exchanged (Table 1). These changes are drawn in green in the schematic in Fig. 3. Figure 4 shows the bandwidth of the DC branch after the above-mentioned modifications, demonstrating the capability of observing fast kHz signals (reaching the 3-dB point at $\sim 3\mathrm{kHz}$). The last board attached to the mechanical structure of the PMT is the temperature sensor and driver board for the vertical-cavity surface-emitting laser (VCSEL), which is used to deliver the AC branch signal of each PMT to the counting house via the optical fiber.¹⁴

Table 1

Values of the replaced components (shown in green in Fig. 3) for the PMT preamplifier of the CPS in order to increase the bandwidth.

Fig. 3

Schematic of the modified PMT preamplifier. Several original components were replaced so as to increase the bandwidth up to kHz sampling rate. They are shown in green, and their values are given in Table 1.

Fig. 4

Measured CPS bandwidth after modifying the DC branch, showing the achieved capability of observing fast kHz signals. The CPS can measure signal to 3 kHz, as seen by the 3-dB point on the plot.

2.3.

Optical Transmitter

The analog output signal of the AC branch from every pixel of the MAGIC camera is transmitted to the counting house using a set of optical fibers, one per pixel. As a result, the data transmission lines are both lightweight and thin. On the other hand, they are also fragile, so spare fibers were installed between camera and counting house as a redundancy measure. The CPS signal takes advantage of one of these spare fibers to deliver its analog signal to the counting house. The electrical signal is converted to an optical signal and successively transmitted by a standard VCSEL via the optical fiber. The conversion board contains a single stage consisting of an operational amplifier in a noninverting configuration driving a VCSEL laser. The schematic is shown in Fig. 5. As a precautionary measure, a Zener diode is in place to limit the output voltage of the operational amplifier to protect the VCSEL from burning in case a bright light shines on the PMT. The power for this board is supplied by a ribbon cable attached to the central PMT. The ground and $+5\mathrm{V}$ DC supply-voltage signal are connected using a standard insulation-displacement connector. It should be noted that the transmitter circuit is AC-coupled to the central pixel signal, and therefore the absolute value of the optical illumination of the PMT cannot be obtained. In Fig. 6, the board implementing the optical transmitter is shown (in green) inside the MAGIC cluster hosting the CPS. Also shown in green is the special socket to connect the optical fiber that transports the CPS optical signal.

Fig. 5

Schematic of the optical transmitter.

Fig. 6

MAGIC cluster hosting the CPS, the location of the optical transmitter board is shown in green. The VCSEL laser for the CPS and its corresponding optical fiber (in orange) are also highlighted.

Fig. 7

Schematic of the receiver.

2.4.

Optical Receiver Mezzanine

A receiver circuit (so-called receiver board) digitizes the optical signal that arrives from the camera at the MAGIC counting house. The circuit diagram of this board is shown in Fig. 7. The low-frequency signal characteristics of the CPS output (few kHz) are incompatible with the standard MAGIC receiver analog channels since they involve typical high-pass filters of several MHz. Therefore, one channel in a receiver board is replaced by the CPS optical receiver, implemented on a mezzanine board connected to a dedicated receiver board. The mezzanine board comprises several stages: the first stage is the photodiode that converts the optical signal coming from the camera to an electric signal. With the proper DC biasing, the photodiode will produce current proportional to the incoming light. Subsequently, the signal goes through a coupling capacitor, which removes the DC component. After the signal is amplified, it is converted to differential and sent both to the Domino readout (as if it were a signal from a standard pixel) and to the CPS PC via a LEMO bipolar connector. Figure 8 shows a diagram with all the above-mentioned stages and how the mezzanine is connected to the receiver board.

Fig. 8

Diagram of the CPS optical receiver mezzanine, connected to the programmable receiver board called monster board. The LEMO bipolar cable transporting the differential CPS optical signal to the dedicated PC-ADC readout is shown in red. The standard Cherenkov pulse is routed to the Domino readout via a separate branch and is shown in blue.

2.5.

CPS Digitizing System

To achieve maximum sensitivity for the CPS, its signal is digitized in a dedicated way, independent of the standard MAGIC readout system. The differential signal produced at the optical receiver mezzanine is digitized by a National Instrument PCIe 6251 M Series ADC card with 16-bit accuracy, named herein PC-ADC. This PC-ADC card is hosted by a dedicated computer in the MAGIC counting house, which is responsible for data taking and storing the digitized CPS signals produced by the PC-ADC.

The PC-ADC operation relies on the MAGIC timing system, specifically on its Microsemi XLi unit, which combines in a single module, a GPS receiver, and a high-precision rubidium oscillator, disciplined by the GPS timing signals. The Microsemi XLi module provides timing signals with accuracy and precision better than 100 ns. In particular, a 1 pulse-per-second (PPS) signal and a 10-kHz signal (in phase with 1 PPS) are delivered to the PC-ADC. The latter is used by the PC-ADC as external reference to sample the CPS signal. To assign an absolute UTC date to each of those samples, the 1 PPS signal is also digitized with the same above-mentioned 10-kHz external sampling reference. A program running on the computer identifies the rising edge of each 1 PPS signal and assigns to the corresponding sample a UTC date based on the NTP-synchronized time of the PC.

3.1.

Periodic Signal: Crab Pulsar Observation

The Crab Pulsar (PSR B0531+21) is the brightest pulsar among the few pulsars to be identified optically. Since its discovery,^15,16 it has been a target of many observational campaigns over the whole electromagnetic spectrum from radio to VHE $\gamma$-ray.¹⁷ The optical Crab pulsation can be detected by the standard MAGIC DAQ. As seen in Fig. 1, the modified DC-branch output signal is also digitized in the standard MAGIC DRS4 channel. A trigger occurs due to a possible $\gamma$-ray event candidate (at the rate of $\sim 200\mathrm{Hz}$). Subsequently, using the dedicated Psearch software, the time stamp of this candidate event is assigned into the appropriate phase of the Crab Pulsar. This method provides the time stamp for both the optical CPS samples and for the VHE $\gamma$-ray events detected by MAGIC, which could play a crucial role in tagging possible coincident optical-gamma emission.

The new CPS is able to continuously sample the optical signal from the Crab Pulsar at the rate of 10 kHz. By folding the light curve of its periodic pulses, we are able to calibrate our instrument, associating the CPS measured voltage with the well-known optical flux. For this reason, the Crab Pulsar was periodically observed in ON mode for about 20 min. A method equivalent to the analysis in Ref. 6 was used to obtain the phase of each sample, relying on the Crab Pulsar ephemerides provided by the Jodrell Bank Observatory.^18,19 Subsequently, all phases were folded into a single phaseogram. The Crab pulsation is well reproduced in the phaseogram presented in Fig. 9(a), clearly showing its main pulse, P1, at phase 0.3, and secondary pulse, P2, at phase 0.7. Using the average flux of the Crab Pulsar in the U band and assuming a pulse shape equivalent to the Crab light curve taken by Aqueye,²⁰ we can derive the relation between the voltage measured by the CPS and the optical magnitudes in the U band. Since the CPS data are affected by high-frequency noise (caused by the frequency of the DC power supply to electronics plus night-sky background), a 1-ms averaging filter is applied to estimate the RMS of the background using off-source data. Figure 9(b) shows the measured light curve of the Crab Pulsar, compared with the equivalent magnitude of 1- and $5\text{-}\sigma$ level fluctuations coming from the background. Using this method, we can infer the sensitivity of the CPS to single optical pulses for different integration times, shown in Table 2. Even if the Crab pulses are well below the $1\text{-}\sigma$ level of our background noise (and therefore individual Crab pulses cannot be detected), the phase-folding analysis technique significantly improves the sensitivity of the CPS.

Fig. 9

(a) Folded light curve obtained from the CPS (black data with error bars) after only 5 min of observation as compared to an equally binned light curve taken by Aqueye (dashed line²⁰). (b) U band magnitude of the same Crab Pulsar light curve as compared with the equivalent flux levels of 1- and $5\text{-}\sigma$ deviations measured from the background noise of the CPS.

Table 2

MAGIC sensitivity to ms pulses as a function of the integration time with a confidence level of 5 standard deviations. These numbers can be considered conservative, as they were calculated from observations close to the galactic plane, a region with high night-sky background (as described in Sec. 3.1).

The folded light curve is then compared to a uniform distribution using several well-known statistical tests, such as the $Z_{10}^2$, $H$, and ${\chi }^2/ndf$ tests, as described in Ref. 21. After 10 s of observation, the significance values for $Z_{10}^2$, $H$, and ${\chi }^2/ndf$ are 6.2, 6.4, and $5.1\sigma$, respectively. Taking the most conservative value of $5\sigma$ among these statistical tests leads to a sensitivity of $1.6\sigma \sqrt{t/s}$, an improvement in sensitivity relative to the old CPS ($5\sigma$ in 20 s).⁶ MAGIC’s sensitivity is thus also comparable to results from other IACTs, such as HESS (4 $\sigma \sqrt{t/s}$)²² and VERITAS ($0.5\sigma \sqrt{t/s}$),⁸ with the disparities attributed to different mirror size and reflectivity, resolution (strongly depending on the field of view of the central pixel), and photosensors efficiency.

3.2.

Isolated Signals: Sky Scan Test

To test the sensitivity of the CPS to isolated optical flashes, its saturation level and different MAGIC sub-systems clock matching, dedicated observations were performed to induce fast optical flashes of known brightness into the field of view (FOV) of the CPS. This so-called sky scan test was done by slewing the MAGIC-II telescope in azimuthal direction over the star Polaris, fixing the zenith angle to that of Polaris. As the typical slewing speed of MAGIC is $\sim 4.7\mathrm{deg}/\mathrm{s}$, stars passing through the CPS FOV ($\sim 0.1\mathrm{deg}$) would produce optical pulses lasting $\sim 20\mathrm{ms}$, smeared by the optical point spread function of the telescope optics. Using this method, optical flashes of known brightness and length were recorded by the CPS, allowing us to directly verify the correlation between the voltage of a pulse and the magnitude of the stars producing such pulses.

As an example, a snapshot from this test is shown in Fig. 10. Figure 10(a) shows the sky map in horizontal coordinates, together with the position and size of the central PMT FOV (red circle) taken from the MAGIC drive system reports.²³ The position of the stars (blue circles) and their magnitudes in the U band (proportional to the marker size) are queried from the SDSS Photometric Catalog²⁴ within Vizier,²⁵ using the Python package astroquery.²⁶ A Gaussian smearing is applied to the stars near the FOV of the PMT to estimate the starlight flux collected, and the PMT voltage is averaged for 1-ms windows to reduce the impact of high-frequency noise. The correlation between the estimated and measured light curves in this test demonstrates the CPS’s ability to response to the ms-duration optical signal in this test.

Fig. 10

Example snapshot of the sky scan test. (a) Sky view of the CPS (red circle) with stars passing through its FOV. (b) The measured voltage from the CPS (red) and the estimated photon flux (blue) showing the correlation whenever the CPS passes through bright stars.

3.3.

Isolated Signals: Irreducible Background Evaluation

Since IACTs do not have domes, several sources within their environment may cause optical flashes mimicking ms-duration signals in the CPS. For this reason, tens of hours of CPS data without known optical transients (corresponding to standard MAGIC dark observations) were taken to study the frequency of such (background) optical flashes. Several optical pulses with varying magnitudes and morphology were clearly detected during these observations. These optical background sources include car flashes, telescope operators activity (flashlights), satellites, and meteors (shown in Fig. 11). By storing the camera reports (PMT DC currents) with 1-Hz frequency, MAGIC is able to efficiently identify most of these sources as background (mainly identifying moving optical transients over the camera). Meteors, on the other hand, are either too faint or travel too fast through the camera, so they are not likely to be detected by these reports. For this reason, they remain as the main background source during MAGIC operation, limiting MAGIC sensitivity to detect with high significance an optical ms transient of astrophysical origin. By studying the frequency of these meteors as a function of their magnitude, time of night, and date of observation, we are able to calculate the expected number of background events during an observation.

Fig. 11

Example of isolated pulses observed at dark patches of the sky, likely to be meteors (given their speed across the CPS and frequency). The time window in each light curve is 100 ms, and the signal is binned and averaged in 1-ms time intervals to reduce the effect of high-frequency noise.