Studies of working point of iRPC with firmware 4.7

authors

L.Balleyguier, G.Galbit, L.Mirabito Institut de Physique des 2 Infinis de Lyon, CNRS, UCBL

Introduction

This note describes the tuning of PETIROC and FEB parameters measured at Dome in Lyon with an iRPC RE3.1 and a telescope, using cosmic muons.

The 2 first sections analyses the minimal values usable in PETIROC for the AUTO_RESET mechanism and the rise time control (CCOMP). The next sections measures the impact of the retriggering for different settings and try to optimise a working point.

The retriggering phenomena is observed at very low rate when we operate the PETIROC asic near its minimal threshold. In cosmic data taking the occurrence is weak, with $\sim ~ 10^{-4}$ orbits affected, but we can expect that it is partially rate dependent and dead time consequences should be limited as much as possible.

The last version (4.7) of the FEBV2 firmware reintroduces the possibility to switch off the input stage of the asics when a retriggering is detected. It permits to limit the dead time introduced by those events to the time of detection of a retriggering and the time where the input stage is blocked. By this way the dead time can be decreased by a factor > 50 and keep the dead time bellow 1 % at 2 kHz/cm $^2$ rate.

Experiment

The synchronous acquisition of tracks with a telescope of 4 chambers around the IRPC gives access to:

Increase rate of $\mu$ (2-3 Hz of reconstructed tracks) and study local performances of iRPC
Estimate precisely the spatial resolution achievable

Setup

Four 2mm gap RPC's are read with the SDHCAL electronic, a PCB of 30 x 50 cm, with 1 cm $^2$ pads. They surround the iRPC as shown on figures

and in photos

The telescope chambers readout

The telescope chambers are covered with a PCB of 1536 pads of 1 cm $^2$ read by Omega asic HardROC2.

Each asic reads 64 channels auto-triggered and stores in an internal buffer frames of 3X64 bits every time one pad is hit during a 200 ns period. The depth of the internal buffer is 127 and when the buffer is full a busy signal is active for this ASIC.

The 24 Asics are chained both for configuration and for readout and controlled by one front end card (DIF/PMR). The DIF/PMR is controlled by the DAQ PC via a USB link. The acquisition loop is handled by the synchronization board MBMDCC:

A window is opened
When one ASIC is busy a signal is propagated to the synchronization board (MBMDCC), the acquisition window is closed and the ASICs data are read and sent via USB to the controlling PC.
if no board is busy the window is closed after a fixed time and the readout occurred

The iRPC readout

The iRPC is read with the FEBV2_R2 connected to the FC7 with latest firmwares.\ A dedicated FMC mezzanine was designed to handle the following signals to and from the MBMDCC:

OUT: Clock at 40 MHz for the 2 systems
OUT: Busy signal if the IPBUS Fifo buffer of the FC7 is full
IN: Window signal
IN: Resynch signal

In IDLE state the FC7 is stuck in orbit gap. When the window starts, the FC7 starts 89 $\mu$ s orbits until the window ends (busy or end at fixed time). It then sends the data to the PC via IPBUS.

Synchronization

The synchronization is achieved with acquisition windows. The telescope takes data continuously during it but the FEB is sampled by orbit (3560 bx active, 127 off) and have consequently 3.4 % dead time.

Looking to the time selected hits in FEB modulo the orbit length we can clearly see the orbit gap inefficiency that is taken in account to correct efficiency measurement.

iRPC hit time modulo orbit length, for hits in time with a telescope track. The hole corresponds to the orbit gap

The time window of the MBMDCC is limited to 0.2 s to keep buffer size below 1 MBytes. The DAQ efficiency is $\simeq$ 66 %, it is limited by the USB readout time of telescope chambers

Operation environment

The gas mixture is the standard CMS one (95.2 % TFE, 4.5 % C4H10, 0.3 % SF6). The humidity is around 35 % and is obtained by the circulation of the mix in cooled water (3C)
The temperature is controlled at 19C ( $\simeq$ 21C input to chamber). Pressure is measured with a BME280 Bosch sensor and the high voltage is continuously corrected with the CMS RPC formula: $HV_{applied} = HV_{eff} \times (0.2 + 0.8 \times \frac{P\times T_0}{T \times P_0})$ where $P_0=990$ mbar and $T_0=293.15$ K
Telescope chambers were operated with $HV_{eff}$ between 9.1 and 9.4 kV.

Track reconstruction

The 4 chambers are read with the same clock used by the FEB and the two systems start their acquisition simultaneously on a window provided by the MBMDCC. But the telescope chambers sample their data with a 5 MHz clock tagging with a 200 ns bxid their data. Synchronous hits in at least 3 of the 4 planes are merged to reconstruct a track that is extrapolated to the iRPC plane.

FEB data are associated if they match the bxid time of the track, taking into account the latency introduced by the FC7. The following figures shows the bxid distribution of FEB hits with respect to the track. Coincidence hits extend on 3 bxid maximal. One can also notice that the level of noise hit is negligible.

FEB tuning and reconstruction

Choice of AUTO_RESET settings

The PETIROC 2C added an AUTO_RESET functionality to the 2B version. In version 2B, the discriminators have to be reset manually after each trigger , common to the 64 channels. It induces a common dead time to all 32 channels. In the 2C version each channel switches with a common programmable period auto_reset with steps of 6 ns.

Unfortunately we observed that low values of auto_reset induced some oscillation and distorts the output signal

From the following figures we observed a pedestal for DAC10B $\simeq$ 324 and a common noise up to 333. It suggest a minimal cut at $\Delta {\rm DAC_{10b}} \ge 10 \simeq 47 fC$ .

SCurve with a Single-channel-on scan for all channels of one ASIC

SCurve with all channels-on scan for all channels of the same ASIC

On the following figures we change the auto_reset from 3 (24 ns) to 15 (96 ns) and repeat the S-Curve scan with all channels on. We observe that the common noise decreases for low value but also large distortion indicating a possible loss of efficiency. We consider that a auto_reset at 8 (54 ns) is a reasonable compromise from dead time and signal quality point of view.

In the following section we will compare results with AUTO_RESET set to 8, our preferred value with respect to these pedestal measurements, and AUTO_RESET set to 4, the value commonly used at CERN for all tests.

Studies of clustering

Cluster reconstruction

By default we activate the pair filtering in the FEB that transmits only pair of hits (High and low radius) compatible with the PCB geometry, i.e. belonging to a $\delta T$ interval compatible with the strip length. We then associate strip hits in cluster if one hit of the cluster satisfies:

the candidate strip hit is adjacent in strip to the hit of the cluster
the absolute time difference between the candidate hit and the cluster hit is less than 3 ns
the distance between this candidate hit and the cluster hit is less than 20 cm

otherwise a new cluster is created.

The position of the cluster is then taken as the one of the earliest detected strip hit. This strip is named central strip in the following.

The position along the strip is named either $Z_s$ or $\rho$ in the various plots.

Cluster properties analysis

We studied the cluster properties at a standard working point with telescope tracking. A very long runs was taken with the following settings:

On Petiroc, CCOMP is set to 15 and the AUTO_RESET to 4
At FPGA Level, the pair filtering is activated, together with the retriggering mitigation (Threshold 3, Decrement count 11, VAL_EVENT off 4)

The HV was set at the estimated working point (7050 V, see bellow). We can measure the number of clusters reconstructed in a time window of 600 ns wide around the track pass through (figure

Number of clusters reconstructed in run 3172, 600 ns around track time.

and then select the events where only 1 cluster is reconstructed. The strip multiplicity is around 3.4.

Strip multiplicity of a cluster selected

We then construct two estimators of the cluster extension. The first one is the maximal extension in $\rho$ of the cluster, i.e. $Max(|\rho_j -\rho_{central}|)$ where j scans all other strips of the cluster. As one can see on figure, the cluster can extend up to 15 cm that cannot be due to a physic phenomena in the chamber.

We then construct a second estimator which is the distance between the central strip of the cluster to the other one:

It is shifted from 0 and can extend up to 10 cm. This effect on side strips comes directly from the principle of the measurement. Lower signal on these strips will trigger later the discriminator, moreover depending on their position along the strip, the attenuation will vary. It is clearly seen on the following figures where the dependency is obvious both on maximum extension and on dispersion.

Cluster ρ size versus the position ρ_ext along the strip

Dispersion in ρ versus the position ρ_ext along the strip

One can conclude that it biases the position reconstruction along the strip but since the position is reconstructed using the central strip and the alignment or calibration is done with it, it has no effect on the $\rho$ reconstruction as it is is shown on figure that shows the difference between the $\rho$ extrapolated from the track and the reconstructed one. Both the position and the dispersion stay stable along the strip.

Distance ρ_central - ρ_ext between central strip and extrapolation versus the extrapolation ρ_ext along the strip

Finally this time walk effect, despite it cannot be used in $\rho$ reconstruction, might be useful to construct a charge proportion estimator and to improve the strip estimator, i.e. $\phi$ resolution. It will need a dedicated study to parameterize it and measure the effect.

Choice of CCOMP setting

The CCOMP mechanism was introduced since version 2B of the PETIROC asic. It allows to smooth the preamplifier rise time. The following table gives the ${\rm DAC_{10b}}$ calibration for different ccomp values. It was done at Lyon with a pulse rise time of 2 ns and is extrapolated to the unique value (15) measured at CERN.

ccomp	Calibration Lyon (fC/DAC)	Expected Calibration CERN	$\Delta {\rm DAC_{10b}} \ge 12$ CERN (fC)	$\Delta {\rm DAC_{10b}} \ge 12$ LYON
3	2.16	1.72	20.65	25.92
4	2.4	1.91	22.94	28.8
5	2.7	2.15	25.81	32.4
6	2.95	2.35	28.20	35.4
7	3.2	2.55	30.59	38.4
8	3.5	2.79	33.46	42
9	3.7	2.95	35.37	44.4
10	4.1	3.27	39.19	49.2
11	4.35	3.47	41.58	52.2
12	4.8	3.82	45.88	57.6
13	5.13	4.09	49.04	61.56
14	5.5	4.38	52.58	66
15	5.9	4.70	56.40	70.8

Calibration of ${\rm DAC_{10b}}$ for different ccomp values. The calibration was done in Lyon and is roughly linear. One point done at CERN gives 4.7 fC/DAC for ccomp=15. A linear extrapolation is done for the other points and cuts are given for $\Delta {\rm DAC_{10b}} \ge 12$

Since the pedestal and the common noise measured seems to be stable with respect to the ccomp value used, we expected to be able to run with the lowest ccomp value available. Unfortunately the crosstalk is heavily dependent on ccomp. We measure the crosstalk looking at the time distance between the central strip associated to a track to others strips reconstructed in less than 20 ns (figure 1{reference-type="ref" reference="fig:stripdt"}). The estimator of the crosstalk is: $E_{xtalk} = \frac{N(\Delta T < -4 || \Delta T > 4 ~ {\rm ns})}{N(-4 < \Delta T < 4 ~ {\rm ns})}$

Distance in time of other strips to the central one for CCOMP=6

At a fixed voltage of 7000 V we measured the efficiency and the crosstalk for ccomp values between 6 and 15 and with a fixed threshold in DAC count (12). It clearly favors a low value with nearly 2 % more efficiency ( $\simeq$ 100 V) but the cross-talk is high and is really stabilized for ccomp greater than 9. Further studies will use ccomp at 15 and ccomp at 10 as a minimal value.

Efficiency and X-talk for different CCOMP settings

Working point using scintillators coincidences

Details of the measurement

In order to repeat the measurements done at CERN we studied the performances using the coincidences of 2 scintillators around the chamber. The top one is 5x5 cm and the bottom one 10x20 cm. The tubes were positioned in parallel , perpendicular to the table where the chamber is positioned, crossing the strips.

The coincidence of the 2 signals is propagated by the FC7 to the RESYNC signal (channel 32) on each FPGA and only orbits with this coincidence are read. The time distance of chamber hits to this signal on high radius side is shown on figure, strips hits are selected if $-855< T^{HR}_{trigger} - T^{HR}_{hit}< -805$ ns.

Time selection of hits $-855< T^{HR}_{trigger} - T^{HR}_{hit}< -805$ ns

The following plots are made at the same working point as before , i.e. AUTO_RESET set to 4, CCOMP to 15 , DAC threshold to 10, pair filtering and retriggering mitigation activated.

The figure shows the number of clusters built in the time windows,

this one, the number of strip hits per cluster,

and this one the number of strip hits per cluster when only one cluster is found in the time window.

We also studied the coincidences at strip level, the figure shows the number of strips selected,

its strip profile

and 2D profile when the strip is in the time window.

The last figure shows the signal of all strips in the readout orbit (89 $\mu$ s), the noise level stays low.

Finally, looking at cluster positions:

All clusters position in time window
Clusters position when only 1 is time selected

One can observe that coincidences include contributions of interactions in the the light guide or the PM tube. We also measured the maximal $\rho$ size and the the $\rho$ dispersion of the clusters and found comparable values to those obtained with the tracking.

Cluster maximal $\rho$ extension

Cluster $\rho$ dispersion

Working point

All measurements were done with ccomp set to 15 on Petiroc, pair filtering and retriggering mitigation activated. The retriggering mitigation was loose (threshold 6, decrement 36 and VAL_EVENT 7) since we cannot measure its effect with just a sampling of orbits when a trigger is seen.

We first set the auto_reset to 4 and made an HV scan with 2 values of thresholds $DAC_{10b}>7$ and $DAC_{10b}>10$ . The results shows a plateau of efficiency around 97 % with a working point at 6970 V at the lower threshold , shifted by 15 V at the higher one

Efficiency $DAC_{10b}>7$

Multiplicity $DAC_{10b}>7$

Efficiency $DAC_{10b}>10$

Multiplicity $DAC_{10b}>10$

We repeated the measurements with AUTO_RESET set to 8 and $DAC_{10b}>10$

Efficiency

Multiplicity (1 cluster)

The working point obtained (6978 V) is compatible with the value obtained with AUTO_RESET=4. Noise, cross-talk, and retriggering must be evaluated to propose a more solid set of parameters. It requires a larger statistic both on signal orbits and background ones that we can obtain running with the telescope readout.

Working point with telescope tracking

Position properties of reconstructed hits and clusters

With the telescope tracking, we can reconstruct hits and clusters in a time window of 600 ns (see figure above) around the time arrival of the track. We can then associate the nearest cluster in the plan to the track. The figures shows the hits and clusters positions selected in the time window:

All strip hits in time window

All cluster hits in time window

and the hits positions outside the window , i.e. a mix of the noise hit and cosmic rays outside the telescope acceptance.

Off-time strip hits

}

Next figure shows the cluster nearest to the track, and we find back the telescope acceptances.

Cluster hits nearest to track extrapolation

}

The figures shows the spatial resolution achieved.

(a) $\rho_{cluster}-\rho_{ext}$ versus $X_{cluster}-X_{ext}$

(b) $<\rho_{cluster}-\rho_{ext}>$ versus strip

(c) $X_{cluster}-X_{ext}$

(d) $\rho_{cluster}-\rho_{ext}$

The dependency strip to strip (b) is very small but is already calibrated. Without calibration (done once on a long run) dispersion up to 2 cm is observed and not yet fully understood. One important point is that the calibration is done at a given gain so a fixed time-walk. Changing the gain moves the mean $\rho$ position reconstructed that is shifted from 0 (figure (d)).

Working point estimation

A first set of measurements is performed with the PETIROC CCOMP parameter set to its maximum (15) to minimize cross-talk. Two values of auto reset, 8 and 4, were used to see their impact on efficiency, noise rate and retriggering loss.

Several estimators are studied:

the efficiency measured as the number of time one strip is found in time window divided by the number of tracks
$N_s$ (count) The cluster strip multiplicity when associated to the track
The crosstalk rate, evaluated with method described in previous section.
The noise rate, evaluated by counting the mean number of strip hits found in the total acquisition window divided by the detection surface and divided by the mean acquisition time.
The spatial resolution in X
The spatial resolution in $\rho$
The spatial bias in $\rho$
The expected loss of efficiency due to retriggering at a background rate of 2000 Hz/cm $^2$

Safe configuration

We first test a safe configuration with maximal ccomp (15), the auto reset set to 30 ns (8) and a $DAC_{10b}$ threshold well above the noise (12 $\simeq 56$ fC).

The auto reset set to 8, implies a large dead time when retriggering occurs with mitigation parameters set to : count=3 and an mute_roc=7. The corresponding minimal dead time is 220 ns.

The figure shows the results of the efficiency scan. The fit working point is $7093 \pm 2.6$ V on the plateau.

Efficiency for auto_reset set to 8 and $DAC_{10b}>12$

The table summarizes the scan for the previous estimators. The noise remains below 1 Hz/cm $^2$ and the retrigering loss is at a level of 0.2 %. The crosstalk is also contained below 3 % and the spatial resolution is 0.6 cm in X and 1.2 cm in $\rho$ .

Summary of all estimators for ccomp=15,auto reset=8 and threshold $\Delta DAC_{10b} \ge 12$

HV	$\epsilon$	$\delta_\epsilon$	$N_s$	$R_{noise}$	$\sigma_x$	$\sigma_\rho$	$\delta_\rho$	$R_{loss}$	$X_{talk}$
7250	99.5	0.3	3.4	0.83	0.53	1.17	1.4	0.21	1.8
7200	99.7	0.3	3.3	0.78	0.52	1.18	1.3	0.2	2.2
7150	98.7	0.5	3.1	0.77	0.52	1.2	1.1	0.18	3.0
7100	98.6	0.3	3.0	0.72	0.53	1.23	1.0	0.17	1.4
7050	97.8	0.4	2.8	0.69	0.53	1.27	0.9	0.17	1.5
7000	96.2	0.4	2.7	0.66	0.52	1.3	0.7	0.16	1.3
6950	92.7	0.5	2.4	0.64	0.53	1.34	0.5	0.15	1.3
6900	87.4	0.6	2.2	0.63	0.54	1.47	0.2	0.15	1.7
6850	79.4	0.7	2.0	0.6	0.54	1.54	0.0	0.15	1.8
6800	68.8	0.8	1.8	0.57	0.55	1.72	-0.3	0.14	2.7
6750	53.6	0.8	1.6	0.56	0.55	1.74	-0.5	0.14	3.4
6700	43.2	0.8	1.6	0.53	0.55	1.94	-0.7	0.12	1.9
6650	27.3	0.7	1.5	0.5	0.58	2.22	-1.0	0.12	11.1
6600	16.6	0.6	1.5	0.46	0.61	2.32	-1.3	0.12	5.1

We repeated the measurement with a slightly lower threshold value of 47 fC ( $DAC_{10b}>10$ ). Results can be seen in the following table and figure. The working point is improved to $7052 \pm 2.6$ V on the plateau with the price of a slightly higher noise (1 %) and retrigerring loss ( $\sim$ 0.25 %). The $\rho$ resolution is also slight;y degraded to 1.3 cm.

Efficiency for auto_reset set to 8 and $DAC_{10b}>10$

Summary of all estimators for ccomp=15,auto reset=8 and threshold $\Delta DAC_{10b} \ge 10$

HV	$\epsilon$	$\delta_\epsilon$	$N_s$	$R_{noise}$	$\sigma_x$	$\sigma_\rho$	$\delta_\rho$	$R_{loss}$	$X_{talk}$
7200	99.3	0.3	3.4	1.0	0.53	1.25	1.3	0.3	3.1
7150	99.6	0.3	3.3	0.91	0.54	1.27	1.3	0.26	1.4
7100	98.2	0.4	3.2	0.86	0.54	1.33	1.1	0.25	2.3
7050	98.1	0.4	3.1	0.82	0.53	1.32	0.9	0.23	1.5
7000	97.1	0.4	2.9	0.77	0.52	1.42	0.7	0.22	2.2
6950	95.4	0.4	2.7	0.73	0.51	1.41	0.6	0.21	1.7
6900	91.9	0.5	2.5	0.7	0.55	1.55	0.4	0.2	1.0
6850	85.3	0.6	2.3	0.67	0.53	1.68	0.2	0.2	1.7
6800	76.3	0.7	2.1	0.63	0.56	1.83	-0.0	0.18	1.3
6750	65.9	0.8	1.9	0.59	0.55	1.99	-0.3	0.18	1.3
6700	50.9	0.8	1.7	0.56	0.57	2.26	-0.7	0.18	3.1
6650	35.7	0.8	1.7	0.52	0.58	2.33	-0.7	0.17	0.4
6600	22.5	0.7	1.6	0.49	0.57	2.51	-0.8	0.17	5.3

GIF++-like configuration

The main difference with the previous configuration is the change of auto reset to 4. It has a direct effect on the dead time induced by the retrigerring sinc we can decrease the mitigation parameters to:count=3 and an mute_roc=4 with a minimal dead time of 120 ns.

We first repeat a scan with a DAC threshold set to 10 (47 fC) and obtain similar working point of $7048 \pm 2.7$ V. The main improvement comes from the retrigerring loss that is contained bellow 0.05 %

Efficiency for auto_reset set to 4 and $DAC_{10b}>10$

Summary of all estimators for ccomp=15,auto reset=4 and threshold $\Delta DAC_{10b} \ge 10$

HV	$\epsilon$	$\delta_\epsilon$	$N_s$	$R_{noise}$	$\sigma_x$	$\sigma_\rho$	$\delta_\rho$	$R_{loss}$	$X_{talk}$
6600	21.3	0.7	1.5	0.54	0.54	2.7	-1.0	0.04	11.7
6650	34.9	0.9	1.6	0.55	0.58	2.3	-0.8	0.04	5.9
6700	50.4	0.9	1.8	0.56	0.57	2.12	-0.6	0.04	4.5
6750	65.8	0.9	1.9	0.58	0.57	1.88	-0.3	0.04	5.0
6800	79.1	0.8	2.1	0.61	0.57	1.76	-0.1	0.04	2.5
6850	86.4	0.7	2.3	0.63	0.54	1.59	0.1	0.04	2.1
6900	91.8	0.6	2.6	0.66	0.55	1.54	0.4	0.04	1.7
6950	96.0	0.5	2.7	0.68	0.52	1.46	0.6	0.04	2.4
7000	97.5	0.4	2.9	0.78	0.52	1.35	0.7	0.04	1.9
7050	97.8	0.4	3.0	0.82	0.52	1.33	0.9	0.04	1.9
7100	99.0	0.3	3.2	0.88	0.52	1.23	1.0	0.04	3.3
7200	99.4	0.3	3.4	0.98	0.55	1.27	1.4	0.05	3.0
7150	99.0	0.3	3.3	0.93	0.54	1.22	1.2	0.05	2.8
7250	99.8	0.3	3.5	1.04	0.53	1.23	1.4	0.05	3.6

We then decreased the DAC threshold to 8 but the working point $7052 \pm 2.9$ V does not significantly improve. The retrigerring loss and the noise stays low but we can observe a small increase of the crosstalk.

Efficiency for auto_reset set to 4 and $DAC_{10b}>8$

Summary of all estimators for ccomp=15,auto reset=4 and threshold $\Delta DAC_{10b} \ge 8$

HV	$\epsilon$	$\delta_\epsilon$	$N_s$	$R_{noise}$	$\sigma_x$	$\sigma_\rho$	$\delta_\rho$	$R_{loss}$	$X_{talk}$
7200	99.9	0.3	3.7	1.24	0.57	1.33	1.4	0.08	5.9
7150	99.2	0.3	3.5	1.15	0.54	1.34	1.3	0.08	4.8
7100	98.6	0.4	3.4	1.09	0.55	1.35	1.1	0.07	3.3
7050	98.1	0.4	3.2	1.05	0.53	1.39	1.0	0.07	3.4
7000	96.5	0.5	3.1	1.04	0.53	1.45	0.7	0.06	1.6
6950	95.2	0.5	2.9	1.01	0.54	1.55	0.6	0.06	2.7
6900	91.8	0.6	2.7	0.98	0.54	1.59	0.4	0.06	2.6
6850	85.9	0.8	2.5	0.95	0.54	1.78	0.2	0.06	2.3
6800	80.6	0.9	2.4	0.9	0.56	1.95	0.0	0.05	3.7
6750	68.1	0.9	2.1	0.87	0.57	2.12	-0.4	0.05	2.5
6700	54.3	0.9	1.9	0.82	0.57	2.4	-0.6	0.05	2.3
6650	40.6	0.8	1.8	0.77	0.62	2.66	-0.6	0.05	6.8
6600	27.0	0.8	1.8	0.71	0.58	2.67	-1.0	0.05	10.5

Eventually we set the DAC threshold to its lowest possible value 7 (33 fC) and repeated the scan. The working point decreases to $7027 \pm 3$ V but the retrigerring loss double ( $\sim$ 0.15 %) and so does the crosstalk ( $\sim$ 5 %).

Efficiency for auto_reset set to 4 and $DAC_{10b}>7$

Summary of all estimators for ccomp=15,auto reset=4 and threshold $\Delta DAC_{10b} \ge 7$

HV	$\epsilon$	$\delta_\epsilon$	$N_s$	$R_{noise}$	$\sigma_x$	$\sigma_\rho$	$\delta_\rho$	$R_{loss}$	$X_{talk}$
7100	99.3	0.4	3.5	1.28	0.53	1.46	1.0	0.14	4.3
7050	97.8	0.5	3.3	1.23	0.52	1.6	0.8	0.13	5.7
7000	98.1	0.5	3.2	1.17	0.55	1.59	0.7	0.17	5.1
6950	95.8	0.6	3.0	1.12	0.53	1.66	0.6	0.15	2.1
6900	94.3	0.6	2.8	1.08	0.55	1.85	0.3	0.15	2.6
6850	89.7	0.7	2.7	1.03	0.55	1.88	-0.0	0.34	2.6
6800	80.2	0.9	2.4	0.91	0.56	2.08	-0.3	0.15	4.0
6750	69.9	1.0	2.2	0.87	0.55	2.18	-0.4	0.12	2.1
6700	59.4	1.0	2.1	0.82	0.6	2.37	-0.8	0.27	3.2
6650	44.0	1.0	1.9	0.76	0.57	2.63	-0.8	0.14	3.0
6600	29.0	1.0	1.8	0.7	0.56	2.61	-1.3	0.15	7.4

Low ccomp test

Calibration tests show that lower value of ccomp have a better sensitivity. With ccomp=10 one can expect 3.2 fC per DAC. After redoing a full calibration (pedestal, time alignment) we repeated the HV scan. With a DAC threshold of 32 fC (10) we obtained similar results (figure 7{reference-type="ref" reference="fig:telv48v5_efficiency"} and table 7{reference-type="ref" reference="tab:telv48v5"}) to those previously obtained with ccomp=15 and DAC threshold to 7. The crosstalk is maybe a bit higher , up to 10 % at higher gain but the retrigerring loss is comparable.

The data taking was stopped due to high summer temperature in the lab , leading to gas temperature up to 27 $^\circ$ C. This induces higher noise, tests will be resumed in September.

Efficiency for ccomp=10, auto_reset set to 4 and $DAC_{10b}>10$

Summary of all estimators for ccomp=10,auto reset=4 and threshold $\Delta DAC_{10b} \ge 10$

HV	$\epsilon$	$\delta_\epsilon$	$N_s$	$R_{noise}$	$\sigma_x$	$\sigma_\rho$	$\delta_\rho$	$R_{loss}$	$X_{talk}$
6550	17.1	1.0	1.7	0.9	0.53	2.33	-0.7	0.16	24.0
6600	30.7	1.3	1.6	0.96	0.56	2.29	-0.3	0.17	24.1
6650	43.2	1.5	1.7	1.0	0.55	1.8	-0.1	0.17	17.0
6700	58.0	1.6	1.9	1.09	0.51	1.86	-0.2	0.18	2.6
6750	72.7	1.4	2.0	1.19	0.54	1.84	0.0	0.19	3.6
6800	79.6	1.3	2.2	1.32	0.55	1.65	0.1	0.18	2.5
6850	88.6	1.2	2.4	1.47	0.49	1.44	0.4	0.19	5.2
6900	92.8	1.1	2.7	1.65	0.51	1.48	0.6	0.19	3.4
6950	96.4	1.2	2.9	1.84	0.48	1.47	0.6	0.2	4.2
7000	98.7	0.9	3.0	2.06	0.51	1.34	0.8	0.19	3.4
7050	97.6	1.1	3.2	2.3	0.47	1.33	1.1	0.2	5.1
7100	98.4	1.1	3.3	2.54	0.53	1.41	1.2	0.21	11.5
7150	98.2	1.1	3.4	2.83	0.52	1.23	1.1	0.21	10.3
7200	100.2	1.0	3.5	3.24	0.53	1.33	1.1	0.22	11.4

Tests were resumed in September and we first repeated the $\Delta DAC_{10b} \ge 10$ scan. The performances are comparable but the fit working point improves by 10 V (systematic error).

Efficiency for ccomp=10, auto_reset set to 4 and $DAC_{10b}>10$ second test

Summary of all estimators for ccomp=10,auto reset=4 and threshold $\Delta DAC_{10b} \ge 10$ , second test in September

HV	$\epsilon$	$\delta_\epsilon$	$N_s$	$R_{noise}$	$\sigma_x$	$\sigma_\rho$	$\delta_\rho$	$R_{loss}$	$X_{talk}$
6550	17.3	0.9	1.6	0.36	0.59	2.65	-0.6	0.22	9.7
6600	28.1	1.0	1.7	0.38	0.48	2.37	-0.3	0.22	4.9
6650	46.2	1.1	1.8	0.42	0.55	2.3	-0.7	0.22	15.2
6700	60.6	1.1	1.9	0.47	0.55	2.16	-0.2	0.22	3.6
6750	74.0	1.0	2.1	0.51	0.58	1.86	0.0	0.23	3.6
6800	82.5	0.9	2.4	0.55	0.53	1.76	0.3	0.23	3.8
6850	91.3	0.6	2.6	0.61	0.53	1.71	0.3	0.22	6.5
6900	93.8	0.7	2.8	0.66	0.5	1.53	0.5	0.23	5.4
6950	97.0	0.6	2.9	0.67	0.51	1.44	0.7	0.26	5.8
7000	98.5	0.5	3.2	0.73	0.53	1.43	0.9	0.26	6.8
7050	99.7	0.5	3.3	0.77	0.52	1.4	0.9	0.28	8.1
7100	99.3	0.5	3.5	0.83	0.52	1.37	1.0	0.28	10.2
7150	99.9	0.5	3.5	0.88	0.52	1.34	1.1	0.29	16.6
7200	99.7	0.5	3.7	0.96	0.53	1.38	1.3	0.3	19.7

We then repeat the test with $\Delta DAC_{10b} \ge 12$ (39 fC). The working point is shifted by 30 V but the crosstalk and the retrigerring loss is improved.

Efficiency for ccomp=10, auto_reset set to 4 and $DAC_{10b}>12$

Summary of all estimators for ccomp=10,auto reset=4 and threshold $\Delta DAC_{10b} \ge 12$

HV	$\epsilon$	$\delta_\epsilon$	$N_s$	$R_{noise}$	$\sigma_x$	$\sigma_\rho$	$\delta_\rho$	$R_{loss}$	$X_{talk}$
6550	13.5	0.8	1.4	0.35	0.56	2.55	-1.1	0.11	49.2
6600	21.2	0.9	1.6	0.38	0.53	2.15	-1.0	0.11	21.1
6650	34.5	1.1	1.5	0.42	0.53	2.06	-0.5	0.11	6.5
6700	48.4	1.2	1.7	0.48	0.54	1.91	-0.4	0.12	15.4
6750	65.1	1.2	1.8	0.55	0.55	1.75	-0.2	0.12	5.3
6800	77.7	1.1	2.1	0.61	0.51	1.6	0.0	0.13	2.5
6850	85.8	0.9	2.2	0.69	0.5	1.55	0.2	0.13	3.8
6900	93.2	0.8	2.5	0.78	0.5	1.38	0.4	0.14	6.1
6950	95.7	0.7	2.7	0.84	0.5	1.33	0.6	0.15	3.1
7000	98.0	0.5	3.0	0.72	0.49	1.32	0.8	0.14	5.6
7050	98.3	0.6	3.1	1.04	0.5	1.3	0.9	0.17	6.4
7100	99.1	0.5	3.2	1.13	0.49	1.25	1.0	0.17	6.4
7150	99.5	0.5	3.4	1.19	0.51	1.23	1.1	0.18	8.5
7200	98.8	0.5	3.4	1.13	0.53	1.21	1.1	0.18	6.5

Eventually we tried to decrease the threshold to 30 fC, i.e $\Delta DAC_{10b} \ge 9$ . The working point is slightly improved but the crosstalk doubles ( $\sim$ 10 %).

Efficiency for ccomp=10, auto_reset set to 4 and $DAC_{10b}>9$

Summary of all estimators for ccomp=10,auto reset=4 and threshold $\Delta DAC_{10b} \ge 9$

HV	$\epsilon$	$\delta_\epsilon$	$N_s$	$R_{noise}$	$\sigma_x$	$\sigma_\rho$	$\delta_\rho$	$R_{loss}$	$X_{talk}$
6500	12.4	0.8	1.7	0.61	0.47	2.91	-0.5	0.49	27.6
6550	23.7	1.1	1.6	0.63	0.59	2.57	-0.3	0.6	11.3
6600	35.9	1.3	1.8	0.64	0.52	2.23	-0.3	0.42	7.2
6650	49.7	1.6	1.7	0.65	0.5	2.18	-0.0	0.9	20.3
6700	62.2	1.3	2.1	0.69	0.58	2.09	0.1	0.22	4.1
6750	74.1	1.1	2.2	0.73	0.54	1.99	0.2	0.22	3.5
6800	84.8	1.0	2.4	0.78	0.52	1.92	0.3	0.23	5.7
6850	90.7	0.9	2.7	0.82	0.54	1.73	0.4	0.24	4.5
6900	94.5	0.8	2.9	0.81	0.54	1.66	0.6	0.35	7.1
6950	96.9	0.7	3.1	0.87	0.52	1.52	0.8	0.38	11.9
7000	98.4	0.5	3.2	0.94	0.53	1.44	0.9	0.27	9.6
7050	99.5	0.5	3.4	0.99	0.51	1.42	1.0	0.25	12.2
7100	99.1	0.5	3.5	1.06	0.52	1.44	1.1	0.25	14.9
7150	99.4	0.5	1.1	0.63	0.56	1.47	1.2	0.36	35.7
7200	99.3	0.5	1.1	0.68	0.61	1.57	1.2	0.39	43.0

Asymmetric threshold

Since the attenuation of the return lines ( $\sim$ 20 %) would require a lower threshold that may induce a higher noise, one can use the pair filtering to keep a high threshold on the direct line and a lower one on the return side. We made two tests (auto_reset=4,ccomp=15) with $DAC^{HR}_{10b}>10 / DAC^{LR}_{10b}>8$ and $DAC^{HR}_{10b}>9 / DAC^{LR}_{10b}>7$ .

Efficiency $DAC^{HR}_{10b}>10 / DAC^{LR}_{10b}>8$

Summary of all estimators for ccomp=15,auto reset=4 and threshold $DAC^{HR}_{10b}>10 ~ \& ~ DAC^{LR}_{10b}>8$

HV	$\epsilon$	$\delta_\epsilon$	$N_s$	$R_{noise}$	$\sigma_x$	$\sigma_\rho$	$\delta_\rho$	$R_{loss}$	$X_{talk}$
7200	100.0	0.3	3.6	1.12	0.53	1.21	2.0	0.07	4.2
7150	99.8	0.3	3.4	1.05	0.54	1.22	2.0	0.07	3.2
7100	98.9	0.4	3.3	1.01	0.53	1.29	1.9	0.06	3.2
7050	98.1	0.4	3.2	0.96	0.53	1.27	1.8	0.06	2.6
7000	97.5	0.4	3.0	0.94	0.51	1.31	1.7	0.06	1.5
6950	95.9	0.4	2.8	0.91	0.53	1.4	1.6	0.05	1.6
6900	93.4	0.5	2.6	0.86	0.54	1.58	1.4	0.05	2.0
6850	87.3	0.6	2.4	0.82	0.54	1.63	1.4	0.05	3.2
6800	78.0	0.7	2.2	0.78	0.55	1.82	1.2	0.04	2.9
6750	65.0	0.8	2.0	0.74	0.55	1.98	1.0	0.04	1.9
6700	53.6	0.8	1.8	0.71	0.58	2.09	0.9	0.04	2.4
6650	38.1	0.8	1.7	0.67	0.55	2.29	0.8	0.04	4.2
6600	23.1	0.7	1.7	0.63	0.55	2.51	0.6	0.04	8.0

Efficiency $DAC^{HR}_{10b}>9 / DAC^{LR}_{10b}>7$

Summary of all estimators for ccomp=15,auto reset=4 and threshold $DAC^{HR}_{10b}>9 ~ \& ~ DAC^{LR}_{10b}>7$

HV	$\epsilon$	$\delta_\epsilon$	$N_s$	$R_{noise}$	$\sigma_x$	$\sigma_\rho$	$\delta_\rho$	$R_{loss}$	$X_{talk}$
7200	99.2	0.3	3.7	1.2	0.56	1.23	2.0	0.08	6.0
7150	99.6	0.3	3.5	1.13	0.53	1.28	1.9	0.08	4.3
7100	99.6	0.3	3.4	1.07	0.55	1.32	1.8	0.07	4.7
7050	98.4	0.4	3.3	1.01	0.53	1.36	1.8	0.07	3.4
7000	98.1	0.4	3.1	0.97	0.53	1.47	1.7	0.07	2.9
6950	96.5	0.4	3.0	0.91	0.54	1.5	1.6	0.07	3.4
6900	93.1	0.5	2.8	0.88	0.54	1.6	1.5	0.07	3.1
6850	87.4	0.6	2.5	0.81	0.55	1.78	1.3	0.07	3.3
6800	81.1	0.7	2.3	0.74	0.55	1.85	1.2	0.07	2.2
6750	69.8	0.8	2.2	0.69	0.56	1.99	1.2	0.07	3.7
6700	56.7	0.8	2.0	0.66	0.61	2.23	0.9	0.06	1.5
6650	40.7	0.8	1.8	0.64	0.57	2.46	0.7	0.06	5.6
6600	28.8	0.7	1.7	0.6	0.59	2.83	0.8	0.06	8.5
6550	16.3	0.6	1.7	0.58	0.57	2.72	0.8	0.05	7.7

One can improve slightly the working point especially with the 9/7 combination where $7039 \pm 2.9$ V is achieved one the plateau. Crosstalk is kept around 3 % and the retrigerring loss around 0.05 %. Further studies will be resume with ccomp=10 in September.

We finally repeat the study with an asymmetric threshold $DAC^{HR}_{10b}>13$ and $DAC^{LR}_{10b}>10$ with pa_ccomp set to 10. The working point ( $7018.3 \pm 2.9$ V) is identical to the one obtained with a symmetric cut at 10 but the crosstalk is really improved. The retrigerring loss at 2 kHz rate stays around 0.3 % and is acceptable.

Efficiency for ccomp=10, auto_reset set to 4 and $DAC_{10b}>10/13$

Summary of all estimators for ccomp=10,auto reset=4 and threshold $\Delta DAC_{10b} \ge 10/13$

HV	$\epsilon$	$\delta_\epsilon$	$N_s$	$R_{noise}$	$\sigma_x$	$\sigma_\rho$	$\delta_\rho$	$R_{loss}$	$X_{talk}$
6550	15.2	0.8	1.5	0.34	0.52	2.44	1.4	0.18	10.8
6600	26.6	1.0	1.6	0.39	0.61	2.4	1.7	0.18	12.6
6650	41.5	1.1	1.6	0.44	0.53	2.08	1.5	0.18	7.6
6700	56.5	1.1	1.8	0.48	0.53	1.9	1.6	0.19	4.7
6750	72.1	1.0	2.0	0.52	0.55	2.05	1.8	0.19	5.3
6800	83.7	0.9	2.2	0.56	0.52	1.81	1.8	0.2	2.9
6850	89.7	0.8	2.5	0.6	0.54	1.58	1.9	0.24	4.2
6900	94.5	0.7	2.6	0.68	0.5	1.47	1.8	0.22	5.9
6950	96.4	0.6	2.9	0.72	0.53	1.32	1.9	0.24	5.0
7000	99.4	0.5	3.1	0.76	0.53	1.25	1.9	0.29	4.5
7050	99.0	0.5	3.2	0.81	0.51	1.26	2.0	0.38	5.8
7100	99.0	0.5	3.2	0.77	0.47	1.21	2.0	0.33	5.2
7150	99.5	0.5	3.2	0.78	0.52	1.23	2.0	0.44	4.8
7200	99.4	0.5	3.2	0.76	0.52	1.24	2.1	0.25	5.0

Comparison with scintillators coincidence triggers

When looking to the efficiency measured using the coincidence of 2 scintillators or the track from the telescope, we notice a shift of 60 V on the working point in favor to the scintillator measure.

One possible explanation may come from the track selection that prefers high energy muons (straight line in 5 chambers) compare to the scintillator studies that allow large multiple scattering up to the photo-multiplier guide line. This last measurement should be repeated with a third coincidence to select muons of higher energy and confirm this hypothesis.

Remark

There is also discrepancies in working point up to 15 V between same measurements repeated at different time. It may comes from the P,T correction that may not be valid for large (>10 degrees) variation or from the quality of the fit. It should be use as a systematic error on our working point prediction. A last point is the momentum of the cosmic muon, despite the track selection, is not on the ultra relativistic plateau of the ionization and it may induce a smaller signal and a worsen working point.