Detector Module Performance

A few runs were dedicated to the comparison of energy loss between pions and protons at a momentum of $350\,\rm MeV/c$. The pions are approximately minimum ionizing with a Landau-distributed energy loss, while the protons deposit much higher energy with almost Gaussian spread. Fig.  shows typical pion (left) and proton (right) signal distributions.

Figure: Typical pion (left) and proton (right) signal distributions measured at a momentum of $350\,\rm MeV/c$. In the proton picture, the x axis range is five times larger in the pion plot. The absolute charge scale has been obtained from internal calibration, which is not very accurate.

The measured energy loss is in good agreement with the restricted Bethe-Bloch theory (see section , p. ), as shown in fig. . Since the pions are approximately minimum ionizing, they are the worst case in terms of the signal-to-noise ratio (SNR). Thus, all further results are obtained with pions except for the angle scan, where both pions and protons are compared.

Figure: Measured energy loss of pions and protons in comparison to calculation from the restricted Bethe-Bloch theory.

Figure: Deconvolution signal-to-noise ratio of different silicon detector modules as a function of the detector bias normalized to the respective non-irradiated depletion voltage.

The APV6 chip has a significantly higher noise contribution than its successor APV25. This reflects in the SNR values of the different modules. Fig.  shows the deconvolution mode SNR results of selected modules vs. the detector bias normalized to the respective non-irradiated depletion voltage. Several conclusions can be drawn from these curves. First of all, the APV25 readout outperforms the APV6. Tab.  shows typical SNR and the electronic noise values obtained with detector modules read out by APV6 and APV25 in peak and deconvolution modes.

Table: Typical noise (ENC) and most probable signal-to-noise (SNR) values of full-size, non-irradiated CMS detector modules measured with minimum ionizing particles (MIPs).

Mode	APV25
	$\rm ENC\,[e]$	$\rm SNR$	$\rm ENC\,[e]$	$\rm SNR$
Peak	$1400$	$16$	$\:\;900$	$25$
Deconvolution	$2250$	$10$	$1300$	$17$

With only one sensor, the capacitive load and thus the noise is lower, leading to a higher SNR as shown for the PD25 module in fig.  and for the multiregion detector in section , p. . Moreover, the SNR curves of irradiated silicon detectors need higher bias voltages and yet do not really saturate (see section , p. ). Although the SNR of irradiated sensors is lower, their performance is still satisfactory for the application in the CMS tracker. From the detector simulation discussed in section , p. , a good approximation of the signal-to-noise ratio around and above depletion has been obtained.

Depending on the APV latency value, the sampled values are stored in the chip pipeline for a defined time. With variation of this latency between 25 and 150 clock cycles, no difference could be spotted in the signal-to-noise ratio. This demonstrates that the pipeline storage capacitors are able to hold their charges without noticeable leakage for at least $3.75\,\rm\mu s$, exceeding the CMS first level trigger latency.

Moreover, the SNR was measured as a function of the particle hit position on the detector along the strip axis. Again, no difference could be observed between hits close to and far from the amplifier chips. Since the particle induced signal is current driven, it does not depend on the series line resistance between hit position and amplifier input, while the noise performance is independent of particles anyway. Thus, the signal-to-noise ratio is not affected by the hit position on the detector.

An angle scan was performed with the Vienna APV25 module outside of the cooling box. The results obtained here are compatible with previous measurements discussed in section , p. , and the same fit functions have been applied. Fig.  shows signal (top) and cluster width (bottom) as functions of the incident angle for both pions and protons, where an angle of zero denotes perpendicular incident.

Figure: Angle scan of the Vienna APV25 module. Both cluster signal (top) and cluster width (bottom) increase with the incident angle.

While the signals of the detector modules were usually read out by a copper cable with a length of $25\,\rm m$, a prototype of the analog optical link (see section , p. ), with $97\,\rm m$ of optical fiber was used instead of the cable for comparative measurements. In general, the same signal-to-noise ratio was measured with cable and optical fiber. This is because the long cable brought a bandwidth limitation which led to a slight signal reduction compared to a short cable. The optical link has a significantly higher bandwidth but contributes additional noise. These effects approximately compensate each other, yielding similar results in both cases.

For a short period, the setup was modified in such a way that the APVs were directly clocked with the $50\,\rm MHz$ PSI frequency and thus, the synchronization module could be abandoned. In peak mode, the performance remained unchanged, while in deconvolution mode a signal loss of approximately $13\%$ was observed due to the static weights of the signal processing algorithm which are laid out for $40\,\rm MHz$ operation (see section , p. ). This loss has been confirmed by simulation.