- Email: [email protected]
Electronics for the CMD-2M detector
Nuclear Instruments and Methods in Physics Research A 494 (2002) 560–564
Electronics for the CMD-2M detector Yu.V. Yudin*, A.A. Ruban Budker Institute of Nuclear Physics, Acad Lavrentiev Prospect 11, Novosibirsk 630090, Russia
Abstract The CMD-2M detector will replace the CMD-2 detector. Many of the detector’s sensitive devices and the relevant electronics will be improved or re-built. Both the front-end, digitizing and data acquisition electronics will be modernized. The new electronics include several types of units which perform analog signal processing together with digitization. Unificated functional units are widely used in the designs. A careful selection of signal processing techniques and circuit design options is needed to make these electronics substantially cheaper than conventional electronics of similar detectors. r 2002 Elsevier Science B.V. All rights reserved. Keywords: Signal digitization; Digitizing electronics; Data acquisition electronics
1. Introduction An upgrade of the VEPP-2M collider and of two detectors installed on it is under way now in BINP. The projected parameters of the new collider (named VEPP-2000) are the following: cm energy up to 2 GeV and peak luminosity about 1032 cm 2 s 1, which is some 10 times the luminosity of the old collider. A new version of the CMD-2 detector, named CMD-2M, will be adequate for the increased collider capabilities. The general structure of the CMD-2M [1] remains the same, but almost all the existing systems of the detector will be re-designed and re-built, and a totally new liquid xenon calorimeter will be installed. The relevant front-end electronics for renewed sensitive devices and LXe calorimeter will be totally new. Taking into account all these principal changes it was decided to re-design the *Corresponding author. E-mail address: [email protected] (Yu.V. Yudin).
rest of the electronics too. It will be replaced with a new one step-by-step.
2. Basic issues for electronics design The sensitive devices of the CMD-2M detector, the data to be derived from their signals and the speed considerations are briefly described below. Drift Chamber (DC) [1] contains 1280 signal wires. The tracks’ co-ordinates in the R–Phi plane are measured by avalanche drift time, and the Z coordinate is measured by charge division techniques. Hence, to obtain tracking information three quantities must be measured for each wire: 2 charges from the ends of the wire and the time interval from signal arrival to a common stop signal. The ‘‘FLT+’’ signal from the First-level Trigger (FLT) is used both to initialize digitization and as the common stop for time measurement. It is aligned in time to the beam crossing moments. For charge measurement a resolution of 12 bits is
0168-9002/02/$ - see front matter r 2002 Elsevier Science B.V. All rights reserved. PII: S 0 1 6 8 - 9 0 0 2 ( 0 2 ) 0 1 5 4 9 - 8
Yu.V. Yudin, A.A. Ruban / Nuclear Instruments and Methods in Physics Research A 494 (2002) 560–564
required (14 bits is desirable). For time measurement a 2.5 ns step and a 10-bit full scale are required. Z-Chamber (ZC) [2] produces two types of signals. 64 signals of anode wires are the fastest and are used for triggering. 512 cathode stripes’ signals carry only amplitude information and are to be digitized at 12-bit resolution. LXe Calorimeter, Barrel Calorimeter and Endcap Calorimeter [3,4] altogether produce 2094 signals, which are used mainly for energy deposition measurements. These signals’ amplitudes are to be digitized at 12-bit resolution. Certain groups of these signals are fed to analog summers, and sum signals are used for triggering. The amplitude and arrival time of sum signals are digitized. Besides, the co-ordinate system of the LXe calorimeter produces 2112 signals for tracking. Their amplitude must be digitized at a resolution of 12 bits. DAQ system. The total number of signal sources in the detector is about 6800, and some 8000 analog quantities are derived from their signals. In a ‘‘good’’ event, usually less than 10% of signal sources operate and the relevant electronics channels contain non-zero data. Therefore, the portion of valuable data related to one event is about 1K words, each of which must include the channel’s address (not less than 13 bits) and datum (10–14 bits). With a reasonable word size of 32 bits the average event size is about 4 kB. The expected ultimate trigger rate will be about 1 kHz, and at this rate the deadtime must be less than 10%. Hence, the duration of digitization and transferring the raw data from ADCs to memory buffer (de-randomizer) must be less than 100 ms. At the same time, at a given trigger rate the average data stream (after de-randomization and zero suppression) will be up to 4 MB/s.
3. Main constraints of the design of the new electronics 1. Having some 8000 analog quantities to be digitized, signal multiplexing allows one to improve the reliability of the electronics and reduce its cost due to substantial reduction of
561
the number of modules and cable interconnections between them. 2. When choosing the option for design of the electronics of each detector’s system we took into account not only this system separately, but also tried to make some optimal unification of the electronics of all systems. Most of the signals in the detector carry just amplitude information. Such signals are usually converted into voltage levels for digitizing. Moreover, the time quantities can be (and often are) easily converted into voltage levels during digitizing. Therefore, for the given set of signals being present in CMD-2M the most effective way of digitization is the following. All analog quantities (amplitudes and times) are converted into voltage levels and digitized by identical ADCs. 3. Since the number of channels in each detector’s system is about one to two thousand, it is not yet cost effective to make a custom ASIC for signal processing. Nevertheless, due to high capabilities of commercially available ICs it is possible to dispose analog signal processing stages and digitizing units in the same modules. This makes the problem of different ‘‘grounds’’ easier to solve, and allows one to reduce the number of cable interconnections. That is why we preferred to place digitizing units (instead of few more analog channels) in new signal processing modules (SPMs). 4. The functions of the DAQ system are distributed as follows. The raw digital data produced by ADC in each SPM are transferred immediately (along with digitization being performed) via an individual serial link to the data receiver. The data receiver is a card installed in the PCIslot of a desktop PC. It accepts the data from 16 SPMs, de-randomizes the data and forwards them to PC’s processor. The PC performs reading data from data receiver cards, zero suppression, etc., and then sends the reduced data via Ethernet for subsequent processing and recording. A functional diagram of the new electronics showing mainly how signal processing and digitizing are carried out is presented in Fig. 1.
562
Yu.V. Yudin, A.A. Ruban / Nuclear Instruments and Methods in Physics Research A 494 (2002) 560–564
Fig. 1. Simplified functional scheme of the new electronics of the CMD-2M detector.
4. Electronics 4.1. DC signals processing A simplified functional diagram of the DC SPM is shown in Fig. 2. After amplification each signal is fed to a gated integrator (labelled ‘‘Q-U’’) and to analog summer. When the summary signal exceeds the threshold, the ‘‘Wire Trigger’’ generates signals for starting charge and time integrators, and also a hit signal for the FLT. So, since the discriminator has operated, the channel has been performing a measurement cycle. When the ‘‘FLT+’’ signal comes, all channels switch into hold mode. In the set of channels triggered by a good event time-to-voltage conversion is stopped by this signal, and charge integration has already been finished, so these channels have stored valuable information. The rest of the channels have just hold pedestals. If the ‘‘FLT+’’ signal has not come before the auto-reset time has expired, all three integrators are discharged, and the channel is returned to the initial state. There are 12 channels placed on 1 card. Each channel produces 3 output voltages, so 36 voltages are to be digitized in the card. They are connected one-by-one to the input of the ADC, and digital
data are sent to DAQ via serial link. Using an ADC with 1MSps sampling rate the total digitizing time will be about 36 ms, and the data transfer occurs at the same time. So, the deadtime necessary for processing DC’s signals related to one event is as short as about 40 ms. That is quite short enough to have the deadtime less than 1%. 4.2. Amplitude signals processing There are 5 sensitive devices which produce amplitude signals (see above). To provide the best performance of these devices, each of them will be equipped with its own option of a ChargeSensitive Preamplifier optimized for the particular device. After amplification all these signals become similar and are to be processed in the same way. Therefore, the modules processing amplitude signals will consist of the same units, but grouped in different sets. The particular set of units in one module corresponds to grouping of the channels for trigger. A functional diagram of the card which performs processing and digitization of amplitude signals is shown in Fig. 3. Each channel includes differential input stage, electronic gain control unit, two shapers and a peak detector. The shortshaped signals from a group of channels go via
Yu.V. Yudin, A.A. Ruban / Nuclear Instruments and Methods in Physics Research A 494 (2002) 560–564
563
Fig. 2. Simplified functional scheme of the signal processing module for DC.
Fig. 3. Functional scheme of the signal processing module for amplitude signals.
analog switches to an analog summer, and the summary signal is fed to the trigger. Gains of analog channels can be adjusted so as to normalize total conversion factors of all the calorimeter’s channels. Any combination of channels can be connected to summer, which allows to calibrate them individually using one common calibration signal. When the ‘‘FLT+’’ signal comes, the control unit generates the ‘‘Gate’’ signal for peak detectors. By the fall of the ‘‘Gate’’ signal (about 5 ms
after the event) the digitizing begins. The outputs of peak detectors are connected one-by-one to the input of the ADC, and digital data are sent to DAQ via serial link. The maximum number of analog channels placed on 1 card is 32 (the option for barrel calorimeter). Using the same digitization and data transfer procedure as for wires’ signals digitization (see above), the deadtime necessary for processing all amplitude signals belonging to one event can be as short as about 40 ms. That is quite short enough to have the deadtime less than 1%.
564
Yu.V. Yudin, A.A. Ruban / Nuclear Instruments and Methods in Physics Research A 494 (2002) 560–564
5. DAQ system During digitization of an event, raw data from SPMs are transferred via serial links to Data Receiver Cards (DRCs). A burst of up to 36 data words at 16 Mbits/s (according to 1 MSps sampling rate of the ADCs) is transferred via each link. The DRC has 16 inputs for serial links, so it receives up to (16 links * 36 words * 2 bytes)=1152 bytes of data per event and stores them in one of the memory pages (at this time another page is available for reading via PCI bus). The DRC is equipped with a simple ‘‘Memory-mapped Target Interface’’, which does not support the burstreading mode. That is reasonable because this mode is also not supported by desktop PC hardware. The number of links coming to one DRC is limited by the real throughput which can be performed by the PC. Without burst mode the real throughput of PCI bus is significantly less than its nominal bandwidth. Besides, the CPU power is shared between a few tasks, which also limits possible readout bandwidth. So, the real readout throughput performed by one PC is expected to be about 10 MB/s. On the other hand, the data stream produced by one SPM will be up to (36 words * 2 bytes) * 1 kHz Trigger rate= 72 kB/s after de-randomization. It was decided to have the number of SPMs readout by one PC not more than 64 (and, correspondingly, each of 4 DRCs installed in one PC has 16 inputs), so that the raw data stream is E4 MB/s, and after zero suppression it will be reduced to about 800 kB/s.
That data stream can be reliably handled by one PC. These data are sent via Ethernet network for further processing and recording. When all the old electronics are substituted by the new one, there will be 332 SPMs totally. To receive their signals, at least 22 DRCs are necessary, which will be placed in 6 PCs. The total data stream of 4 MB/s will be sent to the host computer via 100 Mbit Ethernet.
6. Conclusion A complete design of signal processing and DAQ electronics for the CMD-2M detector is proposed. Due to unification of functional units and optimal distribution of DAQ functions, the whole electronics is substantially cheaper than conventional solutions. Prototypes of the frontend electronics and of the DC’s SPM are tested now and perform as expected.
References [1] D.N. Grigoriev for the CMD-2 collaboration, CMD-2 detector upgrade, HEP: eConf 010430:T09, 2001. [2] E.V. Anaskin, et al., Nucl. Instr. and Meth. A 323 (1992) 178. [3] A.A. Grebenuk for the CMD-2 collaboration, Nucl. Instr. and Meth. A 379 (1996) 488. [4] D.N. Grigoriev, et al., IEEE Trans. Nucl. Sci. NS-42 (4) (1995) 505.
Electronics for the CMD-2M detector Yu.V. Yudin*, A.A. Ruban Budker Institute of Nuclear Physics, Acad Lavrentiev Prospect 11, Novosibirsk 630090, Russia
Abstract The CMD-2M detector will replace the CMD-2 detector. Many of the detector’s sensitive devices and the relevant electronics will be improved or re-built. Both the front-end, digitizing and data acquisition electronics will be modernized. The new electronics include several types of units which perform analog signal processing together with digitization. Unificated functional units are widely used in the designs. A careful selection of signal processing techniques and circuit design options is needed to make these electronics substantially cheaper than conventional electronics of similar detectors. r 2002 Elsevier Science B.V. All rights reserved. Keywords: Signal digitization; Digitizing electronics; Data acquisition electronics
1. Introduction An upgrade of the VEPP-2M collider and of two detectors installed on it is under way now in BINP. The projected parameters of the new collider (named VEPP-2000) are the following: cm energy up to 2 GeV and peak luminosity about 1032 cm 2 s 1, which is some 10 times the luminosity of the old collider. A new version of the CMD-2 detector, named CMD-2M, will be adequate for the increased collider capabilities. The general structure of the CMD-2M [1] remains the same, but almost all the existing systems of the detector will be re-designed and re-built, and a totally new liquid xenon calorimeter will be installed. The relevant front-end electronics for renewed sensitive devices and LXe calorimeter will be totally new. Taking into account all these principal changes it was decided to re-design the *Corresponding author. E-mail address: [email protected] (Yu.V. Yudin).
rest of the electronics too. It will be replaced with a new one step-by-step.
2. Basic issues for electronics design The sensitive devices of the CMD-2M detector, the data to be derived from their signals and the speed considerations are briefly described below. Drift Chamber (DC) [1] contains 1280 signal wires. The tracks’ co-ordinates in the R–Phi plane are measured by avalanche drift time, and the Z coordinate is measured by charge division techniques. Hence, to obtain tracking information three quantities must be measured for each wire: 2 charges from the ends of the wire and the time interval from signal arrival to a common stop signal. The ‘‘FLT+’’ signal from the First-level Trigger (FLT) is used both to initialize digitization and as the common stop for time measurement. It is aligned in time to the beam crossing moments. For charge measurement a resolution of 12 bits is
0168-9002/02/$ - see front matter r 2002 Elsevier Science B.V. All rights reserved. PII: S 0 1 6 8 - 9 0 0 2 ( 0 2 ) 0 1 5 4 9 - 8
Yu.V. Yudin, A.A. Ruban / Nuclear Instruments and Methods in Physics Research A 494 (2002) 560–564
required (14 bits is desirable). For time measurement a 2.5 ns step and a 10-bit full scale are required. Z-Chamber (ZC) [2] produces two types of signals. 64 signals of anode wires are the fastest and are used for triggering. 512 cathode stripes’ signals carry only amplitude information and are to be digitized at 12-bit resolution. LXe Calorimeter, Barrel Calorimeter and Endcap Calorimeter [3,4] altogether produce 2094 signals, which are used mainly for energy deposition measurements. These signals’ amplitudes are to be digitized at 12-bit resolution. Certain groups of these signals are fed to analog summers, and sum signals are used for triggering. The amplitude and arrival time of sum signals are digitized. Besides, the co-ordinate system of the LXe calorimeter produces 2112 signals for tracking. Their amplitude must be digitized at a resolution of 12 bits. DAQ system. The total number of signal sources in the detector is about 6800, and some 8000 analog quantities are derived from their signals. In a ‘‘good’’ event, usually less than 10% of signal sources operate and the relevant electronics channels contain non-zero data. Therefore, the portion of valuable data related to one event is about 1K words, each of which must include the channel’s address (not less than 13 bits) and datum (10–14 bits). With a reasonable word size of 32 bits the average event size is about 4 kB. The expected ultimate trigger rate will be about 1 kHz, and at this rate the deadtime must be less than 10%. Hence, the duration of digitization and transferring the raw data from ADCs to memory buffer (de-randomizer) must be less than 100 ms. At the same time, at a given trigger rate the average data stream (after de-randomization and zero suppression) will be up to 4 MB/s.
3. Main constraints of the design of the new electronics 1. Having some 8000 analog quantities to be digitized, signal multiplexing allows one to improve the reliability of the electronics and reduce its cost due to substantial reduction of
561
the number of modules and cable interconnections between them. 2. When choosing the option for design of the electronics of each detector’s system we took into account not only this system separately, but also tried to make some optimal unification of the electronics of all systems. Most of the signals in the detector carry just amplitude information. Such signals are usually converted into voltage levels for digitizing. Moreover, the time quantities can be (and often are) easily converted into voltage levels during digitizing. Therefore, for the given set of signals being present in CMD-2M the most effective way of digitization is the following. All analog quantities (amplitudes and times) are converted into voltage levels and digitized by identical ADCs. 3. Since the number of channels in each detector’s system is about one to two thousand, it is not yet cost effective to make a custom ASIC for signal processing. Nevertheless, due to high capabilities of commercially available ICs it is possible to dispose analog signal processing stages and digitizing units in the same modules. This makes the problem of different ‘‘grounds’’ easier to solve, and allows one to reduce the number of cable interconnections. That is why we preferred to place digitizing units (instead of few more analog channels) in new signal processing modules (SPMs). 4. The functions of the DAQ system are distributed as follows. The raw digital data produced by ADC in each SPM are transferred immediately (along with digitization being performed) via an individual serial link to the data receiver. The data receiver is a card installed in the PCIslot of a desktop PC. It accepts the data from 16 SPMs, de-randomizes the data and forwards them to PC’s processor. The PC performs reading data from data receiver cards, zero suppression, etc., and then sends the reduced data via Ethernet for subsequent processing and recording. A functional diagram of the new electronics showing mainly how signal processing and digitizing are carried out is presented in Fig. 1.
562
Yu.V. Yudin, A.A. Ruban / Nuclear Instruments and Methods in Physics Research A 494 (2002) 560–564
Fig. 1. Simplified functional scheme of the new electronics of the CMD-2M detector.
4. Electronics 4.1. DC signals processing A simplified functional diagram of the DC SPM is shown in Fig. 2. After amplification each signal is fed to a gated integrator (labelled ‘‘Q-U’’) and to analog summer. When the summary signal exceeds the threshold, the ‘‘Wire Trigger’’ generates signals for starting charge and time integrators, and also a hit signal for the FLT. So, since the discriminator has operated, the channel has been performing a measurement cycle. When the ‘‘FLT+’’ signal comes, all channels switch into hold mode. In the set of channels triggered by a good event time-to-voltage conversion is stopped by this signal, and charge integration has already been finished, so these channels have stored valuable information. The rest of the channels have just hold pedestals. If the ‘‘FLT+’’ signal has not come before the auto-reset time has expired, all three integrators are discharged, and the channel is returned to the initial state. There are 12 channels placed on 1 card. Each channel produces 3 output voltages, so 36 voltages are to be digitized in the card. They are connected one-by-one to the input of the ADC, and digital
data are sent to DAQ via serial link. Using an ADC with 1MSps sampling rate the total digitizing time will be about 36 ms, and the data transfer occurs at the same time. So, the deadtime necessary for processing DC’s signals related to one event is as short as about 40 ms. That is quite short enough to have the deadtime less than 1%. 4.2. Amplitude signals processing There are 5 sensitive devices which produce amplitude signals (see above). To provide the best performance of these devices, each of them will be equipped with its own option of a ChargeSensitive Preamplifier optimized for the particular device. After amplification all these signals become similar and are to be processed in the same way. Therefore, the modules processing amplitude signals will consist of the same units, but grouped in different sets. The particular set of units in one module corresponds to grouping of the channels for trigger. A functional diagram of the card which performs processing and digitization of amplitude signals is shown in Fig. 3. Each channel includes differential input stage, electronic gain control unit, two shapers and a peak detector. The shortshaped signals from a group of channels go via
Yu.V. Yudin, A.A. Ruban / Nuclear Instruments and Methods in Physics Research A 494 (2002) 560–564
563
Fig. 2. Simplified functional scheme of the signal processing module for DC.
Fig. 3. Functional scheme of the signal processing module for amplitude signals.
analog switches to an analog summer, and the summary signal is fed to the trigger. Gains of analog channels can be adjusted so as to normalize total conversion factors of all the calorimeter’s channels. Any combination of channels can be connected to summer, which allows to calibrate them individually using one common calibration signal. When the ‘‘FLT+’’ signal comes, the control unit generates the ‘‘Gate’’ signal for peak detectors. By the fall of the ‘‘Gate’’ signal (about 5 ms
after the event) the digitizing begins. The outputs of peak detectors are connected one-by-one to the input of the ADC, and digital data are sent to DAQ via serial link. The maximum number of analog channels placed on 1 card is 32 (the option for barrel calorimeter). Using the same digitization and data transfer procedure as for wires’ signals digitization (see above), the deadtime necessary for processing all amplitude signals belonging to one event can be as short as about 40 ms. That is quite short enough to have the deadtime less than 1%.
564
Yu.V. Yudin, A.A. Ruban / Nuclear Instruments and Methods in Physics Research A 494 (2002) 560–564
5. DAQ system During digitization of an event, raw data from SPMs are transferred via serial links to Data Receiver Cards (DRCs). A burst of up to 36 data words at 16 Mbits/s (according to 1 MSps sampling rate of the ADCs) is transferred via each link. The DRC has 16 inputs for serial links, so it receives up to (16 links * 36 words * 2 bytes)=1152 bytes of data per event and stores them in one of the memory pages (at this time another page is available for reading via PCI bus). The DRC is equipped with a simple ‘‘Memory-mapped Target Interface’’, which does not support the burstreading mode. That is reasonable because this mode is also not supported by desktop PC hardware. The number of links coming to one DRC is limited by the real throughput which can be performed by the PC. Without burst mode the real throughput of PCI bus is significantly less than its nominal bandwidth. Besides, the CPU power is shared between a few tasks, which also limits possible readout bandwidth. So, the real readout throughput performed by one PC is expected to be about 10 MB/s. On the other hand, the data stream produced by one SPM will be up to (36 words * 2 bytes) * 1 kHz Trigger rate= 72 kB/s after de-randomization. It was decided to have the number of SPMs readout by one PC not more than 64 (and, correspondingly, each of 4 DRCs installed in one PC has 16 inputs), so that the raw data stream is E4 MB/s, and after zero suppression it will be reduced to about 800 kB/s.
That data stream can be reliably handled by one PC. These data are sent via Ethernet network for further processing and recording. When all the old electronics are substituted by the new one, there will be 332 SPMs totally. To receive their signals, at least 22 DRCs are necessary, which will be placed in 6 PCs. The total data stream of 4 MB/s will be sent to the host computer via 100 Mbit Ethernet.
6. Conclusion A complete design of signal processing and DAQ electronics for the CMD-2M detector is proposed. Due to unification of functional units and optimal distribution of DAQ functions, the whole electronics is substantially cheaper than conventional solutions. Prototypes of the frontend electronics and of the DC’s SPM are tested now and perform as expected.
References [1] D.N. Grigoriev for the CMD-2 collaboration, CMD-2 detector upgrade, HEP: eConf 010430:T09, 2001. [2] E.V. Anaskin, et al., Nucl. Instr. and Meth. A 323 (1992) 178. [3] A.A. Grebenuk for the CMD-2 collaboration, Nucl. Instr. and Meth. A 379 (1996) 488. [4] D.N. Grigoriev, et al., IEEE Trans. Nucl. Sci. NS-42 (4) (1995) 505.