Download as pdf
Download as pdf
You are on page 1of 3
Digital Peak Detector with Noise Threshold Valentin T. Jordanov, Member, IE Absiract- A novel real time digital peak detection technique uses noise threshold to eliminate noise sensitivity and to provide high throughput. The peak detector at any moment can be in one of only two operating modes — tracking maximum and tracking ‘minimum, The peak detector instantaneously detects and captures the peak values ofthe digitized signal when the operating ‘mode changes. The maximums are detected atthe transition from tracking maximum 0 tracking minimum. ‘The minimums at detected at the reverse transition. The transitions between the ‘operating modes exhibit a hysterezis that is determined by a digitally set noise threshold, 1 IxtRopucTON EAL time digital spectrometers uilize peak detectors to «defect and extract the peaks from continuous discrete signals, Typically these discrete signals are generated by time- invariant digital signal processors working at high sampling rates [1,2.3]. A distinctive feature of the real time digital peak defectors is their ability to almost instantaneously detect and capture peak values. These peak detectors are self triggering and may not require an external strobe or reset signal [3.4.5] ‘The noise is a major obstacle that limits the performance of the digital peak detectors especially in the case of flat topped pulse shapes. The random noise, electromagnetic pick-up, ‘microphonics, ADC and analog circuit signal distortions cause flat top amplitude variations that are the same as the variations of the base line. Ifthe peak detector is sensitive to noise than a ‘multiple peaks can be detected for a single detector event. In order to achieve high throughput rates, the real time digital peak detectors must provide a fast local peak detection, That is, the peaks must be detected every time the slope of the discrete pulse signal changes. It is obvious that the highest throughput and the complete noise immunity can not be achieved at the same time. Therefore, the objective of this work is 10 develop an optimal high throughput digital peak defector that is immune to the peaks caused by the noise 0. To illustrate the real time digital peak detector operation we vill use a simple example of discrete pulse signal shown in Fig, LL There are two detector events that result in two partially REAL TIME DIGITAL. PEAK DETECTION Patt Pending Valentin T.Sordanov is with YANTRA, Durham, NH 3824, USA (ema: Jendanow ee or. Dave L. Hall and Mat. Kastner are with Canberra Indutes, Meriden, CT. (4650, USA (0-7803-7636-6/03/$17.00 ©2003 IEEE. 140 _Authoraed leensed use nite to: nda nate of Tecnology Hyderabad, Down ‘FE, Dave L. Hall, Mat. Kastner ‘overlapping pulses. Note that the flat top of each pulse (second trace) is fiee of pile-up but the valley between the two pulses does not recover to the base line. The sampling frequency in this ease is much higher than the width of the shaped pulses so the closely spaced discrete points form (visually) a continuous Tine, XQ ee ee Fig. 1. Parialy overlapping pulses with peak values tee of pile-up, ‘A simple approach to eliminate the noise sensitivity of the peak detector is to use a low-level discriminator with a threshold set above the base line noise level. In this ease the absolute maximum is detected for the portion of the signal that is above the threshold. Fig. 2a shows a block diagram of a real- time digital peak detector based on a low-level discriminator. This is a simple configuration that uses two comparators (CMP1 and CMP2) one edge register with enable (PREG) and ‘output latch (MAXL). The operation of the peak detector is illustrated with the \waveforms in Fig, 2b, When the discrete pulse signal is below the threshold PREG output is zero. Once the threshold becomes hhigh PREG output can change. When the discrete pulse signal value (input signal) is larger than PREG then PREG is updated with the input signal, MAXL is enabled during peak detection and tracks the peak value. When the input signal becomes smaller than the threshold the MAXL latches the maximum value and the rest ofthe circuit resets to its initial state. It is clear that this type of peak detector limits the throughput due to inability to resolve partially overlapping pulses. A digital peak detector that is capable to resolve joded on June 162022 at 069402 UTC tom IEE Xplore. Resbicons apply partially overlapping pulses is based on zero-crossing (ee lo ol>_ofMax Discrete - — omps =_ “ Poak Detect signal is compared against the delayed signal by the comparator (CMP). CMP functions as differentiator and zeto- crossing detector. The transition of the CMP output are used to defect and capture maximum and minimum values. At the low- to high transition a minimum is detected and captured. The ‘maximum is detected at the other transition of the CMP output. The digital zero-crossing detectors are very fast but they hhave a major drawback — high sensitivity to noise. False peaks are detected even when the digitized signal changes with one least significant bit (LSB). Thete are modifications of the digital zero-crossing peak detectors that use either timing conditions or sign bit filtering techniques to improve the noise immunity [3,5]. Although these methods provide improved performance the optimal setup is difficult. In order to optimize the throughput performance and to climinate the noise sensitivity of the peak detector a novel peak detector configuration was developed. » Fig. 2 Digital peak dtectr seo lowsleveldisritinater 3 block gram, b waveforms The zero-crossing peak detection is a well established method 10 detect local peaks in both analog and digital domains. The digital peak detectors based on this method utilize the zero-crossing of the first numerical derivative (finite difference) of the discrete pulse signal. In general, there are three operating modes ~ wacking maximum (first finite difference > 0), tacking minimum (first finite difference < 0) and tracking constant (first finite difference = 0). Practically, the third mode is combined with one of the other two modes reducing the operation to only two modes ~ tacking maximum and tracking minimum. ‘The maximum values are detected at the transition from tacking maximum to tracking minimum. The captured ‘maximum values represent the maximum of the discrete signal in the tracking maximum interval. The minimum values are similarly detected at the transition from tracking minimum to tacking maximum and they represent the minimum of the discrete signal in the tracking minimum interval Fig3 shows a block diagram of the zero-crossing peak detector and illustrates the operation, The register (DFF) delays the input signal by one sampling period. The input (0-7803-7636-6/03/$17.00 ©2003 IEEE. 41 _Authoraed leensed use nite to: nda nate of Tecnology Hyderabad, Down peo op My Discrete Puise Sia —e = wax L po} MX, ook Detect, a) aan sa » Fig. Zero crossing peak detection) block diagram, b) waveforms, ‘The digital peak detector with noise threshold combines the noise immunity of the low level discriminator peak detector with the speed and peak resolving ability of the zero-crossing technique. The block diagram of the digital peak detector with noise threshold is depicted in Fig. 4a and the operation is illustrated in Fig. 4b. Similarly to the low-level discriminator method a PREG register is used to track and capture the ‘maximun/minimum values. The peak detection is controlled DyGItAl PEAK DETECTOR WITH NOISE THRESHOLD joded on June 162022 at 069402 UTC tom IEE Xplore. Resbicons apply by a comparator CMP with registered output (DFF). The noise threshold is applied directly t0 one of the inputs of a ‘multiplexer MUX. The threshold is negated by the negating function (NEG) and is applied to the other input of MUX. ‘Thus, the ourput of MUX alternates between positive and negative threshold values depending of the peak detector mode of operation, ‘The peak detector at any moment can be in one of only two operating modes ~ tracking maximum and tracking minimum, ‘The MIN and MAX value are detected at the mode change transitions. The operating mode is determined by the output of DFF (peak detect signal) ~ when it is HIGH the peak detector tracks maximum and when LOW the peak detector tracks ‘minimum, The maximums are detected at the transition from tracking maximum to tacking minimum, The minimums are defected atthe reverse transition, = a a ar b) Fig. Peak detector with oie tresbold: a) block diagram, b) waveforms. (0-7803-7636-6/03/$17.00 ©2003 IEEE. a ‘The noise threshold function utilizes a subtractor (SUB) and CMP. The PREG value is subtracted from the discrete pulse signal and the resulted difference(A) is compared against the threshold (B). (B) is positive when the peak detector tracks ‘minimum and negative when the detector tracks maximum, The sign bit of the (A) is used as zero-crossing indicator and acts as ‘enable signal of PREG. The XOR gate allows proper tracking of minimum and maximum values. As the waveform indicate the threshold adds a hysterezis that eliminates false peak detection caused by the noise. Note that the noise level at (A) is the same as the noise level of the discrete pulse signal IV. CONCLUSION We have developed a real time digital peak detector that exhibits high noise immunity. The peak detection method ‘optimizes the throughput at given noise level. The high throughput is due to ability to resolve partially overlapped pulses. The peak detector is self'triggered and does not require ‘any external gaiting signal. The peak detector has been implemented as a part of digital pulse processor based on field programmable gate arrays V. REFERENCES [1] V7 Jowanov and GF. Knol, "Digital Synthesis of Pulse Shapes in Real Time for High Resolution Radiation Spectoscopy”, Nnc- Jasin: sand Meth A345, pp. 337-348, 1994 and eerences thesia [2] VT. Jodaov, G:F. Knoll, A.C. Huber and JA. Pants, Di Techies for Real Time Pulse Shaping in Radiation Measurements Nuc Incr ond Met, A3S3, pp. 261-268, 1994 [8] VT. Jordanoy, "Some Digital Technignes for Real Tine Processing of Pulses rom Radtion Detectors" PhD. dssratio, The University of Michigan, An Arbor, Mach 1994 Is] _V-Jordanov and G:F. Knol, "Digital Pulse Processor Using A Moving Average Techniqic", IEEE Trans Nnc! Sc, vol 40, no 4, 9p. 764769, Aug 1998, [5] F. Hilserath et al,“ A single chip pulse processor for nuclear spoctscopy”, IEEE Trans Nl Se, vol 32, pp. 148-189, Feb 1988, _Authoraedleeneed use Hite to: nda nate of Tecnology Hyderabad, Download on June 16,2022 a 06:94:02 UTC ta IEEE Xplore. Restone opp

You might also like