Here’s the page we think you wanted. See search results instead:


Contact an Expert

Signal processing can reduce data bandwidth through peak detection

In some applications, information about peak signal levels is sufficient. A signal-processing technique called peak detection reduces all acquired data to the times and amplitudes of the peaks (or troughs) in a signal. Capturing and reporting only this information significantly reduces the amount of data that must transferred across the system bus.

Detecting the peaks
Peak analysis is performed on the fly with firmware that can use multiple field-programmable gate arrays (FPGAs). FPGAs can be used to determine vital peak information—in both amplitude and time—by running an advanced signal-processing algorithm called peak fitting directly on the digitized waveform data. This method produces excellent precision and, because it is performed on the fly, also provides a dramatic improvement in measurement rates.

Understanding the algorithm
Nine samples are used to determine if a sample corresponds to a peak: This includes the sample itself and four samples on either side (i.e., four earlier and four later in time). Through digital signal processing, the sample will be deemed a peak if, within the nine analyzed samples, the following conditions occur:

  • A rising edge is found before the sample that exceeds the programmable delta rise (ΔRise).
  • A falling edge is found after the sample that exceeds the programmable delta fall (ΔFall).
  • The sample is the maximum of all points falling between the two defined rising and falling edges. Additional checks confirm that only one peak is found within the interval between the rising and falling edges.

Using this type of signal-processing algorithm, peaks can be found on the fly in signals with frequencies up to half the sampling rate.

Utilizing peak interpolation
To improve the measurement, peak time and amplitude are determined using an interpolation routine that uses digital signal processing to fit a 12-bit quadratic spline to the three points around a verified peak. This resulting measurements offer improved resolution on the time and amplitude axes. To meet specific application requirements, software commands can be used to balance the increased resolution (in time and amplitude) against the heavier signal-processing burden, increased memory usage and reduced measurement throughput.