Computationally efficient arc detector with coherent sampling
Switching interference is a primary artifact which affects the accuracy of arc detectors. To address switching interference, conventional arc detectors employ computationally intensive techniques which are often designed specifically for a target application. Thus, conventional arc detectors require a significant amount of hardware to accurately detect arc faults, which can increase costs of the power systems and prohibit wide deployment of arc detectors. With improved signal processing, a unique method for arc detection can accurately detect arc faults efficiently while tolerate switching interference from an inverter of the power system. Specifically, the method provides accurate but efficient arc detection by using a small Fast Fourier Transform with coherent sampling that is accomplished with a common clock generator in combination with signal conditioning. The overall system implementing the method is also programmable to suit a variety of target applications.
1. A method for arc detection with improved distinction between an arcing and a non-arcing spectrum, the method comprising:
receiving, by an arc detector, switching frequency information of a pulse width modulation controller of an inverter of a power system;
sampling, by an analog-to-digital converter, a first signal on a first power line to generate a plurality of first digital samples at a sampling frequency determined based on said switching frequency information for coherent sampling;
transforming, using a digital signal processor, a first window of the first digital samples to a first frequency-domain representation; and
classifying, by the arc detector, whether an arc fault is present in the power system based on the first frequency-domain representation.
2. The method of claim 1 , wherein:
the analog-to-digital converter and a part of the pulse width modulation controller for generating a pulse width modulation waveform having the switching frequency information are clocked by a same clock signal for coherent sampling.
3. The method of claim 1 , wherein:
the analog-to-digital converter and a part of the pulse width modulation controller for generating a pulse width modulation waveform having the switching frequency information are clocked by clock signals generated from a same root clock signal for coherent sampling.
4. The method of claim 1 , wherein:
a ratio of the switching frequency information and the sampling frequency is equal to a ratio of a number of cycles of a switching waveform of the pulse width modulation controller in the first window and a number of samples in the first window of the first digital samples for coherent sampling.
5. The method of claim 4 , wherein:
the number of cycles of the switching waveform is a power of two; and/or the number of samples is a power of two.
6. The method of claim 1 , wherein the first window of the first digital samples is filtered by a comb filter prior to transforming using the digital signal processor, wherein the comb filter attenuates spectral energy corresponding to the switching frequency information.
7. The method of claim 1 , wherein the first window of the first digital samples is processed by a windowing function prior to transforming using the digital signal processor.
8. The method of claim 1 , further comprising:
performing a moving average of the first frequency-domain representation before classifying whether the arc fault is present in the power system.
9. The method of claim 8 , wherein a window used for the moving average is programmable.
10. The method of claim 1 , further comprising:
averaging values for a particular bin over a plurality of frequency-domain representations to generate a time-domain average of a power spectrum of the power system; and
classifying, by the arc detector, whether the arc fault is present based further on the time-domain average.
11. The method of claim 1 , further comprising:
classifying, by the arc detector, whether the arc fault is present in the power system based further on a second frequency-domain representation generated based on a second window of second digital samples of a second signal of a second power line in the same power system as the first power line.
12. The method of claim 1 , wherein classifying, by the arc detector, whether the arc fault is present in the power system based on the first frequency-domain representation comprises: determining, on a bin by bin basis, whether a power of the first frequency-domain representation is above a programmable threshold.
13. An apparatus with arc detection with improved distinction between an arcing and a non-arcing spectrum, the apparatus comprising:
an arc detector for receiving switching frequency information of a pulse width modulation controller of an inverter of a power system;
an analog-to-digital converter for sampling a first signal on a first power line to generate a plurality of first digital samples at a sampling frequency coherent with a switching frequency of a pulse width modulation controller of an inverter of a power system;
a common clock generator generating clock signals for coherent sampling for both the analog-to-digital converter and a part of the pulse width modulation controller for generating a pulse width modulation waveform having the switching frequency;
a digital signal processor for transforming a first window of the first digital samples to a first frequency-domain representation; and
a classifier part for classifying whether an arc fault is present in the power system based on the first frequency-domain representation.
14. The apparatus of claim 13 , wherein:
the sampling frequency is configurable based on the switching frequency for coherent sampling.
15. The apparatus of claim 13 , further comprising:
a programmable comb filter for filtering the first window of first digital samples prior to transforming using the digital signal processor, wherein the programmable comb filter attenuates spectral energy corresponding to the switching frequency.
16. The apparatus of claim 13 , further comprising:
a windowing function for processing the first window of first digital samples is processed prior to transforming using the digital signal processor.
17. The apparatus of claim 13 , wherein the classifier part determines whether each value in a plurality of selected bins in the first frequency-domain representation are above a respective threshold.
18. The apparatus of claim 17 , further comprising:
registers for storing programmable thresholds corresponding to the plurality of selected bins.
19. An apparatus for arc detection with improved distinction between an arcing and a non-arcing spectrum, the apparatus comprising:
means for receiving switching frequency information of a pulse width modulation controller of an inverter of a power system;
means for converting a first signal on a first power line to generate a plurality of first digital samples at a sampling frequency coherent with a switching frequency of a pulse width modulation controller of an inverter of a power system;
means for generating clock signals using a common crystal for coherent sampling for both the analog-to-digital converter and a part of the pulse width modulation controller for generating a pulse width modulation waveform having the switching frequency;
means for applying a frequency-domain representation transform to a first window of the first digital samples to a first frequency-domain representation; and
means for detecting whether an arc fault is present in the power system based on the first frequency-domain representation.
20. The method of claim 1 , wherein the digital signal processor used for transforming a first window of the first digital samples to a first frequency-domain representation is a Fast Fourier Transform module.