TIDUEZ9B July   2022  – April 2023

 

  1.   Description
  2.   Resources
  3.   Features
  4.   Applications
  5.   5
  6. 1System Description
    1. 1.1 Isolated Current Measurement
    2. 1.2 Band-Pass Filter
    3. 1.3 Analog-to-Digital Conversion
    4. 1.4 Arc Detection Algorithm
    5. 1.5 Key System Specifications
  7. 2System Overview
    1. 2.1 Block Diagram
    2. 2.2 Design Considerations
      1. 2.2.1 Current Transformer Circuit
      2. 2.2.2 Analog Band-Pass Filter
      3. 2.2.3 Analog-to-Digital Conversion
      4. 2.2.4 Power Supply
      5. 2.2.5 Debugging and Status Indication Options
    3. 2.3 Highlighted Products
      1. 2.3.1 TPS259474
      2. 2.3.2 TPS562202
      3. 2.3.3 TPS745
      4. 2.3.4 OPAx322
      5. 2.3.5 ADS8363
      6. 2.3.6 REF5025
      7. 2.3.7 TMDSCNCD280049C
  8. 3Hardware, Software, Testing Requirements, and Test Results
    1. 3.1 Hardware and Software Requirements
      1. 3.1.1 Hardware
      2. 3.1.2 Software
        1. 3.1.2.1 Arc Detection Theory
        2. 3.1.2.2 Software Implementation
    2. 3.2 Test Setup
      1. 3.2.1 ControlCARD Configuration
      2. 3.2.2 Setup for Hardware and Software Validation
      3. 3.2.3 Setup for Arc Testing
    3. 3.3 Test Results
      1. 3.3.1 Test Results of Hardware and Software Validation
      2. 3.3.2 Testing With Arcs
  9. 4Design and Documentation Support
    1. 4.1 Design Files
      1. 4.1.1 Schematics
      2. 4.1.2 BOM
    2. 4.2 Tools and Software
    3. 4.3 Support Resources
    4. 4.4 Trademarks
  10. 5About the Author
  11. 6Revision History

3.1.2.1 Arc Detection Theory

The arc detection algorithm implemented in this design is an FFT-based arc detection algorithm. Arcing present in a PV system creates random noise current in the cabling used for the PV string. The current noise of the arc itself has a Gaussian distribution with a spectrum extending to several MHz. Because of the geometry of the cabling in a typical PV system, the noise current density above 200 kHz varies significantly with frequency. For this reason, a general frequency band between 10 kHz and 100 kHz is selected for arc detection. This is defined by the analog band-pass on the board. Since there are other noise sources like the inverter switching frequency with in this band, the software allows for further restrictions of this frequency band, by adjusting the ArcTuningParams present in the software. A description of the different tuning parameters is found in Table 3-5. The algorithm performs an FFT of the sampled signal and sums up the noise in the specified frequency band. This computed arc noise can be observed in the variable AD_result. This is performed for every 1024 samples. Since arcing causes noise over the whole frequency band, the AD_result value increases when arcing is present.

Table 3-2 Elements of ArcTuningParams structure and Their Description
NAME DESCRIPTION
float32 ArcTuningParams.B Analysis Bandwidth: bandwidth of the analysis frequency.
float32 ArcTuningParams.I Min Frequency: initial frequency of the band.
float32 ArcTuningParams.F Filter Weight: weighting for each half of the band.
float32 ArcTuningParams.D Bin Discard Factor: number of spurious peaks to be discarded.
int16 ArcTuningParams.T Threshold for arcing: Not used by default
int16 ArcTuningParams.C Clipping Level: Not used by default
float32 ArcTuningParams.ADSampleRate Sampling Rate: set to 250000 by default
  • I - Min Frequency. The initial frequency of the band for arc-noise computation. The I value specifies the starting frequency of the band where the arc-noise is computed.
  • B - Analysis Bandwidth. This bandwidth must be a number greater than 1.0 for proper operation. The computation of arc-noise is carried out till the frequency B + I. For instance, if I is set to 30000, and B is set to 30000, then the arc-noise is computed using the frequency components in the range of 30 kHz to 60 kHz and anything outside this band is not used.

Additional parameters used in the computation of the arc-noise are D and F. These parameters are used to eliminate spurious peaks that may occur that one may sometimes perceive as arc:

  • D - Bin Discard Factor. This parameter specifies the number of spurious peaks to be discarded, it controls how aggressive the filtering is and must be a number be between [0.0, 1.0] for proper operation. Peaks within the band of frequencies set by B and I are eliminated for arc-noise computation based on the value set by D stating with the highest peaks. If D is higher more peaks are discarded, and if D is lower fewer peaks are discarded for computation. Note that changing D has an impact on the cycles consumed by the arc-noise computation algorithm. One needs to try different D values probably in the increment of 0.1 and see for real-world arcs how the arc-noise computation varies.
  • F - Filter Weight. This parameter specifies the weighting for the first half of the analyzed frequency band and must be a number greater than 0.0 for proper operation. The first half of frequencies set by B and I are weighted by the filter weight while the second half of the band is always weighted by 1.
  • T - Threshold. This parameter is currently not used and is a placeholder for a threshold value for AD_result where arcing is detected if this threshold is surpassed. Changing T has no effect.
  • C - Clipping Level. This parameter allows a variable for detecting arc, but is not implemented currently. Changing C has no effect.
  • ADSampleRate - Sampling Rate. This is the sampling rate of the system. In this design this rate is fixed to 250 kSPS and therefore this is set to 250000.