Hi TI,
I have the following project description and I wonder if you could help me select a suitable digital signal processor.
Description:
1. We have 100 hall effect sensors that can output 3-axis magnetic data. The sensor has a maximum SPI clock speed of 10 MHz.
2. Each axis magnetic data point is 18 bit.
Ideally: 100 sensors x 3 axes x 18 bits x 200 Hz sampling rate = 135 kbytes/s without considering muxing, initialization between SPI devices, etc.
3. We would like to display the magnetic field 2D map on a monitor with resolution of 800x800 pixels,
4. The chosen DSP should have a low cost evaluation board (< $150) that we can quickly prototype the system
5. It must also have very fast floating point and matrix calculation capabilities since I need to convert the magnetic 2D map to displacements (using Deep Neural Network)
5. The processor just needs to read the SPI data from the
Question:
1. Would you suggest FPGA for faster processing, or you think DSP is sufficient. If neither, what would be the best device for this application?
2. Besides the design requirement I mentioned above, another issue is that the the sensor can achieve the prescribed sampling rate (either 100 Hz, 200 Hz, or 1000 Hz) only in the continuous mode determined from the control register.
If we use one DSP for multiple sensors, we need to switch between SPI devices (100 of them), and I think this will significantly slow down the overall sampling rate. Is parallel processing from, say, FPGA the only solution?
I know you have a search tool for this type of question, but since I don't have much experiences with DSP, and this design might bottleneck most of the embedded processors out there (except FPGA), I really appreciate if you could provide me some guidance on the matter.
Thank you,
Khoi Ly
PhD student at University of Colorado Boulder
Intelligent Robotics Laboratory
