Honored Contributor II
06-27-2018 11:20 AM
Hello!I'm looking for some advice with regards to the digital platform I will use for a graduate school project. I have also posted this question on the Digilent forums, so my apologies if you frequent both. brief project overview The aim of this project is to measure the dynamic response of a structure under artificial and/or operational stimulus, and semi-actively tune a dynamic vibration absorber to minimize said response. My chosen methodology requires computing the FFT of multiple sensor signals, performing a simple search on the processed data, and then executing a decision making algorithm to select which frequency the dynamic vibration absorber is tuned to. The worst-case scenario may require computing the real FFT of 4-6 signals over a running window of 1-2 seconds every 5-10ms, with a 24 bit ADC sampling at 10KHz. This is a bit of a contrived scenario, however useful for establishing an upper limit on the performance requirements. The simplified diagram below illustrates roughly what the process will look like. It would be nice to have the response time from electromechanical system output, A, to input B less than 10ms. my background Degree(s) in mechatronics, some experience with the MSP430, and some experience with the DE1 SoC development board for course work. A few courses in digital design and computer science using C++. chosen hardware… This is where you come in. So far I'm leaning towards an FPGA + ARM development board with the addition of an ADC development board so I can get right to development. I'm leaning towards an FPGA, because I want to learn.... these options look attractive:
- de10-stanard (http://www.terasic.com.tw/cgi-bin/page/archive.pl?language=english&no=1081) ( ARM + FPGA )
- altera arrow sockit (http://www.terasic.com.tw/cgi-bin/page/archive.pl?language=english&no=816) ( ARM + FPGA )
- zedboard (http://zedboard.org/product/zedboard) ( ARM + FPGA )
- nxp mimxrt1050 (https://www.mouser.ca/new/nxp-semiconductors/nxp-mimxrt1050-evk/) ( MCU with some hardware acceleration )
- omap-l138 development kit (http://www.ti.com/tool/tmdslcdk138) ( ARM + DSP )
- ad7768 (http://www.analog.com/en/design-center/evaluation-hardware-and-software/evaluation-boards-kits/eval-...) (beast)
- ebay adc module (https://www.ebay.com/itm/ads1256-module-24-bit-8-channel-adc-data-acquisition-board-and-high-precisi...) ( may struggle to meet throughput )
- raspberry pi ad/da module (https://www.seeedstudio.com/raspberry-pi-high-precision-ad%2fda-board-p-2765.html) ( may struggle to meet data transfer needs )
- The embedded community seems to be quite saturated with regards to Zedboard, and Xilinx support material and knowledge. Correct me if I'm wrong? But since most of my learning will be from googling, forums, and some digital design books, this seems the more direct route. I have no FPGA experts around me at the moment to ask questions.
- The Zedboard can connect right up to the AD7768 development board via the FMC connector and I've seen a few forum posts on integrating the two boards.
- I cannot find an equivalently powerful ADC development board for the DE10-Standard that would be easily implementable out of the box. There are a few HSMC adapter boards that might allow connecting to the AD7768 board, however I'm not ready to troubleshoot that setup, while also learning a new platform.
- No dedicated RAM for the FPGA other than the 4.9Mb of block ram. If I'm buffering 4 signals at 10,000 samples/s for a capture length of 2 seconds, that's cutting it really close for buffering. Especially if I want to store other processed data, and run a softcore MCU. Actually, it won't all fit worst-case scenario. Any thoughts on this? This is why the DE10 seemed so attractive with it's 64MB of SDRAM.
- I would like to use the DE10 because the FPGA has direct access to dedicated external RAM, however I cannot find a high precision ADC dev board with the sampling rate I need that will plug right in. If I had unlimited time, yeah I would go with the DE10-Standard board and stick it out, but there is also a bit of a time crunch (4-6 months). So getting started ASAP is also desirable.