DSP on Beaglebone

Question

I have Beaglebone running Ubuntu. We want to constantly choose from 3 built-in ATD converters with a speed of 100KS / s, and in each sample window we will run the DSP cross-correlation algorithm. When we find the correlation value above the threshold value, we will send this value to the PC.

I'm worried about the planning process in Ubuntu. If our process is replaced and the ATD sample becomes available during this time, the process will skip the sample. We must ensure that our process captures every sample and stores it in memory.

With that said, there is a way to initiate interrupts on Beaglebone so that if the ATD sample is ready, the sample will be stored in our program’s memory, even if the program does not have a processor at that time

Thanks!

Answer 1

You might be able to run EDMA or use PRUSS. It's probably best to ask for [e-mail]. There are no DSPs on BeagleBone.

Answer 2

This is not exactly the answer to your question, but hopefully it explains how this process works. Since you did not specify what equipment you use to convert AD, this is probably the best thing you can do:

With audio equipment that faces the same problem, the solution comes from the hardware and drivers working together: whenever the equipment fills a sufficient amount of buffer, it signals the driver (through an interrupt or some similar mechanism). In some cases, it is also possible that the driver tried out hardware or something like that, but this is a less efficient solution, and I'm not sure if anyone does this more (maybe on cheaper hardware?). From there, the driver process can invoke the right in the end-user process, or it can simply mark the corresponding end-user process as “runnable”. In any case, control must be transferred to the end-user process.

For this to happen, the end-user process must work with a higher priority than all the others that occupy the central processors at this moment. To guarantee that your process will always be the first in the queue, you can run it with high priority, with the appropriate permissions, you can even run very high priorities.

The time it takes for a top-priority process to go from runnable to running is sometimes called OS latency, although I'm sure there is a more specific technical term. Linux latency is about 1 ms, but since it is not a "hard" real-time operating system, it is not a guarantee. If it is too long to process your chunks of data, you may have to mask some of them in your driver.