Improve _normalizeAngle speed.#530
Draft
jaguilar wants to merge 1 commit intosimplefoc:devfrom
Draft
Conversation
Previously _normalizeAngle used fmodf, which is expensive on Cortex-M series chips (40-100 cycles). For normalizing the angle, we can use a trick of just subtracking 2 pi times the number of full rotations from the angle. This method does lose precision much faster than the fmodf approach. There will be significant error if the total number of radians passed in is greater than 10^5. However, we can counteract this by not passing in such a large number of radians -- basically making sure that we are not using an accumulating radian count in any location where the angle will eventually need to be normalized. This should typically be the case since accumulating radian use cases are definitionally not in need of normalization. We should consider adding an assert here to ensure that we catch any existing cases where we should be passing the mechanical or electrical angle and are instead passing an accumulator. Performance test (on STM32G474, with certain other optimizations): Before: foc_nanos:13737 After: foc_nanos:12550 (approx 9% reduction in loopFOC time)
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Description
Previously _normalizeAngle used fmodf, which is expensive on Cortex-M series chips (40-100 cycles). For normalizing the angle, we can use a trick of just subtracking 2 pi times the number of full rotations from the angle.
This method does lose precision much faster than the fmodf approach. There will be significant error if the total number of radians passed in is greater than 10^5. However, we can counteract this by not passing in such a large number of radians -- basically making sure that we are not using an accumulating radian count in any location where the angle will eventually need to be normalized. This should typically be the case since accumulating radian use cases are definitionally not in need of normalization. We should consider adding an assert here to ensure that we catch any existing cases where we should be passing the mechanical or electrical angle and are instead passing an accumulator.
Performance test (on STM32G474, with certain other optimizations):
Before: foc_nanos:13737
After: foc_nanos:12550 (approx 9% reduction in loopFOC time)
Type of change
How Has This Been Tested?
I built an arduino-foc-based test program based on this commit and observed that the motor still worked fine. I also benchmarked the code and observed it to be faster. I have not tested all possible control modes so it is not impossible that this approach is unacceptable in some configurations. Please review carefully and suggest anything further that I should check.
Test Configuration/Setup: