- Ask a related questionWhat is a related question?A related question is a question created from another question. When the related question is created, it will be automatically linked to the original question.
Hi Yu Zhu,
Check the below screenshot from AWR294x Technical Reference Manual (Rev. D).
You can use HWA_configRam function to configure the Advanced Statistics RAM. Here is an example of how to use HWA_configRam -
You can check hwa.h file for the HWA_RAM_TYPE.
Regards,
Samhitha
thank you for your reply,But I have other questions.
According to the datasheet, I have a 64*64 matrix that needs to search for local max, and the hwa settings are as follows:
LM_THRESH_MODE=3 means that the threshold is selected from the Advanced Statistics RAM in the column and row dimensions. According to the 2-Dimensional Maxima introduction, for each iteration, the maximum value observed and the corresponding sample index are stored in per-iteration RAMs. Similarly, for each sample index, the maximum value observed over all the iterations and the corresponding iteration index are stored in per-sample RAMs. I have a few questions:
1. Does per-iteration RAM correspond to MAXVAL_ARRAY_DIM1, and per-sample RAM corresponds to MAXVAL_ARRAY_DIM2?
2. Does MAXVAL_ARRAY_DIM2 store the maximum value of the entire column (64) or the local maximum value (3), as shown below:
3. Why does MAXVAL_ARRAY_DIM2 have a total of 9 bins, why not 3 bins, and which dimension do these 9 bins refer to?
Please explain the above questions,Thank you!!
Hi Yu Zhu,
Can you confirm if you want to compute 2-Dimensional Maxima or 2D local maxima?
Through 2D maxima, you will be able to compute per-iteration MAX value and Per-sample Max Value.
Local Maxima Engine is useful in a peak detection step where each sample/bin (Cell Under Test (CUT)) is compared against detection threshold(s), and also compared against the neighboring samples and the CUT is declared as a valid local peak if the sample amplitude exceeds the detection threshold and is “more than or equal to” the neighboring cells. The output of the local maxima computations is stored into the destination memory as a bit pattern, where each bit indicates whether the specific sample/CUT was detected as a valid local peak or not.
Does per-iteration RAM correspond to MAXVAL_ARRAY_DIM1, and per-sample RAM corresponds to MAXVAL_ARRAY_DIM2?
If you are computing 2D maxima, here is the screenshot describing the contents of Advanced Statistics memory output -
If the Maxima Arrays from Advanced Statistics RAM are selected for determining thresholds, then your understanding is correct. Advanced statistics should be used if there are different thresholds for different rows and columns of the 2D matrix.
Does MAXVAL_ARRAY_DIM2 store the maximum value of the entire column (64) or the local maximum value (3), as shown below:
If you are computing 2D maxima, MAXVAL_ARRAY_DIM2 has the maximum values corresponding to each sample index. If you are computing 2D local maxima, you use these arrays to configure the row/column thresholds. Refer to section 28.10.3 Local Maxima Engine – Output Write Pattern to understand the output pattern of the Local maxima engine.
Why does MAXVAL_ARRAY_DIM2 have a total of 9 bins, why not 3 bins, and which dimension do these 9 bins refer to?
Figure 28-53 is just an illustration and not a one-to-one mapping of any example.
I suggest you to refer to sections 28.9 Core Computational Unit – Statistics and 28.10 Core Computational Unit – Local Maxima Engine of the TRM for more details.
Regards,
Samhitha
