As you given that the Confidence level range is from 0 to 99 & its consider better if it’s close to 99. But Despite extensive testing indicating good signal strength, why do my recorded values consistently stays up to 11 & not varying? Also tell me in which factor the confidence value depends?
I wouldn’t use the confidence level for anything, it’s never proven to have any real impact on quality. The idea was to simply put a single value on the AoA graph SNR (a higher peak on the DoA graph should mean better confidence), but in practice it’s never really been useful.
That’s why in the KrakenSDR we depreciate the use of the confidence value, and in the app instead use the full 360 degrees of values on an activation grid. The grid naturally filters poor quality results.
The confidence value is just a legacy value kept from the KerberosSDR days.
Understood. May I ask how do you filter the poor quality results?
With the grid system they’re not filtered out.
What happens is that with poor quality results (multiple peaks/no peaks / low DoA SNR) the grid cells get activated less, and the activation energy is more spread out amongst the grid cells.
Compared with a good result (single peak, high DoA SNR), which highly activates cells only in the direction of the peak.
So that provides a natural sort of filtering.
Just to clarify our understanding: If I take a CSV capture with 10 rows and sum the values for each angle column (1-360), then create a circular heatmap where the intensity of color represents the total sum per angle, with greater sum values appearing more intense than smaller ones, is this akin to the filtering concept you mentioned? My assumption is that by summing, we can emphasize signals in specific angles and, conversely, filter out entries in angles where a signal may not truly exist due to poor estimation. I may be mistaken, and I would greatly appreciate your guidance, possibly with an example, on how to effectively filter out undesirable data in the 360-degree grid, as you mentioned
Yes summing works, assuming that you are in a fixed position. Of course if you are moving and taking data from different GPS positions, then simply summing rows like that will not work.
1 Like
Thanks, yes the position of RX is fixed. So is this a procedure that you are usually use to filter out bad estimations?
Yes if you are at a fixed position it essentially reduces to simple averaging over time. You can also look at adding moving average, or decaying older values if you have moving targets.