Hi,
I have drawn CIR data from DW1000 (ACC_MEM) and tried to understand the first path. But still have some problems, as follows:
The index of the first path is 16-bit data, the unit is ns, and the lower 6 bits represent the fractional part. Can I understand that the sampling precision is 1/64 ns, that is, 64 Ghz, and then the accuracy of the DW1000 is 3*10^ 8m/s * 1/64ns=0.46875cm?
Is the index sampling speed of CIR 1 ns? When the indexes of the two peaks differ by one, can it be understood that the two paths differ by 30 cm (1 ns * 3 * 10 ^ 8 m / s)? But why is the index of the first path in the CIR data measured at 8m larger than that at 10m? (FP_INDEX read at 8m is 750.7500, FP_INDEX at 10m is 747.078. It is represented by the red line in the figure)
If it is my understanding above, should it be within 30cm, FP_INDEX should not move left and right? But FP_INDEX will, and within the same distance, the range of movement can be ±5. So how should FP_INDEX be understood?
Close. The precision is 4.69 mm, the accuracy is more like 5 cm. If you measure the exact same distance multiple times you will get some variation in the reading.
The CIR index is 1.001602564ns (1/(499.2 MHz*2)) so yes about 30 cm per index.
However CIR index has nothing to do with travel time of the signal. That would only be possible if the system knew the exact moment that the signal was transmitted. No matter the range the CIR index for the first path will be around 750 unless you have a very weak direct path and a strong reflected path.
When a pre-amble is detected the chip will start accumulating data assuming the set PRF and SFD. It will set the zero point in this accumulator such that it expects the peak to be around index 750. However it’s setting that based on a fairly weak signal before it can accumulate anything, if the reflection is a lot stronger than the direct path it may get it wrong. After the accumulator has finished the first path time can be far more accurately detected based on the accumulated results and so FP_INDEX can be found.
For normal ranging you don’t need to worry about any of this, the packet RX timestamp contains everything you need to know, it’s effectively FP_Index after adjusting for the zero point of the accumulator. FP_INDEX and the CIR memory are only really of use if you want to better understand the noise and reflections within your environment.
Thank you for your reply, let me understand the meaning of FP_INDEX.
Recently, I have conducted many experiments to show that in the case of LOS, that is, the maximum amplitude point in CIR is the first path, the measured result is more accurate. But in the case of most NLOS (Figure 3), it is clear that the first path (the red vertical line in the figure) can be identified by the LDE algorithm, but the measured result is still passed through the maximum amplitude point (Figure 7 Black squares) to calculate, why is the LDE algorithm not calibrating RX_STAMP?
The content described in the AP006 Part3 manual only tells the user how to use the metrics provided by the DW1000 , and gives the decision to the user. It can also be considered that the LDE algorithm does not feed back RX_STAMP.
For a simple example: when NLOS occurs, the first path is obtained according to the LDE algorithm, and prNlos is calculated with reference to Fig. 8, and it is considered that prNlos is calculated as 1, then the Confidence Level CL obtained according to Fig. 4 is 0, that is to say the measurement result is not credible. Then, if the LDE algorithm feeds back the correct first path (which is actually recognized) to RX_STAMP, the result will be correct, but the Confidence Level CL is 0, which is contradictory. Conversely, if the LDE algorithm does not feed back the results, then the result at this time is indeed wrong, and the Confidence Level CL is indeed 0, which makes sense.
If the above statement is ok, according to Figure 3, where (FP_INDEX=745.2188, PP_INDEX=749), the result of the ranging is based on the untrustworthy result of the error obtained by PP_INDEX (as can be seen from CL=0.19913), then how Corrected?
For example, calculate PP_INDEX-FP_INDEX=3.7812, and then subtract 3.7812ns*C=113.436cm from the distance according to question 2 (1ns=30cm). This is obviously impossible.
Hello!
My question is the folliowing: if LDE can recognize FP (cause there is the value register FP_INDEX, so for sure RX timestamp is recognized) why it can’t use that value to compute ranges?
Actually, I know that this way is uncorrect cause I have done lots of measurements and I have found RS = PP_IND-FP_IND with very low values even if ranging error was about 7m.
I can’t get the solution of the problem: in some cases RS = PP_IND-FP_IND has very high values and ranging error is 0.1m, in another ones RS has low values and ranging error is 7m.
Thanks,
Alessio
Actually, my basic doubt is the following: in NLOS scenario with important range errors, when I extract CIR, the reported first path is the “real” first path (the one that I would have in LOS case) or the one that is considered wrongly as first path by the chip and that leads to ranging error?
If the second option is true (it seems the most logical one), how can be considered realiable all the metrics described in APS006 part3? Why the wrong detected first path can’t be also peak path? Theorically, wrong first path should be the path which has suffered less enviromental reflections and so the one with the biggest amplitude.
I really can’t get an explanation.
Thanks,
Alessio
[color=#333333]Thank you for your reply, let me understand the meaning of FP_INDEX.[/color]
[color=#333333]Recently, I have conducted many experiments to show that in the case of LOS, that is, the maximum amplitude point in CIR is the first path, the measured result is more accurate. But in the case of most NLOS (Figure 3), it is clear that the first path (the red vertical line in the figure) can be identified by the LDE algorithm, but the measured result is still passed through the maximum amplitude point (Figure 7 Black squares) to calculate, why is the LDE algorithm not calibrating RX_STAMP?[/color]
[color=#333333]The content described in the AP006 Part3 manual only tells the user how to use the metrics provided by the DW1000 , and gives the decision to the user. It can also be considered that the LDE algorithm does not feed back RX_STAMP.[/color]
[color=#333333]For a simple example: when NLOS occurs, the first path is obtained according to the LDE algorithm, and prNlos is calculated with reference to Fig. 8, and it is considered that prNlos is calculated as 1, then the Confidence Level CL obtained according to Fig. 4 is 0, that is to say the measurement result is not credible. Then, if the LDE algorithm feeds back the correct first path (which is actually recognized) to RX_STAMP, the result will be correct, but the Confidence Level CL is 0, which is contradictory. Conversely, if the LDE algorithm does not feed back the results, then the result at this time is indeed wrong, and the Confidence Level CL is indeed 0, which makes sense.[/color]
[color=#333333]If the above statement is ok, according to Figure 3, where (FP_INDEX=745.2188, PP_INDEX=749), the result of the ranging is based on the untrustworthy result of the error obtained by PP_INDEX (as can be seen from CL=0.19913), then how Corrected?[/color]
[color=#333333]For example, calculate PP_INDEX-FP_INDEX=3.7812, and then subtract 3.7812ns*C=113.436cm from the distance according to question 2 (1ns=30cm). This is obviously impossible.[/color]
[color=#333333]Sorry to ask you questions again[/color]