对于用于测距的dw1000 设备 我知道能过够获得其cir 那么如何获得组成cir的反射信号呢？比如说多个反射路径的信号？ 有人知道吗？如果了解的话能否告知我，感谢你!
Admin edit: Moved your post to the relevant category
The CIR is the sum of all the reflected paths. You can’t separate it into individual signals and reflections.
Thank you very much for your reply!
I have some more questions, thank you for your reply! Regarding the CIR (Channel Impulse Response) of UWB (Ultra-Wideband), what is its system expression? Is it the sum of reflected paths and some random variables? Is it that signals passing through obstacles are all attenuated and gone, leaving only reflected signals at the receiver? Or is it some other situation? If it is the sum of reflected paths and random variables, then if I obtain its CFR (Channel Frequency Response), is it roughly the reflected signals at different frequencies? I have another question: I have seen many UWB NLOS (Non-Line-of-Sight) identifications based on CIR, but most of them set up different obstacles, different distances, and obtain CIR through experimental measurements. They use some DL (Deep Learning) and ML (Machine Learning) methods for NLOS identification. So, how is CIR used for NLOS identification in a dynamic UWB positioning system? What is the best identification method in real-time UWB positioning systems currently? I only know that UWB seems to directly give out the calculated distance. I have quite a few questions, thank you for your reply. Thank you, if you can provide help, I would be very grateful
I don’t work for Qorvo and to the best of my knowledge they have never provided any official information beyond that in the user manual.
As far as I know the CIR is the correlation result of between the received radio signal and the expected ideal signal.
In a NLOS situation the direct signal is often still received but greatly attenuated. This results in the system triggering off a reflection but a small peak indicating the direct path is visible in the CIR before the detected edge. You could potentially train some model to detect these small peaks and calculate the timing of the line of sight signal.
thank you！thank you！
How is the time of a signal calculated? I know that the DW1000 device, which is based on time-of-flight (TOF) ranging, directly provides time values for TWR (Two-Way Ranging) calculations. Can models make this time more accurate? If so, is there any relevant material available? Thank you very much! Thank you so much!
The basic two way range is:
A Transmits
some time passes while the signal travels
B receives.
B takes time to receive and process the message.
B transmits
some time passes while the signal travels
A receives.
Device A can measure the time between its initial transmission and its final receive.
To calculate range you need to know the time the signal took to travel. The total time minus the other (hopefully known) times.
If you want to improve the accuracy you need to either improve the accuracy of your message receive times, your control of when a transmit starts or your measurement/control of the time B takes between receiving and transmitting.
So what you are asking is where can I apply a model to improve one or more of those steps.
Improving the leading edge detection of the CIR data improves the reception time measurement. In theory a better model could improve calculating that time.
Transmit time measurement/control is very much a function of the device design. I suppose antenna and PCB design will also factor in a little. Power supply voltage and temperature will factor in a little. Antenna delay time is clearly also a factor here too.
All of these are also a factor on receive time accuracy too.
So a model of temperature and voltage effects on antenna delay (using that term to include delays inside the IC and PCB, not just the antenna, since they are hard to separate out) would improve that.
Receive detection is also somewhat dependent on signal levels. Which is in turn dependent on antenna gain which is dependent on antenna orientation. So antenna gain models could improve accuracy. Also depending on the antenna design and construction signal propagation time within the antenna may also be dependent upon antenna orientation. This would give a non-signal strength related antenna orientation range effect.
Of course antenna models are only of use if you can control or know the orientations.
Control of the turn around time on B will also involve the same issues with the receive and transmit time control/measurement.
Assuming you had perfect transmit and receive time control/measurement then both the turn around time on B and the total time on A need to be measured. How well you can do this will be dependent on their clock accuracies. Since the turn around time on B is going to be significantly longer than the travel time the system is very sensitive to differences in the relative clocks of the two devices. Turn around time is typically in the ms region. This means an error of 1/100,000 in that time measurement will give ~10ns of total time error and a final range error of around 1.5 meters.
Most systems measure the differences between the two devices clocks on each packet and use that to correct the turn around time. Or they use DS (double sided) TWR to in effect make two measurements, one plus the error, one minus it, so the error cancels out.
I suppose you could instead apply a model based on past measured clock errors and temperatures (clock speed is highly temperature dependent) to attempt to give a more accurate estimate of the clock errors and so improve accuracy that way. It wouldn’t improve average range error but could potentially decrease noise in the measurements.
Thank you for all your responses and for your assistance. Regarding the CIR signal, if it operates at 64 MHz with 1016 sampling points, how many of the CIR data points would be related to the environment? This typically occurs around the FPIndex. Where can I determine that it is likely the multipath echo signal? Is it a part after the FP?
Everything can be split in to background noise, direct signal and multipath signal.
Background noise will be roughly constant the whole time and is very environment dependent.
Direct signal will give you the first path peak. This should be the first thing in the CIR data that isn’t noise*.
Multipath signal will be everything after the direct signal that isn’t the normal background noise.
*There are some odd situations where this isn’t the case. But they generally require you to be in very large enclosed environments where reflections can bounce around for a long time.
Thank you for your reply! My scenario involves the identification and positioning of UWB systems in some enclosed tunnel and mine environments. Will this lead to a lot of multipath signals? Additionally, I have another question for you. I don’t know if you are familiar with it. For signals that are determined as LOS (Line of Sight) when they are actually NLOS (Non-Line of Sight), and vice versa, what are the main causes of this issue? How are they generally distinguished from each other? Thank you!
Anything other than an open field will give multipath. The question is whether it will matter much, most refection paths are only a little longer than the direct path or require large numbers of reflections. Since each reflection will decrease the signal strength it only takes a couple before it’s too weak to care about, how many exactly does obviously depend a lot on what it’s reflecting off.
However in a mine a environment you may have issues with long reflection paths.
I’ve seen it twice, once in a huge hanger environment with metal doors at both ends and once in a 2 mile perfectly straight tunnel. Both had long sight lines and walls that would give strong reflections rather than absorbing the signal.
The UWB signal is a series of pulses, if a reflection path is long enough then the reflection of the first pulse can arrive around the same time as the direct signal of the second pulse. This often shows up as a weak pulse in the CIR data before the main pulse which is still at around 740. If that weak pulse is strong enough to cross the detection threshold then you get a range that is shorter than truth.
Everything would work perfectly in other indoor environments but this one building in certain spots some of the ranges would measure exactly 34 meters too short. It took a while to figure out what was going on.
In terms of distinguishing NLoS from LoS, I’ve neve found a method that seems particularly reliable. The manual says to look at the difference between the first peak power and total power. That is certainly an indicator but isn’t particularly reliable. I spent a fair amount of time trying to find a simple quality indicator as to whether a signal could be trusted or not. Short of pulling the full CIR data and analysing it (which is painfully slow both in processing time and time spent reading data out of the device) I never found a method I considered reliable. In the end I went with the brute force approach; measure rapidly and with lots of anchors. I then filter out measurements that don’t fit at the position calculation stage.
With enough input data you can filter out range outliers, other than the reflections mentions above multipath will always be longer so generally if you are getting two different ranges then the shortest is correct. And then if you calculate an overdetermined position you can detect if any range measurements don’t seem to fit with the others and reject them.
thank you bro！what can i say！thank。