On 10th June 2020, Dr Ramsey Faragher presented a webinar on a new method for processing GNSS radio signals called Supercorrelation. There were so many questions during the webinar that we couldn't get to everything in the time we had so Ramsey has kindly answered the remaining questions for us below. If you didn't catch the webinar at the time you can watch it back here and then read the Q&A below.
Does supercorrelation help with low signal situations? I mean help with getting a lock in low-signal situations (only reflections available) as opposed to getting a very precise position when line of sight signal is available.
There are two parts to answering this question:
1. Supercorrelation boosts sensitivity by around 7-20dB depending on the existing capabilities and performance of the receiver that the technology is going into. We are able to confidently track signals right down to 4dBHz with a 1-second-long Supercorrelation. The longest Supercorrelators we have tested with real data were 5 seconds long. The boost in sensitivity can often allow a weak and obstructed line of sight signal to be found, even if it was not apparently present according to a standard receiver.
2. Since the question specifically states “only reflections available” I will clarify that if all of the signals detected are reflected and no LOS are available then standard Supercorrelation will not see those reflected signals at all if they are coming into the antenna from a direction that does not correspond to the location of the satellite in the sky. However performing a Skyscan reveals where all the energy is coming from, and allows you to still make use of the measurements from those reflected paths if you have the means to do so. For example you could match them to buildings around you if you have a 3D building model, and then employ shadow matching or 3DMA. If low-accuracy positioning is acceptable for the use case (e.g. confirming location indoors to simply update a weather report on a smartphone) then weak non-line of sight signals can still be used, and the receiver would be able to warn that the positioning accuracy is being limited by the use of non-line-of-sight signals.
What are the input parameters to generate the skyscan energy maps?
Skyscans are created using a bank of Supercorrelators, each tuned to couple strongly to energy coming from different azimuths and elevations. So you are solving the same problems that are described in the talk for creating a Supercorrelator, but for a whole range of possible positions for the satellite across the sky.
When will we see this in nav Receivers on ships?
We are engaged with a number of GNSS hardware manufacturers and OEMs to bring the benefits of our technology to their customers as quickly as possible. Please ask your current GNSS receiver manufacturer to let you know their current level of progress in integrating our technology into their chipset.
When will we see in likes of iPhone?
FocalPoint are currently working with a number of major smartphone chipset providers and handset manufacturers. We are hoping to see smartphone deployments of S-GNSS within the next 2-3 years depending on the different chipset cycles between different manufacturers. Apple acquired the Intel GNSS chipset at the end of 2019 so they did very recently become a company that can directly deploy S-GNSS technology themselves directly.
Is there any limitation for the supercorrelation, especially in some very poor skywindow like the urban area in Hong kong
The key requirement for Supercorrelation to function is that the receiver’s antenna must be moving through space. The minimum speed requirement is a function of the desired length of the Supercorrelator, but for our usual settings, the minimum speed required is 5cm per second, which is roughly 20x slower than walking pace.
Thank you Dr. Ramsey for the presentation, very exciting. Question about latency : what is the net effect on latency compared to conventional GNSS of longer correlation period and improved accuracy of Supercorrelation?
The current latency is 0.5 seconds for our typical settings. However in our next generation of the Supercorrelator we expect to bring this down to a few milliseconds. Note also that the latency in the position estimate for existing smartphone receivers is typically about 1 second, because they typically average together about a second’s worth of GPS measurements before providing an output to the user in an attempt to reduce errors in the navigation solution. Such averaging is not required for Supercorrelation.
In a mobile phone the clock source will not be perfectly stable due to the thermal dynamics, how does this affect super correlation over 1sec
Yes we have seen great variations in the performance of smartphone oscillators. Not just from thermal dynamics but from other unpredictable factors. In some smartphones we have tested our technology on there can be regular discontinuous jumps of 100Hz or more. We had to develop specific clock modelling and signal processing techniques to account for these problems, including what we call the Ultracorrelator, which we did not have time to cover in the talk, but the patent is in the public domain if you would like to read how that works.
Hi Ramsey. Thanks for a very clear description. Do you have a rough estimate in MIPS (or similar) of the added processing needs. I am aware that we already use a huge amount in multipath mitigation, so I am looking for the difference between S-GPS and what we save from the conventional
Yes we have a variety of tools to answer this question for each chipset company, as they all have different existing capabilities before we add in our technology. In the best cases we can actually reduce the overall processing load because of the existing code and overhead that can be removed once Supercorrelation is added. For more thorough information about your chipset in particular please do get in touch.
Delighted that FPP have been given the DofE Award! My question is whether super correlation can be switched on/off by the user?
It is unlikely that this will be a user selectable option, but it is possible in principle.
Could the same method by applied to MMS GNSS receiver?
Supercorrelation can be applied to the ranging signals of all GNSSS, on all frequencies. It could also be applied to many terrestrial radio signals too.
How does the supercorrelation differ from autocorrelation?
Autocorrelation refers to when you correlate a signal with itself, in order to study particular properties of that signal. The supercorrelation process involves changing the correlator sequence stored locally to account for a set of error sources in order to provide a better estimate of the incoming radio signal from the satellite than you get by simply using the textbook description of what has been broadcast.
When do you expect this to become available in a product for the mass market?
FocalPoint are currently working with a number of major smartphone chipset providers and handset manufacturers. We are hoping to see smartphone deployments of S-GNSS within the next 2-3 years depending on the different chipset cycles between different manufacturers.