OK. I see what you are saying now. Don't mind me I am a little slow. Is there any sort of way to work out the needed integration time based on the SNR of the incoming signal? I am assuming there must be. Also, Do you know what the range of the GPS signal recieved power is? I have seen -130 as a spec for the minimum power but I have seen nothing about the maximum. Just wondering. I appreciate all of your help as this has been bothering me for a few days now. --Ryan>No. If the noise is random, then the digitized noise+signal will be >slightly non-random. When the amplitude of the noise just happens to be>very small in a given sample instant, the sample will reflect the signal.>Over a long enough time, the correlation between the digitized signal and>the SS code will be significant enough to detect reliably. It's like a >casino making reliable profits over time by skewing the odds of each beta>tiny bit in their favour. > >-- >Matt > > >
GPS Quantization Levels
Started by ●November 4, 2005
Reply by ●November 9, 20052005-11-09
Reply by ●November 9, 20052005-11-09
Hi Ryan The defined range for the GPS signal is between -130 dbm to -110 dbm. But genrally you dont oberve a signal stronger than -114 dbm, even in open sky. If you are talking about integration time in carrier loop , low SNR will require large Integration time,assuming to be non-coherent integration. Infact you can go for adaptive integration time in carrier loop depending upon the signal strength. Regards sandeep>OK. I see what you are saying now. Don't mind me I am a little slow.Is>there any sort of way to work out the needed integration time based onthe>SNR of the incoming signal? I am assuming there must be. Also, Do you >know what the range of the GPS signal recieved power is? I have seen-130>as a spec for the minimum power but I have seen nothing about themaximum.>Just wondering. I appreciate all of your help as this has beenbothering>me for a few days now. > >--Ryan >
Reply by ●November 9, 20052005-11-09
Thanks. Actually I am doing the acquisition and tracking in the frequency domain which ends up as a coherent integration. I was just trying to maybe put some numbers on things instead of just determining them experimentally. Thanks again.>Hi Ryan > The defined range for the GPS signal is between -130 dbm to >-110 dbm. But genrally you dont oberve a signal stronger than -114 dbm, >even in open sky. >If you are talking about integration time in carrier loop , low SNR will >require large Integration time,assuming to be non-coherent integration. > >Infact you can go for adaptive integration time in carrier loopdepending>upon the signal strength. > >Regards >sandeep > > > > > >>OK. I see what you are saying now. Don't mind me I am a little slow. >Is >>there any sort of way to work out the needed integration time based on >the >>SNR of the incoming signal? I am assuming there must be. Also, Do you >>know what the range of the GPS signal recieved power is? I have seen >-130 >>as a spec for the minimum power but I have seen nothing about the >maximum. >>Just wondering. I appreciate all of your help as this has been >bothering >>me for a few days now. >> >>--Ryan >> > > >
Reply by ●November 10, 20052005-11-10
For anyone who cares: I tested the theories put in this discussion out and they seem to make sense. First I plotted the Signal(GPS Signal created in MATLAB[another whole discussion could have gone into that]), Noise, and Quantized 2-bit Signal+Noise. From this graph it is hard to tell whether the correlation between Signal and Signal+Noise still exists. So, to put some numbers on it, and considering the signals are aligned properly in time... If I correlate the noise vs. the noise I get a value of sum(real(noise(1:50).*real(noise(51:100))))) = -2.276863328559156e-009 If I correlate the noise vs. the signal I get sum(real(noise(1:50).*real(noise(51:100)+Signal(1:50)))) = -2.292580903252614e-009 And if I correlate the (signal + noise) vs. the signal I get sum(real(A.data(1:50)).*real(noise(51:100)+Signal(1:50))) = -1.156741657174452e-005 So what does this show? I believe this shows that the quantized Signal+Noise does indeed retain some properties of the original signal due to a correlation existing between the two. I also believe that the correlation will get better over longer data lengths. Thanks to all that have posted in this thread as it has been very informative to me and hopefully you. If you would like the diagrams I have for this or have any comments about what I have discussed above please post as I would like to see this discussion through with some proof. --Ryan
Reply by ●November 10, 20052005-11-10
"effimofunk" <rfrankel@ufl.edu> wrote in message news:AIKdnXx_mJRIeu_eRVn-hg@giganews.com...> For anyone who cares: > I tested the theories put in this discussion out and they seem to make > sense. First I plotted the Signal(GPS Signal created in MATLAB[another > whole discussion could have gone into that]), Noise, and Quantized 2-bit > Signal+Noise. From this graph it is hard to tell whether the correlation > between Signal and Signal+Noise still exists. So, to put some numbers on > it, and considering the signals are aligned properly in time... > > If I correlate the noise vs. the noise I get a value of > sum(real(noise(1:50).*real(noise(51:100))))) = -2.276863328559156e-009 > > If I correlate the noise vs. the signal I get > sum(real(noise(1:50).*real(noise(51:100)+Signal(1:50)))) = > -2.292580903252614e-009 > > And if I correlate the (signal + noise) vs. the signal I get > sum(real(A.data(1:50)).*real(noise(51:100)+Signal(1:50))) = > -1.156741657174452e-005 > > So what does this show? I believe this shows that the quantized > Signal+Noise does indeed retain some properties of the original signal due > to a correlation existing between the two. I also believe that the > correlation will get better over longer data lengths. > > Thanks to all that have posted in this thread as it has been very > informative to me and hopefully you. > > If you would like the diagrams I have for this or have any comments about > what I have discussed above please post as I would like to see this > discussion through with some proof. >Sounds good Ryan , I did a quick and dirty sim too : if you have perfect phase recovery, zero doppler and you are only looking at one satellite then just using the sign bit loses you not quite 2 dB of SNR (in AWGN) compared with using lots of perfect quantisation steps. If you go to 2 bit samples then you lose about 1 dB SNR (according to my rough model if you arrange things so that the S+N actually gives you +/-3 at the ADC o/p a fair percentage of the time). Many people would say that the extra power needed for processing the second bit isn't worth it but the increased tolerance to narrowband interference is. Going to even more bits seems a complete waste of time until you want to do some really expensive anti-jamming. I'm sure that this has been documented in stacks of books and papers before now and they show, not only good mathematico-physical proofs but, in some cases, interesting tricks. Best of Luck - Mike
Reply by ●November 10, 20052005-11-10
"if you have perfect phase recovery, zero doppler and you are only looking at one satellite" <--- NICE! The difference between going from one to two bits has shown through processing real GPS signals through my system an improvement of ~5ft. In most cases this is probably negligable but it some maybe not. I have a reference that quotes similar values to the ones you posted so nice job on the simulations. On another note: would adding doppler or more satellites degrade your SNR? I could possibly see that adding more satellites MAY technically degrade your SNR but adding doppler (since you have to figure out what it is closely anyways) does not seem to degrade SNR in my simulations. (i.e. when I create multiple satellite signals all with a doppler shift and a delay to the first bit transition my results remain tighly grouped).>Sounds good Ryan , >I did a quick and dirty sim too : if you have perfect phase recovery,zero>doppler and you are only looking at one satellite then just using thesign>bit loses you not quite 2 dB of SNR (in AWGN) compared with using lots of>perfect quantisation steps. >If you go to 2 bit samples then you lose about 1 dB SNR (according to my>rough model if you arrange things so that the S+N actually gives you +/-3at>the ADC o/p a fair percentage of the time). Many people would say thatthe>extra power needed for processing the second bit isn't worth it but the >increased tolerance to narrowband interference is. Going to even morebits>seems a complete waste of time until you want to do some really expensive>anti-jamming. > >I'm sure that this has been documented in stacks of books and papersbefore>now and they show, not only good mathematico-physical proofs but, in some>cases, interesting tricks. > >Best of Luck - Mike > > >