Some technologies detect a carrier that is below the level of the noise. I would think that to pull them out you would have to do some sort of synchronous detection. But how do you get in sync if you can't see them? I believe GPS is this way. The data is code division multiplexed. I understand that to pull the signal out you have to do some trial and error but since you don't even know where the chips are, the alignment problem is very critical. How do they get started? I want to detect a low level carrier at 60 kHz with a very low modulation rate. To pull it out of the noise I am thinking of sampling synchronously and accumulating the phases into bins. The noise should average out and the signal should add. I am thinking this could be effective because the integration time can be pretty long, at least 1,000s of samples in each bin. The trick is, how do I get my sample rate close enough to make this work without being able to "see" the received signal in the noise? I could use a crystal oscillator at an exact multiple of the carrier but I am concerned that drift might be too high unless I use a fairly expensive unit. So I am thinking I should use a higher frequency and adjust that to the correct sample rate with a DCO. So I could set it to a nominal rate for a starter, but I'm not sure that will be good enough. How is this done in other applications? Rick
Synchronization With Carrier in the Noise
Started by ●November 21, 2012
Reply by ●November 21, 20122012-11-21
On Wednesday, November 21, 2012 3:24:01 PM UTC-5, rickman wrote:> Some technologies detect a carrier that is below the level of the noise. > > I would think that to pull them out you would have to do some sort of > > synchronous detection. But how do you get in sync if you can't see them? > > > > I believe GPS is this way. The data is code division multiplexed. I > > understand that to pull the signal out you have to do some trial and > > error but since you don't even know where the chips are, the alignment > > problem is very critical. How do they get started? > > > > I want to detect a low level carrier at 60 kHz with a very low > > modulation rate. To pull it out of the noise I am thinking of sampling > > synchronously and accumulating the phases into bins. The noise should > > average out and the signal should add. I am thinking this could be > > effective because the integration time can be pretty long, at least > > 1,000s of samples in each bin. > > > > The trick is, how do I get my sample rate close enough to make this work > > without being able to "see" the received signal in the noise? I could > > use a crystal oscillator at an exact multiple of the carrier but I am > > concerned that drift might be too high unless I use a fairly expensive > > unit. So I am thinking I should use a higher frequency and adjust that > > to the correct sample rate with a DCO. So I could set it to a nominal > > rate for a starter, but I'm not sure that will be good enough. > > > > How is this done in other applications? > > > > RickI am aware of applications that use brute force and parallel processing. In your case it may be necessary to interpolate by some integer, execute a bunch of integrations at different frame widths, and pick the one that peaks up the highest. John
Reply by ●November 21, 20122012-11-21
"rickman" <gnuarm@gmail.com> wrote:> Some technologies detect a carrier that is below the level of the noise.Nonsense.> I would think that to pull them out you would have to do some sort of > synchronous detection.Synchronous detection is equvalent to narrow band filter.> But how do you get in sync if you can't see them?Search in one way or another.> I believe GPS is this way.No.>The data is code division multiplexed.No.>I understand that to pull the signal out you have to do some trial and >error but since you don't even know where the chips are, the alignment >problem is very critical. How do they get started?Search for maximum through all possible time and frequency shifts. There are bazillion methods to do that with more or less efficiency; depending on the signal structure and available hardware.> I want to detect a low level carrier at 60 kHzHow low is low?> with a very low modulation rate.How very low is very low?> To pull it out of the noise I am thinking of sampling synchronously and > accumulating the phases into bins. The noise should average out and the > signal should add. I am thinking this could be effective because the > integration time can be pretty long, at least 1,000s of samples in each > bin.Samples of what? In which bin?> The trick is, how do I get my sample rate close enough to make this work > without being able to "see" the received signal in the noise? I could use > a crystal oscillator at an exact multiple of the carrier but I am > concerned that drift might be too high unless I use a fairly expensive > unit. So I am thinking I should use a higher frequency and adjust that to > the correct sample rate with a DCO. So I could set it to a nominal rate > for a starter, but I'm not sure that will be good enough.That is minor technical problem. Either use DCO or interpolate input samples to desired sample rate.> How is this done in other applications?In many other applications it is done in many other ways. Vladimir Vassilevsky DSP and Mixed Signal Consultant www.abvolt.com
Reply by ●November 21, 20122012-11-21
rickman <gnuarm@gmail.com> wrote:> Some technologies detect a carrier that is below the level of the noise. > I would think that to pull them out you would have to do some sort of > synchronous detection. But how do you get in sync if you can't see them?From http://www.kowoma.de/en/gps/signals.htm it seems that the first code to be detected is at 1.023MHz on a 1575.42MHz carrier. Even is it takes a lot of samples, it isn't so long in human terms.> I believe GPS is this way. The data is code division multiplexed. I > understand that to pull the signal out you have to do some trial and > error but since you don't even know where the chips are, the alignment > problem is very critical. How do they get started?> I want to detect a low level carrier at 60 kHz with a very low > modulation rate. To pull it out of the noise I am thinking of sampling > synchronously and accumulating the phases into bins. The noise should > average out and the signal should add. I am thinking this could be > effective because the integration time can be pretty long, at least > 1,000s of samples in each bin.You carrier is 25,000 times lower in frequency, so it might take about 25,000 times longer to lock. According to the article, only 37 of the 2**1023 possible codes are suitable for detection. Presumably that means good characteristics in terms of the possible phase shifts. GPS allows for 32 satellites.> The trick is, how do I get my sample rate close enough to make this work > without being able to "see" the received signal in the noise? I could > use a crystal oscillator at an exact multiple of the carrier but I am > concerned that drift might be too high unless I use a fairly expensive > unit. So I am thinking I should use a higher frequency and adjust that > to the correct sample rate with a DCO. So I could set it to a nominal > rate for a starter, but I'm not sure that will be good enough.> How is this done in other applications?-- glen
Reply by ●November 21, 20122012-11-21
On 11/21/2012 5:24 PM, Vladimir Vassilevsky wrote:> "rickman"<gnuarm@gmail.com> wrote: >> Some technologies detect a carrier that is below the level of the noise. > > Nonsense. > >> I would think that to pull them out you would have to do some sort of >> synchronous detection. > > Synchronous detection is equvalent to narrow band filter. > >> But how do you get in sync if you can't see them? > > Search in one way or another.Oh, really? I hadn't thought of that. Search... I'll try to remember that one!>> I believe GPS is this way. > > No.Yes.>> The data is code division multiplexed. > > No.Yes. Ok, enough of your nonsense... http://en.wikipedia.org/wiki/Code_division_multiple_access "One of the early applications for code division multiplexing is in GPS" I don't like to use Wikipedia as a primary source so... "Code Division Multiple Access (CDMA) A method whereby many radios use the same frequency, but each one has a unique code. GPS uses CDMA techniques with codes for their unique cross-correlation properties." http://www8.garmin.com/aboutGPS/glossary.html>> I understand that to pull the signal out you have to do some trial and >> error but since you don't even know where the chips are, the alignment >> problem is very critical. How do they get started? > > Search for maximum through all possible time and frequency shifts. There are > bazillion methods to do that with more or less efficiency; depending on the > signal structure and available hardware.Yes, of course, in other words, DUH. Get what I"m saying? Your reply is a stupident reply.>> I want to detect a low level carrier at 60 kHz > > How low is low?Don't know yet, I haven't gotten the front end built to receive it. It will be a loop antenna tuned to the carrier frequency with a high Q... don't ask "how high is the Q".>> with a very low modulation rate. > > How very low is very low?1 bps>> To pull it out of the noise I am thinking of sampling synchronously and >> accumulating the phases into bins. The noise should average out and the >> signal should add. I am thinking this could be effective because the >> integration time can be pretty long, at least 1,000s of samples in each >> bin. > > Samples of what? In which bin?Sampling the 60 kHz carrier synchronously. I'm thinking 4 samples per cycle (so 4 bins) and each phase will be added across all the cycles for some period, perhaps 50 ms. The carrier should add up and the noise should average out.>> The trick is, how do I get my sample rate close enough to make this work >> without being able to "see" the received signal in the noise? I could use >> a crystal oscillator at an exact multiple of the carrier but I am >> concerned that drift might be too high unless I use a fairly expensive >> unit. So I am thinking I should use a higher frequency and adjust that to >> the correct sample rate with a DCO. So I could set it to a nominal rate >> for a starter, but I'm not sure that will be good enough. > > That is minor technical problem. Either use DCO or interpolate input samples > to desired sample rate.Again, DUH. HOW does the DCO get adjusted if the sample rate is not close enough to be coherent. I guess it could be swept very slowly.>> How is this done in other applications? > > In many other applications it is done in many other ways. > > Vladimir Vassilevsky > DSP and Mixed Signal Consultant > www.abvolt.comYou are always a wealth of information. Thank you for your reply. Maybe once in awhile you could try not acting like a total jerk. If you don't have anything useful to say... To the rest of the folks in this group, I'm assuming I'm not the first person to try to do something like this. I also expect this sampling technique has a name and has been described and analyzed somewhere. Any pointers? Rick
Reply by ●November 21, 20122012-11-21
On Wed, 21 Nov 2012 17:46:35 -0500, rickman <gnuarm@gmail.com> wrote:>On 11/21/2012 5:24 PM, Vladimir Vassilevsky wrote: >> "rickman"<gnuarm@gmail.com> wrote: >>> Some technologies detect a carrier that is below the level of the noise. >> >> Nonsense. >> >>> I would think that to pull them out you would have to do some sort of >>> synchronous detection. >> >> Synchronous detection is equvalent to narrow band filter. >> >>> But how do you get in sync if you can't see them? >> >> Search in one way or another. > >Oh, really? I hadn't thought of that. Search... I'll try to remember >that one! > > >>> I believe GPS is this way. >> >> No. > >Yes. > > >>> The data is code division multiplexed. >> >> No. > >Yes. > >Ok, enough of your nonsense... > >http://en.wikipedia.org/wiki/Code_division_multiple_access >"One of the early applications for code division multiplexing is in GPS" > >I don't like to use Wikipedia as a primary source so... > >"Code Division Multiple Access (CDMA) >A method whereby many radios use the same frequency, but each one has a >unique code. GPS uses CDMA techniques with codes for their unique >cross-correlation properties." >http://www8.garmin.com/aboutGPS/glossary.html > > >>> I understand that to pull the signal out you have to do some trial and >>> error but since you don't even know where the chips are, the alignment >>> problem is very critical. How do they get started? >> >> Search for maximum through all possible time and frequency shifts. There are >> bazillion methods to do that with more or less efficiency; depending on the >> signal structure and available hardware. > >Yes, of course, in other words, DUH. Get what I"m saying? Your reply >is a stupident reply. > > >>> I want to detect a low level carrier at 60 kHz >> >> How low is low? > >Don't know yet, I haven't gotten the front end built to receive it. It >will be a loop antenna tuned to the carrier frequency with a high Q... >don't ask "how high is the Q". > > >>> with a very low modulation rate. >> >> How very low is very low? > >1 bpsIs that the chip rate or the information symbol rate? If that's the chip rate you're going to have a very hard time making that work. If it's the symbol rate, then you need to use a high enough chip rate so that the symbol synchronization complexity and acquisition time can be made practical.> >>> To pull it out of the noise I am thinking of sampling synchronously and >>> accumulating the phases into bins. The noise should average out and the >>> signal should add. I am thinking this could be effective because the >>> integration time can be pretty long, at least 1,000s of samples in each >>> bin. >> >> Samples of what? In which bin? > >Sampling the 60 kHz carrier synchronously. I'm thinking 4 samples per >cycle (so 4 bins) and each phase will be added across all the cycles for >some period, perhaps 50 ms. The carrier should add up and the noise >should average out. > > >>> The trick is, how do I get my sample rate close enough to make this work >>> without being able to "see" the received signal in the noise? I could use >>> a crystal oscillator at an exact multiple of the carrier but I am >>> concerned that drift might be too high unless I use a fairly expensive >>> unit. So I am thinking I should use a higher frequency and adjust that to >>> the correct sample rate with a DCO. So I could set it to a nominal rate >>> for a starter, but I'm not sure that will be good enough. >> >> That is minor technical problem. Either use DCO or interpolate input samples >> to desired sample rate. > >Again, DUH. HOW does the DCO get adjusted if the sample rate is not >close enough to be coherent. I guess it could be swept very slowly. > > >>> How is this done in other applications? >> >> In many other applications it is done in many other ways. >> >> Vladimir Vassilevsky >> DSP and Mixed Signal Consultant >> www.abvolt.com > >You are always a wealth of information. Thank you for your reply. >Maybe once in awhile you could try not acting like a total jerk. If you >don't have anything useful to say... > > >To the rest of the folks in this group, I'm assuming I'm not the first >person to try to do something like this. I also expect this sampling >technique has a name and has been described and analyzed somewhere. Any >pointers? > >RickYou just need to look around at very low symbol rate communication systems. I know of an application where this was done (that I can't elaborate about), and SS was applied (maybe not CDMA), and the people that did it basically bamboozled the users as to why it was done. The customer thought it was to improve sensitivity (which SS/CDMA doesn't really do), but it was really because it's the easier way to simplify synchronization. You still have to get chip synchronization, but it's much easier at the higher rate. Eric Jacobsen Anchor Hill Communications http://www.anchorhill.com
Reply by ●November 21, 20122012-11-21
On Wednesday, November 21, 2012 5:46:46 PM UTC-5, rickman wrote:> On 11/21/2012 5:24 PM, Vladimir Vassilevsky wrote: > > > "rickman"<gnuarm@gmail.com> wrote: > > >> Some technologies detect a carrier that is below the level of the noise. > > > > > > Nonsense. > > > > > >> I would think that to pull them out you would have to do some sort of > > >> synchronous detection. > > > > > > Synchronous detection is equvalent to narrow band filter. > > > > > >> But how do you get in sync if you can't see them? > > > > > > Search in one way or another. > > > > Oh, really? I hadn't thought of that. Search... I'll try to remember > > that one! > > > > > > >> I believe GPS is this way. > > > > > > No. > > > > Yes. > > > > > > >> The data is code division multiplexed. > > > > > > No. > > > > Yes. > > > > Ok, enough of your nonsense... > > > > http://en.wikipedia.org/wiki/Code_division_multiple_access > > "One of the early applications for code division multiplexing is in GPS" > > > > I don't like to use Wikipedia as a primary source so... > > > > "Code Division Multiple Access (CDMA) > > A method whereby many radios use the same frequency, but each one has a > > unique code. GPS uses CDMA techniques with codes for their unique > > cross-correlation properties." > > http://www8.garmin.com/aboutGPS/glossary.html > > > > > > >> I understand that to pull the signal out you have to do some trial and > > >> error but since you don't even know where the chips are, the alignment > > >> problem is very critical. How do they get started? > > > > > > Search for maximum through all possible time and frequency shifts. There are > > > bazillion methods to do that with more or less efficiency; depending on the > > > signal structure and available hardware. > > > > Yes, of course, in other words, DUH. Get what I"m saying? Your reply > > is a stupident reply. > > > > > > >> I want to detect a low level carrier at 60 kHz > > > > > > How low is low? > > > > Don't know yet, I haven't gotten the front end built to receive it. It > > will be a loop antenna tuned to the carrier frequency with a high Q... > > don't ask "how high is the Q". > > > > > > >> with a very low modulation rate. > > > > > > How very low is very low? > > > > 1 bps > > > > > > >> To pull it out of the noise I am thinking of sampling synchronously and > > >> accumulating the phases into bins. The noise should average out and the > > >> signal should add. I am thinking this could be effective because the > > >> integration time can be pretty long, at least 1,000s of samples in each > > >> bin. > > > > > > Samples of what? In which bin? > > > > Sampling the 60 kHz carrier synchronously. I'm thinking 4 samples per > > cycle (so 4 bins) and each phase will be added across all the cycles for > > some period, perhaps 50 ms. The carrier should add up and the noise > > should average out. > > > > > > >> The trick is, how do I get my sample rate close enough to make this work > > >> without being able to "see" the received signal in the noise? I could use > > >> a crystal oscillator at an exact multiple of the carrier but I am > > >> concerned that drift might be too high unless I use a fairly expensive > > >> unit. So I am thinking I should use a higher frequency and adjust that to > > >> the correct sample rate with a DCO. So I could set it to a nominal rate > > >> for a starter, but I'm not sure that will be good enough. > > > > > > That is minor technical problem. Either use DCO or interpolate input samples > > > to desired sample rate. > > > > Again, DUH. HOW does the DCO get adjusted if the sample rate is not > > close enough to be coherent. I guess it could be swept very slowly. > > > > > > >> How is this done in other applications? > > > > > > In many other applications it is done in many other ways. > > > > > > Vladimir Vassilevsky > > > DSP and Mixed Signal Consultant > > > www.abvolt.com > > > > You are always a wealth of information. Thank you for your reply. > > Maybe once in awhile you could try not acting like a total jerk. If you > > don't have anything useful to say... > > > > > > To the rest of the folks in this group, I'm assuming I'm not the first > > person to try to do something like this. I also expect this sampling > > technique has a name and has been described and analyzed somewhere. Any > > pointers? > > > > RickIs you application by any chance LORAN or some VLF system? John
Reply by ●November 21, 20122012-11-21
On Wed, 21 Nov 2012 15:23:55 -0500, rickman wrote:> Some technologies detect a carrier that is below the level of the noise. > I would think that to pull them out you would have to do some sort of > synchronous detection. But how do you get in sync if you can't see > them? > > I believe GPS is this way. The data is code division multiplexed. I > understand that to pull the signal out you have to do some trial and > error but since you don't even know where the chips are, the alignment > problem is very critical. How do they get started?With a search. Demodulate the stuff coming from the antenna against a pseudo-noise sequence, look for a peak, shift your timing a bit, repeat. GPS uses a short sequence, so the search can be short.> I want to detect a low level carrier at 60 kHz with a very low > modulation rate. To pull it out of the noise I am thinking of sampling > synchronously and accumulating the phases into bins. The noise should > average out and the signal should add. I am thinking this could be > effective because the integration time can be pretty long, at least > 1,000s of samples in each bin.Yes, this is sensible.> The trick is, how do I get my sample rate close enough to make this work > without being able to "see" the received signal in the noise? I could > use a crystal oscillator at an exact multiple of the carrier but I am > concerned that drift might be too high unless I use a fairly expensive > unit. So I am thinking I should use a higher frequency and adjust that > to the correct sample rate with a DCO. So I could set it to a nominal > rate for a starter, but I'm not sure that will be good enough. > > How is this done in other applications?Search for it. Start your radio, run your oscillator down far enough so that you're guaranteed to be at 60kHz, look for carrier. Then step the oscillator up and repeat until you're guaranteed to be at 60kHz. If you don't find the carrier, repeat. -- My liberal friends think I'm a conservative kook. My conservative friends think I'm a liberal kook. Why am I not happy that they have found common ground? Tim Wescott, Communications, Control, Circuits & Software http://www.wescottdesign.com
Reply by ●November 21, 20122012-11-21
rickman <gnuarm@gmail.com> wrote:> Some technologies detect a carrier that is below the level of the noise. > I would think that to pull them out you would have to do some sort of > synchronous detection. But how do you get in sync if you can't see them?(snip)> The trick is, how do I get my sample rate close enough to make this work > without being able to "see" the received signal in the noise? I could > use a crystal oscillator at an exact multiple of the carrier but I am > concerned that drift might be too high unless I use a fairly expensive > unit. So I am thinking I should use a higher frequency and adjust that > to the correct sample rate with a DCO. So I could set it to a nominal > rate for a starter, but I'm not sure that will be good enough.Reminds me a little bit of floppy disk data separators. Though the signal might not be below noise, there is noise (though part of that is the actual data). It used to be the best choice was an analog PLL to lock onto the clock pulses. To lock faster, the write clock was fed to the PLL when it wasn't reading so it would stay close to the right frequency. Otherwise, averaging decreases noise by a factor of sqrt(N) where N is the number of samples being averaged. You can figure out from that how many you need to get above the noise, then add a margin on top of that. -- glen
Reply by ●November 21, 20122012-11-21
