What is the minimum tolerable SNR in today's (digital) communication systems for successful recovery of the message? I've heard that regular commercial systems can recover the message with a SNR as small as -60 dB (using robust coding techniques). Is it correct? And if yes, how can you observe and distinguish even the presence of a signal when it's burden in a noise with 106X more power?
5
-
\$\begingroup\$ Do you have a technical source for this claim? \$\endgroup\$– Andy akaCommented May 14, 2013 at 8:26
-
\$\begingroup\$ @Andy I heard a former professor of communication systems (university of Washington) said it. \$\endgroup\$– ZorichCommented May 14, 2013 at 8:32
-
\$\begingroup\$ If this cannot be substantiated by an article that shows a worked example and explains what type of error correction algorithms were in place (e.g.), you might not get a decent answer unless someone is familiar with the technicalities of this claim. It's easy to make claims like this. In radio, if you take the full bandwidth of all signals you can simultaneously receive and compared that power with one signal you are looking for you could make a similar claim!! \$\endgroup\$– Andy akaCommented May 14, 2013 at 8:38
-
\$\begingroup\$ I think he was talking about convolutional codes. \$\endgroup\$– ZorichCommented May 14, 2013 at 10:40
-
\$\begingroup\$ There's some people doing insane things with PSK, I don't have details as it's a friend who is into radios who has mentioned it. Worth a google though. \$\endgroup\$– John UCommented May 14, 2013 at 10:49
Add a comment
|
2 Answers
\$\begingroup\$
\$\endgroup\$
1
What your professor is almost certainly referring to is SSDS (Spread Spectrum Direct Sequence) encoding technicals (yes they are convolutional) and they trade off data rate for a given bandwidth for the ability to recover signal in deep noise. CDMA systems from Qualcomm use CDMA but not so much for signal in noise recovery but for code diversity in band, meaning multiple independent transmissions overlapped in band.
CDMA uses the concept of process gain to reflect the ability to recover from noise and it counts on the fact that the "pseudo noise" (aka PN) has a distinct matched filter. When the matched filter is applied any noise in band will be mapped out of band and any PN will be mapped to a modulated signal. the process gain (measured in dB) tells you how much excess that you can have in S/N.
However, 60dB is trivial in context to the the NASA PN codes that are used in it's deep space network. These codes can have repeat lengths of 1000's of chips. A chip is s sub-unit of modulation, it is the ratio of chips to bits that determines the process gain.
An excellent book on this subject is "spread spectrum systems" by Dixon.
GPS is also based upon CDMA.
This work was first developed by Claude shannon and was top secret for many years post war.
answered May 14, 2013 at 10:58
-
\$\begingroup\$ Thank you very much! Do you mean -60dB or +60dB?(in your 3rd paragraph) \$\endgroup\$– ZorichCommented May 14, 2013 at 11:28
\$\begingroup\$
\$\endgroup\$
3
GPS has that kind of ability to recover from negative SNR; however, it does it by correlation. The signal is very repetitive and predictable, so autocorrelation is applied. This magnifies repeating elements of the signal and subtracts noise.
However, that yields a very low data rate and relies on the reciever already guessing what the data should look like. The relationship of link SNR to data rate still holds.
-
\$\begingroup\$ Of course it is done (as I said in the question) by applying robust coding techniques that , from information theoretic point of view, add lots of redundancy to the transmitted signal and the message is recovered using error correction capabilities of the code. \$\endgroup\$– ZorichCommented May 14, 2013 at 8:47
-
\$\begingroup\$ Then what is the question, if you already know this? \$\endgroup\$– pjc50Commented May 14, 2013 at 9:44
-
\$\begingroup\$ What I want to know is if it is possible to have such values for SNR .Can you explain more about GPS detection capability?(or refer to a source). Thanks \$\endgroup\$– ZorichCommented May 14, 2013 at 10:37