Idea for weak signal decoding at or below noise floor

[Beamin]

I'm sure this is probably already invented all my ideas/inventions/discoveries are, as I realized at a young age.


You have a transmitter and receiver with your clocks synchronized such off the WWV time signal, GPS, cell towers, atomic clocks, etc. The transmitter works in short bursts that repeat data many times. The receiver knows that only certain sets of bits or characters are valid and looks up each segment against a look up table and the clock signal and has a check sum at the end of each word like 10101010 repeating.


So no one burst of data can get through because the noise is often over the signal but little chunks can get through. So after getting a few "chunks" of data enough times the receiver can piece together whats its transmitting by over lapping what it hears.

Does this have a name, or does it not work like this?

[MasterT]

CDMA.

[langwadt]

CDMA.

similar but not quite the same, in cdma each bit becomes a sequence and at the receiving end you correlate with the same sequence

[Wirehead]

Look up convolutional coding (and Viterbi decoding ( = probability decoding)); and then take a look at WSPR

[MiDi]

A look at autocorrelation and cross-correlation could be useful.

[rf-fil]

Check out hybrid ARQ (HARQ) protocol that's used in the WCDMA and LTE mobile systems. The same packet gets repeatedly re-transmitted until the receiver decodes it correctly (checksum is OK). The smart thing about this is that the receiver doesn't discard the failed packets but keeps them in a buffer. Subsequent packets get constructively combined with previously received packets, while the random noise doesn't. It also changes up the interleaving / puncturing pattern in-between re-transmissions, but I don't want to overcomplicate this ;-).

[hagster]

It is often better to think about the energy rather than power required to send data bits. Remember "energy = power x time". You can either send the data quickly at high power or slowly at low power as you are suggesting.

The best performance you can theoretically acheive is governed by the Shannon Limit. Most modern modulation schemes(with error correction codes applied) operate very close to this limit. Hence you cant really beat them by much at all.

There are of course other properties of modulation schemes that make the better or worse for given applications. The cell phone industry for example is highly driven by spectral efficiency for instance. I.e they want to cram as much data through their allocated bandwidth as possible.

[Kalvin]

Here is a nice summary of required signal to noise ratio for various amateur (digital) transmission modes:

http://www.pa3fwm.nl/technotes/tn09b.html

I guess the synchronous TX & RX mode would go somewhere between the Coherent BPSK on VLF requiring roughly -1 dB  Eb/N0 and WSPR +5 dB Eb/N0. If the transmission is synchronous enough between the TX and RX, and the transmission is repeated multiple times, that would improve the received Eb/N0 as the receiver would be able to acquire more information for each transmitted bit or symbol.

[Circlotron]

Sounds similar to averaging a repetitive but noisy signal on a scope.

[Kalvin]

If the receiver (detector) has some statistical knowledge about the message sent, the receiver can improve its performance as it can reject all those messages which contain invalid symbol sequencies.

[hagster]

If the transmission is synchronous enough between the TX and RX, and the transmission is repeated multiple times, that would improve the received Eb/N0 as the receiver would be able to acquire more information for each transmitted bit or symbol.

Yes you increase the Eb/No because you transmitted more Eb.

[Kalvin]

Although the TX may be able to send the message at very precise times, the transmitted signal will always contain some timing jitter due to the changing RF propagation conditions.

[Beamin]

So is my idea just what CDMA does and not anything new? I got the idea when I saw that wspr needed a gps signal and its not just for reporting its location. I would call my discovery TIME delay multiple Access or TDMA for short, and make each tower send the calls to a satellite link so no copper lines to the tower reducing installation and maintenance costs as well as local tel co fees. It would then send the signal back down to a tel switch in the central part of the country and the phone company do the rest. Why did I think of this in the late 80's?

[coppercone2]

I think this has to do woth randomness and the material offered by a boring ass random signals and systems class would elaborate on this. I think it has to do with noise curves

[hagster]

By reducing time uncertainty you can eliminate a lit of random data that could randomly just look like the correct signal. As such you can reduced the "detect" threshhold to get improve sensitivity. However most modulation schemes already use clock recovery systems that work at a slower rate and get the same benefit by other means.

[Sparker]

So, you retransmit every bit 6 times, but at different time intervals. Or you could make every bit 6 time larger and transmit it once. The total energy of one bit is the same in either case, which will ultimately define your error rate, depending on your error correction coding and modulation type.  So far nothing new, except that now you also occupy a larger part of spectrum, which might or might not be desirable.

I'm not an expert in this, but maybe if you were going to suppress some specific noise type, a special bit organization would make sense. But in most cases our noise can be modeled as AWGN.

Also another concept which wasn't mentioned is data scrambling. You can rearrange your transmitted bits before transmission and rearrenge them back so that grouped up errors will be evenly spread along your bit stream.

[coppercone2]

I actually bought a psudeorandom noise generator a while back to play with this stuff. Not sure if that project will ever get off the ground.

[radioactive]

https://en.wikipedia.org/wiki/Complementary_sequences
https://en.wikipedia.org/wiki/Barker_code

Here are a couple of links related to complementary code keying that you might find interesting.  The idea here is to find sequences with perfect auto-correlation properties not only for good SNR, but for good spectral properties  (side-lobe reduction).  There is also Gold sequences with similar properties. https://en.wikipedia.org/wiki/Gold_code   Good for DSSS (direct sequence) type modulation.  I think what you are describing also sounds a lot like forward error correction.  You might like this site.  http://complextoreal.com/  Dr. Langton has some awesome tutorials (including modulation and FEC) on her site.

[hagster]

I actually bought a psudeorandom noise generator a while back to play with this stuff. Not sure if that project will ever get off the ground.

You dont need hardware to test these ideas out. You can test it in gnuradio or with Matlab(Octave) or other software.

[coppercone2]

im afraid i cant do that dave
on the other hand this defrosting noise generating equipment sitting in my lab right now is very interesting 

[rhb]

This has been a major interest of mine for over 30 years.  If you actually want to learn about the subject get a copy of "Random Data" by Bendat & Piersol.  An old edition will do.  All of this is based on the work of Norbert Weiner in the early 40's.

The general idea is the basis of all spread spectrum systems.  This was all TOP SECRET during the 50's and 60's.  Now it is wide spread and is the basis of WiFi, etc.

There is an amateur CW mode which was called "coherent CW" 30 years ago, but now goes by QRSS.  It's clocked CW at very slow speeds and very low power levels. In addition to the time you also need the phase delay of the communication path and a very stable receiver as the bandwidth is *very* narrow.

My particular interest is very broadband pulses at FCC part 15 legal levels emitted at random times.  So unless you know which times to integrate over, the probability of intercept is very near zero.  The hard part is an efficient, very wide band antenna.  My standard for success would be to be able to reliably send a 2 KB message point to point over 24 hours at under 10 mW to antipodal distances.

WSPR and the other Joe Taylor developed modes are fundamentally frequency shift keying with precise timing and mathematical bells and whistles mostly from Bendat & Piersol,  but with different jargon attached.

B&P and Octave or MATLAB are the tools for learning this.

The essence of direct sequence spread spectrum  (DSS) is the mathematical property that the autocorrelation of a random sequence is a spike at T0 and the cross correlation of two random  series is zero.

[cdev]

antipode meaning the exact opposite end of the Earth?

There are a lot of broadband antennas to choose from. You choice would depend  on whether you wanted unidirectional or relatively omnidirectional, what polarization you wanted, etc.

[rhb]

That is the correct definition of antipodal.

Yes, there are lots of broadband antennas, but for the example I cited, you really need a frequency independent antenna.  They exist.  But I'm not aware of any that range from DC to daylight.

It's not too hard at VHF and above.  It's the LF to HF range that gets challenging.  There are 6 octaves from 1 MHz to 32 MHz and  11 octaves from 1 MHz to 1 GHz. Or to put it another way, it's 6 octaves from 1-32 MHz, but one octave from 32-64 MHz.  You can always achieve a match at the price of low radiation efficiency.

And, yes, I am probably crazy.  I just spent $800 on antenna theory books because I did not like any of the broadband HF antenna designs I saw for sale.

I have no idea if what I outlined would work.  I also know that neither does anyone else, though I'm certain there is no shortage of opinions.   So it seems like an interesting project.  It *will* work at some power level.  The question is how low a power level will still work.

[cdev]

Above the MUF propagation is very dependent on being line of sight or at least semi close to it, except under fairly unusual conditions.

Very low frequency signals may be able to go very far but only signals with very low bandwidth and high power and/or very large antennas could be expected to be reliably received at the other end of the earth.

[OwO]

That is the correct definition of antipodal.

Yes, there are lots of broadband antennas, but for the example I cited, you really need a frequency independent antenna.  They exist.  But I'm not aware of any that range from DC to daylight.

OFDM instead of wideband pulses might be able to cope with antennas with bad frequency response and phase response.

Since you have stationary nodes and doppler isn't a big problem, OFDM with an extremely long period (on the order of seconds or minutes) might be doable, and will ease timing synchronization (you can use e.g. NTP and align OFDM period to minute boundary just like WSPR does).

I'm working on hardware FFT acceleration to make high-order high-bandwidth OFDM possible without a high end PC.