EMRFD Message Archive 12011

Message Date From Subject
12011 2015-12-15 08:43:38 Tayloe, Dan (Noki... More sensitive receiver archtiecture
I saw this article that discusses a new receiver approach that allow the creation of super sensitive receivers that better hear signals down in the noise.


Anyone figure this out?  

From what I gather, they are breaking up a wide swath of spectrum into a number of smaller pieces, each piece with its own detector.  Then they combine all the detected pieces together to recreate the detected signal.  That seems reasonable.  Heck, SDR does that by breaking the signal into thin FFT bins, and then combining bins back together to get the desired recovered bandwidth.  The thing that sounds unreasonable is the claim for a breakthrough in sensitivity gain.  For some reason noise seems to just disappear leaving the desired signal much easier to see.

First of all, noise is a signal too.  There is nothing special about reducing the bandwidth.  When you do that, you get less undesired noise, but also less desired signal.  It seems to me that the S/N ratio stay the same, it is just that both are smaller in magnitude.

Even if the noise was not an external signal, but rather internally generated by the receiver, the use of multiple receivers, each focused on a different segment of the spectrum is not going to help.  I would expect the wide band noise figure of a receiver to be the same as the individual noise figure of the individual receivers.  We still get undesired noise (a signal) on top of the real signal that we want to hear. In both the wideband and sum-of-narrow-band approaches, the S/N stays the same.

I can see that if a very fast, very wideband detector is hard to build, then using multiple narrower, slower detectors in parallel might be used to create something that cannot be done other wise.  However I completely miss the more general application of somehow "eliminating noise" by this approach as if noise is not a signal and can just be somehow eliminated.  

Is there something fundamental I am missing here?  

- Dan, N7VE
12012 2015-12-15 09:38:53 Philip Pollock Re: More sensitive receiver archtiecture
I am reminded of a technique I learned when studying applied physics.

Epoch testing. Take a narrow bandwidth, sum the output of numerous small time periods, the epoch.
Noise being random cancels it's self out during the summation. The signal buried in the noise is always there and adds to it's self, to rise above the noise. I have only had experience of this technique in radio astronomy.
It is quite a sight to see a star appear as a large spike signal, above the noise when previously it was down in the noise.

Maybe someone is thinking along the line of Epoch testing?

Philip EI8JT

12013 2015-12-15 12:28:05 Eamon Egan Re: More sensitive receiver archtiecture
Another idea, or perhaps a different way of explaining it?

Look at how noise blankers work. Perhaps they are a crude instance of this approach? If I understand correctly, a noise blanker looks for events of energy above a certain level, in a wider-band part of the circuit, which represent a time-limited but wideband signal likely to be noise. This is used to trigger blanking for a limited period of time.

In an approach with much more DSP finesse involved, one could perhaps exploit finer statistical properties of wideband noise, related to its origin in discrete events, to cancel out its effect on a narrowband signal.

I think I have already speculated far too much prior to reading the article in question....

- Eamon, VE2EGN, AB1NK

12014 2015-12-15 13:40:28 Graham / KE9H Re: More sensitive receiver archtiecture

No laws of physics are being broken.

The give up is in modulation bandwidth.

If the signal is a discrete frequency, and the noise is equally distributed across the receive spectrum, then narrowing the receive bandwidth does improve S/N, provided the entire signal is still inside the reduced bandwidth. Signal stays the same, noise scales downward with bandwidth.

Now take the case of a modulated signal. Any modulation sidebands outside the narrow receiver bandwidth are wiped-off.  They might appear in another narrow receiver (called a 'bin' in an FFT type receiver) but once again, the rate at which the signal can change (and therefore the modulation carried) is limited to the bandwidth of the bin.

Now I have a spectrum display which is FFT based.  So, what I am looking at is a CW signal on a display that is, say 4096 bins, that are individually 1 Hz wide.  So my display is 4.096 kHz wide. 4096 each, 1 Hz wide receivers, whose signal levels are plotted side by side on a display

The S/N inside each individual bin, relative to a 400 Hz wide CW receiver filter, is 10*log(400/1) = 26 dB better.

So, I can see signals in my display, that I can not hear or demodulate with my ear, listening through a 400 Hz filter.
I can see signals that are below the noise floor of the audio channel.

But, I can't demodulate the signal on the display, any faster than about 1 Hz modulation.  So not useful for demodulating fast CW.
But, it is still useful for slow processes, like turning an antenna system (or adjusting a phased array antenna system) to peak the signal, hopefully bringing it above the necessary signal level to demodulate in the audio channel.

By example, the bin widths in the panadaptor display on a FLEX-6000 radio are down in the single digit Hertz range when you zoom in, and they are continuously explaining to customers why they can clearly see weak signals that they can not understand until they are 10 dB or more out of the noise on the panadaptor.

--- Graham / KE9H


12015 2015-12-15 13:44:44 Chris Trask Re: More sensitive receiver archtiecture
>I saw this article that discusses a new receiver approach that
>allow the creation of super sensitive receivers that better hear
>signals down in the noise.
>Anyone figure this out?


It looks similar to the matched filter receivers used in high-resolution radars, the difference here being that the signal processing is done at optical frequencies. I seriously doubt that it would have any practical application for voice or any other form of coherent communication.

I spent many years trying to devise a method for suppressing interfering signals for DSB, only to find that the Soviets had reduced it to practice to overcome radar jamming back in the 60's. American engineers copied it and claimed they had devised it first. Made a fully working one in an afternoon. It does reduce QRM by as much as 3dB but doesn't help at all for signals buried in noise.

I did make a tracking commutative filter about 40 years ago that would extract a single frequency tone buried in 60dBv of noise, but it wouldn't help voice communication at all.


When the going gets weird, the weird turn pro
- Hunter S. Thompson
12016 2015-12-15 13:48:46 Steve Dick Re: More sensitive receiver archtiecture
I’ll take a stab at it. The article appears to be leaving something out.  Seems they are developing a whole bunch of frequency bins optically, then doing sampling multiple times within effectively the same the same frequency bin and doing coherent integration. This causes the desired signal to grow while noise, which tends to be gaussian and is incoherent, will cancel out. The thing that is not clearly explained is how “it can see a very fast, faint signal while observing over a much larger time interval”. This implies that the transient signal they are looking for is sampled multiple times.  But the sampling rate can’t exceed Nyquist, no matter what techniques they are using. So you can indeed do this with a very large FFT, depending on the bandwidth of their frequency bins they are using.
12017 2015-12-15 20:57:31 farhanbox@gmail.c... Re: More sensitive receiver archtiecture

i have come to believe that shannons law is the equivalent of laws pf thermodynamics. 

when messing with signals and noise, you need either power or time on your side. all our modea from morse to jt65 are ways to slice and dice shannon. not overcome it.

- f

------ Original message------

12018 2015-12-15 22:28:34 victorkoren Re: More sensitive receiver archtiecture
I think that they are working with technology and frequencies that the limiting noise is the receiver noise and not the external noise. They just use multiple receivers and average their output signals together so they lower the total received noise.
This is comparable to improving the receiver noise figure with one major difference: it also improves the local oscillator phase noise, a point that you would not have by just improving receiver noise figure.
Victor - 4Z4ME 
12019 2015-12-16 02:03:13 Nick Tsakonas Re: More sensitive receiver archtiecture

this post adds a few more paragraphs to the originally posted article (without clearing things out more) and also contains a link to the paper - someone with a subscription on Science magazine can read it.


12020 2015-12-16 04:35:12 farhanbox@gmail.c... Re: More sensitive receiver archtiecture

using multiple receivers will work only id each has a different physical noise source. a bit like the radio telescope arrays. once you have a sampled signal, the noise is already staked in. whichever way you slice it, the noise wont go down. your only hope is to narrow the bandwidth to reduce the noise.

this is entropy at work. it is a brutal reality that is linked to the one way flow of time. the moving waterfall having writ moves on. all your cycles and all your algos..

------ Original message------