On Thu, 09 Jun 2016 20:52:19 +0300, Evgeny Filatov <filatov.ev@mipt.ru> wrote:>On 09/06/2016 15:37, Bob Masta wrote: > >(snip) > >> Hmm, sounds like one of those dither cancellation systems >> where the known dither source can be subtracted at the other >> end. >> >> For that matter, it also sounds like an encryption system >> where the transmitter and receiver are both using a known >> (pseudo) random stream. Without the known encryption >> source, any other receiver gets 100% noise; with the source >> they get 100% signal. >> >> Best regards, >> >> >> Bob Masta >> >> DAQARTA v9.20 >> Data AcQuisition And Real-Time Analysis >> www.daqarta.com >> Scope, Spectrum, Spectrogram, Sound Level Meter >> Frequency Counter, Pitch Track, Pitch-to-MIDI >> FREE 8-channel Signal Generator, DaqMusiq generator >> Science with your sound card! >> > >I was thinking about a system where the sole source of noise is the >transmitter (e.g. you cool the detector of the receiver with liquid >nitrogen, so it's not noisy). And some kind of a feedback loop so that >the transmitter makes use of the noise to transmit less power. It's >still far-fetched or impossible, but I thought it might make for a fun >mind exercise. :)It is interesting. Any signal detector is subject to thermal noise due to electron motion at some non-zero Kelvin temperature (which is a lot colder than liquid nitrogen). So cancelling thermal noise suggests the signal itself can be reduced to the small quanta of an electron or two moving up a shell or two, or something like that. So in the limit as your noise canceller cancels more and more thermal noise, making the entire channel appear to be at 0 Kelvin, you will still have quantization noise associated with the particle quanta. So you can't really make the signal infinitely small in practice, since there is a quantum limit at some point, even if you successfully cancel all of the thermal noise.>Of course, you are correct about the situation when both the receiver >and the transmitter know the dither (or pseudo-random stream), which is >a standard and well-researched case. > >Gene >
Online · Live + on demand
February 9-11, 2027
Price goes up Jan 1: register now to lock in today's rate
Register Now →
signal transmission with cheats
Started by ●June 9, 2016
Reply by ●June 9, 20162016-06-09
Reply by ●June 9, 20162016-06-09
On 09/06/2016 21:09, Eric Jacobsen wrote:> On Thu, 09 Jun 2016 20:52:19 +0300, Evgeny Filatov > <filatov.ev@mipt.ru> wrote: > >> On 09/06/2016 15:37, Bob Masta wrote: >> >> (snip) >> >>> Hmm, sounds like one of those dither cancellation systems >>> where the known dither source can be subtracted at the other >>> end. >>> >>> For that matter, it also sounds like an encryption system >>> where the transmitter and receiver are both using a known >>> (pseudo) random stream. Without the known encryption >>> source, any other receiver gets 100% noise; with the source >>> they get 100% signal. >>> >>> Best regards, >>> >>> >>> Bob Masta >>> >>> DAQARTA v9.20 >>> Data AcQuisition And Real-Time Analysis >>> www.daqarta.com >>> Scope, Spectrum, Spectrogram, Sound Level Meter >>> Frequency Counter, Pitch Track, Pitch-to-MIDI >>> FREE 8-channel Signal Generator, DaqMusiq generator >>> Science with your sound card! >>> >> >> I was thinking about a system where the sole source of noise is the >> transmitter (e.g. you cool the detector of the receiver with liquid >> nitrogen, so it's not noisy). And some kind of a feedback loop so that >> the transmitter makes use of the noise to transmit less power. It's >> still far-fetched or impossible, but I thought it might make for a fun >> mind exercise. :) > > It is interesting. Any signal detector is subject to thermal noise > due to electron motion at some non-zero Kelvin temperature (which is a > lot colder than liquid nitrogen). So cancelling thermal noise > suggests the signal itself can be reduced to the small quanta of an > electron or two moving up a shell or two, or something like that. So > in the limit as your noise canceller cancels more and more thermal > noise, making the entire channel appear to be at 0 Kelvin, you will > still have quantization noise associated with the particle quanta. > > So you can't really make the signal infinitely small in practice, > since there is a quantum limit at some point, even if you successfully > cancel all of the thermal noise. > >> Of course, you are correct about the situation when both the receiver >> and the transmitter know the dither (or pseudo-random stream), which is >> a standard and well-researched case. >> >> Gene >> >Indeed, liquid nitrogen usually means 77 K, which is "much" hotter from absolute zero. But since statistics of e.g. number of particles at a given quantum level with energy E is determined by the exponential term exp(-E/[KT]) (at least in the limit of high T), for many practical uses T = 77 K is "a lot" better than T = 300 K. It's more like threshold behavior, so you'd want to justify too much cooling. At least, I've seen liquid-nitrogen-cooled detectors, but I haven't seen liquid-helium-cooled detectors (which perhaps speaks more about my limited experience, but yet). Perhaps you are correct about physical limits. I haven't been thinking about that. E.g. there are single-photon detectors, but the problem is, you won't detect every single photon. More likely you'd lose the most of them, and only detect a few of the many. Gene
Reply by ●June 9, 20162016-06-09
On Thu, 09 Jun 2016 21:27:57 +0300, Evgeny Filatov <filatov.ev@mipt.ru> wrote:>On 09/06/2016 21:09, Eric Jacobsen wrote: >> On Thu, 09 Jun 2016 20:52:19 +0300, Evgeny Filatov >> <filatov.ev@mipt.ru> wrote: >> >>> On 09/06/2016 15:37, Bob Masta wrote: >>> >>> (snip) >>> >>>> Hmm, sounds like one of those dither cancellation systems >>>> where the known dither source can be subtracted at the other >>>> end. >>>> >>>> For that matter, it also sounds like an encryption system >>>> where the transmitter and receiver are both using a known >>>> (pseudo) random stream. Without the known encryption >>>> source, any other receiver gets 100% noise; with the source >>>> they get 100% signal. >>>> >>>> Best regards, >>>> >>>> >>>> Bob Masta >>>> >>>> DAQARTA v9.20 >>>> Data AcQuisition And Real-Time Analysis >>>> www.daqarta.com >>>> Scope, Spectrum, Spectrogram, Sound Level Meter >>>> Frequency Counter, Pitch Track, Pitch-to-MIDI >>>> FREE 8-channel Signal Generator, DaqMusiq generator >>>> Science with your sound card! >>>> >>> >>> I was thinking about a system where the sole source of noise is the >>> transmitter (e.g. you cool the detector of the receiver with liquid >>> nitrogen, so it's not noisy). And some kind of a feedback loop so that >>> the transmitter makes use of the noise to transmit less power. It's >>> still far-fetched or impossible, but I thought it might make for a fun >>> mind exercise. :) >> >> It is interesting. Any signal detector is subject to thermal noise >> due to electron motion at some non-zero Kelvin temperature (which is a >> lot colder than liquid nitrogen). So cancelling thermal noise >> suggests the signal itself can be reduced to the small quanta of an >> electron or two moving up a shell or two, or something like that. So >> in the limit as your noise canceller cancels more and more thermal >> noise, making the entire channel appear to be at 0 Kelvin, you will >> still have quantization noise associated with the particle quanta. >> >> So you can't really make the signal infinitely small in practice, >> since there is a quantum limit at some point, even if you successfully >> cancel all of the thermal noise. >> >>> Of course, you are correct about the situation when both the receiver >>> and the transmitter know the dither (or pseudo-random stream), which is >>> a standard and well-researched case. >>> >>> Gene >>> >> > >Indeed, liquid nitrogen usually means 77 K, which is "much" hotter from >absolute zero. But since statistics of e.g. number of particles at a >given quantum level with energy E is determined by the exponential term >exp(-E/[KT]) (at least in the limit of high T), for many practical uses >T = 77 K is "a lot" better than T = 300 K. It's more like threshold >behavior, so you'd want to justify too much cooling. At least, I've seen >liquid-nitrogen-cooled detectors, but I haven't seen >liquid-helium-cooled detectors (which perhaps speaks more about my >limited experience, but yet).Oh, so your magic conceptual noise canceller had a lower limit? ;)>Perhaps you are correct about physical limits. I haven't been thinking >about that. E.g. there are single-photon detectors, but the problem is, >you won't detect every single photon. More likely you'd lose the most of >them, and only detect a few of the many.The human eye can detect a single photon, btw. It's pretty cool.
Reply by ●June 9, 20162016-06-09
On 09/06/2016 21:43, Eric Jacobsen wrote:> On Thu, 09 Jun 2016 21:27:57 +0300, Evgeny Filatov > <filatov.ev@mipt.ru> wrote: > >> On 09/06/2016 21:09, Eric Jacobsen wrote: >>> On Thu, 09 Jun 2016 20:52:19 +0300, Evgeny Filatov >>> <filatov.ev@mipt.ru> wrote: >>> >>>> On 09/06/2016 15:37, Bob Masta wrote: >>>> >>>> (snip) >>>> >>>>> Hmm, sounds like one of those dither cancellation systems >>>>> where the known dither source can be subtracted at the other >>>>> end. >>>>> >>>>> For that matter, it also sounds like an encryption system >>>>> where the transmitter and receiver are both using a known >>>>> (pseudo) random stream. Without the known encryption >>>>> source, any other receiver gets 100% noise; with the source >>>>> they get 100% signal. >>>>> >>>>> Best regards, >>>>> >>>>> >>>>> Bob Masta >>>>> >>>>> DAQARTA v9.20 >>>>> Data AcQuisition And Real-Time Analysis >>>>> www.daqarta.com >>>>> Scope, Spectrum, Spectrogram, Sound Level Meter >>>>> Frequency Counter, Pitch Track, Pitch-to-MIDI >>>>> FREE 8-channel Signal Generator, DaqMusiq generator >>>>> Science with your sound card! >>>>> >>>> >>>> I was thinking about a system where the sole source of noise is the >>>> transmitter (e.g. you cool the detector of the receiver with liquid >>>> nitrogen, so it's not noisy). And some kind of a feedback loop so that >>>> the transmitter makes use of the noise to transmit less power. It's >>>> still far-fetched or impossible, but I thought it might make for a fun >>>> mind exercise. :) >>> >>> It is interesting. Any signal detector is subject to thermal noise >>> due to electron motion at some non-zero Kelvin temperature (which is a >>> lot colder than liquid nitrogen). So cancelling thermal noise >>> suggests the signal itself can be reduced to the small quanta of an >>> electron or two moving up a shell or two, or something like that. So >>> in the limit as your noise canceller cancels more and more thermal >>> noise, making the entire channel appear to be at 0 Kelvin, you will >>> still have quantization noise associated with the particle quanta. >>> >>> So you can't really make the signal infinitely small in practice, >>> since there is a quantum limit at some point, even if you successfully >>> cancel all of the thermal noise. >>> >>>> Of course, you are correct about the situation when both the receiver >>>> and the transmitter know the dither (or pseudo-random stream), which is >>>> a standard and well-researched case. >>>> >>>> Gene >>>> >>> >> >> Indeed, liquid nitrogen usually means 77 K, which is "much" hotter from >> absolute zero. But since statistics of e.g. number of particles at a >> given quantum level with energy E is determined by the exponential term >> exp(-E/[KT]) (at least in the limit of high T), for many practical uses >> T = 77 K is "a lot" better than T = 300 K. It's more like threshold >> behavior, so you'd want to justify too much cooling. At least, I've seen >> liquid-nitrogen-cooled detectors, but I haven't seen >> liquid-helium-cooled detectors (which perhaps speaks more about my >> limited experience, but yet). > > Oh, so your magic conceptual noise canceller had a lower limit? ;) > >> Perhaps you are correct about physical limits. I haven't been thinking >> about that. E.g. there are single-photon detectors, but the problem is, >> you won't detect every single photon. More likely you'd lose the most of >> them, and only detect a few of the many. > > The human eye can detect a single photon, btw. It's pretty cool. > >Right... just not _each_ single photon (I suppose!). ;) Gene
Reply by ●June 10, 20162016-06-10
On 09.06.2016 18:16, Eric Jacobsen wrote:> >> That >> suggests the SNR limit to uncoded transmissions as low as -3 dB (which >> is already normally impossible). > > -3dB? Checked the Shannon limit lately? ;) > >That might just mean that the proposed scheme is not physically possible. If you make weird assumptions you are going to get weird results... I think that's because there's no way to reduce thermal noise (save for cooling your circuit/whatever). Gene
Reply by ●June 14, 20162016-06-14
I think you can do a lot better than -3dB, by encoding your message into lo= wer order bits of the signal. If we model the signal as a sequence of 16-bit integers, and assume the noi= se is full scale, we can agree that we'll only try to encode one bit of inf= ormation per sample, in the least significant bit. As the omniscient transm= itter, I only have to add enough signal to ensure the correct parity of the= LSB, so my transmission can be 96dB below the noise. It's interesting that (assuming fixed resolution at the receiver) there sti= ll seems to be a tradeoff between transmission power and data rate. But we = can do way better than the Shannon limit. For one thing we can get an error= rate of exactly zero! Thanks for the interesting thought experiment! -Ethan
Reply by ●June 15, 20162016-06-15
On 14.06.2016 21:26, ethan@polyspectral.com wrote:> I think you can do a lot better than -3dB, by encoding your message into lower order bits of the signal. > > If we model the signal as a sequence of 16-bit integers, and assume the noise is full scale, we can agree that we'll only try to encode one bit of information per sample, in the least significant bit. As the omniscient transmitter, I only have to add enough signal to ensure the correct parity of the LSB, so my transmission can be 96dB below the noise. > > It's interesting that (assuming fixed resolution at the receiver) there still seems to be a tradeoff between transmission power and data rate. But we can do way better than the Shannon limit. For one thing we can get an error rate of exactly zero! > > Thanks for the interesting thought experiment! > > -Ethan >You are welcome. The only problem is, that it violates the second law of thermodynamics. If we've managed to reduce the thermal noise, means we've reduced the temperature, and achieved that without spending any energy, and without connecting the Tx or Rx to a colder body. But that violates the principle that "heat cannot spontaneously flow from cold regions to hot regions without external work being performed on the system". I didn't realize that at the time of the writing, otherwise I wouldn't bother to write it down. ;-) Gene
Reply by ●June 15, 20162016-06-15
On Thursday, June 9, 2016 at 5:48:07 AM UTC-4, Evgeny Filatov wrote:> Hello everyone! > > There's a rather bizarre thought experiment, that's hopefully > interesting to discuss. > > Suppose I'm a transmitter which is about to transmit a frame of data. > According to all textbooks, when the frame of data arrives to the > receiver, the data will be superimposed by noise. In the simplest case, > AWG noise. > > Now, suppose I (the transmitter) have managed to acquire support of a > supernatural being, which has provided me with the information about > future noise samples to originate at the receiver side. (That is, > there's an Universe which enables cheating.) > > Now the question is, is it possible to use that information to transmit > less power while keeping the same SNR (or improving it) at the receiver > side? > > In the case of uncoded transmissions it's only needed to cancel the > noise at the receiver while transmitting arbitrarily low signal power. > The benefits are clear. > > However there's something like a fundamental limit which requires to > transmit the power equal to the noise power. Which suggests that > something like Shannon capacity limit is still in place. > > That leads to some questions I don't have ready answers on. > > Would coded transmissions improve the situation? > > Also, what happens to the Shannon capacity limit when noise is known to > the transmitter (but is unknown to the receiver)? > > GeneIn information theory this is called "Channel coding with side information at the transmitter". Look up "Writing on Dirty Paper" by Costa and Cover at Stanford, here's a handy Wikipedia link. https://en.wikipedia.org/wiki/Dirty_paper_coding From here it becomes important whether the side information is available causally or not. The Costa/Cover paper assumes it's not, that's available (infinitely) ahead of time. In this case then the capacity is the same as if the known noise at the transmitter did not exist. All this is an interesting exercise in information theory, so you can have fun up until you try to implement it. And yes, the solution uses coset codes, etc. so a form of coded dithering. If you need coding recommendations I'm sure Steve Pope has a lot to choose from. For a while it was trendy to look at (coset) partitions of LDPC codes, convolutional codes, etc. I probably have a paper or two on this subject too. Now I'm showing my age. Julius
Reply by ●June 15, 20162016-06-15
On Wed, 15 Jun 2016 10:21:59 -0700 (PDT), julius <juliusk@gmail.com> wrote:>On Thursday, June 9, 2016 at 5:48:07 AM UTC-4, Evgeny Filatov wrote: >> Hello everyone! >>=20 >> There's a rather bizarre thought experiment, that's hopefully=20 >> interesting to discuss. >>=20 >> Suppose I'm a transmitter which is about to transmit a frame of data.=20 >> According to all textbooks, when the frame of data arrives to the=20 >> receiver, the data will be superimposed by noise. In the simplest case,= >=20 >> AWG noise. >>=20 >> Now, suppose I (the transmitter) have managed to acquire support of a=20 >> supernatural being, which has provided me with the information about=20 >> future noise samples to originate at the receiver side. (That is,=20 >> there's an Universe which enables cheating.) >>=20 >> Now the question is, is it possible to use that information to transmit= >=20 >> less power while keeping the same SNR (or improving it) at the receiver= >=20 >> side? >>=20 >> In the case of uncoded transmissions it's only needed to cancel the=20 >> noise at the receiver while transmitting arbitrarily low signal power.=20 >> The benefits are clear. >>=20 >> However there's something like a fundamental limit which requires to=20 >> transmit the power equal to the noise power. Which suggests that=20 >> something like Shannon capacity limit is still in place. >>=20 >> That leads to some questions I don't have ready answers on. >>=20 >> Would coded transmissions improve the situation? >>=20 >> Also, what happens to the Shannon capacity limit when noise is known to= >=20 >> the transmitter (but is unknown to the receiver)? >>=20 >> Gene > >In information theory this is called "Channel coding with side information = >at the transmitter". Look up "Writing on Dirty Paper" by Costa and Cover at= > Stanford, here's a handy Wikipedia link. > >https://en.wikipedia.org/wiki/Dirty_paper_coding > >From here it becomes important whether the side information is available ca= >usally or not. The Costa/Cover paper assumes it's not, that's available (in= >finitely) ahead of time. In this case then the capacity is the same as if t= >he known noise at the transmitter did not exist.=20 > >All this is an interesting exercise in information theory, so you can have = >fun up until you try to implement it. And yes, the solution uses coset code= >s, etc. so a form of coded dithering. If you need coding recommendations I'= >m sure Steve Pope has a lot to choose from. For a while it was trendy to lo= >ok at (coset) partitions of LDPC codes, convolutional codes, etc. I probabl= >y have a paper or two on this subject too. Now I'm showing my age.=20 > >JuliusThat's good for non-flat or interference channels, but I thought the OP was looking at flat AWGN, which is a bit harder to treat with side information. DPC was a common example reference when MIMO and S-T codes were being developed for 802.11n. It's funny how much stuff overlaps...as it's akin to water filling from the modulation perspective.
Reply by ●June 15, 20162016-06-15
On 15/06/2016 20:21, julius wrote: (snip)> > In information theory this is called "Channel coding with side information at the transmitter". Look up "Writing on Dirty Paper" by Costa and Cover at Stanford, here's a handy Wikipedia link. > > https://en.wikipedia.org/wiki/Dirty_paper_coding > > From here it becomes important whether the side information is available causally or not. The Costa/Cover paper assumes it's not, that's available (infinitely) ahead of time. In this case then the capacity is the same as if the known noise at the transmitter did not exist. > > All this is an interesting exercise in information theory, so you can have fun up until you try to implement it. And yes, the solution uses coset codes, etc. so a form of coded dithering. If you need coding recommendations I'm sure Steve Pope has a lot to choose from. For a while it was trendy to look at (coset) partitions of LDPC codes, convolutional codes, etc. I probably have a paper or two on this subject too. Now I'm showing my age. > > Julius >Wow! Julius, thanks for pointing that out! Would surely love to read that paper. As I'm not trying to implement a system like that, I'd rather not bother anyone right now. ;-) Gene







