Author Topic: Maximum Practical Datalink Spectral Efficiency  (Read 2432 times)

0 Members and 1 Guest are viewing this topic.

Offline aeronaughtTopic starter

  • Newbie
  • Posts: 2
  • Country: us
Maximum Practical Datalink Spectral Efficiency
« on: June 12, 2019, 06:43:28 pm »
Hey guys!

I have a question that I thought maybe some folks here might be able to help shed some light on:

Q: What is the maximum spectral efficiency that can be reasonably accomplished by modern technology/engineering/science/magic today?


I have a basic understanding of Shannon's Channel Capacity Limit. Ideally, infinite transmit power leads to infinite signal-to-noise, which leads to log2(1+ infinite) spectral efficiency  :-+. Now, as we do not live on the planet Ideal, there will be a number of factors which will prevent us from simply jamming more and more power into larger and larger Tx and Rx antennas, ad-infinitum. Eventually, non-linearities in the transmitter and receiver are going to prevent you from obtaining ever more channel capacity, regardless of the amount of power or antenna gain you attempt to dump into the link.

Unfortunately, I can't find a good source to help me determine where this limit might be.


Assumptions for a random thought experiment I've been conducting with a friend of mine:

Frequency: X band, 8.0 GHz
LOS link: 16 km (10 mi)
Transmitter: 15 W of Tx power, through a 0.1 sq. meter antenna.

Receiver system has a larger number of identical 0.1 sq. meter antennas attached to dedicated ADC's, somewhere between 1 and 200 separate antennas. The system might periodically use some sort of pilot signal to synchronize all receivers, allowing for coherent summation of signal power across all receive antennas. 

The SNR you obtain from the above scenario is somewhat ridiculous, somewhere between approximately 45 and 65 dB, corresponding to a spectral efficiency of around 15-22 bit/Hz. Of course, it would be relatively easy to image even larger numbers. If we used 150 W of Tx power through 1 meter antennas, we can obtain 100 dB of SNR, good for 32 bit/Hz!

Now, I'm sure that someone somewhere using infinite funding could make a 30 bit/Hz datalink out of unobtainum if they really, really wanted to, but what I'm not sure about is the practicality. I'm assuming that most components going into typical RF devices don't have anywhere near the linearity or phase noise required to achieve this kind of performance.

So, assuming a reasonable cost (<$50k per Rx or Tx module), and a reasonable amount of volume (say, 1,000 - 10,000 modules), and a reasonable development budget (say, well less than $100 million dollars), what kind of spectral efficiency could reasonably be obtained with today's technology?

In addition to any answers folks might come up with, I would be supremely grateful if anyone has any books they think might be helpful on this topic.
Specifically, I would love to be able to calculate the expected link performance given actual device datasheets and see where they stack up in this example.

Thanks very much in advance!

Best regards,
Andrew





 

Offline coppice

  • Super Contributor
  • ***
  • Posts: 10031
  • Country: gb
Re: Maximum Practical Datalink Spectral Efficiency
« Reply #1 on: June 12, 2019, 07:24:57 pm »
An important thing to remember is that the channel capacity theorem applies to a signal in AWGN. That covers most natural sources of noise, but unless you are in deep space, or have incredibly tight beam antennae,  there are many sources of human generated noise. Much of that is quite structured. For most practical purposes the maximum power you can use is just below the point where war breaks out, and everyone is upping their transmit power to maintain their throughput in the face of the energy being spilled into their receiver by everyone else. In those situations, like in deep space, where you generally only see AWGN, spectral efficiency is seldom important, as you can use all the spectrum you like, and that wastes less energy that going crazy with the transmit power level.
 

Offline aeronaughtTopic starter

  • Newbie
  • Posts: 2
  • Country: us
Re: Maximum Practical Datalink Spectral Efficiency
« Reply #2 on: June 12, 2019, 07:53:59 pm »
That's an excellent point coppice.

However, we are talking about a theoretical 50 dB of SNR over thermal. I'm not sure how much background noise there will be in band at this frequency, but I would be surprised to see more than 10-20 dB, which still leaves us in a range of spectral efficiency that is seldom explored. How much noise do you expect to see?

Another thought we had is that we have potentially dozens of independent receive channels. Assuming we don't saturate any individual receiver, I would be willing to bet that most of the background noise could be isolated and removed through the use of clever signal processing.
 

Offline ogden

  • Super Contributor
  • ***
  • Posts: 3731
  • Country: lv
Re: Maximum Practical Datalink Spectral Efficiency
« Reply #3 on: June 12, 2019, 07:54:32 pm »
So, assuming a reasonable cost (<$50k per Rx or Tx module), and a reasonable amount of volume (say, 1,000 - 10,000 modules), and a reasonable development budget (say, well less than $100 million dollars), what kind of spectral efficiency could reasonably be obtained with today's technology?

Around 10 bits/Hz, obviously for single TX/RX channel (no MIMO). Check Ceragon microwave radio datasheets. They are pretty much showing what kind of spectral efficiency could reasonably be obtained with today's technology.
 

Offline mark03

  • Frequent Contributor
  • **
  • Posts: 749
  • Country: us
Re: Maximum Practical Datalink Spectral Efficiency
« Reply #4 on: June 13, 2019, 05:14:49 pm »
So, assuming a reasonable cost (<$50k per Rx or Tx module), and a reasonable amount of volume (say, 1,000 - 10,000 modules), and a reasonable development budget (say, well less than $100 million dollars), what kind of spectral efficiency could reasonably be obtained with today's technology?

Around 10 bits/Hz, obviously for single TX/RX channel (no MIMO). Check Ceragon microwave radio datasheets. They are pretty much showing what kind of spectral efficiency could reasonably be obtained with today's technology.

Interestingly, this is almost the same as the final evolution of telephone modems (33.6 kbps in ~ 3 kHz bandwidth; higher speeds used the fact that the underlying network was 64 kbps digital).
 

Offline ogden

  • Super Contributor
  • ***
  • Posts: 3731
  • Country: lv
Re: Maximum Practical Datalink Spectral Efficiency
« Reply #5 on: June 13, 2019, 05:30:04 pm »
Interestingly, this is almost the same as the final evolution of telephone modems (33.6 kbps in ~ 3 kHz bandwidth; higher speeds used the fact that the underlying network was 64 kbps digital).

Yes. Still benchmark. If you are looking where to invest your millions, better look somewhere else. Pretty much everything that can be done with modulation and coding is done already. What's left - MIMO, smaller distances/cells, higher frequencies and combination of all that.
 

Offline TheUnnamedNewbie

  • Super Contributor
  • ***
  • Posts: 1211
  • Country: 00
  • mmwave RFIC/antenna designer
Re: Maximum Practical Datalink Spectral Efficiency
« Reply #6 on: June 14, 2019, 09:44:37 am »
Using MIMO, I believe Ericson have reported efficiencies in the 50 bit/Hz in practical, 1.5 km link setups at E-band (70 GHz).

I have heard people talking about 4096 and even 8192 QAM in research backhaul-point-to-point links, which is about 12-13 bit/Hz.

Regarding linearity and phase noise: I think phase noise might be a bigger issue - if you really have lots of money, you throw DSP at the problem to predistort - both AM-AM and AM-PM.

Nobody really tries to get that much more data in except for the people working at wireline (I seem to recall Bell Labs doign about 20 bit/Hz on their XGfast platform). Keep in mind that the SNR is in the logarithm of the Shannon capacity - to get a bit better you need exponentially more effort. Much, much easier to increase the bandwidth part, either by actually going wider band (pushing to millimeter-waves) or 'virtual' with MIMO.

Getting 3dB more output power requiers 2 times the power comsumption (ideally, but it will be more probably) and will give you a small increase in datarate. Just using a second transmitter/receiver at a different band in theory doubles the datarate for the same doubling in power.

I think esp. at millimeter waves you will seldom see anyone go to 15 W of power for backbone point-to-point (but then again, I have seen research in TWT amplifiers at 100+ GHz to do just that). You get more out of getting a number of transmitters in a MIMO/beamforming array than just wasting all that power on one single transmitter.
The best part about magic is when it stops being magic and becomes science instead

"There was no road, but the people walked on it, and the road came to be, and the people followed it, for the road took the path of least resistance"
 

Offline ogden

  • Super Contributor
  • ***
  • Posts: 3731
  • Country: lv
Re: Maximum Practical Datalink Spectral Efficiency
« Reply #7 on: June 14, 2019, 11:40:19 am »
Using MIMO, I believe Ericson have reported efficiencies in the 50 bit/Hz in practical, 1.5 km link setups at E-band (70 GHz).
I have heard people talking about 4096 and even 8192 QAM in research backhaul-point-to-point links, which is about 12-13 bit/Hz.

Hyperlinks "I believe" and "I have heard" do not work ;) Could you please provide better sources, especially about 50bits/Hz @E-band?

Ceragon already have 4092QAM support in its IP-50 backhaul radios. 4096QAM (12bits/symbol) increase to 8192QAM (13bits/symbol) gives 1.08x capacity gain, less than 1%. Even 2048 to 4096QAM is 12/11 = 1.09x, nothing close to spectacular. To handle higher modulations you need better linearity, lower phase noise, not to mention better SNR. 4096QAM can be used on short backhaul links, but there you better use mm-Wave.
 

Offline coppice

  • Super Contributor
  • ***
  • Posts: 10031
  • Country: gb
Re: Maximum Practical Datalink Spectral Efficiency
« Reply #8 on: June 14, 2019, 11:42:47 am »
Using MIMO, I believe Ericson have reported efficiencies in the 50 bit/Hz in practical, 1.5 km link setups at E-band (70 GHz).
I have heard people talking about 4096 and even 8192 QAM in research backhaul-point-to-point links, which is about 12-13 bit/Hz.

Hyperlinks "I believe" and "I have heard" do not work ;) Could you please provide better sources, especially about 50bits/Hz @E-band?

Ceragon already have 4092QAM support in its IP-50 backhaul radios. 4096QAM (12bits/symbol) increase to 8192QAM (13bits/symbol) gives 1.08x capacity gain, less than 1%. Even 2048 to 4096QAM is 12/11 = 1.09x, nothing close to spectacular. To handle higher modulations you need better linearity, lower phase noise, not to mention better SNR. 4096QAM can be used on short backhaul links, but there you better use mm-Wave.
Did you miss "MIMO"?
 

Offline TheUnnamedNewbie

  • Super Contributor
  • ***
  • Posts: 1211
  • Country: 00
  • mmwave RFIC/antenna designer
Re: Maximum Practical Datalink Spectral Efficiency
« Reply #9 on: June 14, 2019, 11:52:23 am »
Using MIMO, I believe Ericson have reported efficiencies in the 50 bit/Hz in practical, 1.5 km link setups at E-band (70 GHz).
I have heard people talking about 4096 and even 8192 QAM in research backhaul-point-to-point links, which is about 12-13 bit/Hz.

Hyperlinks "I believe" and "I have heard" do not work ;) Could you please provide better sources, especially about 50bits/Hz @E-band?


https://www.ericsson.com/en/press-releases/2019/5/deutsche-telekom-and-ericsson-top-100gbps-over-microwave-link

I can't give you more on the other stuff since it is what I hear from certain industry partners I work with, without necessarily having official statements out there. I work in the millimeter-wave communication industry myself so I do not necessarily get a lot of information like this in public statements.
The best part about magic is when it stops being magic and becomes science instead

"There was no road, but the people walked on it, and the road came to be, and the people followed it, for the road took the path of least resistance"
 
The following users thanked this post: ogden

Offline ogden

  • Super Contributor
  • ***
  • Posts: 3731
  • Country: lv
Re: Maximum Practical Datalink Spectral Efficiency
« Reply #10 on: June 14, 2019, 12:10:26 pm »
Did you miss "MIMO"?

Sorry? What you are talking about? - Fact that I did not retype "MIMO" mentioning 50bits/Hz E-band or what? If yes - then better delete your pointless post now, so I can delete this one.
 

Offline ogden

  • Super Contributor
  • ***
  • Posts: 3731
  • Country: lv
Re: Maximum Practical Datalink Spectral Efficiency
« Reply #11 on: June 14, 2019, 12:24:17 pm »
https://www.ericsson.com/en/press-releases/2019/5/deutsche-telekom-and-ericsson-top-100gbps-over-microwave-link

I can't give you more on the other stuff since it is what I hear from certain industry partners I work with, without necessarily having official statements out there. I work in the millimeter-wave communication industry myself so I do not necessarily get a lot of information like this in public statements.

Thank you, very interesting info. Not so long time ago 1Gbs E-band was "big thing" :D Do you know something about "line of sight MIMO"? Is it so that antennas are so good that they just run 4 parallel cross-polarized beams w/o any actual MIMO DSP math in the modems?
 

Offline coppice

  • Super Contributor
  • ***
  • Posts: 10031
  • Country: gb
Re: Maximum Practical Datalink Spectral Efficiency
« Reply #12 on: June 14, 2019, 02:24:33 pm »
Did you miss "MIMO"?

Sorry? What you are talking about? - Fact that I did not retype "MIMO" mentioning 50bits/Hz E-band or what? If yes - then better delete your pointless post now, so I can delete this one.
The potential of MIMO, given enough room for the antennae, is so huge that 50 bits/Hz just takes effort, but little innovation, to achieve. So far MIMO has mostly been restricted in its effectiveness because the applications so far using it are mostly small devices.
 

Offline TheUnnamedNewbie

  • Super Contributor
  • ***
  • Posts: 1211
  • Country: 00
  • mmwave RFIC/antenna designer
Re: Maximum Practical Datalink Spectral Efficiency
« Reply #13 on: June 14, 2019, 03:34:35 pm »
https://www.ericsson.com/en/press-releases/2019/5/deutsche-telekom-and-ericsson-top-100gbps-over-microwave-link

I can't give you more on the other stuff since it is what I hear from certain industry partners I work with, without necessarily having official statements out there. I work in the millimeter-wave communication industry myself so I do not necessarily get a lot of information like this in public statements.

Thank you, very interesting info. Not so long time ago 1Gbs E-band was "big thing" :D Do you know something about "line of sight MIMO"? Is it so that antennas are so good that they just run 4 parallel cross-polarized beams w/o any actual MIMO DSP math in the modems?

I'm not someone working on MIMO myself, but I have seen reports (again, unfortunately no official statement) of antennas with gain north of 45 dBi on both ends. Things like active stabilizing because the wind sway of the supports brings them out of alignment.
The best part about magic is when it stops being magic and becomes science instead

"There was no road, but the people walked on it, and the road came to be, and the people followed it, for the road took the path of least resistance"
 
The following users thanked this post: ogden

Offline NiHaoMike

  • Super Contributor
  • ***
  • Posts: 9320
  • Country: us
  • "Don't turn it on - Take it apart!"
    • Facebook Page
Re: Maximum Practical Datalink Spectral Efficiency
« Reply #14 on: June 14, 2019, 06:18:34 pm »
I have managed to send 600Mbps through a single pair of telephone line a few hundred feet long using a $20 pair of hacked Homeplug adapters. Those adapters used 12V output amplifiers so there's quite a bit of room to further boost SNR while staying within the low voltage limit. (More for longer range than more bandwidth, so using a higher voltage is more interesting to telecommunications companies than hobbyists.)
I'm not someone working on MIMO myself, but I have seen reports (again, unfortunately no official statement) of antennas with gain north of 45 dBi on both ends. Things like active stabilizing because the wind sway of the supports brings them out of alignment.
Even a fairly small dish antenna like the ones used for residential satellite communications are on the order of 30dBi or more. The really big ones used for radio astronomy can do as much as 80dBi.
https://en.wikipedia.org/wiki/Parabolic_antenna#Gain
Cryptocurrency has taught me to love math and at the same time be baffled by it.

Cryptocurrency lesson 0: Altcoins and Bitcoin are not the same thing.
 

Offline ogden

  • Super Contributor
  • ***
  • Posts: 3731
  • Country: lv
Re: Maximum Practical Datalink Spectral Efficiency
« Reply #15 on: June 14, 2019, 06:47:43 pm »
Even a fairly small dish antenna like the ones used for residential satellite communications are on the order of 30dBi or more. The really big ones used for radio astronomy can do as much as 80dBi.

My question was - are antenna beams narrow enough so they can get away w/o any actual MIMO DSP math in the modems? Antenna gain alone does not answer this question. What we need to know is radiation pattern at 1.5km distance.

The potential of MIMO, given enough room for the antennae, is so huge that 50 bits/Hz just takes effort, but little innovation, to achieve.

DSP engineering quote of the year :) Search patents for just MIMO keyword. If this is not convincing enough then I simply give up.
« Last Edit: June 14, 2019, 10:50:39 pm by ogden »
 

Offline TheUnnamedNewbie

  • Super Contributor
  • ***
  • Posts: 1211
  • Country: 00
  • mmwave RFIC/antenna designer
Re: Maximum Practical Datalink Spectral Efficiency
« Reply #16 on: June 15, 2019, 07:39:11 am »
Even a fairly small dish antenna like the ones used for residential satellite communications are on the order of 30dBi or more. The really big ones used for radio astronomy can do as much as 80dBi.

My question was - are antenna beams narrow enough so they can get away w/o any actual MIMO DSP math in the modems? Antenna gain alone does not answer this question. What we need to know is radiation pattern at 1.5km distance.

The potential of MIMO, given enough room for the antennae, is so huge that 50 bits/Hz just takes effort, but little innovation, to achieve.

DSP engineering quote of the year :) Search patents for just MIMO keyword. If this is not convincing enough then I simply give up.

I seem to recall a rule of thumb that 1 degree beamwidth equals about 46 dB with a parabolic dish. Going by that assumption, the antennas would have to be about 20 meters away from each other at 1 km to have 6 dB (assuming both antennas have that gain) less received power than the neighbouring antennas, so I am going to go with these needing quite a lot of DSP to get going. It could be that they actually get even narrower beamwidths, which would help, but I am not sure. (I don't know if the 45 dBi I was given included efficiency of the antenna etc - which at 70 GHz is going to be nothing to write home about).

Would be interesting to find out, but I don't know how much these people can share. I'll look up some publications by those groups later to see if that has any info.



The thing that is impressive here is that, to my knowledge, these are not big parabolic dishes, but arrays.
The best part about magic is when it stops being magic and becomes science instead

"There was no road, but the people walked on it, and the road came to be, and the people followed it, for the road took the path of least resistance"
 
The following users thanked this post: ogden

Offline coppice

  • Super Contributor
  • ***
  • Posts: 10031
  • Country: gb
Re: Maximum Practical Datalink Spectral Efficiency
« Reply #17 on: June 15, 2019, 01:38:01 pm »
Even a fairly small dish antenna like the ones used for residential satellite communications are on the order of 30dBi or more. The really big ones used for radio astronomy can do as much as 80dBi.

My question was - are antenna beams narrow enough so they can get away w/o any actual MIMO DSP math in the modems? Antenna gain alone does not answer this question. What we need to know is radiation pattern at 1.5km distance.

The potential of MIMO, given enough room for the antennae, is so huge that 50 bits/Hz just takes effort, but little innovation, to achieve.

DSP engineering quote of the year :) Search patents for just MIMO keyword. If this is not convincing enough then I simply give up.

I seem to recall a rule of thumb that 1 degree beamwidth equals about 46 dB with a parabolic dish. Going by that assumption, the antennas would have to be about 20 meters away from each other at 1 km to have 6 dB (assuming both antennas have that gain) less received power than the neighbouring antennas, so I am going to go with these needing quite a lot of DSP to get going. It could be that they actually get even narrower beamwidths, which would help, but I am not sure. (I don't know if the 45 dBi I was given included efficiency of the antenna etc - which at 70 GHz is going to be nothing to write home about).

Would be interesting to find out, but I don't know how much these people can share. I'll look up some publications by those groups later to see if that has any info.



The thing that is impressive here is that, to my knowledge, these are not big parabolic dishes, but arrays.
The cellular industry has driven some pretty impressive developments in what can be achieved with a compact low cost antennae elaborately formed from simple stamped sheets. That said. the need for extensive DSP, and tight constraints on receiver front end performance, are a given for this type of system. Forget any idea of focussing an RF beam at an antenna so well that it won't significantly spill into adjacent ones. The antennae needed to do that at such low frequencies are largely impractical, even at the high end of the RF band. Even at optical frequencies, where its easy to make a diffraction limited beam, the atmosphere can still fuzz it up enough to cause trouble at the receiver. However, even small amounts of physical isolation can substantially reduce the processing problems at the receiver, so they have a big payoff.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf