Great series Emil, thanks for all this lecturing. For future insipration series, Iam thinking about relation between QAM types, band ranges and SNR in terms of beamforming outputs. I am thinking about, if we need to adapt modulation methods for future high band developments. Have a good day.
Thanks you very much. A more general question - in your lecture we speak about 3 benefits of MIMO systems: - beamforming gain - diversity gain - spatial multiplexing gain Can we say that there is a 4th gain, namely the increase of the effective antenna area (reception case)? Or is that fact already in one of the 3 aforementioned benefits? What do you think?
What you describe is what I would call receive beamforming gain. By combining the received signals coherently from them, you get an SNR gain that is proportional to the number of antennas, and therefore proportional to the total effective antenna area.
Yes, the same thing happens at the receiver side, with the only difference that the signals at the different receive antennas are not added up over the air, but in the receiver processing.
Dear Professor, very nice and educated video. I like the moment when you place cub wrapped in foil which creates multipath causing distortion in constructive interference. It illustrates this effect very well. In real propagation environment when we have many multipaths (and especially large angular spread) the effective antenna array gain is reduced when we have classical type (analog) of beamforming. It requires to have digital beamforming and more advanced beamforming techniques to work efficiently in real channel. I am not sure if you can illustrate it in this setup.
Since this is an analog beamforming setup, it isn’t easy to show the benefit for digital beamforming using it. In principle, one could measure the received signal power for the individual antennas in SISO mode and then add them up to represent what could theoretically be achieved using digital beamforming with maximum ratio transmission. As I talked about in the first video, the transmit power per antenna varies as I turn on more antennas in this setup, so this makes it hard to make a fair comparison between this setup and ideal digital beamforming.
Thank you so much Dear Prof. Emil; Please, I would to ask you question if you have idea about it; I read papers that focused in using horn antennas operating at millimeter wave bands such as W-band and so on; my question is; what's the real usage scenarios for using such types of antennas? For example, at the base station and/or mobile end users microstrip patch antennas are used (omnidirectional or phased array for beam-forming); I confused about that! Please, if you see my this question, I need precise response from your expertise via on the field applications, or practical ideas! Best regards;
Horn antennas are designed to have a specific strong directivity, so they are useful for fixed installations where you know where the signal is coming from or where it should go. This might be a fixed communication link, measurements where you want to determine signal strengths from specific directions, or radar guns. In a base station, you typically want a flat array instead of a horn for deployment reasons, and you want the directivity to be controllable - so it is better to create the directivity through superposition of signals from an array of weakly directive antennas than to have one antenna with fixed directivity.
Thank you for the wonderful video series on antenna array beamforming measurements. At mmWave bands, is there any specific benchmark for beamforming side lobe level? or SLL less than -10 dB is a good value generally for mmwWave arrays?
Yes, I think so. When using a ULA, you typical may get an SLL of around -12 dB. I’m not so concerned about sidelobes, since the interference will anyway be diffused by the propagation environment. But it sometimes matter.
Yes, it is the beam pattern in the far-field that is measured and shown. The key properties in the far-field is that the beam pattern only depends on the angle, not the distance. The Fraunhofer distance, measured based on the total size of the array, is an indicator of this. The array has the width 2*lambda in the horizontal plane, so the Fraunhofer distance becomes 2*2^2*lambda = 8 cm.
Thanks for all of your interesting videos. A more generel question: in mMIMO, how are these "orthogonal" pilots sent? How is the orthogonality achieved? Different frequencies for every UE? Same f, but time shifted? I read often about orthogonal pilots.
Yes, orthogonality can be achieved by sending the pilots on different OFDM subcarriers (frequencies) or at different times. However, it can also be achieved by sending sequences that span the same time/frequency resources. For example, one pilot can be +1 +1 and the other can be +1 -1. The inner product of these sequences is zero, which is the definition of orthogonality. It is further explained in Chapter 3 of my book Massive MIMO networks, massivemimobook.com
Hi, a more general question. When we have 2 cases 1. A very dominant LOS path and 2. No LOS path but with 2 very good reflectors: is it possible to get a higher bit rate in the second case? ( The first case uses beamforming and the second uses spatial multiplexing over the 2 paths) Or will the direct path alsways provide a higher bit rate? Thanks a lot.
Yes, you can get a better bit rate in the second case, if the SNR is sufficiently high. If the SNR is high and similar in the two cases, the second case will achieve almost twice as high rate.
Yes, one can control both phase and amplitude in each antenna branch. We only modified the phases in this video series. Do you have suggestions of some tapering scheme that could be tested/verified in a setup like this?
@@WirelessFuture thanks for response. Some tapering schemes could be tested like Chebyshev. Is it possibile to setup this configuration to have larger array like 8x8?
Hello Prof., do you have a good link about "oversampling in MIMO"? I read that it is used for improvements, but I don't know anything about that topic . Thank you very much.
Hi! I don’t have any reference in mind for that. The main point is that the hardware is nonlinear and the analog filters are not ideal, so the received signal has a slightly wider bandwidth than intended. With oversampling, you have a larger ability to suppress these things by digital receiver processing.
When you say "it's more directive than it looks", that made me think the side-lobes would also be stronger than they appear. That is, i'm comparing lobes to nulls. But you're not comparing lobes to nulls. You're comparing larger lobes to smaller lobes.
Many thanks. You give an equation here. That equation refers to what single element type (dipol, minopol, patch,..)? I think the type of the single elements determine the equation(?) Thanks.
Yes, the equation “4*cos(phi)” refers the radiation pattern of a single element, of a hypothetical kind. It should be replaced with another expression that represents the particular type that is used in the array.
There is no RIS in this video. However, there are many repositories on my GitHub that simulate RIS/IRS-aided communications. It is based on deriving performance equations and then simulating it, so nothing complicated. github.com/emilbjornson?tab=repositories
Hi, when you have 8 Antennas and want to have 2 beams in different directions, would the System use all 8 Antennas to create 2 Small beams or would the System use 4 Antennas for 1 beam and the other 4 Antennas for the 2 beam?
All 8 antennas to create 2 small beams, because this makes the beams twice as narrow and leads to twice the received power compared to the other option.
@@WirelessFuture Hi, a last question to that point. In the baseband the signal is created for the 2 beams. The power off the transmitter off the 8 Antennas is limited. Is the beam forming gain better with the 1/2 power for every beam.
@@RenoBlade2The received signal power is proportional to “transmit power per beam * number of antennas”. So when transmitting two beams, the transmit power per beam will reduce by 1/2 since the total power is limited, but the beamforming gain remains proportional to the number of antennas. If you instead transmit each of the beams using only half of the antennas, the received signal power per beam will reduce by 1/2 * 1/2 = 1/4.
When you say db representation suggests less directivity than actual, do you mean the main lobe should look BIIGGER than the side lobes on a linear scale? Then why not use a linear scale?
The peaks of the side lobes are around 13 dB (20 times) weaker than the main lobe. So if you would use decibel scale, they are barely visible. The reason to use decibel scale is that the power variations are also huge in the wireless communication, so a relatively weak side lobe can still interfere with other systems.
@@johnaweiss The frequency spectrum is normally reused by many different systems, and there are no strict boundaries between WiFi networks or different cells in a cellular network, so it is a good practice to try to spread out the interference to avoid pointing a side lobe towards another system. Yes, the dB-scale remind us about that also seemingly small gains can have a substantial impact on other systems. For example, if 20 antennas you might get a 20 times (13 dB) beamforming gain in the main lobe. The largest side lobe is roughly 13 dB weaker than the main lobe when using uniform linear arrays (irrespective of the number of antennas). Hence, the side lobe will be equally strong as if we transmitted with a single antenna. The beamforming gain is lost but the interference can still be substantial.
@@johnaweiss The array in the video is a uniform linear array with 4 antennas (4 columns). Uniform linear arrays are effective for beamforming, however, their side lobes are ≈13 dB weaker than the main lobe. Other arrays give similar side lobe levels so it isn't a good or bad thing, but a consequence of physics. If we add more antennas, (e.g., from 4 to 20), both the main lobe and the large side lobe become equally much stronger. However, their relative difference is always 13 dB so more antennas is always better, but there will always be some signal leakage in undesired directions.
![Maximum ratio and zero-forcing beamforming [Part 4, Fundamentals of mmWave communication]](http://i.ytimg.com/vi/w46c0QhByPg/mqdefault.jpg)
Hi, I would like to thank you for those videos!!
Great series Emil, thanks for all this lecturing. For future insipration series, Iam thinking about relation between QAM types, band ranges and SNR in terms of beamforming outputs. I am thinking about, if we need to adapt modulation methods for future high band developments. Have a good day.
Thanks you very much. A more general question - in your lecture we speak about 3 benefits of MIMO systems:
- beamforming gain
- diversity gain
- spatial multiplexing gain
Can we say that there is a 4th gain, namely the increase of the effective antenna area (reception case)? Or is that fact already in one of the 3 aforementioned benefits? What do you think?
What you describe is what I would call receive beamforming gain. By combining the received signals coherently from them, you get an SNR gain that is proportional to the number of antennas, and therefore proportional to the total effective antenna area.
Thanks for another great tutorial. I have a question. Can we use beamforming methods on the receiver sign as well?
Yes, the same thing happens at the receiver side, with the only difference that the signals at the different receive antennas are not added up over the air, but in the receiver processing.
Dear Professor, very nice and educated video.
I like the moment when you place cub wrapped in foil which creates multipath causing distortion in constructive interference. It illustrates this effect very well. In real propagation environment when we have many multipaths (and especially large angular spread) the effective antenna array gain is reduced when we have classical type (analog) of beamforming. It requires to have digital beamforming and more advanced beamforming techniques to work efficiently in real channel. I am not sure if you can illustrate it in this setup.
Since this is an analog beamforming setup, it isn’t easy to show the benefit for digital beamforming using it. In principle, one could measure the received signal power for the individual antennas in SISO mode and then add them up to represent what could theoretically be achieved using digital beamforming with maximum ratio transmission.
As I talked about in the first video, the transmit power per antenna varies as I turn on more antennas in this setup, so this makes it hard to make a fair comparison between this setup and ideal digital beamforming.
Thank you for response. Right, digital BF in this setup is complicated.
Thank you so much Dear Prof. Emil;
Please, I would to ask you question if you have idea about it;
I read papers that focused in using horn antennas operating at millimeter wave bands such as W-band and so on; my question is; what's the real usage scenarios for using such types of antennas?
For example, at the base station and/or mobile end users microstrip patch antennas are used (omnidirectional or phased array for beam-forming); I confused about that!
Please, if you see my this question, I need precise response from your expertise via on the field applications, or practical ideas!
Best regards;
Horn antennas are designed to have a specific strong directivity, so they are useful for fixed installations where you know where the signal is coming from or where it should go. This might be a fixed communication link, measurements where you want to determine signal strengths from specific directions, or radar guns. In a base station, you typically want a flat array instead of a horn for deployment reasons, and you want the directivity to be controllable - so it is better to create the directivity through superposition of signals from an array of weakly directive antennas than to have one antenna with fixed directivity.
Thank you for the wonderful video series on antenna array beamforming measurements.
At mmWave bands, is there any specific benchmark for beamforming side lobe level? or SLL less than -10 dB is a good value generally for mmwWave arrays?
Yes, I think so. When using a ULA, you typical may get an SLL of around -12 dB. I’m not so concerned about sidelobes, since the interference will anyway be diffused by the propagation environment. But it sometimes matter.
Nice video! Is there a minimum distance for the beamforming concept to work like you explained here?. Is the Fraunhofer distance relevant here?
Yes, it is the beam pattern in the far-field that is measured and shown. The key properties in the far-field is that the beam pattern only depends on the angle, not the distance. The Fraunhofer distance, measured based on the total size of the array, is an indicator of this. The array has the width 2*lambda in the horizontal plane, so the Fraunhofer distance becomes 2*2^2*lambda = 8 cm.
Thanks for all of your interesting videos. A more generel question: in mMIMO, how are these "orthogonal" pilots sent? How is the orthogonality achieved? Different frequencies for every UE? Same f, but time shifted?
I read often about orthogonal pilots.
Yes, orthogonality can be achieved by sending the pilots on different OFDM subcarriers (frequencies) or at different times. However, it can also be achieved by sending sequences that span the same time/frequency resources. For example, one pilot can be +1 +1 and the other can be +1 -1. The inner product of these sequences is zero, which is the definition of orthogonality. It is further explained in Chapter 3 of my book Massive MIMO networks, massivemimobook.com
Hi, a more general question. When we have 2 cases 1. A very dominant LOS path and 2. No LOS path but with 2 very good reflectors: is it possible to get a higher bit rate in the second case? ( The first case uses beamforming and the second uses spatial multiplexing over the 2 paths)
Or will the direct path alsways provide a higher bit rate? Thanks a lot.
Yes, you can get a better bit rate in the second case, if the SNR is sufficiently high. If the SNR is high and similar in the two cases, the second case will achieve almost twice as high rate.
Dear Professor, is it possible to control the amplitude in this setup to test tapering and different types of BF techniques?
Yes, one can control both phase and amplitude in each antenna branch. We only modified the phases in this video series.
Do you have suggestions of some tapering scheme that could be tested/verified in a setup like this?
@@WirelessFuture thanks for response. Some tapering schemes could be tested like Chebyshev. Is it possibile to setup this configuration to have larger array like 8x8?
Hello Prof.,
do you have a good link about "oversampling in MIMO"?
I read that it is used for improvements, but I don't know anything about that topic .
Thank you very much.
Hi! I don’t have any reference in mind for that. The main point is that the hardware is nonlinear and the analog filters are not ideal, so the received signal has a slightly wider bandwidth than intended. With oversampling, you have a larger ability to suppress these things by digital receiver processing.
When you say "it's more directive than it looks", that made me think the side-lobes would also be stronger than they appear. That is, i'm comparing lobes to nulls. But you're not comparing lobes to nulls. You're comparing larger lobes to smaller lobes.
Yes, we compare the main lobe to side lobe, and want to emphasize the decibel scale that makes the side lobes look stronger than they actually are.
Many thanks.
You give an equation here. That equation refers to what single element type (dipol, minopol, patch,..)?
I think the type of the single elements determine the equation(?)
Thanks.
Yes, the equation “4*cos(phi)” refers the radiation pattern of a single element, of a hypothetical kind. It should be replaced with another expression that represents the particular type that is used in the array.
Hi Emil, How did you simulate RIS? Is it possible to do it in the MATLAB?
There is no RIS in this video. However, there are many repositories on my GitHub that simulate RIS/IRS-aided communications. It is based on deriving performance equations and then simulating it, so nothing complicated.
github.com/emilbjornson?tab=repositories
@@WirelessFuture I follow your RIS videos too. I anticipated a quick reply on a recent video. Thank You.
Hi, when you have 8 Antennas and want to have 2 beams in different directions, would the System use all 8 Antennas to create 2 Small beams or would the System use 4 Antennas for 1 beam and the other 4 Antennas for the 2 beam?
All 8 antennas to create 2 small beams, because this makes the beams twice as narrow and leads to twice the received power compared to the other option.
@@WirelessFuture Hi, a last question to that point. In the baseband the signal is created for the 2 beams. The power off the transmitter off the 8 Antennas is limited. Is the beam forming gain better with the 1/2 power for every beam.
@@RenoBlade2The received signal power is proportional to “transmit power per beam * number of antennas”. So when transmitting two beams, the transmit power per beam will reduce by 1/2 since the total power is limited, but the beamforming gain remains proportional to the number of antennas. If you instead transmit each of the beams using only half of the antennas, the received signal power per beam will reduce by 1/2 * 1/2 = 1/4.
When you say db representation suggests less directivity than actual, do you mean the main lobe should look BIIGGER than the side lobes on a linear scale? Then why not use a linear scale?
The peaks of the side lobes are around 13 dB (20 times) weaker than the main lobe. So if you would use decibel scale, they are barely visible. The reason to use decibel scale is that the power variations are also huge in the wireless communication, so a relatively weak side lobe can still interfere with other systems.
@@WirelessFuture Which other systems? So we use db to remind us to attend to the effects of the side lobes?
@@johnaweiss The frequency spectrum is normally reused by many different systems, and there are no strict boundaries between WiFi networks or different cells in a cellular network, so it is a good practice to try to spread out the interference to avoid pointing a side lobe towards another system.
Yes, the dB-scale remind us about that also seemingly small gains can have a substantial impact on other systems. For example, if 20 antennas you might get a 20 times (13 dB) beamforming gain in the main lobe. The largest side lobe is roughly 13 dB weaker than the main lobe when using uniform linear arrays (irrespective of the number of antennas). Hence, the side lobe will be equally strong as if we transmitted with a single antenna. The beamforming gain is lost but the interference can still be substantial.
@@WirelessFuture Confused. You're describing 20 antennas vs uniform linear arrays? And uniform linear arrays are ineffective for beamforming?
@@johnaweiss The array in the video is a uniform linear array with 4 antennas (4 columns). Uniform linear arrays are effective for beamforming, however, their side lobes are ≈13 dB weaker than the main lobe. Other arrays give similar side lobe levels so it isn't a good or bad thing, but a consequence of physics.
If we add more antennas, (e.g., from 4 to 20), both the main lobe and the large side lobe become equally much stronger. However, their relative difference is always 13 dB so more antennas is always better, but there will always be some signal leakage in undesired directions.