While researching 5G, I came across massive MIMO. Thing is to understand massive MIMO, I need to understand MIMO, which I am somewhat confused about.

First thing I'm confused about is the spatial multiplexing. I understand the base station can send two signals with the same frequency at the same time in multiple different directions and they won't interfere because they take different paths, but how do they not interfere when they reach the mobile? For instance, if my phone has two antennas, my phone is pretty small so how won't the signal interfere?

Under that question is another question. I read some stuff about signal vectors being orthogonal and such, but I have no idea what that means. What are the vectors? How are they orthogonal? Also don't understand the matrix \$H\$ or \$H^{-1}\$ is that's used to compute the CSI using pilot signals, very confused about all the equations I see.

Finally I'm kind of confused as to how the capacity is increased in MU-MIMO. Let's say I have 2 antennas at the base station and two users. Let's say transmitting each bit takes 1 millisecond. If I have 8 bits to transmit to each user, then using MIMO it will take me 4 milliseconds to transmit the data to user 1 (since each antenna sends 4 bits) and 4 milliseconds to transmit the data to user 2, so a total of 8 milliseconds. On the other hand, if I dedicated 1 antenna for user 1 and 1 antenna for user 2, it would still take me 8 milliseconds to transmit the data for both, and that's better because user 2 doesn't have to wait... so where is the increase?

Do you guys know any books or articles that explain massive MIMO comprehensively but also very easy to understand? Most of the stuff I read is very confusing. I'm only an undergrad ECE student, I have a hard time making sense of academic papers published about this topic.

Hope this is the right site to post these questions. Sorry if anything is not clear, I'll try to clear things up, and thanks in advance.

  • \$\begingroup\$ Example: sins are orthogonal to cosines. Thus Amplitude modulation of a sin and of a cosine achieves DOUBLE the datarate, yet the matched-filters have zero correlation. This requires exact 0degree and 90degree mixers and LOs. \$\endgroup\$ Apr 27, 2017 at 2:53
  • \$\begingroup\$ @analogsystemsrf I can see why that's the case but 1) how does that work for frequency modulation? and 2) besides sine and cosine how can you have any more orthogonal signals? As far as I understand all the pilot signals are supposed to be orthogonal, so I'm guessing it's more than 2. EDIT: I think these would be answered if I understood OFDM \$\endgroup\$
    – J. Doe
    Apr 27, 2017 at 3:00
  • \$\begingroup\$ GSM uses symbols that are orthogonal: either 1/2 sincycle or 1 full sincycles. Sketch these and you'll see their correlation is zero, hence they are orthogonal. These 2 basis functions, plus their flipped-polarity versions, form the 4 symbols used in GSM. Regarding OFDM, consider 1Hz and 2Hz and 3Hz and 4Hz and ........, all exactly N# cycles in a symbol time. They all are orthogonal to each other. \$\endgroup\$ Apr 27, 2017 at 3:11
  • \$\begingroup\$ I'm not expert, but with respect to the performance increase of MU-MIMO over SU-MIMO, I believe part of the advantage comes from the ability of the base station to provide multiple sets of spatial streams per base station. For example, a 10x10 base station could simultaneously devote 3x3 streams to one device, 2x2 to another, 2x1 to another and still have 3x4 left for new clients. This is as opposed to currently where a 10x10 device would be downgraded to a 3x3 device when transmitting to that device. \$\endgroup\$
    – lswim
    Apr 27, 2017 at 3:22

1 Answer 1


I'm going to try and answer the latter half of your question because I kept on running out of comment room:

MU-MIMO offers improvement because most mobile devices (called User Equipment (UE) in most documents I've seen) don't support as many spatial streams as the base station. In SU-MIMO, only one UE can transmit at a time. So if you have a 10x10 (10 upload x 10 download) base station, and a 2x2 client, the base station can only use two of its spatial streams. This difference is illustrated from this white paper:

R&S Figure 16 R&S Figure 17

So to return to your example, if both UEs and base stations are 2x2, your logic holds. However, if each UE is 1x1 OR if the base stations is 4x4 then MU-MIMO is twice as fast; 8 vs 16 ms for the former and 4 vs 8 ms for the latter.

As someone with a passing interest in MIMO, I'm not sure I'm qualified to take a stab at the mathematics of MIMO, however, I highly recommend you read white papers or application notes from manufacturers of chips or test equipment first. Although they're written to encourage you to buy their product, they often give more background than academic papers because the people buying the equipment are often not experts (yet):

  • \$\begingroup\$ +1 thank you very much for clearing that up and for the resources. I'll wait to see if there are more answers before selecting the answer. \$\endgroup\$
    – J. Doe
    Apr 27, 2017 at 14:01

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.