Diagnosing speed limitations right next to towers

[xdavidx]

Are there any AT commands to find the MIMO status of LTE antennas? This is for an EM7565 modem in a WE1326 router that I just set up.

I'm using some small omni antennas for testing before my directional antennas arrive. I've driven to different towers with the router being powered on an inverter to see which bands are being used and what kinds of speeds I'm able to get. The speeds were slower than I was expecting for being right next to the towers, so I'm trying to determine if this is a result of:

1. A general limitation of the antennas.
2. The modem not operating them in MIMO mode.
3. The physical separation (width of the WE1326 case) not being far enough for MIMO.
4. A speed limit of the tower.
5. Some setting in the router/modem.

I'd be inclined to think it was a tower limitation, but I'm seeing the same approximate peek speed in multiple towers (at lesat 4) that I've tested -- about 50 to 60 Mbps download.

I am testing these over the wifi connection of the router with an android tablet. And taking into account non-optimal wifi on these devices, I'm using 802.11AC to connect to the 5.8 GHz wifi radio of the router. 802.11n on the 2.4GHz wifi radio becomes a bottleneck at these speeds. I've tested two 802.11AC devices talking to each other through the 5.8GHz radio on the router and I'm able to get anywhere from 120 to 200 Mbps. And that is with 2 devices competing with each other over the same frequency. If only one device was talking to a hardwired device or talking to a device over a different frequency, the speeds would be higher. So, in theory, my cellular speed tests over the 5.8 GHz wifi radio shouldn't be the bottleneck, but I will be trying a hardwired laptop at some point.

I also had some bottlenecks with using the google search web page speed test (search for "speed test" and it appears above the search results), so I switched to using the speedtest.net app. I've been able to get up to 130 Mbps from one of the towers using this app on a Pixel 1 cell phone (although that phone was on T-Mobile instead of AT&T, which is what I'm using for the router).

stats_right_next_to_tower.png

I also tested a block away with a clear line of sight, since being right under the tower isn't the optimal location. I suppose even a block away isn't optimal, since other tower interference can creep in easier and I'm not at the height of the transmitters, but the phone was able to do it and the phone has similar specs to the router modem (CAT-12 and 3x Carrier Aggregation).

In theory, even at 64 QAM, I should be able to get 300 Mbps out of 40 MHz of bandwidth with 2x2 MIMO. And the modem is capable of 256 QAM.

I suppose other possibilities are that the AT&T transmitters aren't doing MIMO on their end. SISO on 40 MHz of bandwidth would still be 150 Mbps.

Of course, this is a real world test, with some level of congestion on the towers too. However, with all towers exhibiting the same max speed, I'm thinking there is more to it than that. Let's say I should be able to get 50% of theoretical in a live test. That would mean 75 Mbps SISO and 150 Mbps MIMO with 40 MHz of bandwidth with 64 QAM. I'm rarely hitting above 60 Mbps.

It seems to me that I should be getting about double what I'm getting (which would be closer to what the phone was able to get), which leads me to believe that MIMO isn't being utilized.

Thoughts?

[JimHelms]

I would test it using an Ethernet cable from the router LAN to PC.

Your stats above show you are receiving MIMO and 3 x Carrier aggregation.

The RxM is the Main antenna port on the modem, and the RxD is the AUX antenna port. Both are receiving the tower signal on all three carriers.

Overall, your signals and quality thereof, look good. May also be tower congestion.

[xdavidx]

Thanks for the feedback, Jim. Sorry, I should have been more specific. I was talking about MIMO spatial multiplexing vs MIMO diversity. I sort of verified that it is doing multiplexing today by taking one of the antennas off (didn't take the router apart and disconnect from the modem, just unscrewed from the case) and saw a performance difference.

The problem was either wifi or congested towers when I tested on Monday, as you surmised. Today, I jumped through a lot of hoops getting ethernet working in a stable way to a laptop (I'll put more details on that in my tutorial that I have yet to post). I then took the whole rig in the SUV to 4 different towers and ran over 200 speed tests and at!gstatus? commands over a period of hours.

The short description of the results is that I was able to get over 100 Mbps download on 3 of the 4 towers, multiple times, with a peak of 145 Mbps. They ranged from an aggregated (3CA) bandwidth of 25 MHz to 40 MHz.

145mbps-2019-05-29 at 5-22-26 PM.jpeg

Towers in smaller towns performed better, with the exception of the last one, which is out in the country (but still some number of houses close by). However, that one was tested well after people were home from work, so it is possible there were just more people on that one by the time I got around to testing it. I could only get 60+ Mbps from that one, peak. That is also the one with the lowest aggregated bandwidth of 25 MHz across 3 carriers.

I also hit my deprioritization limit on my AT&T account in the middle of testing today, so a congested tower may affect me in more ways than just radio contention.

The variables that changed from Monday vs today (Wednesday) are:

1. Using 1 Gbps ethernet connected to a laptop for most tests instead of 802.11AC on a tablet. (more on that below)
2. Different day and different time of day.
3. Speedtest.net Windows app running on Windows 10 instead of running on an android tablet.
4. 5.8 GHz external antennas instead of 2.4 GHz. (more on that below, and this has nothing to do with the laptop speeds over ethernet)

I was able to get the same speeds today on the tablet as the laptop and the link speed was showing 433 Mbps most of the time for the tablet. I originally had the 2 LTE antennas taking the place of the 5.8 GHz antennas and the SMA ends for the 5.8 GHz antennas wrapped in tape inside the case, sitting right next to each other. 5.8 GHz still worked fine and I was able to send 120-200 Mbps from one device to another with them like this, but that was with those devices sitting inches from the router.

When I was doing my testing on Monday in the SUV, the tablet was a few feet from the router, so the link speed could have been quite a bit less and that could have been the limiting factor. Without antennas hooked up and probably without MIMO working properly on the 5.8 GHz radio (due to no antennas plugged in and the SMA ports right next to each other), the signal would have been quite a bit weaker every foot further away from the router.

Between the first round of testing and today, I swapped things around so that the 2.4 GHz SMA connectors are inside the case and the 5.8GHz connectors are once again through the case, with the external antennas attached to them.

However, Monday was a holiday, with many people home during the middle of the day, so that could very well be the reason for the slower speeds too.

What my tower map looks like, with my house in the middle (not all are AT&T and I have yet to verify exactly which ones are, other than those I've sat right next to for testing):

towers_FL_WY_ST_MA_20190530.jpg

I tested a tower in a bigger town first today (A), then a smaller town (B), then a smaller town yet (C) (although those 2 towers are along an interstate freeway too, so passers by could have been hopping on and off too). I then went back to the tower in the bigger town (A) after people would have been home from work and it was slower again, similar to what I saw on Monday.

The main thing is that I was able to achieve speeds closer to what I thought I should be able to get. We'll see what directional antennas do for filtering out signals from other directions and possibly being more efficient with MIMO multiplexing. I'll test the new antennas near the towers as well as from my house. (A) isn't a clear line of site, but the other 3 are, not counting trees.

[swwifty]

A few things to note here on your testing.

1. If you have clear LOS to a tower, spatial multiplexing is basically impossible. You want to actually have some obstruction(s) to help separate the signals and hence double your throughput. This is why you might be seeing slower download speeds than expected.

2. RSRQ is a value that shows cell tower load, but it is only accurate when your tx power is 0. If RSRQ is lower (in negativity that is) than -10.6 than the tower is highly congested. -10.6 represents all resource blocks in LTE being utilized. Ideally you'd see something like -7 or -8 when tx power is 0. If you see -16 when tx power is 0, then the tower is very busy.

[xdavidx]

Thanks for the feedback, swwifty.

I did see a big decrease in performance with one antenna disconnected (from the SMA connector, not from the modem), so I'm pretty sure it is doing spatial multiplexing. Now, how well it is doing it is a good question. I'd have to disconnect from the modem and run a lot of tests to see if it is averaging half speed. One thing I've found is that, at least in my environment, where there are many towers around, results are highly variable (both the speed tests and the modem stats).

I watched the video you linked to in one of your posts where they showed -10.6 (I thought it was -10.8) as being a fully loaded system with no interference. My understanding was that once you add in interference from other towers operating on the same frequencies, then it is a function of the two. So you could have no other towers and -10.8 would mean your tower is maxed out with connections. Or, your tower could be half loaded, but you could have interference from other towers, and you could still have that -10.8 value (or lower, meaning higher absolute value). Or you could have both a fully loaded tower and interference and have a lower value. So I've been watching that number to try to gauge things, but since the other towers can play a role, it is difficult to use that number to understand tower saturation. If my understanding above is correct, a -10.8 (or lower) doesn't necessarily mean a fully loaded tower.

Having said all that, I don't understand how transmit power factors into this. Are you saying that I should only look at the stats when I am not transmitting data? The transmit power value is almost always non-zero. I've seen it show "--" a couple times, but that's rare. Is there a reference that explains how transmit power figures into RSRQ?

[swwifty]

Thanks for the feedback, swwifty.

I did see a big decrease in performance with one antenna disconnected (from the SMA connector, not from the modem), so I'm pretty sure it is doing spatial multiplexing. Now, how well it is doing it is a good question. I'd have to disconnect from the modem and run a lot of tests to see if it is averaging half speed. One thing I've found is that, at least in my environment, where there are many towers around, results are highly variable (both the speed tests and the modem stats).

I watched the video you linked to in one of your posts where they showed -10.6 (I thought it was -10.8) as being a fully loaded system with no interference. My understanding was that once you add in interference from other towers operating on the same frequencies, then it is a function of the two. So you could have no other towers and -10.8 would mean your tower is maxed out with connections. Or, your tower could be half loaded, but you could have interference from other towers, and you could still have that -10.8 value (or lower, meaning higher absolute value). Or you could have both a fully loaded tower and interference and have a lower value. So I've been watching that number to try to gauge things, but since the other towers can play a role, it is difficult to use that number to understand tower saturation. If my understanding above is correct, a -10.8 (or lower) doesn't necessarily mean a fully loaded tower.

Having said all that, I don't understand how transmit power factors into this. Are you saying that I should only look at the stats when I am not transmitting data? The transmit power value is almost always non-zero. I've seen it show "--" a couple times, but that's rare. Is there a reference that explains how transmit power figures into RSRQ?

LTE has a lot of mechanisms to help deal with self-interference. It's a long topic I can't go into, but each cell typically tries to re-use the same frequency (frequency re-use of 1) for spectral efficiency, and because these carriers paid millions for their spectrum they use.

I've attached a research paper that explains RSRQ in much more detail than I can here. It outlines why RSRQ values are only accurate when read with a TX power of zero.

[xdavidx]

LTE has a lot of mechanisms to help deal with self-interference. It's a long topic I can't go into, but each cell typically tries to re-use the same frequency (frequency re-use of 1) for spectral efficiency, and because these carriers paid millions for their spectrum they use.

I've attached a research paper that explains RSRQ in much more detail than I can here. It outlines why RSRQ values are only accurate when read with a TX power of zero.

Thanks for the paper. I went down the rabbit hole with the firewall on Windows, so I can keep it and any other programs from sending anything during my testing. I still don't get a "0" value for TX Power, but I do get -- more often now, which I assume to be not transmitting.

[swwifty]

Thanks for the paper. I went down the rabbit hole with the firewall on Windows, so I can keep it and any other programs from sending anything during my testing. I still don't get a "0" value for TX Power, but I do get -- more often now, which I assume to be not transmitting.

hmm that's strange that you don't see that value of 0 tx power very often. I use my 4G connection full time and see it fairly frequently while looking.

If you're using GoldenOrb (rooter) you can look on the diagnostics page under modem and see a nearly real time view of the at!gstatus command. It's possible you're just not seeing it in between refreshes of that command.

[xdavidx]

hmm that's strange that you don't see that value of 0 tx power very often. I use my 4G connection full time and see it fairly frequently while looking.

If you're using GoldenOrb (rooter) you can look on the diagnostics page under modem and see a nearly real time view of the at!gstatus command. It's possible you're just not seeing it in between refreshes of that command.

We are using different routers and modems, so maybe the firmware of one or the other impacts what it shows for that value. I've never seen 0 for tx power and I've run the command close to 1000 times in my testing since setting up the router (not exaggerating).

I am using GoldenOrb and using the generic modem firmware that came with my EM7565.

The router is unplugged right now, but I'll check the diagnostics page during my testing tomorrow and see if I can catch any 0 values.

[xdavidx]

For anyone else reading this topic, even though I mentioned it in another topic, I took my 15 dBi flat panel, directional antennas and my rig to one of the towers (tower B from above) and tested a couple blocks away. I got about the same peak download that I got with the omni antennas (from that tower). The main difference was that uploads were, on average, a little faster. With the omnis they were generally in the 30-50 Mbps range and with the directional antennas, they were often in the 40-60 Mbps range. That makes sense.

The SINR was better, which also makes sense. I was hitting high 20s sometimes for the SINR. If only it looked that good from my house. But tomorrow is the big day to test the antennas on my roof, so I'll see what tomorrow brings. The only negative is that it is a Saturday, so there will be more people competing for resources on the towers.

After my testing so far, my best guesses for not achieving theoretical speed results right next to the towers are:

1. A combination of other users on the same cells that I'm connected to and other towers using the same frequencies is reducing how much data I can get out of each MHz of bandwidth.
2. I am not getting the full aggregate of bandwidth that the modem is capable of from the tower channels available. The modem can handle 60 MHz. I'm getting a max of 40 MHz from a couple towers (20 MHz, 10 MHz, 10 MHz with 3CA), 30 MHz from another (15, 10, 5) and 25 MHz from another ( 10, 5, 5).

I am getting a max of about half the theoretical (close to 150 out of 300 Mbps) from one of the 40 MHz towers and about 2/3rds the theoretical of the 30 MHz tower (close to 150 out of 225 Mbps). This is assuming 64QAM. Maybe the 30 MHz tower is using 256QAM and I'm again getting half the theoretical (close to 300 Mbps). Hard to say without knowing which modulation the tower and modem are doing.

I suppose, in theory, there could be something limiting me to not being able to go faster than 150 Mbps in the modem/router/laptop, but I don't know what that would be. The router has a USB 2 port, but I'm not using that. I'm using an internal adapter. If the adapter is somehow using USB 2 internally, that would still allow for 480 Mbps theoretical. Even if I was only able to get half of that, I'd still be able to get up to 240 Mbps. It seems more reasonable that A and B above are combining to limit my speeds, even with a very high quality connection to the tower.

I may just take my omni antennas on the road sometime and try to find out if there are any AT&T towers that provide greater bandwidth. If I'm able to find a 60 MHz tower and get close to 225 Mbps, then it will lend credence to the guesses above and rule out a hard 150 Mbps limit somewhere in my data pipeline.

[swwifty]

Your testing is very interesting, but also highlights the need for everyone to do their own testing as things can be very different in any situation (even day to day, etc.)

[xdavidx]

Your testing is very interesting, but also highlights the need for everyone to do their own testing as things can be very different in any situation (even day to day, etc.)

Definitely. Even second by second (at least in my environment).

[unbunchster]

stand under tower

you will not connect to tower. this is not how this works.