Dear all,

I'm hoping you can provide me with some recommendations.

Assume the following situation: a (remote-controlled) robot has to provide the operator with data from sensors mounted to it. The operator is in a fixed location, and we can assume maximum distances of 300 meters as well as Line-Of-Sight communications, with few if any obstacles.

Robot operation itself is accomplished using a 2,4 GHz WLAN system (tried and trusted); no problem here.

The problem is that the sensors attached to the robot produce a LOT of data. In consequence, I need a stable link capable of carrying about 1GBit (more would be better) from the robot to the OCU (Operator Control Unit). There will be some traffic from the OCU to the sensors as well. While negligible in bandwidth, I do require this broadband link to be bi-directional.

I am hoping that two NetMetal 5 units would be an adequate solution to my problem and have purchased two units.

I have added RB11e-HacD cards to both of them, to supplement the on-board 802.11ac WLAN for an (eventually) aggregated setup.

In my initial tests (using just the internal WLAN), I have outfitted both units with omnidirectional 5GHz Antennas and configured them to my best knowledge (80 MHz Bandwith Ceee), WLAN AC only, 3 chains and all MCS-Modes enabled. The units are approximately 15 meters apart. The Antennas of both units are currently aligned in parallel. (Chain 1: -45°, Chain 2: 90° (vertical), Chain 3: +45°)

I have then proceeded to run some bandwidth tests (UDP packets of varying sizes) using the built-in tools (AP = Server, Station = Client).

Throughput as reported by the client can go as high as 300 MBit, but I'm unfortunately not seeing anything higher. CPU-Load seems to top out at approximately 50%.

So far, I have tried configurations 802.11 and NV2, with comparable results. Setting all Antennas to 90° didn't change anything.

Could anyone make some suggestions as to what else may need to get adjusted in order to get more usable bandwidth or am I actually topping out the bandwidth server?

Thanks!

Colin

im happy not to be your neighbor, using omni antenna and 80 mhz channel leaves anybody close to you unable to use that chunk of spectrum

You will never see 1Gbit/s stable link with standard radios. You need Airfiber. Good luck with that. Anyway 1Gbit/s from some sensor it somehow to much.
Btw, from my tests, 3 chains is worse than 2.

@chechito: it is a research project, so it will be running only temporary and when it is, it will be in an open, unpopulated area far away from any place where it could disturb anyone.

@ InoX: Thanks for the suggestion. Unfortunately, I don't have the budget for Airfiber and secondly, I can't use directional antennas on the robot; I must use Omnis there.
Howcome you have had worse success with 3 chains than with 2?

Regarding the bandwidth: a simple Microsoft Kinect One is sufficient to max out a USB 3.0; producing way more than 1 GBit. You won't even be able to connect a mouse or keyboard to the same bus. My project calls for using exactly this camera as one of the sensors, in addition to some HDR cameras, Laser-Scanners and other things. So unfortunately, it is really easy to create large amounts of data rather quickly.
I will have the ability to cache it on the robot's computer and may be able to add some compression (difficult) or reduce frame-rate (not desirable). Nevertheless, I still need a pipe between the robot and its base that is as wide as possible to be able to transmit the data as quick as possible.

Any configuration recommendations are welcome.

Cheers

Colin

Forget about 1Gbps.

Most you could (theoretically) achieve is 860Mbps, and that's assuming you're on a place with all spectrum free. Even with that radio datarate UDP rate will be lower due to framing overhead (TCP will reach 60% of that in the best scenario at most).

How come you cannot use a directional antenna? I'f you're into robotics a directional "tracking" antenna should be something achievable by you

Best scenario in terms of Radio performance and thus throughput will be using a small aperture sector antenna (45º, like the AM-5AC22-45) with a netmetal and its power cranked to the minimum needed as you want the best snr possible.

Equip the robot with a small dish or panel directional antenna rotatable 360º. A QRT-AC when it's released could be interesting to test as is a rather light and compact integrated unit.

Problem with AC is a mesh network with multiple sectors or cells is a no go, as as soon as you're using 80Mhz on one the rest will have no spectrum to work with. So you can have just one AP transmiting.

Regarding the "tracking" directional antenna in the robot, you can use either GPS coordinates to track it, and also use the mikrotik api to "sense" the wireless signal from the robot.

I would use omnidirectional antenna on robot and directional automatically rotated antenna on the base. It would be much easier and cheaper and more reliable than moving with the antennas on the robot.

About what distance Do You Need the 1 GBit?

Dear all,

thank you very much for your replies.

Here are my comments:
@mistry7: Basically, I need as much bandwidth as I can squeeze out. I'd be happy if I could get into the area of 500 MBit per WLAN.

@pukkita: I know that the maximum (theoretical) bandwidth of the NetMetal is 867 MBit. However I am running mine with an additional WLAN-Card and hope that by aggregating both, I can get into the area of 1GBit of aggregated traffic.
I understand that I will never see the theoretical bandwidth, but in my current test-setup, I believe I should be seeing more bandwidth at 15-20 meters distance than what I have now.

Directional Antennas with a pan-tilt head to keep pointing them to the operator are great when your robot is large enough to carry them. Mine however is only about the size of a small lawnmower and is very lightweight.The additional hardware I am installing for the sensors (LIDAR, Kinect One, cameras) and on-board computers for processing and batteries to power everything already push the envelope of what it can carry.

I will however consider sector-Antennas for the operator-side, but won't using a 2x2 Sector Antenna require a MIMO Omni as well, due to polarization?

Good idea about ttrying to optimize the SNR, I will look at that.

I have to demonstrate my sensors on a (given) mobile platform at ranges of up to 500 m Line of Sight, so I never even considered using MESH.

What about the Wireless protocol itself? Should I focus on staying with 802.11 or should I try NV2?

Cheers,

Colin

You can even have more transmitters on different frequencies and some router that will be fast enough to be able to steer the packet splitting. You can create bonding to nearly double the throughput. Maybe rb800 with two or three ac cards. I believe it can make the gigabit over wifi then.

Also take care you will be streaming by big packets and not small ones. Check the sensors and their specifications to be sure.