I just received a brand new WRT54GL from Linksys ($50) that I'll be converting over, and I'd like to add an external unit at the shack too. From what I've been reading, it looks like one of the Ubiquiti systems would be the best way to go, as they are supported with the new firmware.

What is the best Ubiquiti unit to purchase? That is, which one has the best performance?

I've seen the UBNT.com specs on the Bullet M2 HP. It looks like the easiest to attach to the Omni directional 15 dBi antenna I'm looking at. But the specs I saw at ubnt.com seem to indicate it has lower receiver sensitivity at the lower data rates than their other models.

Is the Bullet less sensitive than the others, or is it just the spec sheet that is worded differently?

Has anyone used the Bullet, and what was their opinion of its performance?

And for those who may have other Ubiquiti models, what was their experience with it, especially the sensitivity?

Jerry

K6ANI

Hello Jerry,

I'm not sure a 'best' Ubiquiti device exists,  its a mixture of variables, just like all radio setups.

What your seeing on the bullet in 802.11g models is not 100% specified, in the Rocket datasheet they spec it as 1-24mbps @ -97dbm while on the Bullet they only spec it at 24mbps @ -83dbm

So what is the Bullet at 1mbps? We can't know for sure to be honest if the sensitivity continues towards -97 @ 1mbps  on the bullet  or if it  stays at -83 because it only has a single antenna and can not use both antennas as a diversity receiver. If it is the diversity receiver than it becomes important to note that both antennas need to pointing at the same signal (this is true in Linksys as well)  for that advantage to be realized.

Once you start talking Ubiquiti to Ubiquiti you will also likely jump into the MCS* table which is 802.11n  and the spec sheets are better written here you can see that the sensitivity DOES go down to -96dbm(+/- 2db) on both a Rocket and a Bullet.

Of note:

The Ubiquiti nodes for BBHN do NOT use TDMA (it would require a central authority controlling timeslots) and instead rely on standard 802.11b/g/n  yes bandwidth goes down over the hops (though with good aiming, multiple devices and eventually multiple bands this may be alleviated depending how a network is laid out)

In a mesh your not actually looking at each and every hop decreasing your bandiwth by half  as eventually a packet will make it far enough away that the local devices (your local node and the first relay node) will both have 'open' channels and the next packet can be sent   You can have multiple packets 'outstanding' in a TCP/IP network until you get to what is called the Receive Window (RWIN) which is "max number of packets the destination has not yet answered.

Now onto TDMA.

TDMA does NOT give you a full duplex link.  TDMA = Time Division Multiple Access.

In TDMA Each node has a set slice of time (1/Number of nodes) worth of time (per second)  to send its data where all nodes in the "receive" area  have to hold off and listen for that node to send traffic  and than  when its the next nodes turn it can transmit (or re-transmit) This does solve the hidden transmitter problem however it also means your max bandwidth is going to be MAX-RFSpeed/number-of -nodes  in any Point to Multi Point (ptmp) cluster. 

By its very definition of protocol it is half duplex link as only 1 node can transmit at a time to keep the channel clear.

For those interested I did some throughput tests. I appended them to this blog entry. Ubiquiti gear does about 10% percent better.

http://kb9mwr.blogspot.com/2013/04/hsmm-mesh-to-become-broadband-hamnet.html

You are correct, I should have verified that by SSHing into the node and issuing the iwinfo command

Perhaps I'll redo the test later down the road, or maybe this thread will inspire someone else do further throughput testing

One could write a book on testing results of they had the time.

i can think of the following test scenarios off the bat

Based on RX signal level

Based on SNR

Based on set distance

etc.

 I belive the Ubiqutit are better all around, and I suspect the lower CPU speed and onboard chips have some decrease on performance overall at the high end, it may not be relevant for average in the field deployments. 

i do suspect that is a large part of (having not witnessed the test and not knowing all the ins and outs of your environment) why you saw a significant slow down between the direct and the 2hop.  I do expect some decrease though (50-60%) as the channel becomes a half duplex link that has to repeat the content (same as 1200 baud through a digipeater) but would expect it to taper off the more hopes you get as you get to a point where one set of nodes is receiving while another is transmitting.

Careful planning of core links as you mention (like the local 2.4 with 5ghz backbone to get to far points) will certainly help this as can using a different channel when you have multiple nodes at a site.

Whomever does test may want to consider iperf, it is made for speed testing and may help remove any web server performance issues as well.  I've used it for unscientific testing of different rf paths but nothing scientific as of yet.

One could write a book on testing results of they had the time.

i can think of the following test scenarios off the bat

Based on RX signal level

Based on SNR

Based on set distance

etc.

 I belive the Ubiqutit are better all around, and I suspect the lower CPU speed and onboard chips have some decrease on performance overall at the high end, it may not be relevant for average in the field deployments. 

i do suspect that is a large part of (having not witnessed the test and not knowing all the ins and outs of your environment) why you saw a significant slow down between the direct and the 2hop.  I do expect some decrease though (50-60%) as the channel becomes a half duplex link that has to repeat the content (same as 1200 baud through a digipeater) but would expect it to taper off the more hopes you get as you get to a point where one set of nodes is receiving while another is transmitting.

Careful planning of core links as you mention (like the local 2.4 with 5ghz backbone to get to far points) will certainly help this as can using a different channel when you have multiple nodes at a site.

Whomever does test may want to consider iperf, it is made for speed testing and may help remove any web server performance issues as well.  I've used it for unscientific testing of different rf paths but nothing scientific as of yet.