06/16/2015
AF-5X - major resolution for new launch
Description

It is not my intent to make this a product comparison post, but I need to address it by way of explaining the issue that was solved.

 

We launched our network up into the mountains with a 20 mile back haul from our main datacenter using 34 dBi dishes and a pair of Mimosa B5c radios earlier this year.  Our understanding going into this was that these radios would be a perfect fit, however, it was a painful learning experience to say the least.

 

The OOBE rule on these radios made it impossible to have enough power to really stretch the 20 miles with any type of reasonable link, and even when we turned up the power to very illegal levels just to test we still couldn't get the type of performance we needed on such a long link.  Also, before anyone starts in with all the things we "could have done" to make it work please know that the company actually flew a rep out to our location for the day to work with us on-site to try and make the link really shine but was not able to do so.

 

At the end of the day the SNR numbers just weren't good enough to make it work at the longer distance and it really didn't matter what frequency or widths we used.  Here's a couple of screen captures from different settings we tried.

6-16-2015 10-40-22 AM.png

 

b5c.png

 

We were finally able to get our hands on a pair of AF-5X radios and made the switch to see what the difference would be; expecting that it would be better, although not as good as it actually was.  Using the exact same dishes as before (standard RocketDish 34dBi w/out the conversion kit) the performance and throughput is absolutely amazing.

 

6-15-2015 8-03-58 PM.png

 

In addition to the overall better stability and throughput we've also seen a marked improvement in Ping times from clients ranging from 9ms to 18ms compared to 16ms to 35ms previously.

 

We need to do a little more refining of this link (Pt. 2 later) as it's much easier now to troubleshoot the link as numbers don't bounce around so much.  We can probably pick up a dB or two with some alignment fine tuning, auto power backoff, maybe turn on GPS sync, etc.  This is a production link so we have to schedule a maintenance window to do these types of tests until we have a redundant one in place.

 

Throughput is nice, but link stability rules them all IMO and as I'm not getting client calls about spotty performance anymore I'm sold.

 

AF-5X - major resolution for new launch

by on ‎06-16-2015 10:01 AM

It is not my intent to make this a product comparison post, but I need to address it by way of explaining the issue that was solved.

 

We launched our network up into the mountains with a 20 mile back haul from our main datacenter using 34 dBi dishes and a pair of Mimosa B5c radios earlier this year.  Our understanding going into this was that these radios would be a perfect fit, however, it was a painful learning experience to say the least.

 

The OOBE rule on these radios made it impossible to have enough power to really stretch the 20 miles with any type of reasonable link, and even when we turned up the power to very illegal levels just to test we still couldn't get the type of performance we needed on such a long link.  Also, before anyone starts in with all the things we "could have done" to make it work please know that the company actually flew a rep out to our location for the day to work with us on-site to try and make the link really shine but was not able to do so.

 

At the end of the day the SNR numbers just weren't good enough to make it work at the longer distance and it really didn't matter what frequency or widths we used.  Here's a couple of screen captures from different settings we tried.

6-16-2015 10-40-22 AM.png

 

b5c.png

 

We were finally able to get our hands on a pair of AF-5X radios and made the switch to see what the difference would be; expecting that it would be better, although not as good as it actually was.  Using the exact same dishes as before (standard RocketDish 34dBi w/out the conversion kit) the performance and throughput is absolutely amazing.

 

6-15-2015 8-03-58 PM.png

 

In addition to the overall better stability and throughput we've also seen a marked improvement in Ping times from clients ranging from 9ms to 18ms compared to 16ms to 35ms previously.

 

We need to do a little more refining of this link (Pt. 2 later) as it's much easier now to troubleshoot the link as numbers don't bounce around so much.  We can probably pick up a dB or two with some alignment fine tuning, auto power backoff, maybe turn on GPS sync, etc.  This is a production link so we have to schedule a maintenance window to do these types of tests until we have a redundant one in place.

 

Throughput is nice, but link stability rules them all IMO and as I'm not getting client calls about spotty performance anymore I'm sold.

 

Comments
by Deleted Account
on ‎06-16-2015 10:11 AM

Great story, thanks for sharing! 

by
on ‎06-16-2015 10:50 AM

 Nice story and nice link!  Thanks for sharing these results.

 

Your aim looks extremely close to me.  For your antenna type, frequency and distance, set the Antenna Value to 31 in the wireless tab without a cable loss, and the EIRP to 51 on both sides (making sure the conducted doesn't go above 20).  Your signal bars will probably fill up perfectly indicating a good aim.  

by
on ‎06-16-2015 10:59 AM

@ClaudeSS I appreciate the feedback and suggestions.  Originally I had moved the Antenna to 31 and upped the full allowable power and the signal bars were -50/-51 - full and great; however the modulation dropped; which an excellent resource told me would be expected as the amplifier can actually start to generate intermodulation noise.  His suggestion was spot on to yours in setting the antenna to 31 but don't use the full power.  In our next maintenance window we're going to try that and will post a follow up.  A bit of a learning curve for us but love the radios.

by
on ‎06-16-2015 02:13 PM

So why is there such a difference in TX and RX when the modulation is the same?

by Ubiquiti Employee
‎06-16-2015 02:31 PM - edited ‎06-19-2015 08:39 AM

Hi @DSnet,

 

It's probably worth keeping in mind two things here (the first is the biggest factor):

 

  1. The 75/25% Duty Cycle setting.  So 75% of the throughput is alloted to one direction, 25% to the opposite direction.
  2. Even at the same modulation rate, there can be a range of decode effectiveness.  It's worth keeping in mind that the AF capacity graph represents actual capacity (unlike other gear) of the connection - taking into account things like decode success rate.

 

I hope that helps,

Brandon

by
on ‎06-18-2015 11:38 AM

@DSnet It was the duty cycle setting as I wanted more RX than TX speeds on the link.  We're planning on adjusting the power/antenna settings to try to improve the overall capacity of the link but we'll still keep the RX set for the higher capacity.

by
on ‎06-25-2015 05:46 PM

Hi eomedia! Great story and thanks for sharing as it's applicable to my present situation.. out of interest, what were your latencies? We're planning to use a bunch of 5x's (around 7 hops - longest link 44km) to get out into the Desert, trying to ascertain the expected latency.. most of these will be in bridge mode with only switches at each relay.

by
on ‎06-25-2015 06:48 PM

@doc_karl  The latencies we've seen have been pretty good - ping times are very decent and clients have reported good performance since we made the change.  In our case the clients are only 2 to 3 hops back to the datacenter so I'm not sure how that would translate over something like 7 hops - but I would imagine it would operate similarily.

 

We're getting ready to expand our hops through the mountains out 4 more hops so once we've done that I'll post an update - but it may be a while before that's done.

by Ubiquiti Employee
on ‎06-26-2015 07:28 AM

Thanks, @eomedia for the story and responses!

 

Also, @doc_karl, this story (thanks, @jjohnson) shows ping times across their AF5X link during live throughput:

http://community.ubnt.com/t5/airFiber-Stories/Rocket-AC-PtP-link-upgrade-to-AF-5X/cns-p/1274191

original.png

by
‎06-26-2015 08:53 PM - edited ‎06-26-2015 08:53 PM

@doc_karl  We just got out second AF5X link (39.503 km/24.56 mile) up and runing this morning so here are some actual ping times going through just the first link, and then going through both. (throughput/capacity charts are reflective of AF5X at start of ping test)

 

LInk 1 - Hop 1 - (31.915 km/19.831 mile)

hop1.png

 

Link 1 Hop & LInk 2 Hop - (71.418km/44.377 mile)

hp2.png