Does 2.4GHz makes sense?

For me, the speed from home is rock solid. Your mileage may vary of course.  :)  Again - this is really just a test 'experiemental' Access Point for us. I really just wanted to see how horribly it would fail, in the face of all the 2.4ghz noise.  And (to my great surprise) it's clicking along nicely. So, one of these SM's is at my house, and my family uses it as our internet all the time - 4 kids.  We have it set at 20 Mbit down and 2 Mbit up and that's exactly what I get at home.

Again - I know they are having to work at it sometimes...   my 'Packets per MCS' is lower than I'd predict solely based on the signals and SNR's I have - so its preferring to use MCS11-12 instead of 13-15 and I do see that in some of the statistics... but I'm frankly shocked that it works at all.  Here's a screen shot and you can see that it spends quite a bit of time at MCS-11 on the uplink to the AP.  And, it's important to state again that there is only 3 clients on this AP right now - two of us at 20m/2m and one client at 3Mbit/500kBit. Both the 20m/2Mbit clients are staff, and the 3m/500k is a customer 7.5km away that we've just moved off of older 2.4 gear, and he's very pleased.

So - we havent really deployed this out on any of our country towers yet. We are still struggling with the 'real time throughput display' on the Access Points.  Without that, we can't wrap our heads around how we are going to manage the calls from customer's and us not being able to look and see what everyone on the Access Point is doing at that moment in real time.  However - as far as the 2.4Ghz goes, I'm expecting it to work even better than it does in town of course, so I'm expecting that it'll be night-and-day better than our older non-cambium-brand radios.  :)

Yes, this is a GPS equipped AP, and I'm using 20 Mhz wide, and I'm using 5ms frame size.  Right now, I'm in 'Flexible Frame' because we don't have any other ePMP or Canopy 2.4Gzhs AP's on that tower.

If your question was leading into latency - this is pinging from my desktop at work - to my SM at home (~4km away).  So, this is from my Windows PC, to the switch - out the AirFiber24 backhaul, up to the building top where this AP is - through that switch/router - to the ePMP Access Point (which takes 1-2ms for me to ping the AP) and then over the air to my house.  It's 5:30 PM so the bazillion routers between here and my house will be reasonably noisy right now...

So - this is GPS/TDD 5ms Frame Time, Flexible Frame.  That's 8ms average from my desktop computer at work, to the SM on my roof.

Ping_to_10.2.4.64_Mar15_16.png

If I switch to TDD Fixed Frame - this is 5ms Frame Time, 75%/25% Fixed Frame.  15ms Average.

PingSM_5msFixedFrame.jpg

And - this is TDD 2.5ms Frame Time, 75%/25% Fixed Frame, back to 8ms average ping from my desktop computer at work, to the SM on my roof at home.

PingSM_2.5msFixedFrame.jpg


As far as Cambium's ''extra'' latency...    where we are located - we are about 30ms to Google, so I'm not too concened if our over-the-air latencyis 8ms or 15ms  - so the difference between the customer being 38ms vs 45ms from Google makes very little difference in real world performance. 

Ping_to_Google_Mar15_16.png

I think what is important in latency is reasonably stable latency, reasonably predicable latency, and I want it to stay reasonably flat as we scale, or reasonly flat if there is some traffic going across the link.  WIth our older (non-Cambium) gear, it is basically WiFi gear and it has only 1-2 ms over the air..... until a customer starts using it, and then it's 80ms or 62ms or 214ms or 12ms or whatever.  And until we connect the 24th client on the AP, and then everyone's latency is ???ms.

3 Likes

Great results! Using Flexible you weren't using GPS... I thought the "magic" of your results came from GPS sync, but if it works so good even in Flexible mode the results are stunning for me. I didn't expect more than 2-3Mbps in such a noisy environment!

Speaking about latency, I noticed less TCP throughput using 5ms frames in some cases where I have lots of wireless hops that add latency, so 2.5ms works better at the cost of 10% throughput. I guess it can be the same for 2.4GHz, but if you're using AirFiber24 and you are so "close" to the gateway, I think you shouldn't notice significant loss using 5ms...

Let us know how it will work away from the city!

About single user graph, I can suggest you to use MikroTik routers and, if you use PPPoE, you can monitor real-time traffic from the router, and also enable graph for them. It works very well!

2 Likes

giuseppe4 wrote: Great results! Using Flexible you weren't using GPS... I thought the "magic" of your results came from GPS sync, but if it works so good even in Flexible mode the results are stunning for me.

Well - CAMBIUM CAN CORRECT ME IF I'M WRONG -  but ePMP in TDD mode is always still 'scheduled' by the AP.  So, you are correct in a way - Fixed Frame ratios can be GPS synced among Access Points, because they all will start their transmit frame segment at the sametime, and they will all transmit for X% of the frame, and receive for X% of the frame, and thus all your ePMP AP's can all be synced. However, even in flexible frame, the AP and SM's are all still 'scheduled' by the AP. 

So, I think that the ''magic'' happening here is that ePMP AP is transmitting to the SM's on it's schedule, and that its not dependent on 'waiting' or 'sensing' for a clear channel', and also that (even in Flexible Frame) the AP is still scheduling everything and knows when the SM's should talk, and again - the SMs go ahead and transmit on their schedule, even if the CSMA mechanism on a WiFi device would have told that WiFi device to wait because there's too much interference.

My SM at home is QOS set to 20 Mbit download, and 2 Mbit upload.  This is what I get from our speedtest server.

speedtest-fromhome_20m_2m_mar15_16.jpg

It is just after 11 PM now - so there would be a decent amount of noise/usage out there right now. I mean, a bunch of those routers across town generating the 2.4ghz interference would be fairly busy right now, and still the ePMP clicks right along at 20 Mbit down and 2 Mbit up - jus tlike it's set.

3 Likes

You’re right! It’s still a “real” TDMA, so this is the reason of such good performances!
I hope Cambium will release a feature to have a TCP speed test in the SM and also from the LAN side, to see exactly the speed the user can see!
I think it could be very very helpful for operators like us!

Anyway, I think Cambium should give you some free units!
You made evidence of the good performances of their ePMP 2.4GHz! :smiley:

I would like to try it, but the entry price is too high for a test, maybe a Lite device as you proposed (I voted for it) could help!
Hope they will listen about it!

2 Likes

The Cambium test is usually right on with real world speeds provided that bandwidth is getting to the AP, speedtest.net is with 10% every time for me. You can use a connectorzed cpe for a lite AP, will do the same work but does not use GPS to get you going, the APs retail for 499, in bulk, 430. And with the deal right now, we’ve been ordering 20 packs for about 6k after shipping ect. GPS is definitely worth the money. Very very worth the money.

3 Likes

GPS really worth the money, I know!

I'll try a connectorized AP without Sync for 2.4 Performances, thank you for your suggestion!

About the Cambium test, I agree with you using 20MHz I have about 10% less, but using 40MHz I had sometimes much different values in real world TCP performances (speedtest.net for example)


@giuseppe4 wrote:
maybe a Lite device as you proposed (I voted for it) could help!  Hope they will listen about it!

I hope so too. The full AP's are about $430 each, although right now, Cambium does have a ''Buy 2, Get 1 Free'' deal on 2.4Ghz Access Points. So, that would be $860 / 3 = $287 each. A ''LITE'' AP in 5Ghz are $196 each - and you can buy them individually and in any quantities you want, and we can upgrade them to full versions via software, without having to climb back up that tower. So, installing a non-GPS (SM) as an AP isn't really something we want to do - since that wouldn't allow the benefits of GPS Syncing all the AP's on the tower, and since it requires another climb (or possibly four more climbs) to upgrade the AP's later.

So, in our situation, we have a tower where we want to put up 4 of the DualFrequency Sectors, each with a 2.4Ghz and 5Ghz GPS AP's on them - so the cash-out-of-pocket difference today is about $1,000 difference over those four sectors on the one tower. It'd be nice to have a lower cost of entrance, even if that means we'd give the rest of the money to Cambium in license upgrades in the next year anyway.

1 Like

Is anyone using the UBNT 120degree sectors and teh epmp 2.4 aps? If so how are they performing for you and can you see any differences between the UBNT and cambium sector antenna other than 90 degree sectors

we've got a couple active,    they have the RF armor shields behind them.    the edges are goofy, but they work ok. i defiantly would not use them without the RF armor shields on them, the FTB isn't very high.    we left them up because they are shielded. the cambium antennas have better characteristics.  KP performance antennas rock. but are not cheep. 

2 Likes

I know this is an old post...but...

Chris, would you use the KP over the Cambium sector? I am going to replace my Ubnt sector with one of these, and want the absolute best performance.

both antennas have thier strengths and weaknesses.  depending on what you need, the KPs could be better than the camibums or the other way around. 

we only use the 2nd gen dual frequency KPs now.  

the KPs have a very very tight roll off on the side, they are at 13db on the 90 degree edge, the cambiums are at 12 at the same point.    at 120 degrees, the KPs are almost gone, and the cambiums are still at 9db gain.   (KP has single sector 120s that will hold up same or better than cambium)  the KPs  we use are not good for a 120 degree sector. 

the KPs have a 7.6 degree vertical fan, great for tight signal spread, not good to have very high and try and hit close in subs.  the 5ghz side is 4.6 degrees.   both frequencies share a ceiling.  

we've got one site almost 500' above the closest targets and it wasn't fun getting that taken care off. 

cambiums are 14 degrees, much better for being much higher over your close subs while ending the signal and your controlled distance. 

cambiums 5ghz and KPs are the same gain, the KPs are .8db louder in center, and use a lower loss jumper so probably 1DB total gain improvement overall.  

the cambium 2.4 antenna has 2 db better FTB than the KP on the 2ghz side.   BUT the KPs tighter signal spread generally gives us betterSNRs when off center from the sector.  cambiums are better when you are more centered. 

the KPs are much larger and heavier and are 3x the price. 

we choose to use the KPs over the cambium because of the small vertical beam width, sharp horizontal beam and dual frequency. 

over all, while in the 90 degree field, and each antennas vertical field the differences are very slight.

the dual freq has saved us $75/month in lease fees... really adds up over time. 

1 Like