Is A5 any better than a standard 802.11 AP?


#1

Hi Mimosa,

First off, let me say that I do like the roadmap you’ve laid out and I’m excited to see what the future will bring with the A5/C5 line, specifically in the micro-pop model.

But, right now I’m struggling to see how the A5-14/18 products offer anything more than a standard off-the-shelf 802.11ac access point with a wave2 chipset could offer.

From what I gather, anyone currently deploying an A5 is using it in wifi-interop mode (aka 802.11). I also gather that the A5-14 is 4 x 90* 5dbi single-stream antennas. So those two things boil down to (a) no hidden node avoidance; (b) each client gets a single stream unless they happen to catch a side-lobe from an adjacent antenna; and © just a standard 802.11 mac layer with all the insecurities and inefficiencies it brings.

There’s no GPS sync, so I would be concerned about these close-proximity 80MHz chunks of self interference. I also understand there’s no MU-MIMO, so I assume it transmits (and listens) only to 1 user on all 4 antennas at the same time.

I’m very keen to try some of the Mimosa Kool-aid because on paper it looks real tasty, but I need to understand that it’s not just marketing vapourware and that all the WISP-centric features are not years away.

Rich


#2

Any comment from Mimosa?. I’m also looking to try an A5 with non-Mimosa CPE’s (most Mikrotik). If we have to use standard 802.11 i can’t imagine an improvement over Mikrotik’s proprietary nv2 or even Ubiquity’s Airmax. Although there is no GPS sync (with Mikrotik) at least there is some form of TDMA?


#3

Hi Letaba, if you’re going to mix Mimosa & Mikrotik, then you’ll have to use standard 802.11. That’s the same with any vendor.


#4

I’ve been wondering the same too lately. Just hoping for GPS sync, but I never thought about the MU-MIMO not working, and figured since I’m using the 2x2 MIMO C5’s, they would only link up at 2x2 to the 4x4 OMNI, but now that you mention it, they would only connect to (1) side or “chain”.

We deployed one recently (2 weeks ago) and connected a C5 right under it. Got 80 Down 60 Up.

The WISPAMERICA youtube video had a tester saying his customers were getting 300Mb/s, on A5-18, how??? 80Mhz channel on 1 chain?


#5

I just looked at one C5 connected to our only Mimosa OMNI AP. I’m really confused now because there are x4 panels, and the clients should only connect to one side (unless you want high mist-matched signals on the Chains), and its showing he has both streams.

I wonder if each panel is utilizing both H/V, giving the 2x2 devices a full 2x2 MIMO instead of 1x1.

But that theory would make the AP 2x2 on each panel, so total of 8x8 MIMO on all 4?


#6

From what I understand, each panel is 1 chain but at close distances the client will pick up sidelobes, or edges of the adjacent antenna and possibly establish a second chain albeit with different signal properties. Looking at some of the screenshots posted by others I see a reasonable proportion of the clients with 2x1 or 1x1.

Jamie made a post about this a while back and you may find the other comments in the thread interesting… A5-360-18 OMNI dBi?


#7

Yes, it is.

We have since some weeks 50 (!) Mikrotik SXT5-Lites (well, even some rb411 and 711 and 911 amongst them) connected to an A5-14.
For test purpose we also have 1x SXT5ac-Lite, a QRTac and 2x Mimosa C5’s connected.

I must say that although most ‘promises’ like tdma, GPS sync, MU-MIMO are still in the air, the legacy 802.11 works good.
It works better then Mikrotik NV2’s tdma.

All the clients are connected within a 30 to 400 meter range. In respect of the panels of the A5 I would say the distribution over the sectors is 25-35-30-10%

Before all these clients were connected to 2 Omnitiks, both in a different frequency, one in 20Mhz bandwidth, one in 40Mhz bandwidth channel.

Little by little we ‘handed’ all these clients over to the A5-14 that works with 40Mhz channel.
Because of the higher gain of the antenna (14dBi versus 8dBi) we could tune down most SXT’s considerably down in output power so we reduced the overall noise in the field a lot.
Since radio output of the AP is done by the 14dBi antenna (each panel is 14dBi, losses are 3dB due the misalignment (=Circ. pol versus HV-pol) but these are compensated by the two panels (streams) almost every unit is able to use.

We noticed that where MT need both chains to be working with no more difference than 3-4dB to use them both, the A5 still uses both chains even if the difference is double that… so that’s a first winner.

By reducing the used spectrum from 40 + 20 = 60 to one 40Mhz channel we could creating some ‘cleaner’ spectrum around the clients. Our area still ‘sees’ many other 5Ghz usage. That’s a second winner.

In reducing the SXT’s output levels a lot (most SXT’s went from their default 27dBm to some 13 up to even 8dBm(!) we also made the overall noise less. That’s a third winner.

Since the A5 has much better chipset compared with its better radio and probably better software we see high mcs rates even with signals in -55 to -60dB regions. On the Omnitiks (or Netmetals out in the fields somewhere else for that part) we needed signal levels in the region of -40 to -45 to reach same high mcs rates… So that’s a 4rd winner.

For whatever reason we where never able to ‘pump’ more than some 50Mb aggregated data over an OmniTik or 60-70 on a Netmetal. On the A5 we can easily push up to 200Mbps to one client which gives 200+Mbps aggregated over the A5 while all 50-52 associated clients are still connected and some even functioning… try that with a Mikrotik device! All our other Mikrotik based networks run NV2 but the Mimosa is definitely outperforming them a lot. This throughput gain is probably because of the better chipset and much more powerfull CPU and the ‘cleaner’ spectrum.
Winner no. 5

The ‘normal’ SXT’s (27dBm/600Mhz cpu) we’d used for the clients are mostly able now to saturate their Ethernet port in traffic (up to 95Mbps). On MT networks we never saw that in PtMP.
We also hooked up one SXTac to see it that would do better but never reached more until we realised it still has a fast-Ethernet port only. 95-99Mbps…
So we hooked up a QRTac that has a gigabit port, and yes, 180, 200Mbps download. (For this test we’ve set the A5 to use 80Mhz wide channel. In 40Mhz the speed was still some 150Mbps…

What we also see as a big benefit is the fact that it seems the A5 adapts better to each CPE’s link characteristics. MT in that case struggles to give a poor link client a good CCQ and it delays the rest of the network (in tdma mode) where the A5 hardly seems to be bothered by one or two poor performing clients. Others still get high throughput.

The A5 works in ‘interop’ mode which means csma rts/cts needs to be activated in all CPE’s (set it with the lowest Hw. Protection Threshold = 256 so it is always ‘on’) to avoid the ‘hidden node’ collisions.

We tried to reduce power of the CPE’s even more, so they all hit the A5 with -60 to -65dB signals and each CPE would still get mcs rates 4-5 (single stream) but the moment two or more CPE’s were tested the overall network speeds collapsed completely.
Probably because the low output levels of the SXT’s avoided each CPE’s signals to reach all other CPE’s in the network, a necessity for rts/cts to work.
Although the overall ‘noise’ would be even lower, the network collapsed due the ‘hidden node’ effect that could not be solved by the rts/cts csma system.We increased the output levels again and it seems the network runs pretty stable now for some days.

Thus now we are thinking of replacing more Mikrotik AP’s for Mimosas… but there are some caveats…

Some of our AP’ are in close range of each other. GPS sync would be ideal to preserve as much as possible spectrum as we can so we can start working more in 40 or even better, 80Mhz wide channels. And to avoid tower or close range own, or 3rd party, network interference.
BUT, that can only be done by tdma protocol and that can only work in a full Mimosa environment… and now the costs are going to fly through the roof…
For the price of one C5 with a G2 (that has advantages of its own like continuously scanning frequency usage with auto channel swap) we can buy 3 SXT’s…

Our funds are simply not deep enough to achieve this overnight. At the same time we have to realise we are presently working in a max 20Mb to the client network. Having all Mikrotik AP’s replaced by Mimosa would cost us something but we can increase these speeds to 50Mbps with burst into 80Mb or so and at the same time have more clients per AP. This would be economically feasible and we would stand out of the competition as a WISP.

In going to a full Mimosa deployment we could serve clients an extra 100Mbps in top speed. That would put us even more out of the competition but against a huge investment level from our side the client is probably not willing to bear…

My overall conclusion for now is that the A5 in itself is worth the extra money. I haven’t seen such a powerful device before. It can stand the comparison with a tdma network easy, even in ‘interop’ mode comparing to tdma…
A full syncd Mimosa tdma network is something to dream of I’d guess…

Our next setup planning is one or two A5c’s back to back each connected to two 2x2 Mimo sector antenna’s in a full HV-pol. Mikrotik environment at medium to long range environment.
The idea is that with 2 x 100º RF element’s sectors we could cover some 200º of range with some 17dBi antennas in one frequency (the A5c’s are 4x4 and with 2xMimo we actually get two 2x2 sectors fully syncd.)
A tower could this way with 2 A5c’s in two frequencies in 80Mhz band serve some 150-160 clients and assign them with 15Mb guaranteed and up to 50-100Mb burst download capacity.
I don’t even see a way of doing something similar with 4 Netmetals. These would need twice as much spectrum and still each AP would not reach such high throughput…

On a scale of 0-10 my satisfaction of the A5 would be a 8. A full Mimosa with GPS sync and tdam would make a 10 but due the price a correction to a 9.
If Mimosa could bring the price down of their C5 CPE by 50% and the GPS/tdma would be out, it will be a 11 score system!

One last remark; Although the A5 is circ. polarised and the Mimosa C5’s we’d used for test purpose (2) are 45º slant, we tested all SXT’s and the QRT in HV-pol setup.
We actually set one SXT in a 45º slant setup too (RF elements have an adapter you can use for that) but we saw no significant difference between a normal fit SXT and this 45ª slant SXT.
The Mimosa C5’s are also not doing any better than the QRT. Which makes sence. The QRT has 24dBi gain and a powerful 720Mhz cpu. and narrow beam. Apart from the slant versus HV there will not be so much difference between the two.


#8

Hi Rudy,

Very interesting feedback - thank you for that. On paper, the A5 seems much like a standard 802.11 AP and at its high price I was reluctant to invest in some test equipment. However, I guess there is some X-factor in that Quantenna chipset and Mimosa software.

One concern I had was with the 4 x single chain antennas getting vastly different RSSI across 2 chains for each client, but from what you say it seems to cope with that well. Also, my understanding (and from the FCC filing) is that each panel is 5dbi, not 14dbi.

I really believe GPS is important, particularly in the target deployment of high-density AP’s in a (sub)urban setting. So I will wait for that to happen. As we all know, other vendors have promised GPS and never delivered, so let’s see if Mimosa can really do it before I get my wallet out.

Another thing - I’ve posted a couple of times about MAC security and never had any response. It does concern me that some kid, or a competitor, could knock my clients offline with commonly known disassociation vulnerabilities. These are the things that differentiate a WISP platform from a 802.11 AP.

As I said before - very excited if everything they promise gets delivered.

Rich


#9

We are now running the A5-14 for some two weeks with some 51 associated MT-SXTs (‘an’ protocol, 40Mhz bandwidth) on average (max 56 on the moment) and so far haven’t had any complaints from any of our users about poor performance or cuts or whatever… We can run several speedtests at the same time and show good speeds.
Peak usage of the AP is around 50Mbps but 60% average top usage would be some 30Mbps I would say.
80% of the clients are allowed to use up to 20Mbps download but with a 6min/30Mb threshols burst of 50Mbps.

I must say I am very satisfied in this ‘interop’ mode. This is plain csma 802.11an network!
I remember in the old days before tdma was introduced in the WISP world csma even with rts / cts this was a crime and in those days all we delivered to the clients was some 2-4Mb!

Regarding the dbi; According the specs we do talk about a 14dBi antenna. Even if this is the resulted 4 panels added up this would still mean a single panel would be some 8,75dBi.

I indeed see that some CPE’s at times only got one chain working but it doesn’t seem to harm the connectivity. After all, even the worst CPE’s still show MCS rate 4 (that’s really an occasional low, 5 is more regular where 7 seems to be the standard low rate.) which then still give enough capacity to have 30-50Mb of usable data flowing to a client that can maximal receive 50Mb anyway…, And the moment CPE demands traffic the MCS goes up anyway…
If I compare the s/n ratio of the A5-14 network (CPE side) with an Netmetal in one of my other networks I see the Mimosa maintains always higher MCS rates anyway…

GPS (and with it tdma, one can’t do without the other) sync would be a big plus, but only in environments where really the max has to be squeezed out of the system in sync with several other Mimosa AP’s to re-use spectrum as much as possible. And only if the extra costs can be justified.
(It needs to see all CPE units of 3rd party to be replaced by Mimosa C5 and that is some quadruple in investment in 40-60 CPE’s big PtMP networks where the extra benefit will be only double, maybe tripled…
It’s sort of like replacing your '91 model Chevvy by a 2016 model Toyota. It’s really worth the money, but yeah, a Ferrari will squeeze the extra performance possible in all circumstances… But is it worth the extra expense? Some will say; “Yes!” but they probably have unlimited funds and/or start new anyway. I’ll guess the bulk of users are very happy with the Toyota.)

Your remark about the mac security I can’t fully weigh it to its merits. If you’re referring a deauth attack, yeah. My knowledge doesn’t go that deep but I believe the ‘deauth’ frame in a network is not encrypted so any intruder could indeed bring your PtMP network down. Correct me if I’m wrong, like said, I am not an real wifi expert.
I’ll guess tdma would come to the savier here, but then the extra pennies have to start rolling…

Don’t know what exactly you mean with, quote; These are the things that differentiate a WISP platform from a 802.11 AP. un-quote?
Most WISP’s imho use 802.11 in one form or another and probably most ever started with simple b/g network without any protection whatsoever. Proprietary TDMA brought some security here as a bonus.

And any danger from any competitor? Really? If a competitor wants to attack or disrupt your network he can expect the favour in return or worse, the regulatory police investigating? In the latter he might loose the license to work any longer in the WISP business. How many operators are willing to jeopardise their business in an attack on a competitor?

But yeah, it still makes a point. Maybe Mimosa can look on the Mikrotik ‘management protection’ encryption key they’d developed to protect against ‘deauth’ attacks? Maybe they are so smart they can make it work on a ‘operators’ level instead of a ‘proprietary level’ only?


#10

Has anyone heard the Cambium ePMP Elevate road map? You can load their firmware on your existing WiFi SUs and change out your APs and now you have GPS sync that really works with beam forming. This is all done through the RF…no truck rolls to the SUs!! It only exists for Ubiquity today but their are plans for Mikrotik also. If that isn’t impressive enough, their next gen AP will offer MuMimo like their PMP450 Medusa does and they are claiming with 100% confidence that all the older SUs that have been upgraded with be backwards compatible. I’m sorry but this is the way to go, no question in my mind. The money saved in truck rolls alone is convincing enough.