How does setting MCS on clients(4.7.RC7) affect AP througthput?

Hello, I am trying to find information on how TDD works in real life depending on the MCS configured on the AP or clients.

For example:
An AP (20Mhz) with 20 clients: 5 SMs MCS15/15, 5 SMs MCS13/12, 5 SMs MCS11/11, 5 SMs MCS11/14.
Internet plans are 12/7Mbps, 25/7Mbps, 35/7Mbps.
All customers with MCS11 modulations have 12/7Mbps and 25/7Mbps plans.

Questions :

  1. To avoid retransmissions from clients with MCS13/12 and MCS11/11 modulations, the ideal is to set these modulations in the SM (it is already possible in 4.7.RC7)…?

  2. What consequences does it have for AP/TDD to have customers set to modulations that allow them to have the contracted plans with guarantees even though they can modulate higher? I mean, of those 20 clients set the MCS that allow to obtain 12/25/35Mbps even if the subscriber can modulate higher, for example: a client with 12/7Mbps that can modulate to MCS13/14, but is set to MCS11/11 to make it more stable. What if we do this for all clients on the AP?

  3. What would be the most efficient way to configure this AP for maximum performance with TDD?

  4. From the point of view of the AP, should the one with the highest performance percentage be configured as max downlink modulation? For example, MCS15(1%), MCS14(15%), MCS13(55%), MCS12(9%)MCS11(20%). In this case MCS13.

@Fedor

Thanks in advance.

I would re read the user and system guides. Most of what you are asking is explained clearly.

That said, retransmissions are not due to MSC level of the link but if the data makes it uncorrupted or not. Limiting the MCS per SM will reduce the number of retransmissions simply because a lower coding rate will cope with more interference.

A radio that can modulate/use a higher coding scheme wil use less air time to send the same amount of data and your AP will have lower latencies as a result of this. Though that is not the whole story of reducing latency.

MCS has nothing to do with how you split the air frame. The air frame has everything to do with max bandwidth per direction. You choose the air frame ratio based on several considerations that determine the policy of your network. Since you are not using 100mbps plans, you can set it to just about anything. But make sure your frame ratio is the same across ALL APs even ones that you think can not hear each other (you would be amazed how much these things can hear)

2 Likes

I’m curious as to the goal of these questions. Are you writing a thesis? Or maybe just trying to optimize your PtMP deployment?

2 Likes

Today it is quite difficult to find places to connect clients with stable and uniform modulations. Normally in the real world you will find situations like the ones described, clients with different coding schemes even having the same SNR (due to local interference on the SM side).
The question here is to understand how it affects AP performance to have that variety of MCS spread across all connected subscribers.
I think it is clear that the higher coding schemes the less airtime you need to send the demanded traffic, but is there any way to predict the total bandwidth that the AP is going to be able to reach when you need to set to lower schemes (for interference) ?

If TDD does time division when clients with lower modulations “talk” they will lower the scheme used by the AP and consequently its total capacity, which, depending on the traffic that is passing, will leave it with 0 airtime.
On many occasions, the same thing happens without doing it voluntarily when we have clients with low SNR that prevent us from modulating well.

For example:
If an AP on 20Mhz/Flexible is passing 70Mbps, this means that if all subscribers have MCS13(800ns/104Mbps), the AP’s airtime is 70%, correct?
Now if half of the SMs have MCS11(800ns/52Mbps) and the rest MCS13, what are we going to see when they talk to each other in airtime?

Regards,

Hi Eric, I am trying to understand how to optimize PTMP solutions in the real world by listening to cases and experiences of others, understanding how they interpret the different scenarios that we have today in cities with a lot of interference. Always with the objective of getting the most out of the Cambium equipment.

Regrads,

The short answer is… Cambium specializes in making wireless equipment for unlicensed bands, and as such, they’ve had many years to refine their product for a wide variety of different operators and use cases. There’s very little in the way of manual or individual (per sm) tuning or optimization in the way of settings (like manually choosing an MCS rate) that is recommended. YES, you can set MCS or modulation rates for an SM experiencing interference, but doing this for a large deployment of 100’s or even 1000’s of SM’s is not sustainable. More often you’ll find simply changing the AP’s channel, channel width, or some other AP related setting will result in better performance for the overall AP.

As to your other question about MCS rates and airtime, yes, you will def see less overall throughput on an AP with lower MCS rates. Cambium provides some tools that you can use to simulate this through LinkPlanner and also through the ePMP Capacity Planning Tool.

In short, as an operator that’s used Cambium for many years now and has deployed 1000’s of SM’s, 99% of the time I typically stick with Cambium’s recommended defaults when it comes to MCS rates and other default RF settings.

3 Likes

ok, I know those tools well, I use them a lot. As I said, I wanted to hear opinions on how other WISP administrators manage their Cambium equipment in conditions of high interference and sparse modulations. If they usually adjust on the client side, which I do, or do they do it on the AP side.
In my case, leaving the clients with MCS by default has not worked better for me, this in situations of high interference has created large amounts of retransmissions, high latency and service degradation.
The retransmissions will create many problems at the TCP level (adjusting tcp windows to the bottom) and the total capacity of the clients will go down and there will be enough bandwidth left without using a worse feeling of navigation.

As is logical, the best option is a clean channel that provides stable coding schemes, but… that in 5Ghz is increasingly difficult.

I also manage some Cambium subscribers, there are currently +7,000 (of a total of almost 9,000) in my network with 240 APs.

Well… right off the bat I would hope that you’re using GPS sync between all your AP’s and sites.

I guess as a WISP I’ve never had to deal with interference at the levels you’ve described. There’s always been an option to change to a different channel, or channel width, or worst case, make physical changes like increase the SM’s gain by using a larger dish, or decrease the AP sector size or use a 30deg horn antenna, or with e3k attach the BSA (which arguably provides some uplink benefit in high interference environments)… lastly, if things are really bad, we just move to a different band altogether like 3GHz/CBRS.

We’ve never had to adjust individual SM settings down to the MCS rate or adjusting TCP windows. Maybe some other operators will chime in.

1 Like

All our new sectrors are epmp3k+RF30º to improve SNR and the SMs are F300-x to use MU-MIMO in 20Mhz channels to archive great throughput.

Thanks Eric for your comments.

perhaps someone from cambium can give their point of view on the subject and what configurations will make our APs more or less efficient in airtime.

the SMs are F300-x to use MU-MIMO

Looks like you are using RF Elements dual horn mount which means both of your 30° horns are pointing exactly the same direction (overlaying each other). According to Cambium that means MU will not work.

According to cambium if you want MU to work with RF Elements horns they can not overlay each other so 2 x 30° horns would have to be offset 30° from each other so that they cover 60°

If you deploy those like that I would love to know what kind of results you get (thinking of doing the very same thing).

Hello, I knew about Cambium’s recommendation for this design but I wanted to try it to be clear.
After several weeks running I see that it works well. I see MU-MIMO traffic from time to time as the APs do not have enough load yet for the AP to start using MU-MIMO more actively.
I have tested them by dual-testing two SMs and here you can see the results. (20Mhz; 70/30)

2 Likes

Well, that’s why I always specify “according to Cambium”. Super interesting. Are only two of them groupable because only two of them are more than 10° apart or do you think more should be groupable ?

Be hard to get much MU at 30° even if they were supposedly “tuned” for ePMP 3000. I think I may try it with a pair of 90° Asyms.

Edit: Also curious if you did 0 - 1 to one horn and 2-3 to the other or 0-2 and 1-3 ?

Thanks !

2 Likes

the grouping of SMs depends on each case. The SMs must have 8º of separation (you mention 10º) and in cases where you have a density of clients, it can happen perfectly. I send you a capture of another AP with RF30º antennas, both in the same direction.


Also curious if you did 0 - 1 to one horn and 2-3 to the other or 0-2 and 1-3 ?

I don’t quite understand this question, I’m not crossing the sectors, both horns are facing the same direction.

1 Like

The SMs must have 8º of separation (you mention 10º)

Well, again just going by what Cambium says A minimum angle diversity is about 10 degrees.

That is really awesome and surprising if it really works though you obviously seeing at least “a little” MU while Cambium stated “you will not see any”. I’m gonna for sure give this a try with a pair of 90° asyms.

Well in the same post Dmitry Cambium says MU won’t work he also says you must connect one horn to chain 0 and 1, connect second horn to 2 and 3. I’m pretty sure someone else in the thread says they connected one horn to chain 0 and 2 and the 2nd horn to chain 1 and 3. I’m I’m asking which you did.

2 Likes

I’m surprised that you’re not seeing more SM’s achieve DS MCS 9 modulation… judging from the downlink SnR’s I’d expect almost every one of the SM’s to be DS MCS 9. I wonder if this is somehow related to the horns overlapping?

Also interesting that there are 21 SM’s shown on your screenshot above, and 2 of them seem happy to group with 16 & 17 of the other’s… BUT most of them are only groupable with 2 or 3 or 4 of the other 20 potential partners.

On this ePMP3000 example (with the Cambium 4x4 Sector) we have 19 clients currently, and most SM’s seem happy to pair with about 60% of the other 18 potential partners, and if it was plotted on a curve, the median distribution seems to be about 9 or 10 or 11 of the other SMs.


I wonder if that’s a symptom of a problem, or if it’s just the luck of the draw of random SM placements?

PS – The example posted shows 193 total potential pairing partners / 18 other SM’s = each SM has an average of 10.72 potential pairing partners… or in other words, each SM is grouped with 60% of the other SM’s.

3 Likes

@brubble1 tanks for the thread.

@Eric_Ozrelic I have configured MCS8 as the encoding scheme for the download of the entire AP. There is a lot of noise in the area and I prefer to keep the modulation stable and lose ~30Mbps of maximum capacity from DS9 to DS8. This was the point of this whole thread initially, how to combat noise situations to lower retransmissions and keep the modulations more stable but with less capacity.

@ninedd You are right, it is something that I am closely following because all these APs have been in production for about 1 month.
The grouping of the SMs in certain sectors is complicated with a low number of clients with these narrow horns. My idea if the grouping does not increase as we put more clients would be to expand to 60º sectors (dual horn) and thus obtain better performance of 1 single channel per AP. But it remains to be seen how it changes with more connected SM.

1 Like

O.M.G. :flushed:

Sometimes you don’t know how good you have it until you see another WISPs SA…

6 Likes

I’d love to see a toggle for “Conservative Modulation” on an SM-by-SM basis, and all it would need to do is:

  1. raise the thresholds where it drops modulation levels by a little, so instead of switching up to MCS6 at -70 dB (for instance), require -68 or -67 dB to switch up, and
  2. keep a timer of modulation shifts, and if too many events happen in too short a time, temporarily lock the modulation at the lowest stable modulation level.

What we’ve seen is that the defaults work well about 95% of the time, but those other 5%…? They do benefit from restricting modulation shifts, and their connections are far more stable when we do. One of those is a F300 PtP which we found was not stable enough for use at an office until we locked down the modulation to a level which prevented data loss (uplink locked down to DS MCS4). The ability to force a lower modulation level has been a very useful tool since the old PMP320 days, and it’s still useful on a small handful of PMP450 and ePMP subscribers today.

3 Likes

Ahh I remember the days of Ubiquit 2.4Ghz NLOS client radios where a single poor client link could bring down the capacity/performance of the entire AP and NLOS meant RSSI/MCS changing when the wind blew or a squirrel farted. Rain could completely wreck a ubiquiti AP because the slightest bit of fade meant client MCS rates started bouncing.

Configuring the max MCS was part of the install back then and always one of the first things we looked at when an AP was performing poorly (2nd only to interference).

We have set the MCS on a few ePMP 2.4Ghz but it was mostly to improve the customers connection and not because it was affecting the AP. It does still affect the overall AP performance to some extent but ePMP’s scheduler / Air Fairness was game changing vs Ubiquiti.

I think, at least for us with mostly LOS 5Ghz now, unless the customer’s link is having problems hunting down setting MCS bouncers manually isn’t really worth the time/effort but something more automatic like you describe could be very useful for squeezing every ounce of capacity/performance out of an ap.

1 Like