How does ePMP Adaptive Modulation work?

The ePMP series can transport data over the wireless link using a number of different modulation modes ranging from 64-QAM to QPSK.


For a given channel bandwidth and TDD frame structure, each modulation mode transports data at a fixed rate. Also, the receiver requires a given signal to noise ratio in order to successfully demodulate a given modulation mode. Although the more complex modulations such as 64QAM will transport data at a much higher rate than the less complex modulation modes, the receiver requires a much higher signal to noise ratio.


The ePMP series provides an adaptive modulation scheme where the receiver constantly monitors the quality of the received signal and notifies the far end of the link of the optimum modulation mode with which to transmit. In this way, optimum capacity is achieved at all times.

How does it work? - If ePMP 3000 single group customers have MCS7 and MCS9 then will the whole group be transmitted according to the lowest MCS7 or will each SM be transmitted according to her modulation?

Hi Modcn. I believe that the whole group gets the lowest common modulation. I also believe that the MU-MIMO takes this into account when deciding what data should be SU and what should be MU. So, if it’s faster to send it in SU mode, it will… and only will use MU mode if that’s beneficial.

BUT, for sure when we do MU-MIMO tests, we get the best MU-MIMO performance when testing to the best MCS9 clients, with the highest qualities. Testing to two ‘perfect’ connections like those, test results OFF THE FREAKING CHARTS!!! We’ve seen 465 Megabit aggregate throughput in a 40Mhz channel.

1 Like

I’m thinking to myself now … PTMP should use Mu-MIMO if possible, regardless of higher modulations in SU mode, but what’s the point in the future (epmp 4000) of high modulations eg 1024-QAM, if there is likely to be someone in the group who will have max 256-QAM and then everyone will get max 256-QAM?

Well, MU-MIMO is smart, and will decide what is the most efficient. For example, a 3000 with 5 clients in 40 Mhz mode… you’ll likely never see any MU-MIMO, because it can handle the data needs of every request handily, without ever needing to MU, so it’ll use SU mode, and “won’t break a sweat” to deliver the data needed.

So MU-MIMO will only kick in when the AP starts to get stressed. Basically - you can think of the 3000 like this… as long as a 2x2 3000L could handle the traffic, then the 3000 will basically NOT have to go into MU-MIMO mode. And at the point where the 3000L would start to be congested… the 4x4 3000 will ‘kick in MU-MIMO’ at around that point.

As I mentioned earlier – for me, the two main reasons for Cambium vs other brands are:

  1. MU-MIMO, and
  2. Backwards/Forwards compatibility

When the 4000 arrives, it’ll be 8x8 MU-MIMO, and if we see 465 Megabit aggregate throughput in 40Mhz channel now on our 3000 MU-MIMO, imagine what the 4000 will be capable of with 1024-QAM and 8x8 MU-MIMO! And – when that arrives, we don’t have to figure out how to do an entire sector forklift, like many other brands.

As for ‘whats the point’ of higher modulations… in our example, the AP might be able to deliver that clients data at 1024QAM and then go service two other 256-QAM client in MU-MIMO. The faster the clients are, the quicker they can get on and off the air too.

Anyway - to answer your core question - I guess I would say this. We can only control what we can control. I think the 3000 MU-MIMO is miraculous (when I see 465 Mbit!!!) and the 4000 will be to - but we can only control the part that’s under our control, so ALWAYS get the best modulation you can, Period. Never connect a Force 300-16 unless it’s solid MCS9/MCS9 anyway. Never connect someone at a -69 if they can be a -59. ‘Good enough’, isn’t good enough. Every client which isn’t a MCS9 is leaving money on the table, and leaving bandwidth/airtime wasted.

Obviously that’s not always possible - but the point is never to do the ‘-67 is good enough’ or ‘MCS7 is good enough’ if it’s possible for that to be a -57 and a MCS9. As long as we do that – then the 3000 MU-MIMO and the 4000 MU-MIMO will deliver their part.


Which will be faster or the same: 5 clients SU mimo or 5 clients MU-MIMO (located in different places, split sectors / horns mode)?

Hi. I’m not sure what you’re asking exactly. MU-MIMO is absolutely the way forward, with dramatically more capacity (speed) than SU mode overall. The #1 platform for Wireless capacity is the Cambium 450m… and if you look at that, wisps can get INSANE spectral efficiency and sector throughput, due to the ability for the AP to transmit to Multiple Users at the same time.

That being said – a Single User, with a perfect connection (100% max modulations) will also be very, very fast. There are (for example) certain radios which have 4096-QAM modulations, and the speeds to a single user will test off the charts fast. BUT, if you load that up with 50 users, that ‘peak speed’ will get slower and slower the more you load it up (in SU-MODE). On the other hand, something like the PMP450M might have clients connected which are only 256-QAM, and if you test to a single client, it’ll be slower than a 4096-QAM client by a long shot. BUT, load the PMP450m up with 100 clients, and the MU capacity really can kick in and it can outshine everything else with it’s overall capacity.

So, it depends what you want I guess… if you want a really impressive SpeedTest,net result… a Single User with a faster Modulation would be the way to do. BUT, if you want the most impressive sector bandwidth with as many clients as possible (high density applications) then MU-MIMO (like a ePMP 3000, ePMP4000, or 450m) is the way to go.

1 Like

@ninedd, I don’t believe your right that everyone gets the lowest common modulation. That would be terribly wasteful. The whole 802.11 family is smarter than that. For example, I’m looking at an ePMP2000 sector with three active clients on it. One has a downlink RSSI of -81 dBm, shows downlink MCS 9 (which, remember, on 11n is MCS1 on two chains), and the ePMP speed test shows 21/6. Another has an RSSI of -56 dBm and shows MCS 15, and the speed test shows 86/20. So the slow user isn’t slowing down the fast one at all.

This actually was one of the benefits of ePMP that they talked about when it was a new product and a certain other company was the market segment leader. On that brand’s PtMP system, each user was allocated an equal share of the bps capacity, so a slow MCS got a longer time slot. On ePMP, each gets an equal time slot, so the higher MCS gets more speed, and you can put on users with weak signals without hurting users with strong ones.

I assume that MAC feature was left intact when they went to the 3000 and 4000 series. I like the 2000s because a) we have too many trees to ever get higher MCSs anyway, and b) it seems to have access to more channels than the -ac or -ax radios have, like “5840” and “5165”. These are not official Wi-Fi channels but they’re legal, within certain OOBE limits. While I don’t have any 3000s, it seems from the specs that they want to stick to Wi-Fi channels.

1 Like

Hi. I didn’t mean to imply that everyone get’s the lowest modulation - I meant this only in context of the original question about MU-MIMO to TWO clients with different modulations.

I might not have explained it clearly or correctly - but certainly no, I did not mean to imply that ePMP uses the lowest MCS to speak to every client. :slight_smile: Just in context of when people are testing MU-MIMO, and therefor ‘forcing’ the AP to use MU-MIMO mode to two SM’s with (for example) MCS6 and MCS9 rates… if they do that, I THINK that the AP will be forced to use MCS6 while sending MU data to these two SM’s.

BUT, normally the AP would avoid doing that. The AP uses the ‘what is more efficient’ ruleset from how I understand it, and would tend not to decide to do that. I was using that as an example of when people try to “test MU-MIMO and don’t get the results they expect”.

For us - the ePMP 3000 (MU-MIMO) are remarkable, and we’ve seen over 465 Megabit Aggregate Throughput in a 40 Mhz channel, which is spectacular.


@ninedd, thanks for clarifying. It would be interesting to know if the different members of an MU-MIMO set could be at different MCSs. I don’t actually see why not, since different waveforms are being created for each, but I don’t know how MU-MIMO and its grouping is implemented. I’m thinking about the Massive systems that do several transmissions at once, in different directions, and that would be quite inefficient with a common MCS, especially in the mobile context (those $$$ Ericsson 2264s, for instance). ePMP does amazing things with an orders of magnitude lower price. :grinning:

I don’t know - you’d have to ask Cambium I guess.I can only tell you what we experience. If we take proper care to get every SM with the best signal, with balanced chains, and if we achieve DS MCS9 - then we have seen over 465 MBit throughput in a 40 Mhz channel on our ePMP3000 APs.

And, if we do that same testing, and force it to use lower MCS clients instead, then the results are lower. In our experience, the best performance is when both members of a MU-MIMO pairing, are both the highest modulations, and we get lower performance if the Modulations are lower.

BUT, you’d have to ask Cambium for more info I guess. :slight_smile:

Now that I think about it, the MU-MIMO set might be limited by the chip set the same way that 11ac paired frequencies are. If you have a 4-chain 11ac radio (I don’t use Cambium’s, since it is crippled in the US market), and one chain is good for MCS2 and the other for MCS8, both run at MCS2. It’s a hardware limitation not found in the Massive gear. An ePMP uses WiFi chips so it may have that limitation in this case too. I suspect the Medusa doesn’t.

I would suspect so - Cambium does exceed MUCH of what the chipset is capable of in many ways, but as you say, at the end of the day they are also still subject to some core 802.11AC limitations.

However - to comment on your example:

  • WISPs shouldn’t be installing MCS2 clients and expecting that to not impact performance, because it certainly will. :slight_smile:
  • If you did, the MU-MIMO logic would only use MU-MIMO for those two clients if it thought that was more efficient than sending them data separately. The 3000 MU-MIMO will only use MU mode, if that’s more efficient than SU mode, and only if it needs to.

So - in my opinion, you a) should always strive for the highest modulations on all clients, and b) MU-MIMO will only kick in when it needs to, and when that’ll give some benefit. It really shouldn’t “screw things up” at any point anyway, and c) as I’ve mentioned, we’ve seen over 465 Megabit throughput in a 40 Mhz wide channel, which is pretty spectacularly efficient.