General Question regarding AP resources.

Okay to adequately understand question I will begain by explaining that I have been learning about.

so i have been doing some research and reading to better understand information theory and the methods used to transmit that information over a noisy communication channel.

In a very general sense the Shannon Limit Formula is thus  C = B log2 (1+ S/N)   where C = Capacity in bits per second and B = Occupied Bandwith Width in Hertz  and S/N is a ratio of received Signal / received Noise (a unitless number).

And

The Nyquist formula  is simular except that it works by multiplying the baseband by 2 and is understood to be the following formula   R  = 2 x B Log2 ( M )  where R = Rate in bits per second and B = Baseband Width in Hertz
and M = Number of unique combinations or distinct Signaling patterns.

Now given that we all understand that channel Noise is not constant and thus fluid and can dynamically change from one moment to the next.

I would like to understand, why when I have 40 Subscriber modules connected to an AP and all have a SNR consistantly above 30dB  why then do some achive MCS 15 both in the downlink and the uplink but then some also appear to have MCS 4 or MCS 7 and there appears to be no cause for the lower bit rate.

as i understand it. its not possible to have a SNR 30 and above and only be able to use MCS 4 or 7 unless there is noise or the system for some reason decided to only use this level. if the channel is only able to achive MCS 4 due to additional noise then why does the AP status page not update the SNR to reflect the actual state of the channel?

edited:

I would also like to know how much AP resources are used for station keeping and how many SMs are actually supported by the ePMP 1000 APs. while using GPS sync lets say that we are not using frequency reuse.

if anyone could help me better understand this please let me know.
Thanks.

What happens when you actually try to push some traffic through these clients that are have a low MCS? Do you find that they ramp up to a higher MCS rate?

Have you checked out the statistics found under Montior -> Performance -> Downlink & Uplink Packets Per MCS to get an idea of the health of the sector?

i have they stay at the same MCS and the preformance chart shows 80% of the packets are sent at these levels

The SNR displayed by the ePMP devices is not accurate.  I have several SM's that see the opposite AP in freq reuse at 25db below, but yet display a 45db SNR which is absolutely impossible when seeing the opposite AP at an RSSI 25db below what it sees the AP to which it is connected.

Using the ePMP 1000 2.4Ghz synced 3.xx firmware for example,  you will notice by doing some quick math that all your AP's will show under the Wireless > Monitor tab that pretty much every SM displays its noise floor is at -97dbm (RSSI minus SNR= -97). This is regardless of the true interference level present. We don't even bother looking at the displayed SNR.

I have a feeling that if you run eDetect or spec-an on these SM's you will find strong local interference causing the low MCS rates.

@CWB wrote:

The SNR displayed by the ePMP devices is not accurate. I have several SM's that see the opposite AP in freq reuse at 25db below, but yet display a 45db SNR which is absolutely impossible when seeing the opposite AP at an RSSI 25db below what it sees the AP to which it is connected.

That is because on a 802.11x atheros chipset, it is able to distinguish between Noise (SNR) and interference (C/I). When it sees energy from another 802.11 radio, it percieves that as interference. Everything else is noise and that is why you see SNR as being high but an actual spectrum analysis or eDetect will show you that there is interference.

Bottom-line, use SNR as a intial reference point, then make sure you perform a spectrum analysis and eDetect to get the next level of interference levels. And finally watch your MCS rate to determine the final signal quality of your link. MCS rate is the most accurate measure of your link health because it is based directly on the radio's ability to successfully transmit packets over the link.

1 Like

@Cambium_Sriwrote:

That is because on a 802.11x atheros chipset, it is able to distinguish between Noise (SNR) and interference (C/I). When it sees energy from another 802.11 radio, it percieves that as interference. Everything else is noise and that is why you see SNR as being high but an actual spectrum analysis or eDetect will show you that there is interference.

Bottom-line, use SNR as a intial reference point, then make sure you perform a spectrum analysis and eDetect to get the next level of interference levels. And finally watch your MCS rate to determine the final signal quality of your link. MCS rate is the most accurate measure of your link health because it is based directly on the radio's ability to successfully transmit packets over the link.

Sri, with this being said, is the ePMP line not capable of displaying CINR instead of SNR? This would be more helpful to us end users. You could take the CINR and match it to the MCS level displayed (using the ePMP CINR and RSSI chart). For example, an SM may be at MCS12 even though SM has an RSSI of -50. You then look to see there is a CINR of 20, which matches to a MCS level of 12 (basically tells you with a glance why you are at MCS12). CINR would also let you know before you run eDetect or spec-an that you have some type of interfering signal.

@Cambium_Sriwrote:

That is because on a 802.11x atheros chipset, it is able to distinguish between Noise (SNR) and interference (C/I). When it sees energy from another 802.11 radio, it percieves that as interference. Everything else is noise and that is why you see SNR as being high but an actual spectrum analysis or eDetect will show you that there is interference.

Bottom-line, use SNR as a intial reference point, then make sure you perform a spectrum analysis and eDetect to get the next level of interference levels. And finally watch your MCS rate to determine the final signal quality of your link. MCS rate is the most accurate measure of your link health because it is based directly on the radio's ability to successfully transmit packets over the link.

Sri, with this being said, is the ePMP line not capable of displaying CINR instead of SNR? This would be more helpful to us end users. You could take the CINR and match it to the MCS level displayed (using the ePMP CINR and RSSI chart). For example, an SM may be at MCS12 even though SM has an RSSI of -50. You then look to see there is a CINR of 20, which matches to a MCS level of 12 (basically tells you with a glance why you are at MCS12). CINR would also let you know before you run eDetect or spec-an that you have some type of interfering signal.

Sri, with this being said, is the ePMP line not capable of displaying CINR instead of SNR? This would be more helpful to us end users. You could take the CINR and match it to the MCS level displayed (using the ePMP CINR and RSSI chart). For example, an SM may be at MCS12 even though SM has an RSSI of -50. You then look to see there is a CINR of 20, which matches to a MCS level of 12 (basically tells you with a glance why you are at MCS12). CINR would also let you know before you run eDetect or spec-an that you have some type of interfering signal.

@CWB wrote:

Sri, with this being said, is the ePMP line not capable of displaying CINR instead of SNR? This would be more helpful to us end users. You could take the CINR and match it to the MCS level displayed (using the ePMP CINR and RSSI chart). For example, an SM may be at MCS12 even though SM has an RSSI of -50. You then look to see there is a CINR of 20, which matches to a MCS level of 12 (basically tells you with a glance why you are at MCS12). CINR would also let you know before you run eDetect or spec-an that you have some type of interfering signal.

The chip does not readily measure interference unless you switch to SA mode. And it isnt that simple to equate to an MCS level. From a pure radio / PHY persepctive, it is true that a certain MCS equates to a CINR level. However, there are additional ARQ and Rate Adapt algorithms that run, determining the most optimal MCS the radio should be using in each direction, and therefore doesn't provide a clear mapping of MCS to a CINR level. These algorithms were built (part of the secret sauce) to make ePMP perform best-in-class under interference. As an example, even though a certain CINR level shouldn't allow effective transmission at MCS12, the radio may still be operating at MCS12 aided by the ARQ algorithm, resulting in reduced or completely eliminating re-transmission in the upper layers.

Running SA is the best way to get an idea of the interference level. Hopefully we will be able to embed the SA in the GUI soon, taking away the pain of running a separate java app. Good news is the new gen AC products already support inline real-time SA

3 Likes

Gotcha. When you are able to embed the SA into the GUI, may I suggest you also try to embed CINR into the GUI as well (since the two are related software wise). Even though CINR will not give an exact representation of MCS due to your secret ingredients and other variables, it gives a much much better picture of issues causing low MCS compared to the current display of SNR values. Same suggestion for the AC ePMP devices.