Technology Blog

RF Basics - Things You Didn’t Know That You Didn’t Know

Guest Blogger

Like many industries, the Wireless LAN community is made up of people with a wide range of education and experience. Even those with decades of experience can still learn and adapt our view of how the invisible medium we work with works.


At Aruba’s Atmosphere 2018 conference I had the opportunity to sit in a session by Eric Johnson (@ej_wireless on Twitter), also known as Dr. RF. It was an enlightening and entertaining discussion where I learned new aspects of topics I thought I learned and have repeated for 18 years. Truthfully, it was a bit humbling to know I didn’t know what I thought I knew.


I’m only going to cover two of many “Eureka Effect” moments.


First – we’ve been using this simple Spectral Mask for OFDM signals for years since it was first introduced in 802.11a.



But in this session, I finally learned where it comes from… especially the ‘shoulders’.


The first part is the digital signal – the OFDM portion. It looks something like this:



Note – this isn’t exactly like what the spectral mask from IEEE shows.


We then need to add green in as the wideband channel noise floor…


Then we finally add in the blue ‘noise’ generated by the Power Amplifier’s Non-Linearity. Based on keeping costs low, thermal noise low, we end up with PA’s with less than perfect amplification.



It might be possible to work this out of a Wi-Fi transmitter, but at a very high cost, both in terms of cost of goods sold, but power consumption and heat generation too – all three of which put this out of the range of even Enterprise-grade equipment, let alone consumer client devices.


This total amount of RF being transmitted results in the now familiar Spectral Mask used by the IEEE specifications. It is the combination of the Red Digital Signal, the Green Background Noise, and finally the Blue Power Amplifier’s resulting noise to get the entire package we are used to seeing.



 It is this ‘shoulder’ noise we are concerned with when adjacent channels are so close as to interfere with other RF signals on our primary channel in question.


This also deals with the reduction of power needed at various MCS rates as to not overpower and distort RF. This means going to higher MCS, with higher modulation schemes results in lower Transmit Power. Here is a table of Aruba’s changes in Tx power based on MCS. Note the nearly 10X difference between BPSK and 256QAM!RFNutsandBolts_Pic6.png


We’ve always known there is a significant difference in the ability to transmit more or fewer bits per symbol based on changing of EVM (Error Vector Magnitude) to suit environmental conditions. But this is exacerbated by the necessary lowering of Tx power based on the above table.



The second major learning experience concerned the supposed religious fervor of many to ‘match’ Tx power of the Access Points with the Tx power of client devices. The idea here it to try and keep both as close to possible to continue to have ‘balanced’ power between the two ends of this communication link.


Personally, I’ve not been a believer in this ‘religion’ – since I’ve seen high Tx power of AP’s and low client Tx power situations work fine in the real world. But it has been a big source of contention with many who have listened to and believe. (If it was true that all communications need a balanced link – then packet analysis would show matched MCS upstream and downstream – and in the real world, this is a rarity.)


Not to mention – the 802.11 protocol requires an Acknowledgement for each Data frame sent. So ‘one-way’ communication isn’t possible. Though it is possible, even highly probable, to have differing MCS and modulation schemes in uplink and downlink communications.


Finally, this session included a spreadsheet explaining the math behind allowing for mis-matched Tx power – yet still having link path loss nearly identical. Here’s one example using 100mw of Tx power (20dBm) with an EIRP of 30dBm at the 4x4:4 AP with a 4dBi antenna, along with a standard 1x1:1 client device transmitting at 25mw (14dBm) with a 0 dBi antenna.



When the Aruba Atmosphere 2018 session presentations are published online, I strongly recommend you take a chance to watch and review this session by Eric Johnson to delve deeper into these, and other RF basics.


I learned an amazing amount of details behind what I thought I knew of RF, and I think you will too!


The session title Basic Radio: RF Nuts and Bolts.

Tags (1)
New Member

Excellent article Keith!  As it relates to your second point regarding AP's superior recieve sensitivity relative to clients that can support an unbalanced link and work great; I just want to point out for the non-expert wireless engineer reading this that in the vast majority of environments channel re-use drives power decisions to avoid CCI and ACI.  

New Member

Hi Keith,

Nice article. Questions, if I may:


1) You said, "This also deals with the reduction of power needed at various MCS rates as to not overpower and distort RF. This means going to higher MCS, with higher modulation schemes results in lower Transmit Power. "

   * Can you please explain why this is the case?  I didn't catch the explanation from the preceding paragraphs. I understood your chart and the comment about "10X", but I didn't understand exactly why you need to turn the power down as the MCS goes up.


2) In your red/green/blue graphic, you mentioned that the blue is the PA noise which forms the 'shoulders' of the spectral mask, and the red is the OFDM signal itself.

   * Does red and blue ever separate more than is shown in your graphic? The question that arose in my mind while reading that section was, "If the PA noise is a constant/proportional amount of interference against the OFDM signal, then why would we then worry about any other type(s) of noise?"


3) It looks like you're saying that the amount of power entered into the Aruba controller GUI/CLI is in EIRP. Is that correct?


4) Why is there a 5dB noise floor differential (between the AP and Client) in your example?


5) Why do we care about the total path loss (which is between the Tx and the NF)? Shouldn't we only care about the path loss between the Client and AP at usable distances (e.g. "Want" and perhaps some small portion of "Don't Want")?


Final thought:

While you don't need a 'balanced' link for a wireless link to operate, having high SNR in both directions will mean that both transmitters will achieve better data rates. If the client has low SNR, then it should roam. If the client has high SNR, and the AP has low SNR, then the client may not roam when it should, which could drive up retransmissions and airtime consumption on the channel. It's my opinion that having bi-directionally good SNR is always best.


Thanks for taking the time to share.



New Member

Hi Keith, nice article! 


At one point you say:


"Not to mention – the 802.11 protocol requires an Acknowledgement for each Data frame sent. So ‘one-way’ communication isn’t possible."


I agree with the statement only partially. It is true that each "unicast" data frame requires an Acknowledgement, but "multicast/broadcast" frames do not. Hence, in theory, it IS possible to have one-way communication from the AP to client. 


Secondly, the ack frames can be sent at the highest Basic rate, which is typically one of the legacy a/b/g rates. Hence, when an 11n or ac data rate is being used in the downstream direction (AP -> client), the client may only need to ack at 6Mbps (or if higher Basic rate is configured, then most generally 24 Mbps). In such situation, having an SNR imbalance has the potential to cause upstream (client -> AP) traffic to not be able to use 11n or ac data rates.


Certainly, the AP's receive sensitivity and antenna gain come to rescue here. But to be able to say that imbalance does not matter at all, sounds a bit odd.


Lastly, I have the same question as Devin, what is the reason for the 5dB delta in the noise floor for the AP vs the client. 




Hi Devin, 


EVM is Error Vector Magnitude. It is a spec in 802.11 because of the mechanism is shown in the preso and Keith’s blog.


For low MCS rates, you can have a lot of deviation from the ideal point and still identify the bits that were sent.


For higher MCS rates the points in the constellation get closer together. This has physical meaning. To keep the transmitted point plus noise injected into the channel from causing bit error the noise at the transmitter needs to be cleaned up. When you back off the radio 1 dB the noise at the transmitter from regrowth drops by 3 dB giving a net 2 dB benefit. To go from 16 QAM to 64 QAM and the same coding rate you need to improve transmitter SNR by 6 dB. This is accomplished (to first order) by dropping tx power by 3 dB.


This also answers the question about red and blue separation.


Aruba controller use EIRP for power settings. That has always been the case.


Lower power consumption designs for LNAs results in higher noise figures and the close proximity of stuff in the phone will cause some noise injection from the digital circuitry.


We avoid this on APs design with board shields and mostly continuous ground planes under antennas and we can allocate more current/power to LNA circuitry


Total path loss is shown to demonstrate link balance. The smaller number is the limiting path.


The discussion is not relevant to roaming. AP spacing will drive SNR at client and AP end. The point of the balance discussion is that a 14 dBm client does not mean 14 dBm EIRP on the AP.

The idea that aloud (high EIRP) AP can result in a client not being to get back even though it can hear the beacon has not been relevant since the introduction of MIMO operation.

Search Airheads
Showing results for 
Search instead for 
Did you mean: 
Read all about it! If it’s happening now, it’s in the community.

Check out the latest blogs from your community team, the community experts and other industry sources.