As we continue to deploy more DOCSIS 3.1 (D3.1) modems, it is a great time to evaluate if you are getting optimal capacity out of your D3.1 orthogonal frequency division multiplexing (OFDM) downstream channels.  In this blog, we cover a few items that can improve speeds and invite you to a deep dive on the topic of an upcoming Cisco Knowledge Network webinar on May 27th.

Like everyone else, I am happy to see some hints of “normalcy” returning following over a year of lockdowns.  I admit that I have enjoyed the reduced travel, which let me spend more time at home.  It was also nice to have cheap gas prices even if I wasn’t going anywhere.  As gas prices rise, it reminds me of times when people would start paying more attention to fuel economy and look for ways to improve it by checking things like tire pressure or storing that set of golf clubs somewhere other than in the car.

It seems like it could also be a good time to evaluate our OFDM downstream capacity and possibly look for ways to optimize it.  Chances are that when we first deployed OFDM in our systems, we only had a handful of D3.1 modems, which came nowhere close to using the entire channel capacity.  Five years later and that may no longer be the case.

Let’s start with some basics first.  DOCSIS 3.1 offers faster speeds by increasing the channel size and providing higher-order modulations.  The larger channel no longer uses a single carrier like in past DOCSIS versions.  Instead, thousands of small subcarriers either 25 kHz or 50 kHz wide fill up the spectrum as shown in the figure below.  The subcarriers are orthogonal to each other to pack the spectrum without needing guard bands between subcarriers in the channel – hence the name orthogonal frequency division multiplexing.  The higher-order modulations mean that each symbol can carry more bits of data.  Therefore, we get a faster bit rate for the same symbol rate.


Subcarriers packed tight without guard bands in an OFDM channel can be 25 or 50 kHz

OFDM has many configuration options compared to single carrier quadrature amplitude modulation channels (SC-QAMs) where we pretty much just selected a modulation of 256-QAM and the number of channels we were going to bond.  The OFDM configurations allow us to adjust for different channel conditions but can also lead to unnecessary overhead that reduced channel capacity.  I continue to see a lot of configurations that are far from optimal.

There is less overhead with 25 kHz subcarrier spacing which permits the channel to run faster.  There are fewer subcarriers dedicated to things like synchronization pilots as well as less impact from cyclic prefix.  However, many operators elect to use 50 kHz spacing.  Early modem firmware had issues with 25 kHz spacing, but these were resolved years ago.

Multiple data profiles allow a converged cable access platform (CCAP) like Cisco’s cBR-8 to use different modulation orders when transmitting to different modems on the same OFDM downstream.  We no longer need to limit the channel modulation to the worst-performing modems as in SC-QAM.  However, the default thresholds used to select the ideal modulation order for a modem (shown in the figure below) are overly conservative based on the performance of low-density parity-check (LDPC) forward error correction (FEC) combined with frequency interleaving.  Relaxing these thresholds allows more modems to achieve the highest order modulation (4096-QAM) with limited downside.  Higher modulations mean higher channel capacity.


Default RxMER selection thresholds based on DOCSIS 3.1 PHY specifications are conservative.

If you have followed my OFDM configuration recommendations, you are probably getting close to optimal channel capacity today. If not, join me as I do a deep dive on optimizing OFDM configuration on a Cisco Knowledge Network (CKN) webinar on May 27.  I will cover the configuration items like cyclic prefix, roll-off, and next codeword pointer modulation which are most responsible for overhead, and show how changes can impact channel speeds.  These aren’t just things that worked in a lab but are based on five years of D3.1 field deployment in operators’ plants.