DOCSIS and Cable Modems – How it works :: Upstream RF

February 05, 2009   Brady   23 Comments

Cable modems use RF (radio frequency) signals to transport data over hybrid-fiber coax (HFC) networks according to the DOCSIS® specification.  This blog will discuss the finer points extracted from the DOCSIS specification related to how cable modems communicate with the headend Cable Modem Termination System (CMTS), allowing two-way transport of Ethernet traffic over a cable TV network.

There are currently three major revisions and one sub-revision of the DOCSIS specification; DOCSIS 1.0, 1.1, 2.0 and 3.0.  With each major revision came significant changes to the cable modem upstream specification because the upstream has typically been the bottleneck in terms of data through-put rates as will be discussed.

The DOCSIS 1.0 and 1.1 specification provided for two upstream modulation profiles; QPSK and 16-QAM.  I covered modulation profiles in detail in my Advanced RF blog post and thus will not review it here, but I will cover a couple of new concepts regarding data communications, which are symbol rate and filter shaping.  Per the DOCSIS 1.0 and 1.1 specification, the allowable symbol rates for upstream data transmission are 160, 320, 640, 1,280 and 2,560 ksym/sec.  Additionally, the specification defines that the cable modem shall use a root raised cosine shaping filter with an \alpha=o.25 or 25%.  Great!  So what does that mean to us in terms of data rate?  Let me explain.  First, let’s visualize the QPSK and 16-QAM constellations, theoretically:

QPSK and 16-QAM Modulation

Upstream Modulation Constellation Diagram

The above two diagrams show the data symbol mappings for QPSK and 16-QAM modulations.  Notice that a QPSK modulation has four symbols each containing two bits, while 16-QAM has 16 symbols each containing four bits of data.  This implies that 16-QAM transports more data than QPSK, but it may not be immediately apparent that 16-QAM is more susceptible to signal impairment because the symbols are closer together and therefore are more difficult for the receiver to demodulate.  So now that I have you understanding that modulation is made up of symbols containing multiple bits of data, we can talk “data rate”, which is very straight forward.

Data rate = #_Bits_per_symbol * Symbol_Rate

Example: For 16-QAM with a symbol rate of 2560 Ksym/sec.

Symbol_Rate = 2560 ksym/sec

#_Bits_per_symbol = 4

Data rate = 4 bits * 2560 ksym/sec = 10.24 Mbits/sec

This data rate, 10.24 Mbits/sec, happens to be the maximum theoretical data rate for DOCSIS 1.1.  The actual usable data rate for subscriber “protocol data units” (PDUs) is roughly 8 Mbits/sec because DOCSIS requires some amount of overhead for DOCSIS protocol messaging, which is communications between the cable modem and CMTS, that will be covered in a later blog.

Now about that root raised cosine shaping filter…  The purpose of that filter is to minimize the RF harmonic energy produced by the cable modem so that it does not produce interference with adjacent devices.  The way it does this is by acting as a low pass filter by removing all of the high frequency components of what would normally be a very “square wave” looking signal.  It is what make DOCSIS upstream signals look like a “haystack” (don’t worry, if you haven’t seen a DOCSIS haystack yet, I’ll show you one in a blog post shortly).  The reason we want to know about the \alpha=o.25 factor is because it allows us to calculate the occupied RF bandwidth for any given symbol rate that a cable modem may be using.  This is done by multiplying the symbol rate by (1 + \alpha).  So in the case of our example problem above:

Occupied Bandwidth = (1 + \alpha) * 2560 ksym/sec

= 1.25 * 2560 ksym/sec = 3.2 MHz

See, its all pretty simple stuff and can now be applied to the rest of the DOCSIS specifications.  So some of the changes that occurred in the newer specs are as follows:

  • DOCSIS 2.0 added higher order modulations including 8-QAM, 32-QAM and 64-QAM.  In addition a higher symbol rate was added at 5120 ksym/sec.  So at 64-QAM and 5120 ksym/sec one can achieve a theoretic limit of 30.72 Mbits/sec in the upstream or ~27 Mbits/sec of user PDU after overhead.  DOCSIS 2.0 also added some other advanced features that I will address in a future blog.
  • DOCSIS 3.0 added a fantastically cool feature called “channel bonding”, both in the downstream and in the upstream.  Channel bonding essentially allows one to take up to four upstream DOCSIS channels and transmit data over them as if they were one single channel.  So for a 64-QAM, 5120 ksym/sec channel as in the DOCSIS 2.0 bullet above, you can bond four together and get a theoretical maximum data rate of 122.88 Mbit/sec.  I will soon be devoting a whole separate blog to DOCSIS 3.0, so stay tuned.

A big incentive for these changes hass been to increase the upstream bandwidth.  This was successful with each revision.  D2.o increased three times from D1.1 (10.24 Mbit/sec to 30.72 Mbit/sec).  D3.0 increased four times from D2.0 (30.72 Mbit/sec to 122.88 Mbit/sec).  What was the driver?  Both competition from telecom operators such as Verizon’s FiOS and AT&T’s U-Verse as well as subscribers using more applications on the upstream.  When DOCSIS was first launched most subscribers were downloading web pages, but today peer-to-peer file sharing, Voice-over-IP (VoIP), and many emerging applications are enabling symmetrical data utilization rather than the previous model of asymmetrical-dominant downstream traffic models.

I have now laid the foundation of RF fundamentals for my DOCSIS blogging moving forward.  Please feel free to post comments on any questions you have or areas where I should elaborate a little more on as I am next going to dive into the heart of the DOCSIS specification and start breaking down cable modem registration and how the DOCSIS specification enables data communication while managing hundreds of cable modems in the rather harsh environment of a two-way, outdoor cable TV plant.

This book is a must read if you want to understand the details of the cable modem PHY!  The price is right too… 

Posted in: DOCSIS 101   Tags: , , , , , , , , , , , , ,

23 Responses to DOCSIS and Cable Modems – How it works :: Upstream RF

  1. Ishwar says:
    February 10, 2009 at 11:09 am

    simply excellent I enjoy reading ur blogs

    Reply
  2. Carol says:
    February 26, 2009 at 3:20 pm

    Great post. I’m learning a lot from your Docsis 101 series. Please keep it rolling!

    Reply
  3. Tom says:
    April 1, 2009 at 9:19 am

    Hi Brady -

    Great stuff! I am confused about one thing though… it seems like QPSK and QAM can be mixed and I am confused. How can a Burst preamble be in QPSK, payload be transmitted in 16QAM over a channel that is specified as 64QAM? What am I missing?

    Thanks!

    Tom

    Reply
    • Brady says:
      April 21, 2009 at 11:43 am

      Hi Tom,

      I think that maybe you are thinking the downstream is configured for 64-QAM, which could be true. The downstream also supports 256-QAM. The upstream can be configured for QPSK, 8-QAM, 16-QAM, 32-QAM and 64-QAM.

      -Brady

      Reply
  4. Chand says:
    April 28, 2009 at 9:18 pm

    thank you, you are doing great. keep going…I’m learning a lot from your blog.

    Reply
  5. Nicolas says:
    May 13, 2009 at 5:51 pm

    For telephony applications with noisy nodes, we will on the edge on service problems… i would like to know if the lowest modulation profile is the better way….

    Reply
  6. Jim says:
    August 19, 2009 at 9:05 pm

    What is the relationship between upstream CER of a CMTS to BER? I am trying to understand is it possible to compare the upstream CER at a CMTS to the pre and post BER using a meter capable of reading BER, MER, EVM etc with a CW 64 QAM generator.

    Reply
  7. Kane says:
    October 26, 2009 at 3:27 pm

    Hi Brady,

    This is great work you are doing.
    Please clarify, the speeds you quote are from the cmts to the cable modem (located in the customers premises) or are they speeds from the cmts to the nearest node ?

    Thanks
    Kane

    Reply
    • Brady says:
      November 6, 2009 at 10:18 am

      Kane,

      The speeds I am quoting in this post are the raw data rates from the cable modem (in a subscribers house) to the CMTS. Actual data rates that a home user will see will depend upon the rates that a cable operator allows the cable modem to operate. These rates are limited by a file downloaded to the cable modem called a configuration file (or config file) during the cable modem registration process. Often times you will hear of these rates as 5/1 Mbps. Such a rate speed means that you can download 5 Mbps and upload 1 Mbps, even though DOCSIS is capable of much more. But remember, you have to share the bandwidth with your neighbors. ;-)

      Regards,
      -Brady

      Reply
  8. okinawajoe says:
    December 1, 2009 at 10:46 am

    Thanks very much for the informative and concise information. I expect you will cover this in one of the later blogs (making my way through) though I am curious what standard “best practice” deployments look like with respect to subscriber oversubscrition ratios and the provisioned service levels. I am interested in the capacity planning aspect to determine how to best qualify service areas for DOCSIS 3.0, and how realistic a 3/1 config is for an 800 user segment running at 64QAM DS / 16QAM US. (current environment) Are there any formulas for deriving best practice data rates from typical peak hour usage … something like a BHCR in the POTS world?

    Reply
  9. Henk says:
    March 25, 2010 at 5:10 pm

    Hi Brady,

    I recently learned about your blogs and think these are great, easy to read and very informative…

    One question on upstream rates from my side:
    Why do you think can it be that my DOCSIS 2.0 CMs (different vendors) can only reach a max upstream of ~7 Mbps (= 10.27 theoretical) with 64QAM US channels on the CMTS and a large enough US (sustain)rate config on the CMs?
    DOCSIS 3.0 CMs on the same channel(s) are doing just fine with 4*24 Mbps.

    Thanks and regards,
    -Henk

    Reply
    • Brady says:
      April 2, 2010 at 5:50 pm

      Hi Henk,

      It sounds like your modems are capped to 7 Mpbs by either the config file or the CMTS. Or they are parked on an upstream channel that is only 3.2 MHz BW at 16-QAM (~7 to 8 Mbps max throughput). Otherwise you should be seeing a lot more than 7 Mbps in the upstream. I have seen a number of modems that will bog down in the mid-teens (14-15 Mbps) even with a 64-QAM upstream and a 6.4 MHz BW, but that is a limitation of firmware. Usually a FW upgrade will fix the problem. 7 Mbps sounds like a configuration problem in the config file or the CMTS.

      Regards,
      -Brady

      Reply
  10. Peter says:
    August 26, 2010 at 2:15 pm

    Any info on whether DOCSIS specifies the frequency resolution for setting the center frequency of a channel. Of course the UCD shows a resolution of 1Hz. My question is if you have some special application where you want to use an arbitrary center frequency, what is the resolution for setting it?

    If the isn’t anything in Docsis about it, is there any controlling document that specifies what the frequency resolotion can be?

    Same question applies for the downstream as well.

    Reply
    • Brady says:
      August 30, 2010 at 8:29 pm

      Hi Peter,

      For the upstream, the DOCSIS 3.0 RFI specifies the upstream signal accuracy as being within 50 ppm (parts per million) over an operating temperature range of 0C to 40C for 5 years from the day the cable modem was manufactured. So how do we translate 50 ppm to frequency? Simple, just use this equation:

      delta f = (center freq * x ppm) / 1e6

      So if your cable modem is transmitting at a center frequency of 35 MHz
      Your x ppm = 50

      Then your allowed cable modem error (or delta f) is:

      delta f = (35 MHz * 50 ppm) / 1e6 = 1,750 Hz = 1.75 KHz (error)

      Then your minimum frequency could be 34.998250 MHz and your maximum frequency could be 35.001750 MHz and you would still be within the DOCSIS specification.

      For the downstream accuracy, the DOCSIS specification just defines this with an straight forward spec. of +/- 30 kHz. So no math involved.

      Hope that helps.

      -Brady

      Reply
  11. End says:
    January 31, 2011 at 4:07 pm

    Hi Brady,
    excellent post!
    One question about the 64QAM for Motorolla CMTS. We reach only max upstream throughput of 4Mbps with 3,2MHz, but changing de bw for 6,4 MHz the throughput turn regular.
    This problem there isn´t for CISCO CMTS.

    Thanks

    Reply
  12. Arthur says:
    February 8, 2011 at 5:39 am

    Hi Brady,

    I’m trying to figure out what docsis version my provider supports.
    Weird thing is, the symbol rate as I can see at my modem diag page, is 2.560Msym. So that should team up with 16QAM for docsis 1.1.
    But my upstream modulation is 64QAM.
    So maybe I have docsis 2.0?
    The only reason my provider is not doing 5.12 Msym could be that they have still a lot of docsis 1.1 modems in the field.
    Is this technically possible? I mean, combine 64QAM with 2.560Msym?

    Cheers,
    Arthur.

    Reply
    • Brady says:
      February 17, 2011 at 1:47 pm

      Hi Arthur,

      If your modem is running 64-QAM in the upstream, then your operator supports at least DOCSIS 2.0 (maybe even DOCSIS 3.0). It is not uncommon for an operator to run a 64-QAM modulation with a low symbol rate. Why? Because using the higher modulation with a lower symbol rate gives higher data rates in the same bandwidth with reasonably good impairment immunity when using the DOCSIS 2.0 pre-equalization and adaptive signal cancellation available in the DOCSIS 2.0 and higher chip sets.

      So your your cable operator is doing the right thing, making the most out of their upstream. I would imagine that they likely have some upstream issue to still resolve before they can go full 64-QAM, 6.4 MHz, but until then they are using a smaller bandwidth (3.2 MHz). It is also likely, as you suggest that they are using A-TDMA where the legacy modem transmit data in 16-QAM on the same channel, which is perfectly normal. The DOCSIS 1.1 modem just uses a different burst descriptor (IUC) in order to do this.

      -Brady

      Reply
  13. NOC Tech82 says:
    February 11, 2011 at 4:11 pm

    Brady,

    Thanks for the all the great HFC information on your sight. I mainly work on metro ethernet & SONET but still get allot of CMTS work at my job and this site has really helped.

    My question:

    When did cable providers first get the chance to offer cable modem returns on the cable network without relying on the telco return architecture? Based off my research it appears this was achieved with the launch of 1.1.

    Thanks

    Reply
    • Brady says:
      February 17, 2011 at 10:54 am

      Hi NOCTech82,

      There were actual pre-DOCSIS deployments in the mid-90′s, but these were done without using an inter-operable specification. The first DOCSIS 1.0 spec was developed in the late 90′s (released in 1999 I believe), but actual deployments to subscribers did not occur until about 1990. It took awhile before the last non-DOCSIS LanCity modems where taken out of service. So the cable upstream was used for data since in a non-telephony method since at least 1994, but I bet there is a geek out there who can date it back even before then. :-)

      -Brady

      Reply
  14. Sam says:
    February 20, 2011 at 6:52 pm

    Fantastic Blog. I’m a network sales engineer just getting up to speed on DOCSIS and this is such a great read! I’ve been reading the specs but the explanation you give creates a visual perspective that helps comprehend the material.

    Thank you! I plan to keep reading.

    Reply
  15. Daswang says:
    December 29, 2011 at 5:22 am

    Hi Brady,

    I’m a new guy for DOCSIS, and I have a question about the symbol rate. The symbol rate is increased a lot from DOCSIS1.x to DOCSIS2.0, and I want to know what is the most important thing which will affect the symbol rate on HFC? Both for downstream and upstream.

    Thanks,
    Daswang.

    Reply
    • Brady says:
      December 30, 2011 at 12:27 pm

      Hi Daswang,

      This is a very good question! Symbol rate and bandwidth are directly related to one another. Also, the higher the symbol rate the higher the data rate (think faster data speeds), so we ideally we want big symbol rates, but that means big bandwidths. There are a couple of problems with big bandwidths. The upstream has a limited usable bandwidth and the larger the channel bandwidth the greater the chance it will be impacted by noise or other upstream impairments such as group delay.

      So what is the relationship between symbol rate and bandwidth? As the symbols are passed through the root-raised cosine filter in the cable modem to filter away high frequency components, it is up-sampled with a resulting factor of 1.25 (rrc shaping alpha = 0.25). So the bandwidth is 1.25x the symbol rate. A symbol rate of 2560 ksym/sec is then 2560 ksym/sec * 1.25 = 3.2 MHz.

      A number of studies have been down which show that two 3.2 MHz channels running at 64 QAM are much more resilient to RF impairments than one 6.4 MHz (5120 ksym/sec) channel at 64 QAM. Why? The biggest impairment is group delay which has a substantial impact over a large bandwidth channel like 6.4 MHz.

      So to answer your question succinctly, the most important thing which will affect the symbol rate on HFC is RF impairments as the symbol rate increases. The smaller the symbol rate, the less impact by RF impairments; similarly the larger the symbol the more impact by RF impairments.

      -Brady

      Reply

Leave a Reply

Your email address will not be published. Required fields are marked *

*

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>

  • Search This Site:

  • Ads

  • Facebook

bradyvolpe.com is Stephen Fry proof thanks to caching by WP Super Cache