DOCSIS 3.0 Upstream Tips
(Extended version from that published in Communications Technology, June 1, 2009)
By John Downey and Brady Volpe
In Part I of this extended edition (original abbreviated version appeared in CT’s March 2009 issue John and I discussed many general DOCSIS upstream issues that should be understood prior to deploying DOCSIS 3.0. In this post we will focus more specifically on DOCSIS 3.0 issues that will occur as you are deploying DOCSIS 3.0 or post -deployment.
DOCSIS 3.0 upstream US Considerations
When DOCSIS 2.0 Upstream speed is exhausted, then DOCSIS 3.0 Upstream can be implemented. Some considerations include:
- Frequency Stacking Levels – What is the maximum output with multiple channels stacked, is it p0wer/Hz, could it cause laser clipping?
- Diplex Filter Expansion to 85 MHz – If amplifier upgrades are planned for 1 GHz, then pluggable diplex filters may be warranted to expand to 85 MHz in the US. We still must address existing CPE equipment in the field and potential overload.
- Monitoring, Testing, & Troubleshooting – Just like DOCSIS 2.0, now test equipment needs to have D3.0 capabilities.
DOCSIS 3.0 Upstream Issues
As with DS issues, there are also Upstream issues that need to be addressed. US bonding has not been pursued at this point because most people haven’t even exploited D2.0 US capabilities. This does not mean we should avoid the potential issues that will arise. Eventually, we will want to offer US speeds greater than what a single channel modem can offer of ~ 25 Mbps. This will require more US spectrum, D3.0 CMs, and CMTS linecards with US bonding capability. Some of the potential issues are:
- Levels – Activating multiple frequencies per US connector on a 3.0 CM has different maximum power per channel vs. a D2.0 CM. Maximum transmit for a D2.0 CM using 64-QAM for one channel is 54 dBmV. D3.0 US channel max power is 57 dBmV when using 32 & 64-QAM, 58 dBmV when using 8 & 16-QAM, & 61 dBmV when using QPSK. D3.0 S-CDMA US max power is 56 dBmV for all modulations. As explained above, it can be seen that the max power for one channel on the connector has been raised by 3 dB over a D2.0 CM, but max transmit per channel for four frequencies stacked using 64-QAM ATDMA is only 51 dBmV & 53 for S-CDMA.
- Upstream Passband – The US upper edge has changed in the D3.0 specification to 85 MHz, whereas it was previously 42, 55, & 65 MHz for Euro-DOCSIS. The option of going higher is good for future spectrum re-allocation and avoiding known bad frequencies on the US. Some things need to be considered though and that includes; diplex filters, line EQs, step attenuators, and CPE overload. If incorporating any of these devices in the plant, they may need to be replaced for the new frequency split. Also, can current customer premise equipment (CPE) like settops and TVs handle a potentially high level of “noise” from a modem at 50 MHz or higher?
- Channel Placement – Since each US channel used for bonding is an individual channel, frequencies can be anywhere in the US passband and do not need to be contiguous. Although, it may be wise to keep relatively close so plant problems like attenuation and tilt don’t cause issues. The CM should have some dynamic range to allow specific channels to be a few dB different vs. other channels. Since the transmitters (channels) are separate, they don’t have to be contiguous and can have different physical layer attributes like; modulation, channel width, TDMA or S-CDMA, etc.
- Total Power – Significant consideration must be given to the total RF power loading that will now be realized with US channel bonding in DOCSIS 3.0 Cable Modems. In previous DOCSIS standards, only one US channel was present. For DOCSIS 3.0, up to four (4) US channels will be transmitting at the same time, possibly with a 6.4 MHz bandwidth each, resulting in nearly 26 MHz of US channel loading. This is a lot of power hitting the return path fiber optic transmitter. The probability of laser clipping is increased, especially if one has legacy Fabry-Perot (FP) lasers in the fiber nodes. It is a good idea to upgrade to Distributed Feedback (DFB) lasers, which have significantly more dynamic range. It’s wise to have a comprehensive plan to monitor laser clipping in the return path. Using a return path monitoring system capable of looking above 42 MHz, like PathTrak, will enable you to see second and third order harmonics. Remember, any burst noise above the diplex filter (i.e. 42 MHz) coming out of the return path receiver in the HE is usually indicative of return path laser clipping.
The blue trace in Figure 1 below shows the case of strong laser clipping, while the green line represents a flat noise floor from the return path laser with no clipping. Note that this return has four bonded channels in the US.
Figure 1
Using Return Path monitoring tools like PathTrak or Cisco Broadband Troubleshooter (CBT) to view 5-65 MHz for apparent laser clipping as in the above figure is critical for ongoing preventative maintenance. Also needed is an analyzer that can read < 5 MHz for AM radio or ham radio ingress which can quickly leak into the network and contribute additional power to the return laser causing clipping, as well as cause problems at the input of the CMTS. PathTrak (RPM-3000 card only) also has the ability to demodulate and display live cable modem US constellations as seen in Figure 2 below.
Figure 2
This monitoring system also provides average, maximum, and minimum MER of DOCSIS 3.0 and lesser CMs on each node. In monitor mode it provides the ability to determine relative node quality. It can determine which node has better CM MER than another and during which time of the day MER is the worst, since historical analysis is built-in.
D3.0 CM Transmit Levels
To address the potential issue where a CM today transmits near max power of 54 dBmV for 64-QAM, the specification has changed the CMTS US port level setting to allow it to be 6 dB lower as shown in figure 3 below.
Figure 3
This means the CMTS can be set lower so modems can be placed on those high value taps without changing HE or plant losses. This is at the expense of lower MER/SNR readings. The lowest setting on the CMTS today is -1 dBmV for 6.4 MHz wide channel. The range allowed on the CMTS is dictated by DOCSIS 2.0 and lower and says -1 to + 29 dBmV for 6.4 MHz and related to channel width, also known as symbol rate or baud. D3.0 identified this potential issue and forced D3.0 CM vendors to support a transmit level of 3 dB higher than the 2.0 spec. Therefore, 64-QAM has to transmit at least a maximum output of 57 dBmV with a single channel.
To keep the CM inexpensive and act as a constant power device, the max output will be dictated by how many US frequencies are active on the port (known as TCS for transmit channel set) and the highest modulation used. Four channels stacked will be 57-6 = 51 dBmV per ch. So, the overall effect from 2.0 to 3.0 is really a 3 dB difference. The spec lowered the nominal setting allowed on the CMTS so cable operators didn’t need to lower tap faceplates or drop padding on every US port on the CMTS to keep D3.0 CMs online if they maxed out. Most systems will leave the default of 0 dBmV configured and adjust padding appropriately. Cisco has a command to keep CMs online that are maxed out, power-adjust continue 4 (default). Many customers set it to 6 dB to give more room to work with.
New Architectures
When fiber is run deeper into the network as in the case of RF over Glass (RFoG) or DOCSIS Passive Optical Networks (DPON), a new conundrum is raised. Many of these architectures will incorporate 32-way optical splitter/combiners. Having a laser Tx in your house combined with 32 other houses feeding 1 Rx in the HE is addressed with lasers timed with the actual traffic from the house; unlike how it is done today where the US laser is on all the time. Having US bonding and/or load balancing presents a potential issue where an US laser could be transmitting the same time as another US laser. This may be address with the fact that even though multiple lasers are transmitting the same instant in time, if they are carrying different frequencies, then it may be acceptable. The next question, will S-CDMA pose the same problems? This multiplexing scheme allows multiple modems to transmit the same instant in time.
Isolation Issues
Figure 4 below depicts possible US isolation issues that can occur when an US frequency is narrowcast and another US frequency is introduced across multiple nodes. Creative placement of pads and filters and/or isolation amplifiers can be used to prevent the signal from back-feeding when architectures like these are implemented.
Figure 4
Because of different combining schemes for different nodes as shown in figure 4 above, two USs which are at the same frequency, could interfere with each other. Upstream 2 is being fed from two nodes, while US0 is from one node and US1 is from the other separate node. The 24 MHz signal for US0 will travel to the US2 combiner and back-feed to the fiber receiver number 2, then possibly back-feed to US1 of the CMTS. This 24 MHz signal will be about 40 dB lower than the expected signal if the isolation in each splitter is at least 20 dB. Some ways to increase the isolation are: use amplifiers, add filters, or selectively move attenuation to different points in the network. This situation may also cause issues with load balancing since one port is shared across two other ports. Can US2 load balance with US0 and a different load balance with US1?
Plant Variations
Figure 5
Figure 5 above illustrates system plant design variations; not the in-home wiring variations! This is the level that a home user device would need to transmit (if attached directly to the tap) to arrive at the associated amplifier at its recommended level. If step attenuators, EQ taps, or line equalizers with reverse padding are used to make end of line CMs transmit higher, it addresses two concerns. Since more attenuation is added where there was once less loss for ingress, it allows “ALL” CMs to have a better CNR and maybe a better MER/SNR. The loss from step attenuators and these other options attenuates the ingress from those low value taps and houses attached to those taps. These options also solve the potential issue of modems “ramping up” to max power when a power outage occurs. A typical unit with 55 dBmV output would overdrive the system if located at the end of the cable span.
Frequency Expansion & Tap Change-outs
If systems are looking to do 1 GHz amplifier upgrades, they should look into removable diplex filters in case we change to 5-85 MHz, which DOCSIS 3.0 supports. Also, the truck roll for this is expensive, so doing some tap change outs now would be optimum. The first tap off of the active used to be a 29 dB tap, then got dropped to 26, then dropped to 23 all because of CM US transmit levels, very close to 55 dBmV or so. Modems farther away on low value taps transmit much less, sometimes as low as 35 dBmV. To get this delta much closer (tighter bell curve of CM transmit levels) we need to either add loss (attenuation) to low value taps or maybe we could just change the first few taps off the actives to cable simulator taps. So the tap would look like a 32 dB tap at 1 GHz, but maybe 17 dB at 5 MHz. This solves two issues:
- DS level and tilt hitting the house
- US max transmit levels
Since these modems would go from say 55 dBmV to 49 dBmV, we have room for 3.0 CMs with US bonding and/or we could add padding at the node to force all the CMs higher again and get better SNR. If CM transmit levels are still a concern, using a D3.0 CM in 2.0 mode would allow higher US Tx level because the use of a single frequency and a D3.0 CM offers 3 dB higher power. Running D3.0 in low modulation schemes allows higher power as well. Using S-CDMA with more codes may also allow higher transmit power, but will depend on implementation.
Summary
It is important to recognize that customers and competition are driving the need for increasingly faster speeds in our DOCSIS networks. DOCSIS 3.0 provides an effective migration path for customer satisfaction, exceeding competitive pressures and enabling the delivery of new services and penetration of new markets. With more than four times the data throughput of its predecessors, DOCSIS 3.0 is apt to be nearly an order of magnitude more complex to deploy in such a manner as to take full advantage of its capabilities. Careful planning, proper plant maintenance and a well-developed continued preventive maintenance program from the onset will help pave the road to a speedy and fiscally fruitful deployment.
Upcoming events can be seen under Broadband Events. Previous events can be seen under the blog.
- If you are watching this on youtube please hit the subscribe button!
- Let us know what you think and remember to share!
- You can find slides at the bottom of the page and some on slideshare.
- Find out about events or articles by following us on Twitter, LinkedIn or Facebook too.
Also available on iTunes, Google Podcasts, Spotify, vurbl see podcasts “get your tech on”.
Hi, like to read your valuable information about DOCSIS.
Here is my question, can the US be improved with the new standard to define more “bits” in a block ? instead of having 8-bit for 256-QAM, how about pump up to 16-bit per block ? would this technically be possible ?
Thanks,
Jim Chen
951-907-3210 (cell)
Hi Maurizio,
Currently the DOCSIS specification states that a minimum of four (4) upstream channels can be bonded. So the implication is that more than four can be bonded. I don’t know of any system that is currently doing more than four, in fact few are actually truly doing four upstream bonded channels – its just getting traction now.
-Brady
Hi Jim,
The idea sounds good, but does not work in application. By moving to a higher-order constellation, it is possible to transmit more bits per symbol. However, if the mean energy of the constellation is to remain the same (by way of making a fair comparison), the points must be closer together and are thus more susceptible to noise and other corruption; this results in a higher bit error rate and so higher-order QAM can deliver more data less reliably than lower-order QAM, for constant mean constellation energy.
For 64- and 256-QAM, you are confined to 6 and 8 bits per block, respectively. That fundamentally limits your data throughput. I also added a document to the Library for your reference on QAM in a digital cable plant. You might find it interesting. Check it out https://volpefirm.com library/ANSISCTE072006.pdf”
-Brady
Hi.Is there an formula to calculate how many end amplifiers should be on a physical port of an node.(ex. we have in some cases 24 end amplifiers and upstream modulation 64qam with 6.4Mhz channel and upstream snr reported by cmts is not greater than 30db)
Hi Gabriel,
You have a lot of amplifiers! With that said, there is no formula. Many operators are in North America are migrating to short runs of amplifiers. The terminology is often called Fiber Node + # of amplifiers or Node + 4. In many cases is it is not uncommon to see no more than four amplifiers after the fiber node and a number of operators are even migrating to Node + 1. Why? Because when you have so few actives on your return plant, you end up with very few upstream impairments. further you end up with fewer subscribers per port. Clearly this does not scale in rural areas where you will need many more actives, but it should give you a sense of what is becoming common.
So when you have 24 amps in cascade and an SNR of <30dB, this is not surprising. You have a lot of opportunities for ingress and other plant noise, group delay, etc. to get in. I would recommend that if at all possible you segment your plant with some fiber nodes and you should see your SNR improve.
-Brady
Hi , my question is about fiber node optimization, My company is injecting 38 db at 42.50 MHz thru their Motorola nodes and actives for optimization of new plant, this is new Motorola gear, is this is a mistake? as I think the diplexers should roll off at 42 mhz
Hi Kevin,
Your question has a number of factor which comes into play. If the nodes and actives are 42/54 MHz split, then there will be some attenuation at 42.5 MHz that must be compensated for during plant optimization. However if the nodes and actives are European split (i.e. 65/85 MHz), then the roll-off does not occur until 65 MHz.
So there is a more involved discussion that would need to take place, but your question is a good one and should be reviewed as it could have plant balancing implications if not properly addressed.
-Brady
Hi Bray Volpe.
I’m reading DOCSIS 3.0 but I don’t understand “How about frequency for US and how about frequnce for DS?)
Hi Phan,
Thanks for reading my articles.
I would like to answer your question, but its very general. I would recommend that you begin by reading the DOCSIS Tutorial Series under Blog This should help answer a lot of your questions.
-Brady
Hi, Normally we do amplifier balancing in DOCSIS 2 with single frequency @ 22 MHz at an RF level of 20 dBmv at Return input. In DOCSIS 3 with four channel bonding , what could be the balancing technique ?? any tilt needed and how ?
Hi,
Same technique since every US ch for D3.0 US bonding still does independent station maintenance and treated independently. But, keep mind that max TX level; for 4-ch US bonding using 64-QAM is only 51 dBmV, so modems could max out quicker. US balancing is still fundamentally based on unity gain. Once balance, CM levels can be changed by just manipulating padding in the HE. Also see https://volpefirm.com/cablelabs-f2f-and-acta-along-with-other-topics/
Hi there. In my system we are currently using 4 upstream channels and are beginning to prepare for our DOCSIS 3.1 roll out. We currently have diplex filters at 42/54. One of our first changes is to upgrade our existing FP laser equipped node transmitters with DFB transmitters. I have read that there are issues with the old FP lasers when using 64 QAM upstreams, especially close to the roll off of 42MHz. Since an optical receiver is basically a laser run backwards, will the same effect be experienced if the FP equipped receivers are not changed? Meaning that the lasers in the nodes would be changed to DFB, but not the receiver?
Hi Ed,
Running near the diplex rolloff can be an issue for that US ch’s MER readings especially for 64-QAM and wider chs like 6.4 MHz. Also worse when running through long amplifier cascades because of the cumulative effect of group delay. Easiest fix is lower the freq or activate pre-eq (equalization-coefficient. Or run narrower ch like 3.2 MHz or lower amplifier cascade. FP laser has nothing to do with the grp delay issue associated with diplex filter edge. With that said, FP lasers have much lower dynamic range vs DFB lasers and can be overdriven much easier and cause non-linear mixing from laser clipping. Digital reverse lasers are even better and also the new architectures coming soon called remote phy will get rid of analog lasers altogether. The optical receiver is a PIN diode or avalanche diode. I think you are using the term receiver to mean the node, but not sure. – JJD
Hi Guys,
Just feel free to answer this question to me as detailed as possible within your valued time.
Hi Guys,
Just feel free to answer this question to me as detailed as possible within your valued time.
I am working on deploying 6.4 MHz ATDMA project. We only have 15-30 MHz spectrum in the region so with this limitation decided to do 2 3.2 ATDMA channels and 1 6.4 ATDMA, I bounded these three channels into one upstream bonding group and a test D3.0 Modem hooked up in Headend came online and working.
My question how the load –balancing with work on upstream channels having different width and bounded in one US bonding group. I am under the impression that they would just load balance equally across all bonded three upstream. Hence D3.0 will not do proper load-balancing. Am I right? If not then how, and what I can do best to optimize my mac-domains that have different width upstreams bonded.
Thanks,
Hi Irfan,
Thanks for reading and the question. Your question was selected to be part of our podcast. You can watch or listen here. https://volpefirm.com/docsis-cmts-best-practices-recommendations/
Thanks or your interest.
Mia