Upstream SNR problem |

You are here

Upstream SNR problem

7 posts / 0 new
Last post
Upstream SNR problem
Image icon upstream 1.png16.19 KB
Image icon upstream2.png14.02 KB


I am currently facing some issues with an unstable upstream SNR.
As the two pictures below show two modems connected on different nodes you can see the difference in upstream SNR.
The first one has many ups and downs and is very unstable ( upstream SNR varies from 25-33 ), this value is common in most of the nodes.
The second one is more stable and has a higher value of SNR, about 36.1 ( we were working on the line when there was a drop that lasted 2 hours ) than the first one.
We have checked the nodes and amps and there are no signs of noises.

DS Modulation : 256QAM
US Modulation : ATDMA
US Channel width 3.2 mhz

What can cause this issue and what can we look after to solve it?

Thanks in advance!



noise is due to multiple reasons.
Abnormal noise for example appeared when you have a:
- problem of transmit power levels of your modems
- cable network not correctly protected from hertzian attacks (old amplifiers, old cable, open wallets, bad connectors/connections...)
- high signal generating noise near your network (lights, GSM/LTE/DVB-c..., police freq, bus, baby phones....)
- reflected noise from modems (more you have modems on the same node more you will have noise)
to solve it, you have to go on site and check noise locally.
You need to correlate modems with problems (high power level, noise...) in order to find addresses where you need to control your network

good luck

In a perfect world, you will

In a perfect world, you will be running 6.4MHz width, 64QAM and the USSNR will be flat 36.

But as we all know, cable networks aren't perfect, and it's easy for noise to get into the system.

DOCSIS is able to handle some noise, but if there is too much you will end up with packetloss / dropouts.

The thresholds for when the loss kicks in will depend on your combination of width and QAM.

Generally if the USSNR is > 27 or so, your end users shouldn't notice any impact, even with 6.4 width and 64 QAM.

If you drop from 6.4 to 3.2, it will boost the USSNR by about 3dB, at the expense of losing around half the channel capacity. Same again if you go from 3.2 to 1.6 MHz width.

If you drop from 64QAM to 16QAM, your system will be able to run OK at lower SNR, again at the expense of channel capacity.

One handy feature is to configure the CMTS to run at high width/QAM values, and automatically drop back when SNR or uncorrectable thresholds are hit.

Also you have control over what frequencies to place the upstreams. I like to use highest possible frequencies as noise is usually worse in the lower ranges. You can hardcode these frequencies or setup a spectrum group on the CMTS to allow it to try and auto-locate the cleanest spectrum.

The CMTS has some other techniques to counter noise. Make sure you have got the "ingress-noise-cancellation" feature activated, and also assuming you don't have some really ancient modems in your plant you should turn on the "equalization-coefficient" feature.

In terms of how to fix the noise, I reckon there are a couple of main tips :

1. Make sure the node is correctly balanced. If the balance isn't right, the amps will can pump up small amounts of noise and make them very big. This includes making sure the coax amps are correct, optical links are correct, and headend/cmts padding is correct.

2. Once the node is confirmed to be balanced, it is usually a divide and conquer type approach. A few different methods available, but the coarsest one that describes the technique is to start at the node and start disconnecting legs one at a time. Keep an eye on the SNR. Once you have found the leg with the noise, you have to proceed down it and repeat the process until you find the source.

Maybe you can check which of

Maybe you can check which of the cm are sending uncorrectable codewords (from scm verbose command). Then you try to tackle these modems by either reseting them and see which of them when you resets improve the US MER

Look at your return

Hook an analyzer up in your headend, set it to look at 5-50mhz, and hook it into your return splitter that should be on the output of your return receiver. If you can hook the video out to a modulator, and put it out on an empty channel in your system, you can use a small battery-powered tv to help you track and eliminate the noise in your return. 90% of ingress and return noise will come out of a customer's house(loose connectors, bad/damaged cable etc). Once in a while you'll find a mainline crack with it, but usually you'll pick those up with your leakage meter before you chase it down on the return.

Along these same lines we

Along these same lines we have been tinkering with connecting the video out from headend spectrum analyser to a security camera box that lets you view the output via a web page (MPEG-4 video stream). Can then view on ipad or phone. Worked OK with initial trial with cheapy ebay box, the video update speed was great, looked pretty much real time. But we are currently looking for one that doesn't need a dodgy unsigned plugin installed in the web browser.

Another tool we wrote/use is a webpage where the field tech can bring up a stats page for any node. It has big font (easy to read) live CMTS USSNR readings, plus 24 hour graphs for each upstream. Sometimes also we show the uncorrectables counters. The page auto-refreshes every few seconds. The techs can unplug / pad legs and watch the SNR in pretty much real time. They use their phones for this, and reckon it works very well. They said they cant imagine doing without this tool now.

They also use their JDSU DSAMs, but some of these other headend-based tools provide them with extra options for tracking down problems.


US noise and low SNR is very easy to find.
SNR is the easiest noise source to track.
Use your US spectrum, and perform isolations if allowed.

Log in or register to post comments