SNR Upstream in frequencies lower than 20Mhz. | docsis.org

You are here

SNR Upstream in frequencies lower than 20Mhz.

6 posts / 0 new
Last post
mrdengue
SNR Upstream in frequencies lower than 20Mhz.

I'm having some issues with SNR stability in low freq.

Here's the deal:

We have 4 Upstream channels for each optical node.
Our 4 channels are in 36.8Mhz, 30.4MHz 24MHz and 17.6MHz.
We have 54 nodes that are using the 17.6MHz freq. In around 70% of those 54 nodes, we experience high noise in the 17.6MHz freq. Which results in SNR lower than 25dB.
We are having issues finding out what causes this, so we spent. around 3 or 4 days sweeping the entire node, looking for the source of noise (which usually is a faulty installation inside customers premises)

So my questions are:

Is there a way to pin point where in the HFC Network are the issues without sweeping the entire node?

On the other hand, is there a way to know how many customers are affected when the SNR is extremely low only in that specific channel (17.6MHz)?

I appreciate any help!

Thanks in advance,
Jorge

mbowe
Some ramblings

Noise is always going to be bad in the low end.
Below 20MHz is always going to be troublesome.

For each US, we graph stats like : SNR, user count, utilisation %, uncorrectable FEC, width, qam

Low USSNR is a warning sign. It doesn't necessarily mean the user will have problems.

Same with correctable FEC. Just a warning sign. User wont notice any problems.

Uncorrectable FEC is the key.

If there is no uncorrectable FEC, then there is no packet loss and users will be happy. But you should probably be allocating some time to investigate the noise.

If there is lots of uncorrectable FEC then you have trouble. They are deadly. Any user on that channel will experience packetloss. You need to fix asap.

How to avoid uncorrectable FEC?
Lower width and qam on CMTS will help.
But you dont want to drop these unless you have to (because you will lose network capacity)
So you can configure the CMTS for auto width and auto qam
Eg prefer to run at 6.4w and 64qam, and have the CMTS auto drop to 3.2w and/or 16qam (good combo for the higher channels)

For the bad/low channels another combo might be 3.2w and 16qam, auto drop to 1.6w and QPSK.

We dont run any channels as low as 17.6MHz, our lowest is around 19MHz.
For the low channels we typically set to 6.4 auto drop to 3.2, and hardcode to 16QAM

For tracing noise sources you have a few options. I am not a field tech but I believe our guys use techniques such as :

Techs go to the node and divide/conquer
They can watch headend-based analog spectrum analyser (eg HP cable analyser)
They can use a basic html page on that we wrote that constantly displays CMTS USSNR levels for each US in that node (updates every few seconds)
They can use fancy app on ipad to view output from a headend based spectrum analyser like VSE-1100
They can take VSE-1100 into field and plug it to legs
They can use ingress functions on DSAM (in conjunction with i-stop probe?)

We find that most common source of noise is loose coax flyleads on customer modems :

Every 30 mins we are storing each customer's modem stats into a database, eg , hit/miss, PADJ, FLAP, USPWR, USSNR, DSPWR, DSSNR
We can draw graphs of each statistic
We have noticed that when a customer has loose flylead there will often be a sudden drop in DSSNR and a corresponding jump in USPWR.
By viewing the graphs of modems in the noisy node, we can look for this particular trend and build a list of addresses for techs to consider as suspect, ideally to visit/fix.
If the graph jumps line up with the noise starting/stopping then you can sometimes send the tech directly to this source premises for a very quick fix.

There are also Proactive Network Maintenance (PNM) software tools available.
These tools are polling "equaliser" data from the modems and can help pinpoint network problems.

Noise can happen if network balancing isnt right.
Incorrectly adjusted they can pump up the noise.

mrdengue
Thanks a lot for this answer.

Thanks a lot for this answer. IT was very helpful.. Here are my comments:

Techs go to the node and divide/conquer (We do that)
They can watch headend-based analog spectrum analyser (eg HP cable analyser) (We do that)
They can use a basic html page on that we wrote that constantly displays CMTS USSNR levels for each US in that node (updates every few seconds) (We do that)
They can use fancy app on ipad to view output from a headend based spectrum analyser like VSE-1100
They can take VSE-1100 into field and plug it to legs (We do that)
They can use ingress functions on DSAM (We do that)

We find that most common source of noise is loose coax flyleads on customer modems We arrived to the same conclusion

We have noticed that when a customer has loose flylead there will often be a sudden drop in DSSNR and a corresponding jump in USPWR. Thanks!!, Ill look. into this!! EDIT: Is this only valid with 2.0 CMs? I mean 3.0 have multiple DS channels
By viewing the graphs of modems in the noisy node, we can look for this particular trend and build a list of addresses for techs to consider as suspect, ideally to visit/fix. Excellent Idea!!.. Ill look into that

There are also Proactive Network Maintenance (PNM) software tools available. We made our own PNM tool, but discarded it as we didn't find a exact way to use this data pinpoint problems. Do you know any PNM tools available for download? .

Noise can happen if network balancing isnt right. We solved some noisy nodes using this

mbowe
With regards to the graphing,

With regards to the graphing, and looking for loose flyleads :

We have a script that runs every 30 mins, it logs into CMTS CLI, and executes a few commands such as
* show cable modem
* show cable modem phy
* show cable modem flap

The data is parsed and stored into database.

For each modem :
* DS data (DSPWR, DSSNR etc) only relates to the primary DS. (CMTS doesnt know the per-DS data, because CMTS remote-query only collects primary DS data)
* US data (USPWR, USSNR etc) relates to each US.

Then we have tools which read this database and plot the graphs for a modem.

When there is a loose flylead :
* The DSPWR does not change much
* The DSSNR drops down
* The USSNR jumps up
* The USSNR often has a tilt (the USPWR is not even across the channels)

Obviously these values can individually change at any time, and that is normal.
But the key is to watch for any DSSNR drop which happened at same time as USPWR jump.
We have found this almost certainly indicates a loose flylead.

When you see such a change happen at the same time as the noise starting, its like gold.
Very good chance it is the noise source.

Even when the change doesn't line up with start / stop of noise, its almost 100% hit rate.
If you send a tech to site, the flylead is always loose and this is worth fixing for sure.

We have recently changed our flyleads to use have plastic thumb grip over the top of the connector (PPC SimplEX).
These make it so much easier for customer to tighten correctly.
Much less likely to be under tightened.

windwaterwaves
When the noise is real bad we

When the noise is real bad we do service impacting testing.. have a tech watch the headend ingress. I use a tool called CMTS Simple Monitoring Tool it will show you the cisco upstream noise in real time. It's a windows app that uses SMTP. msarmento.com.br then go half way out segment, pull a pad if it drops you know it's further out. We have hunted down some real problems using this method. If you are fast enough, say 3-5 seconds you may not drop modems...

schg
Has anyone tried SCDMA-D3

Has anyone tried SCDMA-D3 instead of ATDMA for the lower frequency channels?
It's being said, that SCDMA-D3 should offer much better impulse noise protection than ATDMA and much better spectral efficiency than the old SCDMA.

Log in or register to post comments