Hanging On To Old DSL Myths?

They Could Be Making Your SOPs More Difficult Than They Need Be

Letter from a reader

Vernon,

I spoke to Don McCarty and he suggested I speak with you. I am looking for some guidance on bonded DSL. We have areas that are vectored and areas that are not. Our SNR standard for vectored is 10 dB and non-vectored is 12 db. It is set to rate adaptive based on SNR. My question is with bonded copper pairs. How much difference between the pairs is considered too much? For example, if I had a 50-Mbps customer, one pair trains at 26 m and the other trains at 24 m. Is the difference a big deal? If not, how much is a big deal? Is it relative to the speed? Is 2 meg not an issue at 100 Mbps bonded but is at 25 m bonded? I have heard that 1 Mbps difference is an issue. I’m just seeking clarity. I also have the same questions about SNR.

Don and I are always glad to hear from our readers. However, your questions make me sad. I am sad because I am again reminded how many myths that were never true, or are no longer true, are still being used as the basis for daily operations. Don’t be insulted by that. More likely than not, the people that designed your DSL networks did exactly what the DSLAM vendor told them to.

I have proven over and again that default profile designs actually hurt subscriber service. For example, if you dropped the target SNRM to reasonable settings and turn on some key features, you would need fewer bonded circuits. You would have never needed Vectoring with the right profile design.

Your current design with single pair, bonded and vectored circuits is the most complicated network I can imagine. Especially trying to make them all work in the same binder group.

The ones with the higher SNRM targets will overwhelm the ones with the lower targets. There is no reason whatsoever to have different SNRM targets for different services. It creates a barfight where the weapons are not equal. That said, it is the most profitable design for the DSLAM vendors. This design causes more noise in the binder group and forces you into Vectoring. Unfortunately, your subscribers are not getting the service they deserve, and your company is wasting millions of dollars a day because of the faulty design of the profiles.

Now that I am off my soapbox, I will answer your questions.

In the beginning and for a few years, the member circuits of a bonded group had to be the same speed. The design could not handle the "traffic cop" issues with much difference between the 2 members. The result was that both member circuits would be set at the rate of the slowest member.

This is no longer true. One member could be at 32K and the other could be 100M and the bonded group will work just fine. Today, that much gap would never be acceptable. Something is wrong with the 32K circuit that needs to be addressed. I like working on bonded circuits because you have a "comparison" member circuit to compare the slow one.

As a rule of thumb:

• If the members are in the same binder group…
• If the members are in the same drop…
• If the members are in the same house wire…
• The difference between the 2 should be no more than 5%. More than that and there is probably something you can fix.

Concerning SNR, the DSLAM’s don’t tell you the SNR. They report the Signal to Noise Ratio Margin (SNRM). The difference is the reference point. The bit rate settings in the profile moves the minimum ratio referenced. More power is required for higher speeds because more bins are being filled and the bins are holding more bits. As with bit rates, the SNRM readings from the member circuits mean nothing to the bonded group. Also, as with bit rates, too much difference in the SNRM means something is going on. I would not assign a standard to it. The difference should be sane; if I had to put a number on it, more than 25% difference would be insane.

I have not quoted any spec. The existing specifications are all based on laboratory results, and more than likely on the same line simulators at every lab. As a former employee with an expired NDA, I can say with full confidence that they do not even resemble real field conditions.

You work for a large provider and they are generally difficult to convince. Lumen has been the only one that I have been able to convince to go against the DSLAM vendor recommendations and use my profile designs. (Dozens of independents use my profile design now.) However, I am willing to try to help if you point me to the right person/people.

Thank you again for writing to us. You can call me anytime you wish if you have more questions: 1.319.238.0285

Vernon

For more information, call Vernon at 1.319.238.0285, email: [email protected], visit https://www.vmaysolutions.com/, and follow him on Twitter @Vernonmay13.