Avalon Networks Avalon Networks

Modem Evaluation on Lines

Modems Tested
Maximum Reach
Maximum reach of the modems was tested by connecting the ADSL units to each of our lines and noting whether the modem connected or not. In our trial, the lines happened to have clustered around round number lengths - 5K, 10K, and 15K. Because of this, some care must be taken in interpretation of these results - some modems may look much worse than others when in fact they're close. We note however for our purposes of a deployment that 15K feet reach is a minimum, so this was not a significant limitation for us.



Throughput Testing Methods
A number of prior tests have been done on ADSL equipment using a variety of thruput testing methods. In general, these tests tend to be done using existing network tools such as FTP and they tended to be unidirectional. In this test, our goal was to stress the ADSL modems in a bidirectional fashon. We wanted to have absolute control of the packet generation including speed in each direction, packet size, and payload. The tool we used that met these requirements was the Fluke xDSL One Touch.



A Fluke xDSL One Touch was connected to the ATU unit at each end of the line under test. Test parameters for our test were as follows: It's important to note with these units that you should not significantly overdrive the unit beyond what you expect for thruput and the test should be run as long as possible. The reason for this is that some units have internal buffers, and this can skew the results.

Example images while running an ADSL test using the Fluke xDSL One Touch.



Modem Speeds at 5K feet




Modem Speeds at 10K feet




Modem Speeds at 15K feet




Reliability Testing
We were interested in evaluating the modems for reliability. In particular, we wanted to find out how often the ADSL modems would retrain. Several of the ADSL modems running early firmware tended to do this, and we wanted to find out if this existed after several updates. In this test, the modems were deployed on a single loop of 5, 10, or 15K feet. Our test was using a modified unix ping program. The measurement system at the POP was a unix based system. At the remote end, we used the ATU-R management when available, and a unix system for bridge-only products that lacked management. Test duration during this experiment was 12-72 hours per line. The following table sumarizes the results of this test:

Modem 5K 10K 15K
Amati 0%   0%
Aware 0%   1%
Netspeed 0% 0%  
Pairgain 768 0% 0%  
Pairgain CRA 0% 0%  
Pairgain Falcon   0% 0%
Paradyne 0%   0%
3COM/USR 0% 0%  
Westell 0%   0%

There was no significant packet loss with any of the modems.

Copyright (c)1998 Avalon Networks

Dave Lacey and Mike Lutz
To contact us check out the contacts page.