DSL system estimation and control
View Patent ↗Methods, apparatus, computer program products and other embodiments model a system's channel and noise characteristics using one or more Hidden Markov Models (HMMs) by modeling the state dynamics for one or more lines and associated profiles and by learning from the line history. Besides channel and noise characteristics, user data activity also can be modeled using an HMM, where all channel, noise and user data activity can be modeled jointly using HMMs as long as the joint states are properly defined. In multi-line systems, multiple lines (for example, a set of lines in a DSL system binder) may be modeled jointly. After learning channel, noise, user data activity and/or other dynamics using HMM techniques and methods, system operation and/or configuration parameters can be controlled based on the estimated HMM, providing flexible solutions for virtually any channel, noise, user data activity and/or other dynamics.
1. A method of controlling an operational line in a DSL system, the method comprising:
constructing a Hidden Markov Model (HMM) to model one or more internal states of the DSL system, the states characterizing at least one of data activity, impulse noise, crosstalk, noise margin, maximum attainable data rate or bit distributions;
collecting update operational data from the operational line;
analyzing the collected update operational data, the analyzing farther comprising:
estimating an observation probability, a state-transition probability matrix, and an initial state distribution of the HMM based on the collected update operational data;
estimating a likelihood of at least one state of the HMM based on the collected update operational data; and
controlling operation of the DSL system based on the constructed HMM and analyzed update operational data, the controlling further comprising:
determining control parameters based on the estimated likelihood of at least one state of the HMM; and
modifying operation of the operational line through application of the determined control parameters to reflect the information available from the HMM.
2. The method of claim 1
wherein the HMM is constructed to model one or more internal states characterizing at least data activity;
wherein the observation probability, the state-transition probability matrix, and the initial state distribution of the HMM are estimated with the Baum-Welch algorithm;
wherein the likelihood of at least one state of the HMM is estimated with the Viterbi algorithm; and
wherein the control parameters are determined based on the estimated likelihood of at least one state of the HMM only for those states with data activity.
3. The method of claim 1 wherein constructing an HMM to model the internal states of a DSL system comprises:
collecting training operational data from the operational line; and
analyzing the collected training operational data to construct the HMM.
4. The method of claim 3 wherein analyzing the collected training operational data comprises using the Baum-Welch algorithm.
5. The method of claim 3 wherein at least one of the collected training operational data or the collected update operational data is checked for compliance with at least one of the following:
a data timeliness rule; or
a data sufficiency rule.
6. The method of claim 1 further comprising updating the HMM based on the collected update operational data.
7. The method of claim 1 wherein controlling operation of the DSL system based on the constructed HMM and analyzed update operational data comprises at least one of the following:
determining the internal state of the DSL system; or
controlling at least one of the following parameters of the operational line:
data rate;
target SNR margin (TSNRM);
maximum SNR margin (MAXSNRM);
minimum SNR margin (MINSNRM);
FEC parameters;
an FEC delay parameter;
an FEC INP parameter;
transmit power;
a carrier mask parameter;
a power spectral density parameter;
a bit loading parameter;
bit allocation;
PSDMASK;
CARMASK; or
BCAP.
8. The method of claim 1 wherein controlling operation of the DSL system based on the constructed HMM and analyzed update operational data comprises forcing a retrain of the operational line.
9. The method of claim 8 wherein forcing a retrain of the operational line is scheduled to occur when at least one of the following conditions is satisfied:
at a predetermined time;
when user data activity is detected to be below a certain threshold.
10. The method of claim 1 wherein constructing the HMM comprises choosing one HMM from a plurality of candidate HMMs, each candidate HMM corresponding to a characteristic line state.
11. A computer program product comprising:
a machine readable medium; and
program instructions contained in the machine readable medium, the program instructions specifying a method of controlling operation of an operational line in a DSL system, the method comprising;
constructing a Hidden Markov Model (HMM) to model the internal states of the DSL system, the states characterizing at least one of data activity, impulse noise, crosstalk, noise margin, maximum attainable data rate or bit distributions;
collecting update operational data from the operational line;
analyzing the collected update operational data, the analyzing further comprising:
estimating an observation probability, a state-transition probability matrix, and an initial state distribution of the HMM based on the collected update operational data;
estimating a likelihood of at least one state of the HMM based on the collected update operational data; and
controlling operation of the DSL system based on the constructed HMM and analyzed operational data, the controlling further comprising:
determining control parameters based on the estimated likelihood of at least one state of the HMM; and
modifying operation of the operational line through application of the determined control parameters to reflect the information available from the HMM.
12. A device for controlling operation of an operational line in a DSL system, the device comprising:
means for collecting operational data from the operational line;
means coupled to the collecting means for analyzing the collected operational data, the analyzing means further comprising:
means for estimating the observation probability, the state-transition probability matrix, and the initial state distribution of a Hidden Markov Model (HMM) based on the collected update operational data;
means for estimating a likelihood of at least one state of the HMM based on the collected update operational data;
means for controlling operation of the operational line based on:
the collected operational data; and
the HMM modeling the internal states of the DSL system, the states characterizing at least one of data activity, impulse noise, crosstalk, noise margin, maximum attainable data rate or bit distributions.
13. The device of claim 12 wherein the HMM modeling the internal states of the DSL system pertains to modeling at least one of the following:
channel characteristics;
non-impulse noise characteristics; or
user data activity.
14. The device of claim 12 wherein the device is configured to construct and update the HMM by:
using the collecting means to collect operational data from the operational line; and
using the analyzing means to analyze the collected operational data.
15. The device of claim 12 wherein the device updates the HMM as necessary based on the collected operational data.
16. The device of claim 12 wherein the collecting means is configured to check for compliance with at least one of the following:
a data timeliness rule; or
a data sufficiency rule.
17. The device of claim 12 further wherein means for controlling operation of the operational line controls at least one of the following parameters:
data rate;
target SNR margin (TSNRM);
maximum SNR margin (MAXSNRM);
minimum SNR margin (MINSNRM);
FEC parameters;
an FEC delay parameter;
an FEC INP parameter;
transmit power;
a carrier mask parameter;
a power spectral density parameter;
a bit loading parameter;
bit allocation;
PSDMASK;
CARMASK; or
BCAP.
18. The device of claim 12 further wherein means for controlling operation of the operational line forces a retrain of the operational line.
19. The device of claim 18 wherein forcing a retrain of the operational line is scheduled to occur when at least one of the following conditions is satisfied:
at a predetermined time;
when user data activity is detected to be below a certain threshold.
20. The device of claim 18 wherein the HMM modeling the internal states of the DSL system is chosen from a plurality of candidate HMMs.
21. The method of claim 1 wherein the HMM model is dependent on time and the states are modeled as a function of a time index;
wherein the collected operational data pertains to data activity and CV counts over time; and
wherein a forward error correction (FEC) parameter is determined as a function of the time index.
22. The method of claim 1 , wherein the collected operational data comprise bit distributions, wherein the internal states of the model correspond to the presence or absence of at least one disturber, and wherein the determined control parameters comprise BCAP.