Gradient based method and apparatus for OFDM sub-carrier power optimization

• US 20060153309A1
• Filed: 01/12/2005
• Published: 07/13/2006
• Est. Priority Date: 01/12/2005
• Status: Active Grant

• Current Assignee: Nokia Corporation
• Original Assignee: Nokia Corporation
• Inventors: Stolpman, Victor J., Tang, Clive K.
• Granted Patent: US 7,453,966 B2
• Time in Patent Office: Days
• US Class Current: 375/260
• CPC Class Codes: H04L 5/0044 allocation of payload

Abstract

In one aspect thereof this invention provides a method to increase data throughput in an Orthogonal Frequency Division Multiplex (OFDM) wireless communications system. The method includes modeling data throughput as a function of received Signal to Noise Ratio (SNR) with a sigmoid function to produce a throughput curve characterized as having a lower knee region and an upper knee region; and performing a gradient ascent search along the throughput curve by varying the power of a plurality of sub-carriers so as to remove power from any sub-carrier having a received SNR to the left of the lower knee region, and to re-allocate the removed power to sub-carriers having a received SNR to the right of the lower knee region.

Claims

1. A method to increase data throughput in an orthogonal frequency division multiplex (OFDM) wireless communications system, comprising:

modeling data throughput as a function of received Signal to Noise Ratio (SNR) with a sigmoid function to produce a throughput curve characterized as having a lower knee region and an upper knee region; and

performing a gradient ascent search along the throughput curve by varying the power of a plurality of sub-carriers so as to remove power from any sub-carrier having a received SNR to the left of the lower knee region, and to re-allocate the removed power to sub-carriers having a received SNR to the right of the lower knee region.

2. The method of claim 1, where different modulation and coding schemes are used in individual sub-carriers at different times.

3. The method of claim 1, further comprising improving error rate performance by disabling any sub-carrier whose power is removed.

4. The method of claim 2, further comprising improving error rate performance by disabling any sub-carrier whose power is removed.

5. The method of claim 1, further comprising, for sub-carriers between the upper and lower sigmoid curve knee regions, increasing the power of the sub-carriers within a total power constraint so as to be to the right of the sigmoid curve upper knee region.

6. The method of claim 1, where performing the gradient ascent search comprises:

initializing the power of all sub-carriers to be equal, such that all elements of a transmitted power vector P⁽ⁿ⁾ are set to have the same value;

evaluating a resultant total data throughput using a current value of the transmitted power vector P⁽ⁿ⁾ to obtain a first data throughput value;

finding a search direction vector g_n;

determining a step size α

_n;

updating the transmitted power vector P⁽ⁿ⁾;

evaluating a total data throughput with the updated transmitted power vector P⁽ⁿ⁾ to obtain an updated data throughput value; and

if the updated data throughput value is improved relative to the first data throughput value, completing an iteration of the gradient ascent search, incrementing n, and returning to finding the search direction vector to perform another iteration, else if the updated data throughput value is not improved relative to the first data throughput value, or if some predetermined number of interactions have been completed, terminating the gradient ascent search.

7. The method of claim 6, where after terminating the gradient ascent search P⁽ⁿ⁾ is a final power vector that dictates the power in the plurality of sub-carriers, each of the sub-carriers is loaded with a data symbol from an associated packet of a set of packets forming an OFDM symbol to be transmitted through a channel, where the transmitted power vector P⁽ⁿ⁾ is maintained unchanged throughout the transmission of the set of packets, and upon completion of a frame of OFDM symbols, returning to initialize the power of all sub-carriers to be equal and repeating the gradient ascent search for a next set of packets.

8. The method of claim 6, where after terminating the gradient ascent search the transmitted power vector P⁽ⁿ⁾ is a final power vector that dictates the power in the plurality of sub-carriers, the method further comprising:

spreading a plurality of data packets across the sub-carriers, where the plurality of packets spread across the sub-carriers form an OFDM symbol;

transmitting the OFDM symbol through a channel;

maintaining the transmitted power vector P⁽ⁿ⁾ unchanged throughout the transmission of the OFDM symbol; and

upon completion of transmission of the OFDM symbol,

returning to initialize the power of all sub-carriers to be equal and repeating the gradient ascent search for a next OFDM symbol.

9. The method of claim 8, where the data packets are spread across the sub-carriers in a manner to ensure that the data packets share similar error probabilities.

10. The method of claim 6, where after terminating the gradient ascent search the transmitted power vector P⁽ⁿ⁾ is a final power vector that dictates the power in the plurality of sub-carriers, the method further comprising:

selecting a time interval corresponding to the time required to transmit at least two OFDM symbols in succession, where the transmitted power vector P⁽ⁿ⁾ will be maintained unchanged throughout the selected time interval;

spreading a first plurality of data packets across the sub-carriers, where the first plurality of packets spread across the sub-carriers form a first OFDM symbol;

transmitting the first OFDM symbol through a channel;

spreading a second plurality of data packets across the sub-carriers, where the second plurality of packets spread across the sub-carriers form a second OFDM symbol;

transmitting the second OFDM symbol through the channel;

continuing to successively spread pluralities of data packets across the sub-carriers and to transmit the resulting OFDM symbols through the channel until the selected time interval is reached; and

returning to initialize the power of all sub-carriers to be equal and repeating the gradient ascent search for a next time interval.

11. The method of claim 10, where the data packets are spread across the sub-carriers in a manner to ensure that the data packets share similar error probabilities.

12. The method of claim 2, where performing the gradient ascent search comprises:

initializing the power of all sub-carriers to be equal, such that all elements of a transmitted power vector P⁽ⁿ⁾ are set to have the same value;

evaluating a resultant total data throughput using a current value of the transmitted power vector P⁽ⁿ⁾ to obtain a first data throughput value;

finding a search direction vector g_n;

determining a step size α

_n;

updating the transmitted power vector P⁽ⁿ⁾;

updating the received SNR for each sub-carrier using the sub-carrier'"'"'s power allocation as indicated by the updated transmitted power vector P⁽ⁿ⁾;

sub-channel frequency function; and

noise power;

comparing the updated received SNR for each sub-carrier to a look-up table created with cross-over points of the sigmoid curves and selecting sigmoid parameters that result in highest throughput for the sub-carrier;

evaluating a total data throughput with the updated transmitted power vector P⁽ⁿ⁾ to obtain an updated data throughput value; and

if the updated data throughput value is improved relative to the first data throughput value, completing an iteration of the gradient ascent search, incrementing n, and returning to finding the search direction vector to perform another iteration, else if the updated data throughput value is not improved relative to the first data throughput value, or if some predetermined number of interactions have been completed, terminating the gradient ascent search.

13. The method of claim 12, where after terminating the gradient ascent search P⁽ⁿ⁾ is a final power vector that dictates the power in the plurality of sub-carriers, each of the sub-carriers is loaded with a data symbol from an associated packet of a set of packets forming an OFDM symbol to be transmitted through a channel, where the transmitted power vector P⁽ⁿ⁾ is maintained unchanged throughout the transmission of the set of packets, and upon completion of a frame of OFDM symbols, returning to initialize the power of all sub-carriers to be equal and repeating the gradient ascent search for a next set of packets.

14. The method of claim 12, where after terminating the gradient ascent search the transmitted power vector P⁽ⁿ⁾ is a final power vector that dictates the power in the plurality of sub-carriers, the method further comprising:

spreading a plurality of data packets across the sub-carriers, where the plurality of packets spread across the sub-carriers form an OFDM symbol;

transmitting the OFDM symbol through a channel;

maintaining the transmitted power vector P⁽ⁿ⁾ unchanged throughout the transmission of the OFDM symbol; and

upon completion of transmission of the OFDM symbol,

returning to initialize the power of all sub-carriers to be equal and repeating the gradient ascent search for a next OFDM symbol.

15. The method of claim 14, where the data packets are spread across the sub-carriers in a manner to ensure that the data packets share similar error probabilities.

16. The method of claim 12, where after terminating the gradient ascent search the transmitted power vector P⁽ⁿ⁾ is a final power vector that dictates the power in the plurality of sub-carriers, the method further comprising:

selecting a time interval corresponding to the time required to transmit at least two OFDM symbols in succession, where the transmitted power vector P⁽ⁿ⁾ will be maintained unchanged throughout the selected time interval;

spreading a first plurality of data packets across the sub-carriers, where the first plurality of packets spread across the sub-carriers form a first OFDM symbol;

transmitting the first OFDM symbol through a channel;

spreading a second plurality of data packets across the sub-carriers, where the second plurality of packets spread across the sub-carriers form a second OFDM symbol;

transmitting the second OFDM symbol through the channel;

continuing to successively spread pluralities of data packets across the sub-carriers and to transmit the resulting OFDM symbols through the channel until the selected time interval is reached; and

returning to initialize the power of all sub-carriers to be equal and repeating the gradient ascent search for a next time interval.

17. The method of claim 16, where the data packets are spread across the sub-carriers in a manner to ensure that the data packets share similar error probabilities.

18. The method of claim 3, where performing the gradient ascent search occurs in stages, where during a first stage the method identifies sub-carriers to be disabled, and where during a second stage the method allocates power to retained sub-carriers at the expense of disabled sub-carriers.

19. The method as in claim 3, where performing the gradient ascent search comprises:

initiating a first stage of the gradient ascent search by initializing the power of all sub-carriers to be equal, such that all elements of a transmitted power vector P⁽ⁿ⁾ are set to have the same value;

evaluating a resultant total data throughput using a current value of the transmitted power vector P⁽ⁿ⁾ to obtain a first data throughput value;

finding a search direction vector g_n;

determining a step size α

_n;

updating the transmitted power vector P⁽ⁿ⁾;

evaluating a total data throughput with the updated transmitted power vector P⁽ⁿ⁾ to obtain an updated data throughput value; and

if the updated data throughput value is improved relative to the first data throughput value, completing an iteration of the gradient ascent search, incrementing n, and returning to finding the search direction vector to perform another iteration, else if the updated data throughput value is not improved relative to the first data throughput value, or if some predetermined number of interactions have been completed,

initiating a second stage of the gradient ascent search by evaluating the received SNR of every sub-carrier resultant from the transmitted power vector P⁽ⁿ⁾;

if the received SNR of a particular one of the sub-carriers is less than a predetermined threshold, identifying that sub-carrier as one to be disabled;

re-assigning power from any identified sub-carriers to remaining sub-carriers; and

returning to finding the search direction vector to form the second stage gradient ascent search for the remaining sub-carriers.

20. The method of claim 19 where after terminating the gradient ascent search the transmitted power vector P⁽ⁿ⁾ is a final power vector that dictates the power in the retained sub-carriers, each of the retained sub-carriers is loaded with a data symbol from an associated packet of a set of packets forming an OFDM symbol to be transmitted through a channel, where the transmitted power vector P⁽ⁿ⁾ is maintained unchanged throughout the transmission of the set of packets, and upon completion of a frame of OFDM symbols, returning to initialize the power of all sub-carriers to be equal and repeating the gradient ascent search for a next set of packets.

21. The method of claim 19, where after terminating the gradient ascent search the transmitted power vector P⁽ⁿ⁾ is a final power vector that dictates the power in the retained sub-carriers, the method further comprising:

spreading a plurality of data packets across the retained sub-carriers, where the plurality of packets spread across the retained sub-carriers form an OFDM symbol;

transmitting the OFDM symbol through a channel;

maintaining the transmitted power vector P⁽ⁿ⁾ unchanged throughout the transmission of the OFDM symbol; and

upon completion of transmission of the OFDM symbol,

returning to initialize the power of all sub-carriers to be equal and repeating the gradient ascent search for a next OFDM symbol.

22. The method of claim 21, where the data packets are spread across the retained sub-carriers in a manner to ensure that the data packets share similar error probabilities.

23. The method of claim 19, where after terminating the gradient ascent search the transmitted power vector P⁽ⁿ⁾ is a final power vector that dictates the power in the retained sub-carriers, the method further comprising:

selecting a time interval corresponding to the time required to transmit at least two OFDM symbols in succession, where the transmitted power vector P⁽ⁿ⁾ will be maintained unchanged throughout the selected time interval;

spreading a first plurality of data packets across the retained sub-carriers, where the first plurality of packets spread across the sub-carriers form a first OFDM symbol;

transmitting the first OFDM symbol through a channel;

spreading a second plurality of data packets across the retained sub-carriers, where the second plurality of packets spread across the sub-carriers form a second OFDM symbol;

transmitting the second OFDM symbol through the channel;

continuing to successively spread pluralities of data packets across the retained sub-carriers and to transmit the resulting OFDM symbols through the channel until the selected time interval is reached; and

returning to initialize the power of all sub-carriers to be equal and repeating the gradient ascent search for a next time interval.

24. The method of claim 23, where the data packets are spread across the retained sub-carriers in a manner to ensure that the data packets share similar error probabilities.

25. The method as in claim 4, where performing the gradient ascent search comprises:

initiating a first stage of the gradient ascent search by initializing the power of all sub-carriers to be equal, such that all elements of a transmitted power vector P⁽ⁿ⁾ are set to have the same value;

evaluating a resultant total data throughput using a current value of the transmitted power vector P⁽ⁿ⁾ to obtain a first data throughput value;

finding a search direction vector g_n;

determining a step size α

_n;

updating the transmitted power vector P⁽ⁿ⁾;

updating the received SNR for each sub-carrier using the sub-carrier'"'"'s power allocation as indicated by the updated transmitted power vector P⁽ⁿ⁾;

sub-channel frequency function; and

noise power;

comparing the updated received SNR for each sub-carrier to a look-up table created with cross-over points of the sigmoid curves and selecting sigmoid parameters that result in highest throughput for the sub-carrier;

evaluating a total data throughput with the updated transmitted power vector P⁽ⁿ⁾ to obtain an updated data throughput value; and

if the updated data throughput value is improved relative to the first data throughput value, completing an iteration of the gradient ascent search, incrementing n, and returning to finding the search direction vector to perform another iteration, else if the updated data throughput value is not improved relative to the first data throughput value, or if some predetermined number of interactions have been completed,

initiating a second stage of the gradient ascent search by evaluating the received SNR of every sub-carrier resultant from the transmitted power vector P⁽ⁿ⁾;

if the received SNR of a particular one of the sub-carriers is less than a predetermined threshold established for that sub-carrier, identifying that sub-carrier as one to be disabled;

re-assigning power from any identified sub-carriers to remaining sub-carriers; and

returning to finding the search direction vector to form the second stage gradient ascent search for the remaining sub-carriers.

26. The method of claim 25 where after terminating the gradient ascent search the transmitted power vector P⁽ⁿ⁾ is a final power vector that dictates the power in the retained sub-carriers, each of the retained sub-carriers is loaded with a data symbol from an associated packet of a set of packets forming an OFDM symbol to be transmitted through a channel, where the transmitted power vector P⁽ⁿ⁾ is maintained unchanged throughout the transmission of the set of packets, and upon completion of a frame of OFDM symbols, returning to initialize the power of all sub-carriers to be equal and repeating the gradient ascent search for a next set of packets.

27. The method of claim 25, where after terminating the gradient ascent search the transmitted power vector P⁽ⁿ⁾ is a final power vector that dictates the power in the retained sub-carriers, the method further comprising:

spreading a plurality of data packets across the retained sub-carriers, where the plurality of packets spread across the retained sub-carriers form an OFDM symbol;

transmitting the OFDM symbol through a channel;

maintaining the transmitted power vector P⁽ⁿ⁾ unchanged throughout the transmission of the OFDM symbol; and

upon completion of transmission of the OFDM symbol,

returning to initialize the power of all sub-carriers to be equal and repeating the gradient ascent search for a next OFDM symbol.

28. The method of claim 27, where the data packets are spread across the retained sub-carriers in a manner to ensure that the data packets share similar error probabilities.

29. The method of claim 25, where after terminating the gradient ascent search the transmitted power vector P⁽ⁿ⁾ is a final power vector that dictates the power in the retained sub-carriers, the method further comprising:

selecting a time interval corresponding to the time required to transmit at least two OFDM symbols in succession, where the transmitted power vector P⁽ⁿ⁾ will be maintained unchanged throughout the selected time interval;

spreading a first plurality of data packets across the retained sub-carriers, where the first plurality of packets spread across the sub-carriers form a first OFDM symbol;

transmitting the first OFDM symbol through a channel;

spreading a second plurality of data packets across the retained sub-carriers, where the second plurality of packets spread across the sub-carriers form a second OFDM symbol;

transmitting the second OFDM symbol through the channel;

continuing to successively spread pluralities of data packets across the retained sub-carriers and to transmit the resulting OFDM symbols through the channel until the selected time interval is reached; and

returning to initialize the power of all sub-carriers to be equal and repeating the gradient ascent search for a next time interval.

30. The method of claim 29, where the data packets are spread across the retained sub-carriers in a manner to ensure that the data packets share similar error probabilities.

31. Apparatus operable to increase data throughput in an orthogonal frequency division multiplex (OFDM) wireless communications system, comprising at least one data processor operating under control of a stored program, and coupled to OFDM transmission circuitry, to model data throughput as a function of received Signal to Noise Ratio (SNR) with a sigmoid function to produce a throughput curve characterized as having a lower knee region and an upper knee region;

and to perform a gradient ascent search along the throughput curve by varying the power of a plurality of sub-carriers so as to remove power from any sub-carrier having a received SNR to the left of the lower knee region, and to re-allocate the removed power to sub-carriers having a received SNR to the right of the lower knee region.

32. The apparatus of claim 31, where different modulation and coding schemes are used in individual sub-carriers at different times.

33. The apparatus of claim 31, where said at least one data processor is further operable to improve error rate performance by disabling any sub-carrier whose power is removed.

34. The apparatus of claim 32, where said at least one data processor is further operable to improve error rate performance by disabling any sub-carrier whose power is removed.

35. The apparatus of claim 31, where said at least one data processor is further operable, for sub-carriers between the upper and lower sigmoid curve knee regions, to increase the power of the sub-carriers within a total power constraint so as to be to the right of the sigmoid curve upper knee region.

36. The apparatus of claim 31, where said at least one data processor is operable while performing the gradient ascent search to initialize the power of all sub-carriers to be equal, such that all elements of a transmitted power vector P⁽ⁿ⁾ are set to have the same value;

to evaluate a resultant total data throughput using a current value of the transmitted power vector P⁽ⁿ⁾ to obtain a first data throughput value;

to find a search direction vector g_n;

to determine a step size α

_n;

to update the transmitted power vector P⁽ⁿ⁾;

to evaluate a total data throughput with the updated transmitted power vector P⁽ⁿ⁾ to obtain an updated data throughput value; and

if the updated data throughput value is improved relative to the first data throughput value, to complete an iteration of the gradient ascent search, to increment n, and to return to finding the search direction vector to perform another iteration;

else, if the updated data throughput value is not improved relative to the first data throughput value, or if some predetermined number of interactions have been completed, to terminate the gradient ascent search.

37. The apparatus of claim 36, where after terminating the gradient ascent search the transmitted power vector P⁽ⁿ⁾ is a final power vector that dictates the power in the plurality of sub-carriers, where said at least one data processor is further operable to load sub-carriers with a symbol from an associated packet of a set of packets forming an OFDM symbol to be transmitted through a channel, to maintain the transmitted power vector P⁽ⁿ⁾ unchanged throughout the transmission of set of packets, and upon completion of a frame of OFDM symbols, to return to initialize the power of all sub-carriers to be equal and to repeat the gradient ascent search for a next set of packets.

38. The apparatus of claim 36, where after terminating the gradient ascent search the transmitted power vector P⁽ⁿ⁾ is a final power vector that dictates the power in the plurality of sub-carriers, and where the at least one data processor is further operable to spread a plurality of data packets across the sub-carriers, where the plurality of packets spread across the sub-carriers form an OFDM symbol;

to transmit the OFDM symbol through a channel;

to maintain the transmitted power vector P⁽ⁿ⁾ unchanged throughout the transmission of the OFDM symbol; and

upon completion of transmission of the OFDM symbol, to return to initialize the power of all sub-carriers to be equal and to repeat the gradient ascent search for a next OFDM symbol.

39. The apparatus of claim 38, where the data packets are spread across the sub-carriers in a manner to ensure that the data packets share similar error probabilities.

40. The apparatus of claim 36, where after terminating the gradient ascent search the transmitted power vector P⁽ⁿ⁾ is a final power vector that dictates the power in the plurality of sub-carriers, and where the at least one data processor is further operable to select a time interval corresponding to the time required to transmit at least two OFDM symbols in succession, where the transmitted power vector P⁽ⁿ⁾ will be maintained unchanged throughout the selected time interval;

to spread a first plurality of data packets across the sub-carriers, where the first plurality of packets spread across the sub-carriers form a first OFDM symbol;

to transmit the first OFDM symbol through a channel;

to spread a second plurality of data packets across the sub-carriers, where the second plurality of packets spread across the sub-carriers form a second OFDM symbol;

to transmit the second OFDM symbol through the channel;

to continue to successively spread pluralities of data packets across the sub-carriers and to transmit the resulting OFDM symbols through the channel until the selected time interval is reached; and

to return to initialize the power of all sub-carriers to be equal and to repeat the gradient ascent search for a next time interval.

41. The apparatus of claim 40, where the data packets are spread across the sub-carriers in a manner to ensure that the data packets share similar error probabilities.

42. The apparatus of claim 34, where said at least one data processor is operable while performing the gradient ascent search to perform the search in stages, where during a first stage said at least one data processor is operable to identify sub-carriers to be disabled, and where during a second stage said at least one data processor is operable to allocate power to retained sub-carriers at the expense of disabled sub-carriers.

43. The apparatus of claim 34, where said at least one data processor is operable while performing a first stage of the gradient ascent search to initialize the power of all sub-carriers to be equal, such that all elements of a transmitted power vector P⁽ⁿ⁾ are set to have the same value;

to evaluate a resultant total data throughput using a current value of the transmitted power vector P⁽ⁿ⁾ to obtain a first data throughput value;

to find a search direction vector g_n;

to determine a step size α

_n;

to update the transmitted power vector P⁽ⁿ⁾;

to evaluate a total data throughput with the updated transmitted power vector P⁽ⁿ⁾ to obtain an updated data throughput value; and

if the updated data throughput value is improved relative to the first data throughput value, to complete an iteration of the gradient ascent search, to increment n, and to return to finding the search direction vector to perform another iteration, else if the updated data throughput value is not improved relative to the first data throughput value, or if some predetermined number of interactions have been completed, said at least one data processor is further operable to initiate a second stage of the gradient ascent search by evaluating the received SNR of every sub-carrier resultant from the transmitted power vector P⁽ⁿ⁾ and if the received SNR of a particular one of the sub-carriers is found to be less than a predetermined threshold, to identify that sub-carrier as one to be disabled;

to re-assign power from any identified sub-carriers to remaining sub-carriers; and

to return to finding the search direction vector to perform the second stage gradient ascent search for the remaining sub-carriers.

44. The apparatus of claim 43, where after terminating the gradient ascent search the transmitted power vector P⁽ⁿ⁾ is a final power vector that dictates the power in the retained sub-carriers, where said at least one data processor is further operable to load retained sub-carriers with a symbol from an associated packet of a set of packets forming an OFDM symbol to be transmitted through a channel, to maintain the transmitted power vector P⁽ⁿ⁾ unchanged throughout the transmission of set of packets, and upon completion of a frame of OFDM symbols, to return to initialize the power of all sub-carriers to be equal and to repeat the gradient ascent search for a next set of packets.

45. The apparatus of claim 43, where after terminating the gradient ascent search the transmitted power vector P⁽ⁿ⁾ is a final power vector that dictates the power in the retained sub-carriers, and where the at least one data processor is further operable to spread a plurality of data packets across the retained sub-carriers, where the plurality of packets spread across the retained sub-carriers form an OFDM symbol;

to transmit the OFDM symbol through a channel;

to maintain the transmitted power vector P⁽ⁿ⁾ unchanged throughout the transmission of the OFDM symbol; and

upon completion of transmission of the OFDM symbol, to return to initialize the power of all sub-carriers to be equal and to repeat the gradient ascent search for a next OFDM symbol.

46. The apparatus of claim 45, where the data packets are spread across the retained sub-carriers in a manner to ensure that the data packets share similar error probabilities.

47. The apparatus of claim 43, where after terminating the gradient ascent search the transmitted power vector P⁽ⁿ⁾ is a final power vector that dictates the power in the retained sub-carriers, and where the at least one data processor is further operable to select a time interval corresponding to the time required to transmit at least two OFDM symbols in succession, where the transmitted power vector P⁽ⁿ⁾ will be maintained unchanged throughout the selected time interval;

to spread a first plurality of data packets across the retained sub-carriers, where the plurality of packets spread across the retained sub-carriers form a first OFDM symbol;

to transmit the first OFDM symbol through a channel;

to spread a second plurality of data packets across the retained sub-carriers, where the second plurality of packets spread across the sub-carriers form a second OFDM symbol;

to transmit the second OFDM symbol through the channel;

to continue to successively spread pluralities of data packets across the retained sub-carriers and to transmit the resulting OFDM symbols through the channel until the selected time interval is reached; and

to return to initialize the power of all sub-carriers to be equal and to repeat the gradient ascent search for a next time interval.

48. The apparatus of claim 47, where the data packets are spread across the retained sub-carriers in a manner to ensure that the data packets share similar error probabilities.

49. A computer readable memory medium storing computer program instructions for increasing data throughput in an orthogonal frequency division multiplex (OFDM) wireless communications system, comprising operations of performing a gradient ascent search along a data throughput curve, where the data throughput curve is modeled using a received Signal to Noise Ratio (SNR) as a sigmoid function to produce the data throughput curve to be characterized as having a lower knee region and an upper knee region, said gradient ascent search operating to vary the power of a plurality of sub-carriers so as to remove power from any sub-carrier having a received SNR to the left of the lower knee region, and to re-allocate the removed power to sub-carriers having a received SNR to the right of the lower knee region.

50. A computer readable memory medium storing computer program instructions for increasing data throughput in an orthogonal frequency division multiplex (OFDM) wireless communications system, comprising operations of performing a multi-stage gradient ascent search along a data throughput curve, where the data throughput curve is modeled using received Signal to Noise Ratio (SNR) as a sigmoid function to produce the data throughput curve to be characterized as having a lower knee region and an upper knee region, said gradient ascent search operating to vary the power of a plurality of sub-carriers so as to remove power from any sub-carrier having a received SNR to the left of the lower knee region, and to re-allocate the removed power to sub-carriers having a received SNR to the right of the lower knee region, where during a first stage of the gradient ascent search an operation identifies sub-carriers to be disabled, and where during a second stage of the gradient ascent search a further operation allocates power only to retained sub-carriers at the expense of disabled sub-carriers.

51. Apparatus operable to increase data throughput in an orthogonal frequency division multiplex (OFDM) wireless communications system, comprising data processor means operating under control of a stored program, and coupled to OFDM transmission circuitry, for modeling data throughput as a function of received Signal to Noise Ratio (SNR) with a sigmoid function to produce a throughput curve characterized as having a lower knee region and an upper knee region;

and for performing a gradient ascent search along the throughput curve by varying the power of a plurality of sub-carriers so as to remove power from any sub-carrier having a received SNR to the left of the lower knee region, and to re-allocate the removed power to sub-carriers having a received SNR to the right of the lower knee region.