Search

Nedeljko Varnica Phones & Addresses

  • San Jose, CA
  • Sunnyvale, CA
  • Allston, MA
  • Cambridge, MA
  • 3329 Shoshana Ln, San Jose, CA 95135

Work

Position: Student

Education

Degree: Graduate or professional degree

Publications

Us Patents

Error Correction Coding For Varying Signal-To-Noise Ratio Channels

View page
US Patent:
7685494, Mar 23, 2010
Filed:
May 7, 2007
Appl. No.:
11/745254
Inventors:
Nedeljko Varnica - Sunnyvale CA, US
Gregory Burd - San Jose CA, US
Zining Wu - Los Altos CA, US
Assignee:
Marvell International, Ltd. - Hamilton
International Classification:
H03M 13/00
US Classification:
714752
Abstract:
An error rate sensitive error correction (ERSEC) system that applies a level of error correction that is inversely related to susceptibility to error as indicated by a signal-to-noise ratio (SNR) profile of a channel. The SNR profile is estimated, detected or retrieved from an external source. The ERSEC system can be used with any channel for which the SNRs vary spatially, temporally or both.

Methods For Generating And Implementing Quasi-Cyclic Irregular Low-Density Parity Check Codes

View page
US Patent:
7805652, Sep 28, 2010
Filed:
Aug 21, 2006
Appl. No.:
11/507161
Inventors:
Nedeljko Varnica - Sunnyvale CA, US
Assignee:
Marvell International Ltd. - Hamilton
International Classification:
H03M 13/00
US Classification:
714752, 714755, 714786
Abstract:
A device includes a supernode generator module, a supernode splitting module, and a control module. The supernode generator module generates S supernodes each comprising d symbol nodes, where S and d are integers greater than 1. The supernode splitting module splits each of the S supernodes into d derived symbols, wherein a total number of symbol edges of the d derived symbols is equal to a predetermined number of symbol edges of each of the S supernodes. The control module generates a quasi-cyclic irregular low density parity check (LDPC) code based on S*d derived symbols, wherein S equals the S supernodes and d equals the d derived symbols.

Low-Density Parity-Check Decoder Apparatus

View page
US Patent:
7827461, Nov 2, 2010
Filed:
Sep 18, 2007
Appl. No.:
11/857305
Inventors:
Nedeljko Varnica - Sunnyvale CA, US
Gregory Burd - San Jose CA, US
Zining Wu - Los Altos CA, US
Assignee:
Marvell International Ltd. - Hamilton
International Classification:
H03M 13/00
US Classification:
714752, 714786, 714799
Abstract:
A low-density parity-check (LDPC) decoder includes a plurality of bit node processing elements, and a plurality of check node processing elements. The LDPC decoder also includes a plurality of message passing memory blocks. A first routing matrix couples the plurality of bit node processing elements to the plurality of message passing memory blocks. A second routing matrix couples the plurality of check node processing elements to the plurality of message passing memory blocks. The first routing matrix and the second routing matrix enable each bit node to exchange LDPC decoding messages with an appropriate check node via a corresponding one of the message passing memory blocks.

Enhancements Of Soft Decoders And Detectors On Asymmetric Channels

View page
US Patent:
7876860, Jan 25, 2011
Filed:
Aug 17, 2007
Appl. No.:
11/893810
Inventors:
Nedeljko Varnica - Sunnyvale CA, US
Gregory Burd - San Jose CA, US
Zining Wu - Los Altos CA, US
Assignee:
Marvell International Ltd. - Hamilton
International Classification:
H04L 27/06
US Classification:
375340
Abstract:
Systems and methods for enhancing soft decoders and detectors on asymmetric channels are provided. The methods include acquiring log-likelihood ratios (LLRS) for error-correction code (ECC) encoded data symbols, selecting a quality measure function and a quality threshold based on the LLRs, applying the selected quality measure function to the LLRs to obtain quality measures, comparing the quality measures to the selected quality threshold, and updating the LLRs for selected ECC encoded data symbols based on the comparisons. The updating may occur by multiplying the LLRs for the selected ECC encoded data symbols by a selected scaling factor.

Error Correction Coding For Varying Quality Channels

View page
US Patent:
7979774, Jul 12, 2011
Filed:
Feb 18, 2010
Appl. No.:
12/708271
Inventors:
Nedeljko Varnica - Sunnyvale CA, US
Gregory Burd - San Jose CA, US
Zining Wu - Los Altos CA, US
Assignee:
Marvell International, Ltd. - Hamilton
International Classification:
H03M 13/00
US Classification:
714752, 714704, 370333
Abstract:
An error rate sensitive error correction (ERSEC) system that applies a level of error correction that is inversely related to susceptibility to error as indicated by a signal-to-noise ratio (SNR) profile of a channel. The SNR profile is estimated, detected or retrieved from an external source. The ERSEC system is used with a channel for which the SNRs vary spatially, temporally or both.

Methods For Generating And Implementing Quasi-Cyclic Irregular Low-Density Parity Check Codes

View page
US Patent:
8051357, Nov 1, 2011
Filed:
Sep 27, 2010
Appl. No.:
12/891251
Inventors:
Nedeljko Varnica - Sunnyvale CA, US
Assignee:
Marvell International Ltd. - Hamilton
International Classification:
H03M 13/00
US Classification:
714752, 714755, 714786
Abstract:
A system includes a supernode generator module configured to generate supernodes. Each of the supernodes includes a plurality of symbol nodes. A supernode splitting module is configured to split each of the supernodes into derived symbols. The total number of edges of the derived symbols in each of the supernodes is equal to a predetermined number of edges of each of the supernodes.

Approximate Soft-Information Computation In Multi-Level Modulation Signaling Schemes

View page
US Patent:
8059763, Nov 15, 2011
Filed:
Nov 7, 2007
Appl. No.:
11/936531
Inventors:
Nedeljko Varnica - Sunnyvale CA, US
Zining Wu - Los Altos CA, US
Assignee:
Marvell International Ltd. - Hamilton
International Classification:
H04L 27/06
US Classification:
375341
Abstract:
Apparatus and methods are provided for calculating soft information in a multi-level modulation scheme using one or more nearest neighbors. The nearest neighbors correspond to signal points in a signal constellation set nearest to the value of a received signal. For the nearest neighbors of a received symbol of information, a detector can determine whether the nearest neighbors have a same bit value at a bit position of the symbol. When the bit values are the same at that bit position, soft information in the form a log-likelihood ratio can be computed based on the nearest neighbors and a predetermined scaling factor. The predetermined scaling factor can be optimized for system performance.

Defect Recovery For Iteratively-Decoded Data Channel

View page
US Patent:
8122314, Feb 21, 2012
Filed:
Nov 7, 2007
Appl. No.:
11/936418
Inventors:
Shaohua Yang - San Jose CA, US
Nedeljko Varnica - Sunnyvale CA, US
Nitin Nangare - Salinas CA, US
Zining Wu - Los Altos CA, US
Assignee:
Marvell International Ltd. - Hamilton
International Classification:
H03M 13/00
H04L 1/18
G06F 11/00
US Classification:
714752, 714751, 714755, 714758, 714759, 714800, 714807
Abstract:
In iterative decoding, a data recovery scheme corrects for corrupted or defective data by incorporating results from a previous decoding iteration. In one embodiment, a final multiplexer selects between the final detector output or a previous detector output based on the absence or presence of defective data. In another embodiment, the branch metrics for the defective data, which otherwise would be combined with a priori LLRs from an outer decoder of a prior stage, are ignored so that the a priori LLRs themselves are used alone. The two embodiments can be used together.
Nedeljko P Varnica from San Jose, CA, age ~50 Get Report