To learn more about our technologies click on any of the links below or scroll down the page.

The 48 bit Reed-Solomon ECC device
The 88 bit Reed-Solomon ECC device
The 144 bit Reed-Solomon ECC device
The speech compression technology
The speech compression (TELP) vocoder
MPEG compression technology
JPEG compression technology
Wavelet compression technology
Audio compression technology

---

THE 48 BIT REED-SOLOMON ECC DEVICE SPECIFICATIONS

Product functions:

• Corrects single bursts up to 11 bits and two 10-bit symbols
• Sector size: 512 bytes
• Overhead (number of redundant bits): 48 bits
• Can be implemented with about 1000 gates

Product Performance:

• Data transfer rate: up to 25 megabytes/seconds
• Provides correction for single and double 10-bit symbols in less than 50 microseconds using DSP with 25 nanosec cycle

THE 144 BIT REED-SOLOMON ECC DEVICE SPECIFICATIONS

Product functions:

• Corrects single bursts up to 65 bits, double bursts up to 17 bits and triple bursts up to 17 bits.
• No microprocessor intervention for correction.
• Maximum code length: 765 bytes.
• Overhead (number of redundant bits): 144 bits
• Code symbol width: 8 bits.
• Degree of code interleave: 3.
• Data transfer rate: 25 MB/sec.
• Correction time for a record size 512 bytes less than 15 microsec (with 40 MHz corrector clock).
• Implements Buffer CRC with transfer rate up to 25 MB/sec.
• Overhead for Buffer CRC is programmable: 4 and 6 bytes.
• Provides 5 bytes ECC for ID protection with "on the fly" correction of single bursts up to 9 bits.
• Implemented with about 11,000 gates.

SPEECH COMPRESSION TECHNOLOGY

GT Technology together with Alaris, Inc. developed new speech compression methods which allows very high speech compession and need less computation power for its implementation than other known in industry methods with the same and even better quality of speech reconstruction. TELP speech compression method TELP algorithm is a CELP-like speach compression method. CELP means Code Excited Linear Prediction. As in all CELP like methods, in TELP the linear prediction filter of 10-th order has been used.

The information sent to a decoder contains:

1. Linear prediction coefficients.
2. Information about excitation for predicting filter.

Usually excitation is chosen by 2 steps. On the first step so called long-term prediction information has to be found, and then short-term prediction information should be found. Long-term prediction may be found from previously used excitation (Adaptive Code Book) or by special Long-Term Prediction Filter (as in True Speech, for example).

Short-term prediction may be found from the special stochastic codebook (as in CELP) or as a pulse sequence (as in True Speech).

Linear prediction coefficients usually encoded in a form of Linear Spectral Parameters.

New features used in TELP are the following: 1. Use of a special codebook instead of a stochastic codebook as inCELP for chosing optimal short-term prediction excitation. This modification allows to reduce computation.2. Special variable-rate coding for Linear Spectral Parameters.

This coding allows at average 20 bits per 30 ms frame instead of 34 bits as in CELP.

3. Special algorithm for sound classification which allows substantially reduce the number of bits for excitation encoding for some sound sequences and for pauses.

THE SPEECH COMPRESSION (TELP) VOCODER

Product functions:

• Based on a proprietary compression algorithms. (Two patents granted).
• Compress speech to the data rate of 2,200, 4,800 and 6,000 bps
• The vocoder uses fixed point arithmetic. It can be implemented on the DSP providing compression as well as decompression in real time.

Requirements for DSP for real time compression:

15 Mips, 16kB ROM, 4KB RAM.
Multiplication with accumulation should be done in 1 cycle.

Requirements for DSP for real time decompression only:

2 Mips 4 kB ROM, 1 KB RAM.
Multiplication with accumulation should be done in 1 cycle.

MPEG COMPRESSION TECHNOLOGY

Proprietary implementation of MPEG1 allows to do real time compression/decompression on Pentium 90 MHz computer.

JPEG COMPRESSION TECHNOLOGY

Developed a Verilog model for JPEG baseline compression/decompression.

WAVELET COMPRESSION TECHNOLOGY

• Proprietary method od coding of quantized values allows to achieve in average more than 100:1 compression ratio with a good quality of the picture.
• Proprietary implementation of filtering allows to achieve real time compression/decompression on Pentium 100 MHz computer.

AUDIO COMPRESSION TECHNOLOGY

Compression ratio 8:1, 12:1 and 15:1.

Requirements for DSP for real time compression/decompression:

1. 11 KHz Mono input files: 15 Mips (compression), 7 Mips (decompression), 11 KB RAM, 25 KB ROM.
2. 22 KHz Stereo input files: 60 Mips (compression), 28 Mips (decompression), 11 KB RAM, 25 KB ROM.

---