EE 445S Real-Time Digital Signal Processing Laboratory - Lecture 8

Before Lecture

• "Mindflex uses your brainwave intensity to control the ball". Headset uses electroencephalogram (EEG) signals. Link sent by a real-time DSP lab student in spring 2011.

Announcements

• Real-Time DSP Lab course alumni Jung "James" Lee, Arthur Ishiguro, Kevin Scholz and Alexander Yeh team with Farhad Abasov to place first in the fall 2011 ECE senior design project competition for their augmented reality application for smart phones. November 22, 2011.

• Analog Devices 20-bit Digital-to-Analog Converter (DAC) AD5791 clocks input data serially at speeds up to 50 MHz. The DAC works internally with 24-bit words, which would correspond to a sampling rate of 2.08 MHz with 24 bits/sample. Applications include "medical instrumentation, test and measurement, industrial control, and high-end scientific and aerospace instrumentation". Interestingly, 2.08 MHz is the sampling rate for ADSL version 1 systems. March 11, 2010.

Lecture

• Lecture on Quantization in PowerPoint format.

Supplemental Information

• Example tradeoffs in sampling rate and resolution for analog-to-digital converters a.k.a. digitizers:
  NI PXIe-5186: two channels, 8 bits/sample, 12.5 GS/s sampling rate, 5 GHz bandwidth (Mar. 2011)
  NI PXI-5154: two channels, 8 bits/sample, 2 GS/s sampling rate, 1 GHz bandwidth
  NI PXIe-5122: 14 bits/sample, 100 MS/s sampling rate
  NI PXI-5192: 16 bits/sample, 15 MS/s sampling rate
  NI PXI-5192: 24 bits/sample, 500 kS/s sampling rate
• Mark A. Castellano, Todd Hiers, and Rebecca Ma, "TMS320C6000 mu-Law and A-Law Companding with Software or the McBSP", Texas Instruments Application Report, SPRA634, April 2000.
• The lecture discusses mu-law companding used in the US and Japan (u = 255) and A-law companding in Europe (A = 87.6). In the companding formulas, the log is the natural logarithm.
• A student asked what happens when a call is placed from the US to Europe. Companding in telephony is to eight bits, with one bit for the sign. Hence, the formulas in terms of |x| in [0, 1] would be used to generate a seven-bit number. To the precision of seven bits, there is no difference between A-law companding (A = 87.6) and mu-law companding (u = 255) for |x| in [0.18, 1].
• There is a mu-law pulse coded modulation (PCM) audio format. The format is an eight-bit floating-point format: 1 bit for the sign, 4 bits for the mantissa, and 3 bits for the exponent. This is a common audio format on Sun workstations.

Last updated 11/23/11. Send comments to bevans@ece.utexas.edu