01 December 2009

Brain Teaser Challenge - "November"†

by Butch Shadwell
Fred Jones was in his last term as an EE undergrad at Whatsamatter U. For his senior project he decided to build a codec that would send 8 bit PCM audio at 2.5 kilobytes per second. It seemed to work pretty well on most male voices, but there was weird distortion with music and some women. Fred may have fallen asleep in his DSP class. What do you think could be his problem? Fred often preferred to be called Stanley Smith for some odd reason, but I am not sure why I am telling you his alternate identification. I have to stop with the clues.

Reply to
Butch Shadwell
904-223-4510 (fax)
904-223-4465 (v)
3308 Queen Palm Dr.
Jacksonville, FL 32250-2328.
www.shadtechserv.com

The names of correct respondents may be mentioned in the solution column.
† Our friend Butch volunteers a bit too much, he never quite catches up to the current month.

2 comments:

Anonymous said...

Thanks a lot for a bunch of good tips. I look forward to reading more on the topic in the future. Keep up the good work! This blog is going to be great resource. Love reading it.
................................
essay term paper-Essay Writing Help

Butch Shadwell said...

Pulse code modulation was the topic last month. "he decided to build a codec that would send 8 bit PCM audio at 2.5 kilobytes per second. It seemed to work pretty well on most male voices, but there was weird distortion with music and some women. Fred may have fallen asleep in his DSP class. What do you think could be his problem? Fred often preferred to be called Stanley Smith for some odd reason, but I am not sure why I am telling you his alternate identification."

The clue referred to Fred's alias, Stanley Smith. And of course the answer to the question was aliasing. At the given sample rate, the highest frequency that can be digitized properly is 1.25kHz. If spectral components higher than this frequency are fed to the analog to digital converter at amplitudes greater than 1/512 of the peak signal voltage, then that energy will interact with the sample frequency and produce noise. The typical answer is to have a filter on the analog input that attenuates these higher frequencies, or you can over sample and filter it after digitization. But I bet you already knew that.