Reducing noise
Mr. Robin,
I read your October article entitled “Digitizing audio,” and have a question on the SNR formula for digital audio. We have two different audio levels (one very low and the other higher), and the quantization error is higher for the lower audio level. Should we associate the same discrete levels to both signals (increase the number of bits for the lower signal) in order to have the same SNR?
Also, in this case, does SNR refer to a reference audio level (i.e. a 0dBm test tone) or to the different audio levels (considering each audio level has the same discrete quantizing intervals)?
Tiziano Colombo
Mr. Colombo:
The analog-to-digital conversion is characterized by two inseparable processes: sampling and quantization.
The sampling process consists of a pulse amplitude modulation (PAM) process, where a sequence of pulses occurring at constant time intervals T=1/Fs (where Fs is the sampling frequency) modulated by the sampled signal. PAM is essentially a distortionless process as long as the sampling frequency is higher than twice the sampled frequency and no aliasing occurs. PAM results in a sequence of pulses whose amplitude is proportional to the sampled analog signal at the sampling instant.
The quantization is a pulse code modulation (PCM) process that helps represent the amplitudes of the successive samples of the analog waveform with binary integers. In this process, an infinite number of possible pulse amplitudes is reduced to a finite number of discrete levels, Q, according to the expression Q = 2n, where n is the number of bits per sample. This is a nonlinear process. It is essentially a sample-and-hold where the instantaneous approximate values, with an uncertainty of +/-Q/2, also called quantizing error or Qe, are held in memory until the next pulse arrives.
Depending on the original analog audio signal amplitude, Qe is perceived as noise or distortion and noise. With large signal amplitudes Qe has a random character and is perceived as noise. In digital systems the signal-to-noise ratio (SNR) is measured with respect to the highest possible digital signal level, known as 0dBFs or zero dB Full Scale. This noise is the result of the Qe so it exists only in the presence of an audio signal. As the level of the analog signal is decreased, the Qe becomes less random and the effect is a distorted representation of the analog audio signal.
Color bars
Dear Michael:
I was wondering why we use only NTSC color bars now, instead of 100 percent color bars. Why do we have 100 percent color bars?
Michael Robin responds:
NTSC and PAL transmitters get overloaded with 100 percent colors bars; so, since the early years of color television, 75 percent color bars have been used, sometimes with the white bar level raised to 100 percent. Digital video equipment, including digital distribution networks and digital transmitters, can accept 100 percent color bars, so there is no problem. The problems appear when NTSC or PAL transmitters are fed with 100 percent color bars from a digital signal source. This raises a different question altogether: Why do we use color bars? In the good old days, color bars were transmitted shortly before the beginning of the daily transmission to allow the viewer to adjust the notoriously unstable NTSC receiver to obtain pleasing colors. To reduce the problems of program interchange, videotape recordings had a “leader” consisting of a 75 percent color bar signal to allow adjustment of Quad VTRs or, later, UMATIC and Betacam VTRs. With digital equipment the need to use a color bars signal is restricted to the adjustment of the A/D and D/A converters for maintenance purposes. So here one would use a 100 percent color bars signal, since the digital equipment is designed to carry 100 percent color bars. In the year 2003, I cannot see the need to transmit on-air color bars.
December Freezeframe:
Q. What unique feature was introduced on the Ikegami DNS-11 and DNS-101 cameras at the 1995 NAB convention?
A. A removable hard disk storage module from Avid called the CamCutter.
Winners:
David Budwash
Roberts Yent
James Allen, KTBC-TV
Test your knowledge!
See the Freezeframe question of the month on page 6 and enter to win a Broadcast Engineering T-shirt.
Send answers tobdick@primediabusiness.com
