Audio Noise-Reduction IC

Single-ended system achieves 25dB of noise reduction without pre-encoding

The SSM2000* is an audio dual-channel noise-reduction IC that reduces noise through a combination of variable filtering and downward expansion, in conjunction with a unique adaptive noise-threshold detector. Without requiring pre-encoding of program material, this combination of techniques yields an overall noise reduction of up to 25dB on a wide variety of program sources: AM and FM radio, open-reel and cassette tape, LPs, CDs, Dolby-B®-encoded programming, broadcast studio-transmitter links and telephone lines, and other audio sources without the need for any additional manual adjustment. The HUSH® Noise Reduction System, as implemented in the SSM2000, has been demonstrated to reduce noise substantially while preserving fidelity and transparency in PC multimedia, intercom systems, teleconferencing systems, mobile communications, automotive audio, home stereos and TVs, and other consumer and professional audio applications.

Analog noise-reduction systems must be able to distinguish between "hiss", or white noise, and the source material, and then attenuate the noise. One method of noise rejection assumes that all signal levels below a pre-selected magnitude are noise, and then attenuates that noise via a voltage-controlled amplifier (VCA). A popular variation of this noise-reduction method is found in Dolby B® cassette tape systems. This double-ended (encode-decode) approach achieves about 9 to 10dB of improvement in signal-to-noise ratio by incorporating a high-pass compressor on the recording side and a high-pass expander during playback.

Another technique senses and rejects noise by measuring the frequency content of the audio signal and then filtering all noise that occurs above the highest audio frequency. This method utilizes a voltage controlled filter (VCF) and is the key to operation in the single-ended DNR system, which achieves about 10dB of noise rejection; performance is limited by the fixed noise threshold, which can be ineffective on very noisy material and can interfere with signal content of noise-free signals.

The HUSH® single-ended system (Figure 1) incorporates aspects of both approaches, resulting in up to 25dB of noise rejection without needing pre-encoding.

Figure 1
Figure 1

The principal key to the SSM2000's ability to reject noise is its ability to recognize the noise floor, whether it be the low-level hiss of an audiotape, or the higher-level noise interfering with reception of weak stations in an automobile radio. This is accomplished in a patented noise threshold detector, which operates on the principle that virtually all program material contains gaps during which the only energy present (especially at frequencies in the 3-kHz to 8-kHz region) is noise. A negative peak follower finds the lowest of these intervals at a given time and subsequently causes the VCA gain to be substantially reduced (downward expansion) for input levels in the vicinity of that threshold. The threshold is constantly updated, adapting to the characteristics of the input.

Another principle is that the highest audio signal amplitudes typically occur at low frequencies (100Hz to 1kHz) and taper off exponentially as frequency increases. This allows the possibility of maintaining wide bandwidth for signal levels well above the noise floor, but significantly reducing the bandwidth for input levels approaching the noise floor rejecting mostly noise, with little effect on signal reception, so long as changes in gain and bandwidth are not excessively rapid. Indeed, the SSM2000 responds well to sudden transitions from low noise levels to normal noise levels.

Both Dolby® and DNR require input line levels that are carefully controlled in order to reach rated performance. This is because the noise floor magnitude is assumed not to change substantially. In reality, the noise floor often changes with different types of audio input, as a function of equipment, recording media, receiver signal strength, and the environment. As noted above, the patented adaptive noise threshold detector in the SSM2000 senses these changes and will adapt the VCA and VCF operation for optimal noise-reduction performance.

Within the chip, both left and right audio paths through the SSM2000 are differential. This greatly reduces control feedthrough and distortion by taking advantage of the inherent common-mode rejection capability of the differential VCFs, VCAs, and inputs to the output amplifiers. The SSM2000 enables additional functions. For example, a TTL-level-controlled MUTE is provided, for full attenuation. In addition, a DEFEAT pin allows a simple on-off comparison for HUSH® noise reduction. An external VCA control port is also provided for gain control; with simple external circuits, it can provide valuable functions such as automatic volume leveling, compression, and road-noise compensation by taking advantage of the magnitude and frequency information available at separate SSM2000 pins. The SSM2000 can accept a wide range of audio line levels, but external circuitry (pre- or post-) can also be used for control.

The SSM2000 was designed by Analog Devices Fellow Derek Bowers at our Santa Clara facility. James K. Waller, Jr., of Rocktron, Inc., Rochester Hills, MN, is the inventor of the HUSH noise-reduction technology.

*HUSH® is a Registered Trademark of Rocktron Corporation; Dolby® is a Registered Trademark of Dolby Laboratories, Inc.; DNR is a Trademark of National Semiconductor Corporation.



John McDonald