Cassette tape has a noise problem. Compact Cassette was designed by Philips in 1963 primarily for voice dictation, not music. The narrow tape, slow speed, and low track width meant that the magnetic noise baked into the recording medium sat roughly 50 dB below the maximum recording level for standard Type I tape. Play a quiet piano passage off a raw cassette and you hear a hiss underneath it. Turn up the volume and the hiss turns up with it.
Ray Dolby had already solved a similar problem for professional reel-to-reel tape by 1965. Within a few years the same underlying principle, applied in a simplified form, would define how cassette players were built for the next three decades.
Companding: The Core Idea
The insight behind Dolby noise reduction is not to filter the noise out after recording. That would also filter out quiet high-frequency content that sits at similar levels. Instead, the system modifies how the signal is written to tape so that when it is played back normally, the noise has been effectively compressed below audibility.
The technique is called companding: compress during recording, expand during playback. Before the signal reaches the tape heads, a compressor raises the level of soft signals, making them louder relative to the tape noise floor. When you play the tape back through a matching expander, it lowers those same soft signals back to their original levels. Because the tape noise was recorded at the same time and is now subject to the same attenuation, it gets pulled down with the quiet content.
A simple full-bandwidth compander would work, in theory. But it produces an audible side effect: the noise floor rises and falls as loud signals come and go. This is called noise modulation or pumping, and it is worse than steady hiss in most listening situations.
How Dolby A Avoided Pumping
Ray Dolby presented the Dolby A system to professional studios in 1965. It is a multiband compander with four frequency ranges: a low-pass band below 80 Hz, a bandpass band from 80 Hz to 3 kHz, a high-pass band above 3 kHz, and a second high-pass band above 9 kHz. Each band is processed independently by its own compressor and expander.
The compander only acts on signals that fall below a fixed threshold of -40 dB. Loud signals pass through unmodified. Quiet signals are boosted by up to 10 dB on the way to tape, then attenuated by 10 dB on playback. The ratio is 2:1, meaning for every 2 dB of input level change below threshold, only 1 dB of level change is written to tape.
Splitting the spectrum into four bands is what defeats pumping. A loud kick drum saturates the low band and suppresses compander action in that range, but the high-frequency band continues to compress independently. The noise modulation in each band is caused only by content in that band, so a loud bass note does not cause high-frequency hiss to pulse in and out. At 15 kHz, where tape hiss is most audible and where signal content is most sparse, Dolby A achieved up to 15 dB of noise reduction.
The Consumer Version: Dolby B
Dolby A required external rack units costing thousands of dollars per channel. For the cassette market, Dolby developed a simpler single-band system called Dolby B, introduced in 1968.
Dolby B only compresses and expands the high-frequency region, above approximately 200 Hz, and only for soft signals. This is where cassette hiss is most noticeable. The compression ratio and amount are less aggressive than Dolby A: about 10 dB of noise reduction at 5 kHz and above. That was enough to drop cassette hiss from obviously audible to barely noticeable in most listening environments.
The simplification came with a tradeoff. Because Dolby B is single-band and only processes high frequencies, pumping is possible when high-frequency content is present and then absent. In practice, music contains enough continuous high-frequency energy that the transitions are inaudible. Dolby B became the standard on virtually every cassette deck from the early 1970s onward.
Dolby C, introduced in 1980, applied the compression in two stages to achieve approximately 20 dB of noise reduction at 5 kHz, double the effectiveness of Dolby B. Dolby S, released in 1989, applied a more sophisticated sliding-band approach that could deliver near-professional noise reduction in consumer hardware.
What "Not Dolby Decoded" Sounds Like
Any tape recorded with Dolby encoding and played back without the matching decoder will sound wrong in a specific way. The high frequencies are boosted by the encoding process and never pulled back down. The result is a recording that sounds harsh and bright, with over-emphasized sibilance and a thin quality in the upper midrange.
This is why every cassette deck with Dolby circuitry had a Dolby on/off switch. If you put in a commercially pressed cassette and engaged Dolby playback, it usually sounded wrong, because commercial releases were sometimes manufactured to sound correct without decoding, relying on the tape's natural HF rolloff to compensate. Decoding a tape not designed for it adds a shelf cut that makes it sound dull. The switch let you match the playback to how the tape was made.
dbx: The Single-Band Alternative
While Dolby worked band by band, dbx (founded by David E. Blackmer in 1971) took the opposite approach: a single wideband compander with a 2:1 compression ratio across the entire audio spectrum. The wider ratio and full-spectrum action meant dbx could achieve approximately 30 dB of noise reduction on reel-to-reel tape.
The cost was more obvious pumping. Because the entire spectrum compresses together, a loud signal in any frequency range can suppress the action and cause the noise floor to shift perceptibly. For studio use where tape noise matters most at the mastering stage, dbx's aggressive noise reduction was worth it. For consumer cassette playback where pumping is more objectionable than a small amount of residual hiss, Dolby's approach generally won.
Why This Still Matters
Cassette tape sales have grown for well over a decade as part of the physical music revival. Anyone releasing music on cassette today is encoding and distributing content that will eventually play on hardware with varying Dolby implementations, or on players without any noise reduction at all. Understanding that the tape is not playing back a neutral recording, it is playing back one half of a complementary process, shapes decisions about what to record at what level and whether to print with NR enabled or off.
Tape plugin emulations often model the saturation, frequency response, and noise character of specific machines. Few include the compander stage. That omission is acoustically significant: a recording made to a Nagra or Studer with Dolby A engaged does not have the frequency response of a recording made without it. The encoded tape is not the same as the source.
The hiss was always there. Dolby's contribution was a method for making both the tape and the listener agree to pretend it wasn't.