dbx 128 - TECHNICAL DETAILS OF dbx PROCESSING; Level Detection Mechanisms

30
Tape Noise Characteristics (Refer to Figure 17)
Asperity noise yields the curve shown by ‘’D’’. The recorded
signals (A’, B’ and C’) are all sine waves at 180Hz. The noise side-
bands created by these signals are illustrated by curves A, B and C.
Observe that the higher the recorded signal level, the higher the
noise sideband level. This level dependent noise is known as tape
modulation noise.
The noise sidebands are masked partially by the recorded signal,
but only for about two octaves on either side of the signal. This
masking is depicted by the shaded box in the chart. The ear is less
sensitive to lower frequencies, so the lower sidebands are masked
sufficiently by the signal. Notice the upper sideband of the +10dBm
recorded signal (curve A) extends beyond the masked area and at a
level which would be audible in a program of 100dB dynamic range.
To negate modulation noise effects, dbx applies pre-emphasis to
the signal before recording and de-emphasis upon playback. The de-
emphasis starts at 400Hz, and reaches a maximum weighting of
-12dB at 1600Hz (see Figure 18). The net result is a reduction in
modulation noise of nearly 12dB with strong low frequency
recorded signals, while the overall record/play frequency response
is flat.
The shaded line at -65dBm indicates the level of steady state
background noise which would be required to mask modulation
noise if pre-emphasis and de-emphasis (or signal weighting) were
not used. With signal weighting, there is no need for this ‘‘noise
perfume” as used in other compander systems.
TECHNICAL DETAILS OF dbx PROCESSING
Level Detection
Regardless of the specific techniques employed by a given
compander-type noise reduction system, some method must
be used to sense the audio input level to the compressor (when
recording) and to the expander (when playing back) (Fig. 18,
6&6A). This level detection circuitry tells the compressing or
expanding amplifier (Fig. 18, 3&3A} what the actual input
signal level is, and then the amplifier’s gain increases or de-
creases to accomplish the required expansion or compression.
In theory, the basic principle of operation is simple, but in
practice, the difficult part is to maintain precise mirror image
encoding and decoding. In other words, whatever degree of
compression takes place during the encoding must be pre-
cisely matched by the same degree of expansion during de-
coding... and at the same point in time with respect to the
program. There are several ways to detect signal level, and
some make it difficult to assure mirror image encode/decode
performance (tracking).
Previous attempts to create compander type tape noise
reduction systems have utilized peak or average level detec-
tion, both of which are sensitive to phase changes. Phase
changes are inherent in tape recording due to characteristics
of the record heads, electronics, and tape, so level detection
schemes which are phase sensitive are subject to mistracking
errors upon decoding. That is, the amount of expansion does
not correspond with the original compression so the retrieved
program does not sound exactly like the original one. Unlike
peak and average level detection systems, the RMS method
sums the squares of the instantaneous energy of all frequency
components present. Therefore, RMS detection is impervious
to phase changes. However, true RMS detection has been
very complex and expensive. dbx equipment uses our own
recently developed and patented analog techniques to achieve
excellent RMS detection at a moderate cost.