Dbx 1981 2

Page 1

Overshadowed by Dolby despite its theoretical superiority, the dbx
noise reduction system now seems poised to take the hi-fi market by
storm. Peter Green was at the press launch and this is what he heard.

with anything like the quality of the original. However
much you spend on your hi-fi, however low the distor-
tion figures, however carefully you tend your records and
however diligently you clean your tape heads, you cannot get
away from the twin evils of surface noise and restricted
dynamic range. ~
Surface noise is caused by the limited quality of the recor
ding medium itself. Imperfections and inhomogeneity in the
particles that make up the tape coating generate the familiar
hissing we know and hate; the roughness of the vinyl surfaces of
record grooves produces its share of hissing and pops. Surface
noise also affects dynamic range. During a live performance it is
possible for sound pressure levels to momentarily hit 120 dB
during music transients; however, background noise levels in
the audience can easily range from 30 to 50 dB and so this
simplistic treatment indicates that live music has a perceived
dynamic range of about 70 to 90 dB. Unfortunately a good
cassette recorder has a dynamic range of only about 45 dB,
while a conventional vinyl record is not much better at 55 dB.

U ntil now it has been impossible to listen to music at home

Recording engineers have to compress the signal by a factor of

2:1 so that the loudest passages are below the level that causes
tape saturation or distortion, and the quietest passages are
above the level of tape hiss or record surface noise (see Fig. 1).
This squeezing of the signal into a restricted dynamic range
makes the music sound flat, unexciting and unrealistic. You
know its a recording,

D Versus d

The noise reduction system that established a virtual
monopoly in the world hi-fi market of the seventies was Dolby
B, a system that reduced high frequency noise (the most objec-
tionable kind) by boosting high frequencies on recording and at-
tenuating them on playback, thus also reducing the noise added
in the recording process.

dbx decided to tackle the other problem - restricted
dynamic range. Since sounds have to be compressed to be
recorded on tape or disc. a system that allowed 1:2 expansion
on playback would recreate the range of the original perfor-
mance (Fig. 2). This idea is quite old but difficult to apply,
because the expander and compressor must track each other
(be exactly complementary), especially on transients - this in
turn requires accurate detection of the signal. The difficulties in

. this and the way that dbx overcame them are ~describsdglater, '
but the important point is that when the recorded signal is ex- '

Fig. 1 (left) When recording tapes or

\ up: msrannm discs, engineers monitor the signal
m. and turn the level up or down (gain
l riding) to avoid distortion or his
'0 T," This compression greatly reduces the
,-- dynamic range to an absolute
,-' hisvsfl or TAPE maximum of about 60 dB.
ENOOOE DECODE

lam 11:2)

Fig. 2 (right) By introducing 2:1 i
compression on recording then m m
accurately and automatically ' i

reversing the process on playback,

the dbx noise reduction system

reclaims the original dynamic range.
As a bonus, tape hiss becoma
inaudible.

< IIAUDIBLE
[L

m

panded the tape noise drops below the level of audibility. dbx
have killed two birds with one stone, together with an assort-
ment of sacred cows; for example, they recommend that you
use chromium dioxide rather than metal tapes, because its
pointless to pay extra money for an improvement of.2 to 3 dB
when the system itself is giving you an extra 30 dB across the
whole frequency range. 1

Paths Of Glory

Figure 3 shows the block diagram of a Type ll dbx noise
reduction system for domestic use (the Type I system has cer-
tain differences in the turnover frequencies of the filters to suit
professional situations). The encoder and decoder each have
two paths - the signal and detection paths.

The music signal to be recorded first goes through a band
pass filter to remove unwanted outof-band components. Type
II values arelq Hz to _l_qg_kHz, Type l are 22 Hz to 27 kHz. The
lower limit is there to prevent subsonic noisarom umerground
trains, traffic vibrations and the like from being recorded; the
upper limit prevents pickup of CB and other interference. The
Type I value is much lower because studios often have very
long leads which are more susceptible to pickup. The signal
passes through a preemphasis network that boosts high fre-
quencies and helps to overcome tape modulation noise (which
is caused by uneven magnetisation due to tape inconsistency),
and into the voltagecontrolled amplifier. The VCA is linear in
dB with control voltage to make things simpler later on; it com-
presses the signal by a 2:1 ratio.

ET I OCTOBER 1981

Page 2

:ucnusn
2:1 mass-on

rue
mum

FEATURE '

acumen
1:2 rxpmmu

amp was _ ch Low use vm DE- EMPHASIS
- mm films" nun monk
' ru/ . '
m-p 0: oat mar 1 In our in our -> 33"
nor in _ 3 i gnarl I
. . ' W me C
i
CONTROL CONTROL
aim 10mm: mu; Imknz mu:
rseuucx rzswonwsnn
nus onscron vapwmsis mo us: amp PASS pas-smurs nus DEIEMOR
- usrwoax nus-r FILTER as, usrwonx
mm
3::an "a, ocr ocT
01" u] I - \ I 3%! l w our
1
ism. aw mm sum mm mm:

Fig. 3 Blodr diagram of the Type II dbx nob: reduction sym

The output of the VCA is passed to the tape recorder, and
also into the detector section. The first stage is a band pass filter
which has a slope of 18 dB/octave above 10 kHz, so that the
detector 'takes less notice of high frequencies. Next comes
another preemphasis network, this time to increase the com-
pression of the signal at high frequencies and avoid the possibili-
ty that the preemphasis in the signal path will cause high-
frequency tape saturation. The RMS detector converts the AC
signal into a DC voltage proportional to the level in dB; since the
VCA also follows this law the whole system is linear in dB.

Mirror, Mirror
The encoder is a compressor with negative feedback; the
higher the signal level the greater the gain reduction. The
decoder must be a mirror image to give the correct tracking and
recovery of the original performance, so it is configured as a
' feedforward system acting as a1 :2 expander. The importance of
this is that if the tape recorder is considered to be transparent"
the VCA control voltage in the decoder is being derived from ex-
actly the same signal as that for the encoder VOK. The detector
path in the decoder is identical to the one already described -
band pass filter, preemphasis network, RMS detector - and so
the control signals will also be identical (to within component
tolerances). The decoder VCA has its control polarity reversed
and gives a complementary gain change to that of the encoder,
the deemphasis network reverses the effect of the encoder pre
emphasis to restore a flat overall frequency response. Its in-
teresting to note that this flat system response is due solely to
the mirror-image nature of the signal processing - the frequen-
cy relsponses of encoder and decoder are complementary, but
not at
Because a high value of loop gain (40 dB) is employed in
the decoder, a low pass filter is needed at the input to prevent
high-frequency oscillation caused by capacitive coupling. This
keeps the system stable.
RMS detectors have been used because tape recorders
arent perfect, or transparent - they introduce huge amounts

Fig4fliaeshiftunhavedmnaliceffed'son lbeshapeofwaveforms!
HI OCTOBER 1S1

of phase shift. This is not noticeable to the ear, but the effect of
passing a 1 V peak-topeak square wave through a 90° phase
shift filter is shown in Fig. 4; cheaper and simpler detectors such
as peak or averaging types would not give identical outputs for

I the two waveforms and the mirror-imaging would be lost. An

RMS detector is the only one that will give the same output level
for both.

Silence Is Golden

If anyone feels thatthe working of the system is hard to
understand, rest assured that a practical demonstration isnt,
The press showing started with a hiss comparison using blank
unencoded tape, first with no noise reduction, then Dolby B,
Dolby C and finally dbx. Hiss was still significant with the two
Dolby systems, although the improvement was noticeable -
but when the dbx cut in, nothing could be heard except the
gasps of amazement from hardened journalists. No hiss what-
soever! Listening to extracts from discs and tapes was a revela-
tion; for the first time in my life I could shut my eyes and believe
the orchestra was really there. I heard musicians fingering their
instruments, I heard someone on the record sniff, and the music
appeared out of a silent background - it was real!

Naturally the impact dbx has on the market will depend on
how many encoded discs and tapes they can release, and how
quickly. But several major cassette deck manufacturers (eg
Technics, Trio, Teac, Yamaha, Marantz) are fitting dbx along-
side Dolby in some of their models, and Dolby must be thinking
hard about their next move. (At ETI were thinking about how
nice a review model would be, hint, hint!) With approximate
RRPs of £120 for the Model 222 (for two head tape decks) and
£170 forthe Model 224(three head decks), adbx noise reduction
system could be the best upgrade you'll ever make.

to
n

5:

mil Damn m

NOISE Lei/n. (an 5m
H

.an ' -. am I run

FREQUENCY IR!) ,
ETI

Fig. 5 A comparison graph of noise reduction systems, issued by dbxi
41