Design Philosophy
Introduction
In this page, the reasoning
behind the design of the audio preamp elements is explained. Some of the
viewpoints (which have been reached over a number of years...) may not be to
every reader's agreement.
This preamp system was
designed to provide features which were not available on commercial products or
other designs. Some of these features may be of interest to other readers too
(!) which may encourage them to spend time building one of these
systems.
Sonic Features
What Can We Hear
?
The performance requirements
needed from audio electronics should be defined by the sound information which
we can actually hear with our ears and how that information is
interpreted.
It is well known that the
frequency range of hearing for audio tones is between 20Hz and 20kHz (although
the high frequency limit tends to reduce with age, or from regular exposure to
high sound levels). All (or most) audio equipment will carry sound within these
limits, although some extend to a higher frequency.
For a single tone, most
people can hear a change in level of 1dB, although at certain frequencies and
under controlled conditions some people can hear changes of 0.25dB. So, it
would seem that if the amplitude response varied less than 0.25dB within
20Hz-20kHz then this would be adequate. As music is a complex signal containing
many frequencies simultaneously, the change in total power heard by the ear
will be greater than that of a single frequency tone. So, for example, even a
0.1dB drop in amplitude response over a sufficiently wide frequency range may
become audible.
Clearly, the amplitude
response should be as flat as practical within the 20Hz to 20kHz
range.
Time Response
Accuracy
A system which is "Time
Accurate" will be able to reproduce the many different sounds from a single
instrument's note (eg. a cymbal or drum being hit) so that they all arrive at
the ear with the same time relationship (ie. together) as when they left the
instrument. Note that few loudspeakers are time accurate across the whole audio
band, especially at low frequencies where the sounds can appear to be 1-3
metres behind the loudspeaker cabinet.
To achieve a flat time
response requires the amplitude ("frequency response") to be flat within the
audio band, with only gentle roll-off filters being used - thus causing the
actual frequency response to extend well outside the audio band (this is
because an analogue filter will cause the phase of the signal to change). With
modern op-amps it is easy to build an amplifier with a very wide bandwidth, but
this will allow "stray" radio interference signals above 20kHz (some shortwave
broadcasts, mobile phones, etc) into the amplifier which will probably cause
effects on the audio. The effects produced will vary, but could be audible
interference or a subtle increase in distortion (due to intermodulation
distortion, or the amplifier oscillating at inaudible frequencies), etc.
Distortion, Noise and
Headroom
Historically, the
measurement of an amplifier's distortion performance has been to measure its
total harmonic distortion - the unwanted harmonically related frequencies which
are "created" by the amplifier when it is fed by a single tone. As music is
rarely a single tone, but a complex arrangement of individual tones forming
"sounds", the total harmonic distortion measurement is useful, but is not the
only relevant measurement. Intermodulation distortion (an amplifier's response
to a number of simultaneous tones) must also be low. This is a measurement of
any unwanted frequencies created by the input signals "modulating" each other.
An amplifier should not add
noise which could be audible to a listener even when a quiet CD etc is being
played loudly. This is relatively easy to achieve.
An amplifier must have
sufficient dynamic range to accurately represent signals from digital sources
(CD, DVD). It must also cope with the wide variation of signal levels from
different source equipment ("hi-fi" levels and "professional" levels etc). From
measurements made on a number of sources in a home, this is of the order of
25-30dB.
The amplifier must have some
"headroom" to allow for signals which are "unexpectedly" louder than the
"loudest expected" signal. For example, with vinyl records, the pickup
cartridge may generate large signals from scratches which the stylus hits,
which should not cause the amplifier to distort the signal further. Even with
digital audio, where in theory the signal cannot exceed a particular level,
some headroom is still needed to allow for "ringing" in the filters with
"clipped" music signals.
Cable
Effects
An amplifier should be
independent of the inter-connection cables which are used. Many people spend a
lot of money on "Super-fi" cables because they believe that they can hear a
difference (and...possibly...the salesman conducts the demonstration in such a
way that a difference can be heard).
Technical parameters of
cables (capacitance and inductance) do vary between cables which can have an
effect on the input/output stages of amplifiers. Although the frequency
response effects will be minor, they will probably have an effect on the time
accuracy of high frequency sounds, so this may be a feature people can
hear.
The cost of a cable is not
the key to a consistently good sound! Remember that in the entire audio chain
from the microphone to the loudspeaker (and also via broadcasters) the audio
signal travels along long distances of "standard" cable (eg. 500 - 1,000
metres). If cables caused a (technically) unknown effect which is audible on
short interconnection cables in the home, then the audio signal heard from
CD/radio/etc would sound appalling by the time it reached the listener, but
(somehow) would still measure perfectly - but this does not happen.
The user should consider the
capacitance of the interconnecting cables used, especially where "long" runs
will be used.
Practical Features
Flexible
Construction
The system design is based
on a number of different PCB cards. This modular approach allows many different
configurations to be built - for example, including a stereo preamplifier with
a large number of inputs, an 8 channel surround system, and a "low cost" two
input source selector and volume control.
This is useful for the home
constructor (with not enough time...) as a smaller or simpler system can be
built initially, which then can be expanded over time.
Cards
This preamp system is based
on three main cards:
- Dual Input
Card - two stereo balanced inputs with XLR input sockets, with preset gain
adjustment jumper plugs, electronic audio switches to two stereo busses and
local voltage regulators
- Dual Output
Card - two stereo balanced outputs on XLR sockets, with selection from two
stereo busses, electronically controlled volume control, jumper plug
mono/stereo selection and local voltage regulators
- Motherboard -
two stereo busses, distribution of DC power to all cards, marshalling the
data/control line signals and slots for up to 12 cards.
Balanced Inputs and
Outputs
Balanced inputs will help to
reject unwanted signals and to minimise any "hum" problems where a large number
of sources are used (radio, CD, VCR etc) or where some source equipment is in
another room. As most hi-fi equipment has unbalanced outputs, then the balanced
input stage must work like an audio transformer and still be efficient at
cancelling out unwanted signals from unbalanced equipment as well (not all
electronic balanced inputs do this, due to unequal source
impedances).
Multi-room
The design allows the system
to be configured and used to feed a number of amplifiers/loudspeakers locally,
and also to feed a house distribution. For the house distribution, the output
card can generate a stereo signal and a mono signal (simultaneously), because
mono sound is often more audible in a kitchen or bathroom, etc. High quality
output transformers can be fitted to the output card specifically to fully
isolate the central audio system from a (large) house
distribution.
Electronic
Implementation - Some Key Points
DC Blocking
Capacitors
In an amplifier, it is
important to stop DC signals from source equipment (CD player etc) being fed
into the amplifier circuitry because, if this is large, it will cause
distortion to occur on loud audio signals, clicks to be audible each time the
volume control is used and may also cause problems if it reaches the power
amplifiers/loudspeakers.
Capacitors are used to
"block" any DC signals. However, these must be "non-polarised" (as their
response to the audio signal must be equally linear for positive and negative
waveforms) and sufficiently large to minimise any low frequency roll-off
(causing amplitude response changes and time accuracy variations) - which means
physically large and expensive! If capacitors are needed throughout the
circuitry, then the total cost increases and the effect of the multiple
capacitors throughout the chain may become audible.
In this design, a single
capacitor is used at the input stage to the amplifier to block any DC. The
electronic design and choice of components is such that no further DC blocking
capacitors or control circuits are then needed throughout the rest of the
preamplifier circuitry.
Choice Of
Op-Amp
The
Burr Brown
OPA134 opamp was chosen for this design as it has:
- low noise (although there are quieter
op-amps, they have other features/problems which may make them not the most
appropriate choice for this design)
- a sensible gain bandwidth (opamps with
higher gain bandwidth are not really appropriate for this use, and might cause
instability problems)
- an output capable of driving low
impedance loads (which allows low value resistors to be used throughout the
design, helping to keep the noise low)
- a very low DC "error" signal at the
output (which removes the need for dc blocking capacitors in the
circuit)
- very low distortion (harmonic and
intermodulation)
- a sensible price!
Click Free
Switching
When a listener changes from
listening to one source to another (eg. CD to radio) there should not be any
audible clicks. Clicks can occur due to small amounts of DC being present on
the audio signal at the switch and also by the audio waveform being "chopped"
as the sources are switched on/off.
Although the volume control
could be turned down before each source selection change, it is easier if the
audio switches themselves are "silent". For this design,
Analog
Devices SSM2402 electronic audio switches have been used which fade the
signal out quickly, and then fade the new signal up - and so are guaranteed to
be "click free" for audio signals.
Radio Frequency
Filtering
As the upper limit of the
amplitude (frequency) response is outside the audio band (to minimise the
amplitude and time accuracy effects) some form of protection against
radio-frequency signals entering the amplifier is needed. In this design, a
"bifilar coil" is used at each of the audio inputs.
A bifilar coil is a small
inductor (ferrite ring) with the two (differential) input wires wound together.
This means that if an audio signal is passed through, because it is
differential (the current flows in opposite directions down the two wires) the
electromagnetic field cancels out and so the ferrite ring/coil has little
effect. However, the bifilar coil will have a much greater effect on any
common-mode radiofrequency signals (which are present on both wires). The
inclusion of small capacitors after the coil will cause the amplifier to reject
these unwanted signals.
4-Layer
PCBs
Four layer PCBs are used for
all of the cards as this allows the use of a large ground plane (helping to
minimise crosstalk - audio signals breaking through audio switches which are
"off" etc) and large tracks/areas of copper for the power supplies to the ICs.
This allows the performance
to be defined at the design stage and is not dependant on "adjust on test"
components etc.
Email | Copyright 1996 - 2004 errock.co.uk -
audio design |
