This is only a preview of the August 1988 issue of Silicon Chip. You can view 40 of the 96 pages in the full issue, including the advertisments. For full access, purchase the issue for $10.00 or subscribe for access to the latest issues. Articles in this series:
Items relevant to "High Performance AC Millivoltmeter":
Items relevant to "Universal Power Supply Board":
Items relevant to "Building the Discolight, Pt.2":
Articles in this series:
Articles in this series:
Articles in this series:
Articles in this series:
|
State of the art
AC MILLIVOLTME
Just how do you measure the extremely low noise
output voltages of modern audio equipment,
particularly power amplifiers and compact disc
players? Our AC millivoltmeter has been designed
for those specialist tasks and can measure audio
signals down into the microvolt region.
By LEO SIMPSON & BOB FLYNN
Today's topline high fidelity
equipment really does stretch the
measuring limits of even the best
audio test equipment. The run-ofthe-mill AC millivoltmeter with a
bottom range of 1 or 3 millivolts full
scale is nowhere in the race.
Just consider a typical CD player
today. It will have a signal to noise
ratio of - 96dB as a minimum and it
might be as low as - 106dB. Take
the - 96dB figure for a moment,
which is with respect to the maximum 2 volts RMS output. To confirm that noise figure, the AC
millivoltmeter must be able to
measure accurately down to below
30 microvolts.
To confirm a noise figure of
-106dB below 2V, it must be able
to measure accurately to below
lOµV.
Modern stereo preamplifiers and
power amplifiers present a similar
challenge. Consider the Sony
TA-N77ES stereo power amplifier
reviewed in the February 1988
issue of SILICON CHIP. It has a
signal to noise ratio of - 120dB 'A'
weighted with respect to its rated
power output (200 watts into 80).
To be able to confirm that, the
millivoltmeter must be able to be
accurately measure below 40µV.
· Or consider the Studio 200
Stereo Control Unit presented in
our June and July issues. It has a
signal-to-noise ratio of - 107dB 'A'
weighted with respect to its rated
output of 1.25V. To confirm that,
you need an instrument capable of
measuring signals down to less than
5µ V! In fact, if the measuring instrument's own internal 'noise
floor' is not to intrude on the
measurement, it must be able to
measure down to around 1µ V.
Now you see why conventional
millivoltmeters are just not in the
race.
In addition, modern hifi equipment may be measured with wide
frequency bandwidth, say to
lOOkHz or beyond, band-limited
(20Hz to 20kHz) or, as already mentioned, with 'A' weighting. As far as
we know, there is no commercial
Most of the parts, with the exception of the resistors on the main attenuator switch, are mounted on this PCB.
TER
Designed specially for audio measurements,
this instrument has a noise floor of less than one microvolt.
equipment available today which is
up to the task.
OK, so we've demonstrated the
problem. Now we present the solution, our new state-of-the-art AC
millivoltmeter. Its performance is
summarised in the accompanying
panel.
Incidentally, our use of the term
"state of the art" may lead some
people to jump to conclusions. They
may associate SOA equipment with
programmable microprocessorcontrolled digital measuring equipment costing tens of thousands of
dollars. But while such equipment
is available, their absolute measuring limits are often pretty ordinary.
As you can see from the specification panel, the new AC
millivoltmeter is designed especially for measuring modern high performance audio equipment. It does
not have the very high input impedance of 10 megohms, typical of
digital voltmeters and some older
designs of AC millivoltmeters. Such
high input impedances are not required, for two reasons.
First, most audio measurements
are made at the output of equipment which has very low impedance. For example, most CD
players, tuners, cassette decks and
preamplifiers have output impedances considerably less than
1k0. For measurements on power
amplifiers we are looking at outputs
with source impedances of a few
milliohms!
Second, even where measurements are being made between
stages of audio equipment, they still
involve low impedances. Accordingly, we have adopted the standard
input impedance used by commercial noise and distortion meters; ie,
lOOkO unbalanced.
The frequency response of the instrument is - 3dB at 5Hz and
160kHz, on the 3V range. We quote
the specific range because the
ultimate bandwidth does vary
slightly, depending on the input
range selected.
Mind you, some power amplifiers
have a small signal bandwidth far
in excess of 160kHz, sometimes to
as high as 1MHz or more. Where
such figures have to be confirmed,
the only practical way is to use an
oscilloscope. For the vast majority
of audio measurements though, the
160kHz bandwidth of the AC
millivoltmeter will be more than
adequate.
More typically, where noise
measurements are to be made,
bandwidth limiting is required. For
unweighted noise measurements
(ie, with a flat frequency response),
it is usual to measure with a frequency response of 20Hz to 20kHz,
at the - 3dB points.
The alternative is to make an 'A'
weighted measurement, with a frequency response specified by the
IHF-A-202 (or EIA RS-490 )
specification, and defined in ANSI
S1.4 (specification for a sound level
meter) as shown in Fig .1. This filter
characteristic is designed to more
Specifications
Input impedance
1 00k0 (unbalanced)
Frequency response
5Hz-160kHz at - 3d8 points (on 3V range)
5Hz-130kHz at -3d8 points (on -30dB
range)
Measuring ranges
1mV to 1 00V RMS f.s.d. in 11 ranges
Noise (ratio) ranges
0dB to -60d8 f.s.d. in six ranges
Noise floor
-64d8 below 1mV (630nV) with 20Hz to
20kHz filter; -68 .5d8 below 1mV (375nV)
with 'A' weighting .
AUGUST 1988
19
~
/,r
.... "'"'"
~r-,....._
"
-10
/
V
-2 I)
II'
II
0
I/
/
-4 0
I
-50
I
j
-6 0
-7 0
The circuitry
/
I
I
100
10
HERTZ
1k
10k
20
Fig.1: this is the A-weighting characteristic used for measuring most audio
equipment today. It tends to approximate the response of the human ear to
very low noise but neglects the more audible effects of mains ripple in power
amplifiers.
or less approximate the sensitivity
of the human ear to low level
sounds. Therefore, both the high
frequency response and low frequency response are rolled off, as
illustrated.
No provision has been made for
DC measurements. The average
digital multimeter is more than adequate for this task, even where
voltage measurements down to a
fraction of a millivolt have to be
made (as for example, when
measuring the DC offset voltage at
the output of an op amp).
Features
Our prototype is housed in a folded metal case with aluminium base
and blue Marviplate (steel) cover. It
measures 235mm wide, 210mm
deep and 120mm high (including
rubber feet). It is mains powered
and is switched on at the rear of the
unit.
On the front panel it has a large
meter movement (nominally 100mm
wide) with scales 0-lV, 0-3.16V,
and decibels, with 0dB referenced
to 0.775 on the 0-lV scale.
There are four rotary switches
and one potentiometer: the Input
range switch, Noise range switch,
Mode switch and Filter selector.
The input selector has 11 ranges
measuring from lmV fsd (full scale
deflection on the meter) to 100 volts
20
SILICON CHIP
the instrument and when very low
noise measurements are being
made; eg, lower than - 60dB with
respect to lmV.
Just to show how good this instrument is, it can measure signal noise
ratios of better than - 120dB with
respect to 1V RMS (or - 126dB
below 2V). Alternatively, for an input reference of 30V RMS (typical
for a 100W amplifier), it can
measure SIN ratios better than
-150dB. In other words, our new
AC millivoltmeter is several orders
of magnitude better than even the
best audio equipment.
fsd. The input divider runs with the
standard 3.16 ratio between
ranges. This odd figure is used
because it is equivalent to lOdB
steps when switching ranges.
The Noise range switch has six
positions, giving settings of 0dB to
- 60dB. It is used in conjunction
with the Set Level potentiometer
which sets the meter's pointer to
the 0dB mark on the scale before
taking a signal-to-noise ratio
measurement.
The Mode switch has three positions: Volts, Set Level and Noise.
These will be explained later in this
article. Finally, the Filter switch
has three positions: Flat (giving the
widest freqency response), 20Hz 20kHz, and A Wt ('A' weighted)
which has already been mentioned
above.
There are two insulated BNC
sockets on the front panel, one for
the input signal and one for the output signal to an oscilloscope or a
frequency meter. The output level
from this socket is around 140mV
RMS, for a full scale deflection of
the meter.
There are two more switches to
be mentioned. One is a toggle
switch which is used to connect the
CRO signal earth to the case of the
instrument or to the mains earth.
The other is a pushbutton switch
used to check the "noise floor " of
The circuitry relies for its performance on a number of carefully
selected op amps. The most important op amp is the input device.
Contrary to what a number of
readers have expected, we have
not used the low noise LM833 in
this application. Instead, we have
used a quieter and more tightly
specced device, the ultra-low noise
OP27. This was first produced by
Precision Monolithics, Inc, USA and
has since been second-sourced by
Harris Corporation and Motorola
Inc.
Not only is the OP27 one of the
quietest op amps currently available, it also has the advantage of a
relatively high input resistance
which is a minimum of 700k0.
Also specified are three LM318
op amps. These have been selected
for their wide bandwidth. The
LM833 dual op amp is featured too,
in the precision rectifier and meter
driver. Apart from those, there are
two LF351 Fet-input op amps and
one LF353 dual Fet-input op amp.
Let's now have a look at how the
circuit works. The easiest way to
understand it is to look at it in the
"Volts" mode first and then look at
the other functions.
In the "Volts" mode, the Mode
selector (S3) is set to Volts, the
Noise (ratio) switch (S4) is set to its
0dB setting and the Filter selector
switch (S5) is set to Flat. This is
depicted on the complete circuit
Fig.2 (right): of the eight op amps
►
specified in the circuit, IC1 (OP27) is
the key to the high performance of
the unit. It has very low noise and
high input resistance.
I
N
... I
(0
0:)
0:)
....
~
en
c:::
~
>
c:::
E
240VAC
-
L--S&.
(Ut//30k)
.
ALL RUISTOIIS 1%
111111()
5.54k -10d8
PRESS FOR
Mlt NOISE
+15V
+15V
1-~ 6
F'' f
NOISE All'l.flER
NOISE
.,.
6
f
NOIS
S-4'
6
f
+15V
10tlf
GND
II
-~~~M
.,.
S5a
SC04-1-11888-1
AC MILUVOLTMETER
-15V
+15V
______·_·:~~!
VOLTS
1
100k:
FlA
3.3k
3.3k
20k
.002
(.0011/.001)
+1_5V
+1JV
20Hz IIGH•PASS FI.TER
+15V
25.5k
* SELECT FOR MIIIIUII OFFSET
18k
'A' WElll1111116 FI.TER
2k
CAL
VR2
50
.
•
MAINS
EAIITN
rJ~,
~~
steps (ie, 3.16 times) by switch sections S4a and S4b. S4a switches the
feedback resistors for IC3 while
S4b switches the feedback resistors
for IC4.
For the OdB setting, both op amps
have a gain of unity. For the - lOdB
setting, IC3 has a gain of + lOdB.
For the - 30 to - 60dB settings, IC3
has a gain of + 30dB (set by the
12k0 and 3900 resistors). The gain
of IC4 varies from unity at the
- 30dB setting to + 30dB at the
- 60dB setting.
The output from IC4 is then fed
via S3b to IC2 and then passes
through to IC7 as before. We'll explain just how noise and ratio
measurements are made in practice, later in the article.
In the Set Level mode, the gain of
IC2 is varied by the Set Level control, VRL The gain can be varied
between 6 and 2L6 times.
Filter stages
Here's what the completed unit looks like inside the chassis. Note that the
power supply is mounted near fhe rear panel, to keep mains hum away from
the meter circuitry. Full constructional details will be published next month.
diagram, Fig.2. The signal to be
measured is fed in via the BNC
socket to the stepped input attenuator SL
The total resistance of the attenuator is very close to l lOkO and
the resistor values are arranged to
give the 1 to 3.16 ratio between
ranges mentioned above.
If set to its highest setting (ie,
lOOV) and with an input signal of
lOOV, the signal at the wiper of Sl
will be lmV. In fact, for each range,
when the maximum input signal is
fed in, the signal at the wiper will
be lmV.
For example, on the 300mV
range, if you feed in 300mV, you
will get lmV at the wiper of SL
This lmV signal is then fed via the
4.7k0 resistor to the non-inverting
(+)input of !Cl, the OP27. This is
arranged to have a voltage gain of
34, as set by the 3.3k0 and 1000
feedback resistors.
So for an input of lmV from Sl,
the output at pin 6 of !Cl will be
34mV. From there the signal is fed
via switch S3a and S3b to IC2. In
the Volts mode, as set by S3c, the
gain of IC2 is 6, as set by the 6.Bkn
22
SILICON CHIP
and 3300 resistors, and the lkn pot,
VRL
For the same lmV input to Sl, the
output from IC2 will now be 204mV
(6 x 34). This signal is fed via switch
S5a, trimpot VR2 and S5b and then
via a 50µF capacitor to the stages
consisting of IC7a and IC7b.
IC7a and IC7b form a precision
full wave rectifier and filter circuit
to drive the meter. The lµF
capacitor across the 160k0
resistor, in the feedback loop of
IC7b, provides a DC-averaged output to drive the meter movement.
VR4 is provided for calibration.
ICB provides a buffered version
of the signal from S5 for viewing on
an oscilloscope. After calibration
has been performed on the instrument, a lmV RMS sinewave fed into
the input of ICl will be displayed on
the oscilloscope with an amplitude
of about 400mV peak-to-peak.
Noise & ratio measurements
In the Noise mode, the output
signal from ICl is fed via S3a to
variable gain amplifiers IC3 and
IC4 which are LM318s. The gain of
these op amps is varied in lOdB
Now let's have a look at the filter
stages, as selected by S5. As mentioned before, there are two filters,
for 'A' weighting and for the 20Hz
to 20kHz bandpass. The 'A'
weighted characteristic (shown in
Fig.l) is provided by a 4-stage
passive filter, consisting of four
capacitors and four resistors. This
is buffered by IC6 which is connected as a unity gain voltage
follower.
The 20Hz to 20kHz filter is provided by IC5a, connected as a
20kHz low pass filter, followed by
IC5b which is connected as a 20Hz
high pass filter. In other words,
IC5a effectively passes all frequencies below 20kHz and IC5b passes
frequencies above 20Hz. Between
the two of them, they provide the
20Hz to 20kHz bandpass.
The 'A' weighting filter has a loss
of about - 3dB and to ensure that
there is no jump in gain when the 20
to 20kHz bandpass or Flat filter
conditions are selected, trimpots
VR2 and VR3 are provided to
equalise the signal levels, at lkHz,
for all three settings of switch S5.
Power supply
The millivoltmeter is powered
from a 30V centre-tapped transformer feeding a full-wave rectifier, two lO00µF capacitors and
PARTS LIST
1 aluminium case with
Marviplate lid, 235 x 21 0 x
117mm (W x D x H)
1 Scotchcal front panel label,
228 x 113mm
1 30V 1 50mA centre-tapped
transformer (Altronics Cat.
No. M-2855)
1 meter PCB, code
04108881, 193 x 98mm
1 power supply PCB, code
04106881, 71 x 52mm
1 MU65 1 OOµA panel meter,
100mm x 82mm (Altronics
Cat. No. Q-0550 or
equivalent)
1 set of metal shields (see
metalwork diagrams, Part 2)
2 insulated panel-mount BNC
sockets (Belling Lee
LX04-0503-ZZOO5N or
equivalent)
2 miniature SPST toggle
switches
1 single pole 12-position
switch (make before break
contacts)
1 2-pole 3-position switch
(make before break contacts)
1 2-pole ?-position switch (2
wafers, make before break
contacts)
1 3-pole 3-position switch (3
wafers, make before break
contacts)
1 momentary contact, miniature
pushbutton switch
4 23mm fluted plastic pointer
knobs (Altronics Cat. No.
H-6050 or equivalent)
1 15mm knob
8 6mm PC standoffs
28 PC pins
two 3-terminal regulators to provide balanced outputs of ± 15 volts
DC. The regulator outputs are further filtered by 100µF and 220µF
capacitors.
There are also 16 0.lµF capacitors dotted around the circuit to
provide extra power supply rail
bypassing.
Three switches remain to be
mentioned. S2 is the momentary
contact pushbutton switch. It shorts
out the 4. 7k0 resistor in series with
the input to IC1. It is used when
making extremely low noise
measurements or when confirming
the "noise floor" of the instrument.
1 3-core mains flex with
moulded 3-pin plug
1 cord-grip grommet
1 2-way insulated terminal
block
1 3-way tagstrip
2 solder lugs
4 rubber feet
Semiconductors
1 OP2 7 ultra low noise op amp
3 LM318 op amps
2 LF351, TL071 FET-input op
amps
1 LF353, TL072 dual FETinput op amp
1 LM833 dual low noise op
amp
1 7815 3-terminal +15V
regulator
1 7915 3-terminal -15V
regulator
4 1N4002 1A silicon diodes
2 1N4148, 1N914 small signal
diodes
Capacitors
2 1OOOµF 25VW PC
electrolytics
2 220µF 16VW PC
electrolytics
2 100µF 16VW PC
electrolytics
3 4 7 µF 50VW bipolar
electrolytics
1 1µF 200VW metallised
polyester (greencap)
1 1µF tantalum or low leakage
electrolytic
2 0.22µF 63VW miniature
metallised polyester
2 0.15µF 63VW miniature
metallised polyester
S6 is a toggle switch which connects the CRO signal earth to the
case of the instrument. This is used
if the equipment being measured
gives an erroneous display on the
CRO, which is likely to occur with
double-insulated audio gear.
S7 is the power switch and is
mounted on the rear panel so that
mains wiring is kept as remote as
possible from the sensitive front
panel circuitry.
Using the millivoltmeter
As a further help to understanding how the circuit works, let's
consider a typical signal-to-noise
16 0 .1µF miniature ceramics or
greencaps
1 .047µF 63VW miniature
metallised polyester
1 .0022µF metallised polyester
(greencap)
3 .001 µF metallised polyester
(greencap)
1 39pF ceramic
1 22pF ceramic
1 12pF ceramic
1 1OpF ceramic
Potentiometers
1 1 kO linear pot, 16mm
diameter, PC mount
1 1OkO trimpot, horizontal mount
2 2k0 trimpots, horizontal mount
Resistors (0.25W, 1 %)
1 X 910k0, 1 X 160k0, 3 X
100k0, 1 x 91 kO, 1 x 75k0, 1 x
68k0, 1 x 56k0, 3 x 51 kO, 1 X
30k0, 2 X 22k0, 2 x 20k0, 2 x
12k0, 3 x 1 OkO, 1 x 7 .5k0, 3 x
6.8k0, 1 X 6 .2k0, 2 X 5.6k0, 1 X
4.7k0, 2 X 3.9k0, 4 X 3.3k0, 1 X
2. 7k0, 1 X 2.2k0, 2 X 2k0, 1 x
1.8k0, 1 X 1.5k0, 1 X 7500, 1 X
3900, 1 X 3300, 1 X 2200, 1 X
1800, 4 X 1000, 1 X 750, 1 X
220, 1 X 200, 1 X 180, 1 X
7.50, 1 X 2.70, 1 X 1.80, 1 X
1.10
Miscellaneous
Insulated hook-up wire, tinned
copper wire, shielded cable,
heatshrink tubing, copper or
brass shim, screws, nuts,
lockwashers, solder, Presspahn
insulating material ( 1 50 x
75mm).
ratio measurement. Say we're
measuring a run-of-the-mill 60 watt
power amplifier. For 60 watts into
an 80 load, it will deliver 21. 9 volts
RMS. This would be confirmed on
the 30V range.
Note that the Filter switch must
be in the 'Flat' position for voltage
measurements. Then the mode
switch would be moved to the 'Set
Level' position and the Set Level
control adjusted to bring the
meter's pointer to 0dB.
The input signal would then be
removed from the power amplifier
and its inputs loaded with a 4. 7k0
continued on page 71
AUGUST
1988
23
,------------------~I
I
STOP
PARITY
START
I
I
m
-~-
m
__
I
TO
DATA BUS
MICROPROCESSOR
DATA BUS
..,.__ _,,. BUFFERS
I
I
...,__.__...__ _._~
SHIFT REGISTER
I
I
I
SERIAL
1----,,-DATA
OUTPUT
TRANSMIT
CIRCUITS
RS-232
INTERFACE
ClRCUITRY
I
PARITY
CHECK
I
_____________
UART_j
CLOCK
(SETS BAUD RATE)
ADDRESS
DECODE - . . , _ _ . J
Fig.9: block diagram of a UART.
It is capable of full duplex operation.
specifies all of those characteristics. They are summarised briefly in Fig.8.
The UART
The main logic functions of the serial interface are
usually taken care of by a special LSI serial interface
chip called a UART, or universal asynchronous
receiver transmitter. A simplified block diagram of a
UART integrated-circuit chip is shown in Fig.9.
Bi-directional data-buffers connect the CPU data
bus to the UART. Inside the UART, there are two
separate sections: one for transmitting, the other for
receiving. The heart of each section is a shift register
that performs the parallel-to-serial or serial-toparallel conversion as required. Other logic circuits
add the stop, start and parity bits in the transmit
mode, or extract and respond to them in the receive
mode. Most UARTs can operate full duplex, meaning
AC millivoltmeter -
SERIAL
--DATA
INPUT
RECEIVE CIRCUITS
cs
L_ _ _
-
SHIFT REGISTER
CONTROL --..;..._- CONTROL
LINES
LOGIC
send and receive operations can take place
simultaneously.
The UART chip is set up and controlled by the host
microprocessor. Special data words transmitted to the
UART specify things like baud rate; 1 or 2 stop bits;
odd, even or no parity; and data word length from 5 to
8-bits. A short initialising subroutine in the main program sets up the UART prior to its use.
Another way to create a serial interface is to do it
with software. A short program can be written to do
the parallel/serial or serial/parallel conversions, deal
with the start, stop and parity bits, and provide the
timing for the desired baud rate.
We will show you how that is done in the next and
final instalment of this series devoted to
microprocessor programming.
lltl
Reproduced from Hands-On Electronics by arrangement.
(c) Gernsback Publications, USA.
ctd from page 23
resistor (or shorted, according to
the manufacturer's specs). The
amplifier's output voltage will then
drop to a very low value.
The next step is to move the
Mode switch to the Noise setting.
The Noise range switch should be
at the 0dB setting. Now we switch
down the input attenuator until a
reading above 1/3 of meter deflection is obtained. If the amplifier is
any good (ie, reasonably quiet), very
little pointer deflection will be obtained even on the lmV range.
At this point, we are measuring a
signal which is better than - 90dB
with respect to the amplifier's
rated output voltage of 21.9 volts.
(Remember each change of range
on the input attenuator corresponds to todB).
To increase the gain of the
measurement, we start rotating the
Noise range switch until the
meter's pointer moves up the scale.
That may be obtained with the
Noise switch on the - 20dB range.
If the pointer is indicating - 4dB,
the overall signal-to-noise ratio of
the amplifier is - 90 + - 20 +
- 4dB = -114dB.
This is a measurement of the
wideband residual noise. For most
hifi equipment this measurement
would be taken with the 20Hz to
20kHz filter selected which will
normally improve the measurement
slightly, to say, - 116dB. If an 'A'
weighted measurement is taken instead, the reading may improve
slightly again, particularly if there
is hum in the residual noise.
The procedure is similar when
measuring separation between
channels of a stereo amplifier, except that the Flat filter condition
would be selected.
Next month, we will conclude the
description with the info on construction and calibration.
lltl
AUGUST 1988
71
|