Modifications to the
16-channel mixer
In February, March, April & May of this year
we published a 16-channel mixer design which
has been very well received by enthusiasts.
Predictably, many people are modifying the
design to suit their own purposes and equally
predictably, some have found that the
performance can he improved.
Whenever we publish a new design, we do so in the expectation that
at least some of our readers will
closely check through the design parameters to see what it achieves, to see
if there are any mistakes and to see if
it can be improved. And every now
and again, some of these readers are
moved to tell us the results of their
endeavours. One such reader is Phil
Denniss and we'll let him take up the
story:
I was filing away my back issues of
SILICON CHIP when I came upon the
articles for your 16-Channel Mixing
Desk that were printed earlier this
year. As I have a great interest in audio
electronics I decided to read through
the articles.
Before pursuing this much further
I would like to say that while I have
not read the entire set of articles, I
found the presentation very good,
particularly the setting up procedure.
But it is a shame to see such a lot of
effort messed up by a fairly small but
important mistake. However, the situation is very easily fixed without any
need to change the circuit board.
Basically, the problem is that the
resistor values chosen do not yield
the specified differential input impedance, nor do they give very good
common mode rejection with any
S1 : 1 : MIC
2 : LINE BALANCED
3 : LINE UNBALANCED
10DpF
10k 1°/,
1k
1%
MIC ZO-....-'r--......- - - l l t - -......---w,1,,----...,_-"1
AND
LINE
INPUT 30---'-::+---+---tlr--4t----"WiAc--+----t--.......- - - - : : I
S1b
2,~
_.
• DENOTES CHANGE FROM ORIGINAL CIRCUIT
"1'
ia'l
0
REVISED PREAMPLIFIER FOR 16-CHANNEL MIXER
Fig.1: revised circuit for the 16-Channel Mixer preamplifier stage. The
component values that have been changed are highlighted with a star.
Compared to the original, this circuit offers better common mode rejection and
therefore is less prone to hum & noise pickup.
source that does not have zero source
impedance; ie, most real sources. I
have done some calculations to establish what the input impedance and
common mode gain are for the general case using the published circuit.
I checked the results by lashing up
the circuit and measuring the relevant circuit parameters.
I initially used an LF351 op amp at
DC without coupling capacitors because it was easier for me.
I found that the 9H2 resistor should
be 1kQ and the 427Q and 483Q values should total 10kQ, to balance the
values in the inverting arm of the
circuit. This will yield an input impedance of 1.07kQ and give a common mode gain determined by the
matching between the resistors
around this stage and the balance in
the source resistance. I reckon there
should be a 100pF capacitor to ground
from the non-inverting input of the
op amp as well, to keep a lid on the
common mode gain at high frequencies, and I feel that the 10kQ resistors
just after the input coupling capacitors do not do anything worthwhile
either, though I might be persuaded
otherwise.
Initially, when considering the
noise performance using the suggested modifications, I reckoned that
maybe it would be 6dB worse because
the input resistance was roughly
doubled, or perhaps 3dB because the
noise currents in the two op amps
were perhaps correlated (it turns out
that they are not).
Well anyway, I figured that I had
better check it out properly with an
LM833, to find out the real answer. It
sure surprised me. I figured that with
the inputs shorted to ground, the difference in S/N between the two circuits would be less than 1.5dB. With
300Q to ground on each input (ie, a
600Q source), the difference would
be about 1.15dB. This is because the
input noise depends hardly on the
NOVEMBER1990
75
11le MB3 I•• n»dium to
,,.•vy
duty two pl•tfonn
mounting br•ckel spec/flc.lly de•igned for •curing
•peak,,n, ate lo wall• and
eel/Ing•, deab or bench
tape. Once mounted lhe
br•cket can be rotated• full
360 degree• u _, u
l»lng •wfve/led up or down
until lhe required viewing or
ll•tening •ngle I• •che/ved.
I~
Both pl•lfor1T111 have
predrllled hole• for
mounting and •II moving
parla and jolnta are
eaai/y re/eaaed or locked
with an large a/Ian key
aupplied with lhe unit
~
ELECTRONICS
CHRISTIAN
BLIND
MISSIQ~ ·J
R,,gtoriKf[;~~e, bt1«
COUPON
Please cut and send to:
CHRISTIAN BLIND MISSION
INTERNATIONAL. P.O. _Box 5,
1245 Burke Road, KEW. Vic. 3101
Phone: (03)817-4566
e
D Please send me further information
about CBMl's work.
As long as it is possible for me. I will help:
D monthly D quarterly D annually
D lo prevent blindness
CJ to restore eyesight
CJ to rehabilitate the blind
Enclosed is my gift of$ _ _ _ _ __
Mr1Mrs, Miss,
Street, __ _
City : __ .
. l'aslcode :
The MB5 i• • heavy duty IWO
platform mounting bracket
de•lgned for securing alTIIIII
TV'• and speaker• lo walls,
ceiling3, desks or bench
Iopa. When mounled It can
be rotated• full 360 degrees
•• -11 n being swlw,lled
up or down to any viewing
or listening angle. Both p/sl
-forlTIII have predrilled holes
a'D , ~
.n.&\.liJ ,I ft
ELECTRONICS
76
for mounting and sreenily
adjusted with •large •llen
keysupplledwlthlheunit.
SILICON CHIP
noise current of the op amp at all but
depends mostly on the input noise
voltage and to some extent on the bias
and feedback resistors.
While studying the LM833 data, I
noted a graph of "Input Referred Noise
Voltage vs Source Resistance" and this
plot clearly shows that the noise does
not rise 6dB above the zero Rs level
until the source resistance exceeds
3kQ. For my circuit this does not happen until the source resistance actually reaches lkQ. At this value of
source resistance, the original op amp
will start to seriously load the source
and this will in turn increase the SIN
ratio more than my proposed circuit
will.
Just to make sure, I decided to lash
up the circuit and measure the noise.
The results were not too satisfactory
because I used a protoboard which
left the circuit open to pick up noise
and hum. The exercise mostly showed
how quiet the LM83-3 is and how easy
it was to pick up a lot af unwanted
garbage. However, it did show that the
equivalent input noise voltage was
about lµV and that the difference in
noise performance between the two
circuits is very small.
Generally the published circuit is
not a very good performer because of
this very problem. The performance
of the preamp is too dependent on the
source impedance and the source
impedance presented to the preamp is
unknown. The ideal solution to this is
to use a circuit that provides much
higher input impedance, such as the
two op amp instrumentation amplifier, so that the source impedance has
much less effect. This will change the
noise performance of the preamp. At
worst, the noise may increase by 3dB
but it may provide better performance
because the preamp will effectively
see a lower source impedance. It will
see only the impedance of the source
and not the input resistors.
Another less important point I
would like to discuss is the extensive
use of electrolytic capacitors for coupling. I appreciate that circuit designers are pretty well up the creek when
it comes to good high value (470nF
and upwards) capacitors, and it is very
hard to find an acceptable alternative
to the old electro in terms of size, cost
and capacitance. Bipolar electros seem
to have disappeared.
I think it would have been safer not
to use so many electros, although this
may compromise frequency response
and/or noise performance. But more
importantly I think you have specified the incorrect polarity for some
of the coupling capacitors in the
mixer and this may degrade the performance of the mixer somewhat. My
reasoning is this: the LM833 has a
fairly high input bias current, 500nA
typical and lµA max according to
the National book, and this can produce quite a high offset voltage (5m V
or more) at the inputs of the op amp.
This is amplified by the op amp (if it
has a DC gain of more than one) and
can reach 100mV or so at the output.
Now the input transistors of the
LM833 are PNP (for low noise I guess)
so the current flows out of the input
and will therefore produce a positive
voltage across the input biasing resistor, or feedback resistor for the inverting input. It is important to note
that the offset so caused may be positive or negative at the output, depending on the configuration of the
stage, but is usually positive at the
input. In the mixer, nearly all the
electrolytic coupling caps have their
positive terminal connected to the
source and their negative terminal
connected to the load.
A quick check over the circuit indicated that the input capacitors of
ICs la, 2b, 6a, 6b, 7b, 8b & 9b, and
the output capacitors for ICs la, lb,
2a, 7a, 8a & 9a, are the wrong way
around. I cannot tell the extent to
which this will affect the performance of the mixer but I think that it is
advisable not to reverse bias electrolytic capacitors if possible. (P. D.,
Chippendale, NSW).
• How do you answer an onslaught
such as this? Well, as we have done,
you publish it.
We quite agree that our single op
amp balanced input will not give as
good common mode rejection as the
classic twin op amp design: In designing the balanced input stage for
the mixer, we were deliberately trying
to minimise the op amp count and
obtain the best signal/noise ratio.
However, as Phil Denniss points
out, the circuit can be modified to
improve the common mode rejection,
particularly with real source impedances such as 600Q, without a significant increase in the residual
noise.
We calculated the noise performance for both single op amp circuits
Fig.2 (left): this is the amended wiring diagram for the
preamplifier board. The changes are all at the top of the
board, near switch S1.
and confirmed that the suggested modifications will provide only a 1.3dB increase in noise while changing the
common mode rejection from -16.2dB for our circuit to
better than -48dB with the modifications.
We also tested the alterations and found the S/N ratio to
be -99dB with respect to a 2V output and 600Q source impedance. The original arrangement produced -90dB under the same conditions, although this measurement was
masked by hum pickup.
The results indicate that Phil's modifications give superior results since the hum pickup is considerably less
due to the improved input balance.
Making the changes
The revised circuit for the preamplifier is shown in
Fig.1. By comparing this to the original circuit on page 61
of the March 1990 issue, you will see where the changes
are. To help you spot the changes, we have highlighted
each changed component value in Fig.1 with a star. Most
of the circuit changes are associated with the section
involving pin 3 of ICla.
To help those who have already built the mixing desk,
or those who intend to build it, we have produced an
amended wiring diagram for the preamplifier board - see
Fig.2. This will take the place of the wiring diagram
shown on page 72 of the April 1990 issue.
Note that the wiring to the switch is now simplified.
The shielded cable from the pole of Slb to point "x" on
the original diagram has been removed entirely. So has
the shielded cable from the very top of the switch wafer. If
you intend making changes to the original switch wiring,
follow the new diagrams very closely.
Six components along the switch end of board are
altered. If you look at the top of the board you will see a
line of 7 components, with two 33µF capacitors at the
end. The changes to these are tabulated below:
The 91Q resistor adjacent to these components is also
Old
1.2kQ
10kQ
560Q
1.8kQ
470Q
13Q
0
New
1.1kQ
10kQ
100pF
10kQ
link
link
changed, to lkQ. Note also the 1. lkQ resistor added across
the switch and the earth from point 12 on the switch to
the board.
In addition, as noted by Phil Denniss, a number of
electrolytic capacitors in the circuit are reversed in polarity. These are now correctly shown on the diagram of
Fig.2. However, capacitors on the other boards should
also be reversed. The capacitors in question are the input
capacitors of IC2b, IC6a, IC6b, IC7b, IC8a & IC9b and the
output capacitors of IC2a, IC7a, IC8a & IC9a.
Ideally, these capacitors can all be bipolar types which
are readily available, although they do cost a little more.~
NOVEMBER 1990
77
