I/O Module (pseudo balanced)

[Martin Czech]

For the further investigation of modules I found it necessary to have
an I/O module that is able to provide external sources (CD-player,
generator etc.) to my testcircuit and the other way signals from the
testcircuit to my mixing desk (then either headphone or loadspeaker). Both
directions use one half of an 5532. Well, and proper 6.3 mm jacks to safely connect to external devices.

The output is a noniverting 1x or (via jumper) 10x amplifier (rfeedback
~20k) with 1uF 10k dc filter and 330 Ohm output resistor. I've tryed a
simple 2x3 diode 1n4148 hard limiter after the output resistor, but it is
rather soft, the maximum amplitude is about 4V (8Vpp).  This is a bit too
high. I'll change that to one diode pair. I think the output protection
is necessary, I often use test clamp cables and other not very safe
connections in the test circuits, a mistake could give enormous spikes
at the output (10x mode !) that could kill my ears or the speakers. The
10x option is quite usefull for determining noise etc.

In the input section I wanted to dry a difference amplifier (the very
simple design with four resistors) like J.H. has proposed a year ago
or so.  I took a 20k (2x10k) resistor from output to inverting opamp
input and 20k from inverting opamp input to - signal, and 10k from +
signal to noninverting opamp input and finally 10k from noninverting
opamp input to signal ground. The idea is simply to reject common mode
signals. The resistors where 1% right out of the box, no specialties. A
5V dc common signal will give 8mV output, ie. dc common mode rejection
is about 55dB. This may sound no good, but we have to keep in mind that
our normal signal level is 1V. Any common mode noise should be 40dB down,
ie. 10mV. Together with the common mode rejection of 55dB this will make
95dB, and this is good enough. I tested common mode rejection also with
ac, since I have not yet an oscilloscope, I had to trust my ears, the
common mode feedthrough was not getting louder from 100Hz up to 16kHz.
The last test was a test lead from + to - input this loop, was then put
near to a 220V ac mains transformer (wall wart). With the 10x output
option the noise was quite loud, twisting the loop into a twisted pair
"cable" made this inductive effect cancel out completely. It is allways
nice to see that theoretical well understood things work in reality, too!
I think I'll use this configuration for ALL my module audio inputs.
This makes (together with microphone cable and stereo jacks (hey,
microphone cable is durable and comes in many colours!)) a pseudo balanced
patching cable scheme (the outputs are not balanced) that should avoid
ground loops and inductive hum completely. It is dirt cheap and needs
no special components. The additional noise of the 5532 can be neglected.

Common mode rejection is good right from the start, cause I used 6 10K
resistors from the same belt, they match within 0.3%. The rejection can
be further improved by making the output-to-inv.-input resistor variable,
instead 2x10k take 10k and 9k , the difference is measured with a pot and
then a fixed resistor about that value is soldered in. The resistors have
this 20k/10k ratio because I want the input impedance to be equal for +
and - input. Differential input impedance is 20k this way.

In the future I may lower this to the usual 10k. An additional passive
lowpass filter should be added in order to keep hf down, before it hits
the input up-amp, fc about 40kHz or so. The SSM2017 app. note shows this.

The only disadvantage I can see now is some uncertaincy in dc
output offset, the 5532 makes about 0.5mV, but this can vary widely
with temperature.  The 20k/10k resistor ratio makes the input offset
voltage worse via input bias currents. Applications with more demanding
offset figures could be made  with an LF411 low offset type (eg. cv
for frequency).

Ok, now I have this I/O thing it is also usefull since it represents
front and backend of any module I'll do (with the restrictions above).
So I have only to build once what is needed several times.  This is also a
good idea for the usual expo converter that I'll use.  Reuseability. Of
course, this is only valid for test purposes, the real modules need
allways their own ;->, this is no good saving advice for christmass.

m.c.