Re: Proper size of capacitors for transient suppression?

Andy Wander wrote:

>
> I understand that the ground is "differential" to the signal
> lead, but a differential input can only CANCEL what is "present"
> on the 2 lines. Induced noise is cancelled by the input circuitry
> seeing the same signal on BOTH lines, and taking the difference
> which should be "zero".

One way of looking at it is to keep in mind that _all_ voltages
are differential. Voltage is _always_ measured between TWO
points. Many of the voltages we look at are measured between
a particular point and this magic thing we call ground, so we
tend to forget that, and start thinking about the voltage on
a single point.

A differential input circuit detects the voltage between its
two input terminals. A single ended circuit detects the
voltage between its input terminal and its own _local_ ground
terminal.

> If the ground lead is grounded, and not tied to an input, then
> any noise induced in it should be(at least partially) shunted
> to ground, and so will not appear at the input circuitry. I
> guess that would only be 100% true with a zero-impedance
> ground-kind of hard to achieve in the real world, huh?

Actually, the case we're discussing here is the exact opposite
of the zero-impedance ground case. We're dealing with things
like encoders that are _not_ grounded at the far end of the
cable. So ground loops are not an issue. Since all voltages
are measured between two points, we simply need to figure out
which two points they are, and insure that the same noise is
induced on both points.

Take non-differential encoder for instance. It has four
terminals: negative supply, positive supply, phase A signal,
and phase B signal. I specifically use the term "negative
supply" instead of "ground", because it is NOT grounded to
the frame of the encoder.

If you want to connect this encoder you need four wires. There
are three ways to do it:

1) use 4 individual wires (perhaps loosely twisted inside an
overall jacket

2) use 2 tightly twisted pairs, with + and - in one pair and
A and B in the other pair.

3) use 2 tightly twisted pairs, with + and A (or B) in one pair
and - and B (or A) in the other pair.

Number 3 is the best choice. The single ended outputs of the
encoder generate signals on A and B that are relative to the
negative supply (and the positive supply, since + and - are
almost certainly bypassed to each other). Likewise, the single
ended inputs at the other end of the cable measure the signal
with respect to the negative (and positive) supply. So twisting
each signal wire with one of the power supply wires is the best
thing to do.

If you have a true differential encoder and differential inputs,
then your best bet is to use three twisted pairs, with + and -
in one pair, A and ANOT in the second pair, and B and BNOT in
the third pair.

If you have a differential encoder but single ended inputs,
you should use wiring arrangement #3 from above, not the three
pair arrangement. Since single ended inputs are referenced
to the supply conductors, the signals should be twisted with
the supply conductor.

I have worked extensively with VFDs using encoders in places
like paper mills, where the encoders can be hundreds of feet
from the VFD, and the noise is horrible - pulse width modulated
motor power at hundreds of amps and 480V. Our standard encoder
cable is three _unshielded_ twisted pairs, with differential
encoders and inputs. The pairs are very tightly twisted, about
3 or 4 twists per inch.

> I have had many experiences(some good, some bad, but almost
> all ultimately successful) in wiring and troubleshooting
> large audio systems to make them work right, after trying
> it the "right" way.

One big difference between connecting audio equipment and
connecting encoders is that individual pieces of audio gear
each have line connected power supplies and individual grounds,
so ground loops are a major issue. Encoders on the other hand
get their power thru the signal cable, and do not have a local
power supply or ground.

Another difference is that audio signals are sensitive to
small amplitude (millivolt) in-band signals like hum. A 60Hz
sine wave at 100mV would generate horrible hum in most audio
equipment, but has absolutely no effect on an encoder where
the signal is a 5 volt square wave. Encoders are more likely
to suffer from 100nS 5V spikes induced by switching of relay
coils or power semiconductors. The techniques for dealing
with fast, high amplitude "edge noise" are different than
those for dealing with low amplitude in-band noise. (Some
techniques, such as differential inputs, work for both.)

> I know that if you "Simulate" a balanced
> line by tying an unbalanced audio output ("+" and "gnd") to
> a balanced input ("+" and "-")using a twisted pair, with an
> overall shield, you CAN get some observable results as far
> as noise reduction-but I have never measured it, and don't
> know "why" it works.

It works because the balanced input amplifies the difference
between its "+" and "-" inputs. The unbalanced output delivers
the desired signal between its "+" and "gnd" outputs. So if
you connect "+" and "-" of the input to "+" and "gnd" of the
output, the balanced input is amplifing the desired signal.

Regards,

John Kasunich