Propeller Head Plaza Technical and scientific discussion of amps, cables and other topics.

---

[ Follow Ups ] Thread: [ Display  All  Email ] [ Propeller Head Plaza ]

---

[ Alert Moderator ]

---

Re: Steve E. or anyone -Grounding...Again

Posted by Jacques (A) on August 28, 2005 at 06:08:09

In Reply to: Re: Steve E. or anyone -Grounding...Again posted by Dan Banquer on August 27, 2005 at 08:43:17:

first, I shall repeat your statement in capitals:
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
! !
! NOTE: THIS APPLIES TO UNBALANCED CONSUMER ONLY. !
! !
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

>. Many of us who work in the electronics industry outside of audio who have observed the posts from audiophiles, designers and dealers scratch our heads in near disbelief.

Oh yes! If the same problems would appear in instrumentation for example,
we should not have pHmeters, mass spectrometers, navigation compasses,
vibration signature analysis based helicopter preventive service computers,
gaz analysers... And most of the sensors used are inherently unbalanced.
However, it works...(well, I know that hospital-grade instrumentation
designed in the '80s was not often designed in order to work in a high RF
level environment: EMI certification standards used at the time were
unaware of the prevalent diffusion of cell phones and wireless networks.
So, nothing is perfect)

>The following is a technique used in instrumentation for low frequency
>applications.
>The chassis will be earth grounded via the earth ground at the three-prong
>outlet. The audio signal ground contained in the chassis is not connected
>to the chassis ground. This will require the design to be electrically
>isolated from the chassis ground
(1) As for low frequency, yes, you are right: the principle is to do things
in order that the current loop that links the source equipt and the load
equipt is contained within the two wires of the interconnect.

(1a)By having a very high resistance between the source signal ground and the noisy earth
plane, any voltage building between them (through stray caps and amins)
will give only a very small common mode current which cannot be detected as
long as the voltage build up won't break up the input devices.

(1b) furthermore, any stray magnetic field (mains xformer etc) will not
build any noticeable perturbation thanks to cancellation of the induced
current (coax or twisted wires)

+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
(2) The earth ground should be a protective ground only, and only when
needed (that is no double insulation). Its purpose is triple:

(2a) create a huge imbalance or a very high current in case the mains goes to contact
the chassis: this way, the breaker will trigger and save your life.

(2b), to give a return path to mains common mode noises from any origin.

(2c),give a voltage reference to the SYSTEM (a system: one or several pieces of
equipement connected together to work together to a defined purpose, say
your sound system as say Jamaicans) in order to avoid dangerous building of
static or any other voltage.

+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Having said that, connecting the system ground in one place (for unbalanced
consumer, I repeat ;-) as you advice is relevant. All topics (1a) to (2c)
are covered this way. So, I agree

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
! !
! However, a problem arise, that I shall adress below with help of Dan's !
! text. !
! !
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
We will now address the high frequency perturbations issue !

(3) Since the source equipment and the load equipement both have capacitances
to their chassis, and therefore to the earth ground, a loop can build up
which can carry RF current if allowed to (I let the reader make calculations
as an exercice for source and load stray capacitance of both 50 pF for a
100mV 5MHz signal superimposed onto the mains, the kind that gives a bad
quality non CE certified switching power supply ).
On audio equipment, depending on the technology, this RF signal will be
low-pass filtered and everything is fine (many tube equipement or well built
solid state) or perturbate the internal devices (bias modification,
enveloppe detection, uprising of > 2nd order nonlinearities, many weird behavior
that are said to give noticeable changes insound)
To get rid of these, we shall use one or several of these 3 ways:

(3a) give the RF sneak currents a new less impedant path (provided through
the Y caps of a common mode mains filter), while at the same time putting
in serial with the intrumentation path a large impedance at RF frequency
(provided by the common mode filter choke)
This way, attenuation up to 60-80 dB is attainable in the low RF range
(5-20MHz).

(3b) do things in order that common mode noise (the same on both
interconnect wires) stays common mode, and that the smallest part of it be
changed into differential mode (not the same on both interconnect wires).
Why? Because any well built audio gear has a decent common mode rejection
(often up to 40dB even at 1MHz), and no differential mode rejection, since
differential signal is what the audio signal is (you can have differential
mode rejection at RF frequency thanks to a cap in parallel with the inputs,
but a cap usefull to attenuate a 100KHz perturbation with wreck havoc at
20KHz and cut trebles...) To stay common mode , the noise has to stay
balanced on both wires when seen from the load. By construction, it is the
case with 100% covered coax, as Dan notices. However, 100% covered is not
practical and any stray capacitance difference between the 2 cables (easy
with a single shield coax) gives diff mode noise from CM. Just use
Thévenin.

(3b1)A real world concern in this matter is the output impedance of each wire.
Most of the time quite zero on the ground output, and a value quite
undefined on the signal output (remember, we are talking of the impedance
at RF frequency, not audio). Although I don't remember having seen that on
any audio schematics, several ways can be explored: as I do with balanced
outputs, put a (very small) common mode xformer/coil (murata SMD, ask me
for the specs) in serial with the outputs then two SMD caps of about 100pF
between each output of the xformer/coil and the chassis). This way, both
impedances are quite the same and no common mode noise changes into
differential mode (in fact, in balanced world, use a twisted wire to
connect the xlr 2 and 3, and connect the common return of the above caps to
the twisted wires shield back to pin 1, which is connected to the chassis
by its own means. This way, you can get rid of 5 times (!) the standard
perturbation up to 2.5 GHz, I measured it).
Or you can explore putting in serial with each line a resistor of same
value (it often exists on the signal output, use the same for the ground).
For best results, small RF caps should be put behing the resistances as
explained above. I didn't try nor measure this way.

(3b2) some ppl count upon the interconnect cable lumped inductance and
capacitance to equalize the impedances as seen by the load equipement. Easy
to visualize, less easy to mathematize: see the cable as a horizontal scale
where the vertical rungs are capacitances and the twin horizontal bars
between those rungs are inductances. The more steps in this scale, the
lesser any impedance unbalance in the two inputs be seen at the output.
Unhappily, this solution is not for me a good one, because it's too much
specific. It works or not or just a little depending on the perturbation at
this time, this period of the day, since the lumped inductances and
capacitances of the cable are so scarcely specified vs frequency (up to 2.5
GHz to work smoothely in an environment with BlueTooth, WiFi, cellphones
etc). It also depends on the souce and load common mode and differential
impedances vs frequency, and as always, well above the audio band up to
microwaves.
Given the oddities many "high-end" equipement can generate when fed with RF
pereturbation well within the specified standards -which should not be-
maybe it explains some odd differences between cables, between hours of
the day, between this equipement working fine with this interconnect cable
and the like
So, I much do do prefer solution (3b1)

(3b3) you can also balance RF impedance at the load side: from the
connector, use the common mode xformer/coil, then the 2 caps with their
common connected to the chassis at the connector. There, the cable's lumped
impedance is of much less importance.

(3c) We can also design things in order that the stray capacitance between
the board and the chassis is _really_ low (less than 1pF !). Or, as often
done in instrumentation, use no chassis at all but a plastic enclosure. It
is efficient, but does not protect against any other sneaky RF path for
example along the interconnect cable (once grounded on the floor, squeeze
it into your hand). I don't like too much this method for metallic
enclosure, since (3c1) you need very long stand-off to have really small
capacitance; you double the stand-off length and the capacitance is only
halfed and so the perturbation. Not efficient, and the suspension is much
less stiff and furthermore (3c2) you need large "isolated bulkhead RCA
jacks" (ref Dan) which means holes about 2 cm diameter, excellent for #GHz
spraying into the enclosure and "wetting" everything inside of it. See also
topic (4b).

(3d) connecting the chassis to the ground at RF frequencies. It seems
weird: the stray capacitance between earthed chassis and signals being the
source of common mode perturbation problem, why lower this capacitance
by paralleling it with new caps? Because without doing it, you are at risk
of new problems. See paragraph 4. And when you do it, you just get most
of the time less than twice the CM noise you should have without. Just
because the CM loop impedance is greater than the stray capacitance at the
frequencies of concern, having a much greater cap parallel to the stray
capacitance does change much the current (too lazy right now to draw a
schematic, the post is long enough...)

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
! !
! !
! Another concern: chassis/ground insulation induced positive feedback !
! !
! !
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
>This configuration also poses an additional problem due the fact that we
> now have two different grounds with two different potentials. In the past
> I have observed this problem when using a certain brand of rotary switch
> for a volume control. The rotary switch was not well isolated internally
> and had enough of a leakage current so that noise was developed when it
> was used. Moving to a different vendor with higher isolation devices
> corrected the problem.
>
(4) I don't think these problems have to do with different ground potentials
(we talked on this issue above) nor to the leakage current of your rotary
switches (well, as long as we use the same vocabulary, for me "leakage
current" is DC or very low frequency. Capacitive leak or stray capacitance
is noticeable for RF). I do think that the culprit is the capacitance
between your rotary switch contacts and its enclosure, metallic axle and
the like. For me, you got chassis/ground insulation induced positive
feedback, exactly what I explain in the reference below (the same as
my yesterday's post).
By changing the switch, it worked again, yes, why not, it should also have
not worked either. That's the problem: this positive feedback will give or
not bad behaviors counting upon too many parameters: RF behavior of your
active elements, paint used inside cabinet (dielectric constant can vary
maybe 10x depending on the chemical used), anything you can think of,
(maybe even your girlfriend's weight as I wrote in the reference :-)
You say "Moving to a different vendor with higher isolation devices
corrected the problem."
OK, but did the manufacturer _specified_ what you call isolation and
what I call stray capacitance?
If he didn't specify it, I wouldn't count upon it, because he can change,
say his axle diameter for example, and you get a completely different figure.
As it is not specified, you cannot do anything against it. He wouln't even
likely notice you of this change since it's not specified.
So you are at risk that if the problem is to arise, it will at a customers'
(Murphy's law ;-).
Same with the paint (you know its dielectric constant vs frequency?), same
with active elements (less likely as long as no "equivalent" parts are
used...) Same with your girlfriend's weight. ooops, sorry.
Anyway, too complicated to manage. Open door to endless problems.
So, I would apply the recipe of connecting signal ground and chassis in
several places through , say 10nF caps, one of them being a serial RC as
damper, as shown in the reference.
Maybe raw common noise would be without any further precaution 2 or 3
times greater, but if at the same time you apply other actions as described
above and described in your post, you get more than 60dB attenuation and you
can forget and forgive this slight increase in CM noise of a few dBs

Comments welcome, and, as Dan says : "I would greatly appreciate it if the
following discussion is limited to the issue of grounding only."

• Ground lift and positive feedback   (Open in New Window)

This post is made possible by the generous support of people like you and our sponsors:

The Cable Cooker

---

[ Top of Page ] [ Contact Us ] [ Support/Wish List ] [ Copyright © 2025, Audio Asylum, all rights reserved ]

---

[ General ] [ Speakers ] [ Tubes ] [ Vinyl ] [ Digital ] [ Hi-Rez ] [ Video Asylum ] [ Cables ] [ Tweaks/DIY ] [ Music ] [ Films ]

---