Are ferrite cable shields on coax a good idea?
Nemo
RG174 cable, which is pretty well shielded (85% coverage or more if I
recall the cable spec correctly). It struck me I might be able to reduce
noise further by adding a ferrite tube to the cable, which I've often
done with unshielded wires. However, when asked to explain this to a
colleague, I found I didn't know as much as I thought I did, and
couldn't convey how ferrites act as common mode chokes to him.
On thinking about how coax works, with the inner and outer current
cancelling each others' fields, I wondered if adding a ferrite might
actually *disrupt* the flow of current round the coax and make
interference worse rather than better - i.e. by increasing impedance to
high frequency currents so that the coax shield was poorer at high
frequency. And reading further about ferrite cable shields, they seem to
be discussed mainly in terms of reducing emissions rather than reducing
incoming interference. Another thing that makes me wonder if mixing
ferrites and coax is bad is: articles discussing coax as a solution to
interference make no reference to ferrite cores. I am now rather
confused on the issue. If anyone has any experience in combining
ferrites and coax I'd appreciate advice on this matter.
--
Nemo
Phil Hobbs
The magnetization in the ferrite depends only on the total current
threading the loop. The normal transmission mode in coax has current
going one way on the centre conductor and an equal and opposite current
on the shield. Capacitive pickup leads to a current flowing in the
shield only, and inductive pickup occurs more or less equally on both
centre conductor and shield. All currents in the coax can be
decomposed into even and odd modes:
i_even = 0.5*(i_centre+i_shield)
i_odd = 0.5*(i_centre-i_shield).
You can obviously make any centre conductor and shield currents you like
from some combination of i_even and i_odd.
If you put a huge common mode choke in the line, it will leave the
differential (odd mode) part of the signal unaffected (no net current,
so no magnetization, hence no inductance), but it will impede the
common-mode (even mode) part.
For capacitive pickup immunity, you have to rely on the shield
preventing any of the ac field from getting inside the coax. This is
true whether or not you have a choke--if your pickup is producing a
differential signal, it'll go right on through the choke, and otherwise
you'll never see it. RG-174 is pretty crappy stuff--double shielded,
braid-and-foil, or flexible semirigid coax would be much much better.
That low shield coverage makes it talk to the whole world, besides the
fact that the shielding relies on good contact being made by tinned
wires just lying on top of each other--not very confidence-inspiring.
For inductive and radiative pickup (i.e. like an antenna), the CM choke
is a big win, because it prevents RF currents from getting into your
board in the first place. For ground loops, you need a whole lot of
iron, so the usual right answer is a 1-ohm resistor in the ground lead
from the connector, with a bypass cap and a differential amp.
Perhaps the source of confusion is the notion that the choke will
open-circuit the shield at AC, leaving the coax vulnerable to pickup
along its whole length. That isn't so, because for even mode signals it
open-circuits the centre conductor as well, and for odd mode signals it
doesn't do anything.
Cheers,
Phil Hobbs
Baron
> Perhaps the source of confusion is the notion that the choke will
> open-circuit the shield at AC, leaving the coax vulnerable to pickup
> along its whole length. That isn't so, because for even mode signals
> it open-circuits the centre conductor as well, and for odd mode
> signals it doesn't do anything.
>
> Cheers,
>
> Phil Hobbs
Hi Phil,
I was with you up to the last paragraph !
Since the currents on the inside of a co-axial cable are separate from
any current on the outside. How can a ferrite tube affect current on
the inner conductor ?
--
Best Regards:
Baron.
Phil Hobbs
It's the total current that matters, not how it's arranged in space.
Magnetism is weird that way.
Cheers,
Phil Hobbs
Joerg
A ferrite dampens the current on anything that goes through it in the
same fashion. So if you run the whole coax through there it will not
disturb the differential signal (which you want to preserve) but will
muffle asymmetrical currents (which you seem to want to get rid of). You
said up to 150khz but didn't say how low it goes. Looks like a 77 or J
material toroid would be great. The thicker the better, but also the
more expensive. You can slide several of those over the cable to improve
things some more. If you must muffle stuff in the kHz range as well
you'd also need an old transformer that you can scrape out and run the
coax through.
Oh, and make sure to secure the ferrite. Ferrite is brittle, if it
clangs against something hard it can shatter.
--
Regards, Joerg
http://www.analogconsultants.com/
"gmail" domain blocked because of excessive spam.
Use another domain or send PM.
Jeroen Belleman
> I've got some very low level signals at up to 150kHz going down some
> RG174 cable, which is pretty well shielded (85% coverage or more if I
> recall the cable spec correctly). It struck me I might be able to reduce
> noise further by adding a ferrite tube to the cable, which I've often
> done with unshielded wires. However, when asked to explain this to a
> colleague, I found I didn't know as much as I thought I did, and
> couldn't convey how ferrites act as common mode chokes to him.
I could go on for hours on this subject. Here's my take on the
matter:
The desired signal over coax is purely odd-mode, i.e., the current
flowing over the inner conductor is accompanied by an equal and
opposite current flowing along the inside of the screen.
Interfering signals ('noise') are even mode. If the cable screen
had been perfect, these would not affect the desired signal. The
screen conductor however has a finite, non-zero, resistance, and
so, at low frequencies, an even mode signal will leak into the
cable.
There are two strategies to reduce this leakage. One seeks to
reduce the ingress of interference by lowering the screen
resistance: Use double-screened cable or double up the screen
by running a copper braid or thick cable in parallel with it.
The other seeks to lower even mode currents by reducing the
interfering source (if it can be identified) or by increasing
the impedance of the loop formed by the cable screen and the
-unknown- return path. (Not by inserting resistance into the
screen, of course! That would be the worst you could do!)
Slipping ferrites over the cable falls in that last category.
The trouble is that the loop referred to in the previous
paragraph usually has an inductance of the order of 10uH or
(much) more. Slipping a ferrite over your cable is not going
to increase that substantially, and therefore the effect is
likely to be minimal.
In an application of these ideas in a particle accelerator
here, I managed to gain an additional 20dB of interference
rejection on some position pick-ups in the machine. This
enabled us to observe beams with a factor of ten lower
intensity, which is just what was needed to measure the
probe beams for the LHC.
Jeroen Belleman
Phil Hobbs
Ferrite can help with RF, but not with ground loops. Also, whether
pickup is purely even mode or not depends on how the two conductors are
loaded.
Cheers,
Phil Hobbs
MooseFET
How about RF ground loops?
>
> Cheers,
>
> Phil Hobbs
Jeroen Belleman
> Jeroen Belleman wrote:
>> [...]
>> Interfering signals ('noise') are even mode. If the cable screen
>> had been perfect, these would not affect the desired signal. The
>> screen conductor however has a finite, non-zero, resistance, and
>> so, at low frequencies, an even mode signal will leak into the
>>
>> Jeroen Belleman
>>
>>
>
> Ferrite can help with RF, but not with ground loops. Also, whether
> pickup is purely even mode or not depends on how the two conductors are
> loaded.
>
> Cheers,
>
> Phil Hobbs
I have this gadget in the lab to measure cable leakage. Basically
it's a brass pipe with a piece of cable mounted on-axis inside it,
so that the space between the cable screen and the pipe wall also
is an approximately 50 Ohm transmission line. So we have a coax
inside a coax. All four ends are looking into 50 Ohms.
The idea is to inject a signal between the cable screen and the
pipe wall, and then look how much of it gets picked up on the inner
line.
For very low frequencies (<100kHz, I may be off a bit; This was 7
years ago), this shows that the leakage path is simply a resistive
divider, with the cable screen resistance and the termination
impedance as the main players. So for a 1m piece of RG223, with
about 7 mOhm of screen resistance, I get about -77dB of ingress.
Above 100kHz, the ingress drops by 20dB/decade, until is gets into
the analyzer noise floor at about -110dB.
At the high frequency end, at some point, leakage would go up again
because of the incomplete coverage of the screen. I cannot see that
however. My setup doesn't work very well beyond 200MHz, because it's
not easy to cleanly couple a signal into the outer line.
Maybe I should dig this out again. Many people might be interested
in some cable leakage data. It's not something commonly detailed
in cable specs. (It's more fun than writing reports, which is what
I should be doing right now.)
Jeroen Belleman
Jeroen Belleman
> On Jan 23, 3:17 am, Phil Hobbs
>
>> pickup is purely even mode or not depends on how the two conductors are
>> loaded.
>
> How about RF ground loops?
Aha! The infamous ground loop. Whatever you choose to call it,
the thing that matters is current flowing over the cable screen.
Normally the screen is grounded at both ends. There may be
something that forces a potential difference, not necessarily DC,
between the two ends. That will cause a current to flow in the
screen and since the screen isn't perfect, some of it will find
its way into your signal.
The nature of the remedy will depend on the coupling mechanism
and the frequency of the agressor signal. The agressor source
and the way it couples are often difficult to identify, if only
because the 'components' of the equivalent circuit are not on
the schematic diagram and of unknown magnitude. Once you get
a model of the way the interference gets into your box, the
fix is usually obvious.
Jeroen Belleman
Phil Hobbs
If you're working near a transmitter, then yep, you have to worry about
those too. Most of the time in my world, ground loops are 60 Hz
affairs--a 1-m loop of cable has about 3 uH of inductance, so you wind
up with millivolts of signal and circulating currents of hundreds of
milliamps. The reactance is of the same order as contact resistances,
which is why jiggling the cable often makes a temporary difference.
I'm most commonly trying to keep the spurs down at -170 dBm or someplace
like that, right next to a synthesizer or acousto-optic cell driver
putting out a watt or so. I have a disgracefully leaky Agilent sweeper
(brand new) that produces -58 dBm in my measurement system at 6 GHz,
with the sweeper completely disconnected. Pathetic.
Cheers,
Phil Hobbs
Phil Hobbs
> Phil Hobbs wrote:
>> Jeroen Belleman wrote:
> >> [...]
>>> Interfering signals ('noise') are even mode. If the cable screen
>>> had been perfect, these would not affect the desired signal. The
>>> screen conductor however has a finite, non-zero, resistance, and
>>> so, at low frequencies, an even mode signal will leak into the
>>> cable. [...]
>>>
>>> Jeroen Belleman
>>>
>>>
>>
>> Ferrite can help with RF, but not with ground loops. Also, whether
>> pickup is purely even mode or not depends on how the two conductors
>> are loaded.
>>
>> Cheers,
>>
>> Phil Hobbs
>
> I have this gadget in the lab to measure cable leakage. Basically
> it's a brass pipe with a piece of cable mounted on-axis inside it,
> so that the space between the cable screen and the pipe wall also
> is an approximately 50 Ohm transmission line. So we have a coax
> inside a coax. All four ends are looking into 50 Ohms.
> The idea is to inject a signal between the cable screen and the
> pipe wall, and then look how much of it gets picked up on the inner
> line.
Cute idea.
>
> For very low frequencies (<100kHz, I may be off a bit; This was 7
> years ago), this shows that the leakage path is simply a resistive
> divider, with the cable screen resistance and the termination
> impedance as the main players. So for a 1m piece of RG223, with
> about 7 mOhm of screen resistance, I get about -77dB of ingress.
> Above 100kHz, the ingress drops by 20dB/decade, until is gets into
> the analyzer noise floor at about -110dB.
RG223 is double-shielded, though, isn't it? At high frequencies the
holes in the shield dominate, particularly if there are any kinks in the
cable. Also you were using brand new coax, I expect--once the braid has
a chance to oxidize a bit, life gets much worse (again, at high
frequency--the strands ought to be continuous throughout the braid).
>
> At the high frequency end, at some point, leakage would go up again
> because of the incomplete coverage of the screen. I cannot see that
> however. My setup doesn't work very well beyond 200MHz, because it's
> not easy to cleanly couple a signal into the outer line.
>
> Maybe I should dig this out again. Many people might be interested
> in some cable leakage data. It's not something commonly detailed
> in cable specs. (It's more fun than writing reports, which is what
> I should be doing right now.)
I'd certainly be interested.
Cheers,
Phil Hobbs
Nemo
To answer your concerns, I don't think ground loops aren't a problem,
but I'll look into that. I usually secure ferrites to cables with some
heat shrink. I suspect the lowest frequency of interest is about 5kHz -
I'll need to check the system specs there, but certainly not DC.
--
Nemo
Joerg
For 5kHz you'd have to loop the cable several times through the ferrite.
Or just use an old audio transformer core or small line transformer core.
MooseFET
> MooseFET wrote:
> > On Jan 23, 3:17 am, Phil Hobbs
>
> >> Ferrite can help with RF, but not with ground loops. Also, whether
> >> pickup is purely even mode or not depends on how the two conductors are
> >> loaded.
>
> > How about RF ground loops?
>
> Aha! The infamous ground loop. Whatever you choose to call it,
> the thing that matters is current flowing over the cable screen.
> Normally the screen is grounded at both ends.
Unless you loop both of the cable's ends back to exactly the same
place, you can't really say that both ends are grounded without
qualifications. Consider a cable that is run between two points that
are 1/2 a wavelength apart at some frequency. The shield in the cable
is never perfect and there are always external sources of radio waves
in the environment.
If the COAX in question is on the order of 10,000 feet and is running
from down in the ocean and onto a ship, "RF" starts to mean some
fairly low frequencies.
Frithiof Jensen
"Nemo" <Ne...@nospam.nospam.nospam.com> skrev i meddelelsen
news:ZjzS7JMr...@furfur.demon.co.uk...
For frequencies that low you may be better off with an "iron" core or maybe
a iron powder toroid.
MooseFET
<frithiof.jen...@diespammerdie.jensen.tdcadsl.dk> wrote:
> "Nemo" <N...@nospam.nospam.nospam.com> skrev i meddelelsennews:ZjzS7JMr...@furfur.demon.co.uk...
>
> > Thank you everyone, loads and loads of useful info there.
>
> > To answer your concerns, I don't think ground loops aren't a problem, but
> > I'll look into that. I usually secure ferrites to cables with some heat
> > shrink. I suspect the lowest frequency of interest is about 5kHz - I'll
> > need to check the system specs there, but certainly not DC.
> > --
> > Nemo
>
> For frequencies that low you may be better off with an "iron" core or maybe
> a iron powder toroid.
Remember that the core is intended to remove things that are
unwanted. The frequencies of the unwanted currents is what really
matters.
Jeroen Belleman
> On Jan 23, 7:45 am, Jeroen Belleman <jer...@nospam.please> wrote:
>> Normally the screen is grounded at both ends.
>
> Unless you loop both of the cable's ends back to exactly the same
> place, you can't really say that both ends are grounded without
Agreed. That was my point. 'Ground' isn't universal. Between two
different points called 'ground' there is always some impedance
and depending on conditions, there may be a sizable voltage.
Jeroen Belleman
Ross Herbert
:On Jan 25, 2:47 am, "Frithiof Jensen"
Yes, and the ferrite on the coax is intended to prevent the shield of the cable
radiating EMI/RFI from the equipment it is connected to where it can be picked
up by other sensitive equipment. The ferrite is not there to prevent the
equipment it is connected to from picking up stray noise.
Eeyore
Only if you need them.
Graham
Eeyore
Joerg wrote:
> Oh, and make sure to secure the ferrite. Ferrite is brittle, if it
> clangs against something hard it can shatter.
Read 'almost certainly will shatter'.
How about iron powder cores ? Not enough permeability I suppose.
Graham
MooseFET
I have used them to prevent the equipment they are on from picking up
noise. If you have a large signal source nearby and a coax that is
some significant part of a wavelength long, you can end up with large
RF currents flowing in the shield and coupling into the core of the
coax.
Ross Herbert
I dare say that a ferrite core around a coax cable might achieve some degree of
"bulk" suppression of noise pick-up but this would not be optimal. The preferred
method of suppressing pickup in sensitive equipment is to do so on each of the
individual input lines at the input. This can be achieved using filtered input
connectors (if the extra expense warrants it) or simply passing the input lines
through individual ferrite beads as close to the input point as possible. Good
quality filtered connectors may include feedthrough capacitors and an inductor
in pi configuration on each pin.
Ross Herbert
<OneBi...@InfiniteSeries.Org> wrote:
:On Mon, 26 Jan 2009 18:49:35 -0800 (PST), MooseFET <kens...@rahul.net>
:wrote:
:
:
:
: When you get up into GHz ranges, they are referred to as 'spurs'.
:
: One reason why there are not very many switchers used in these bands.
:Linear supplies make a friggin' rack module pretty heavy.
:
: We need a switcher that has a super-smooth output that has no HF
:switcher noise included with it. So what? Super caps, computer grade
:cap banks?
:
: How do we reduce switcher noise so that they do not pose a problem with
:GHz range RF bands?
In my experience the average switchers pump out wideband noise consisting of
many harmonics and they only vary in the degree and level of noise produced. As
such, it is very difficult to suppress pickup on low level input lines in
sensitive equipment. Filtering individual input lines will usually help but the
best solution is to stop it from being produced in the first place, but that
requires special attention in the design of the switcher. It is for this reason
that military and medical grade switchers cost more than the run-of-the-mill
switchers.
MooseFET
> On Mon, 26 Jan 2009 18:49:35 -0800 (PST), MooseFET <kensm...@rahul.net> wrote:
>
> :On Jan 26, 4:23 pm, Ross Herbert <rherb...@bigpond.net.au> wrote:
> :> On Sun, 25 Jan 2009 12:18:33 -0800 (PST), MooseFET <kensm...@rahul.net>
> wrote:
> :>
It got me enough to be worth the effort. Stopping the RF current from
flowing in the shield and eating the RF power in the process worked
well. Non-lossy parts tend to reflect the RF power and send it
looking for another way to cause trouble.
> but this would not be optimal.
Not creating the RF in teh first place would be the right answer.
> The preferred
> method of suppressing pickup in sensitive equipment is to do so on each of the
> individual input lines at the input.
Yes and also to keep it off the power lines etc.
>This can be achieved using filtered input
> connectors
Filtered connectors aren't enough, often.
> (if the extra expense warrants it) or simply passing the input lines
> through individual ferrite beads as close to the input point as possible.
The ferrite beads method usually works out to cost more in
production. It takes manual labor.
Joerg
I rarely use iron powder much these days. It's more expensive than
ferrite and ferrite usually does the job.