Sensor cable design

[Doug D.]

I'm having difficulty finding a suitable cable for a wired sensor application, and I was wondering if anyone could check my design approach. Basically, my control circuit powers a wired sensor located about 2 meters away. The sensor circuit uses a single-ended op-amp to amplify a 5 MHz transducer signal, and then sends it back to the control circuit where it is digitized.. So, I need 3 wires between the control circuit and the sensor: Vcc, ground, and signal.

Since the cables from multiple sensors will be running next to each other, I thought I should use twisted-triad wire to reduce cross-talk. External EMI shouldn't be a problem, so I don't think shielded cables are necessary. However, I haven't been able to find twisted-triad wire without a jacket or shield. What kind of wire should I use for this? If I used two twisted-pair wires for each sensor, what would I connect the extra wire to?

The sensor has a Vcc bypass capacitor, so I think most of the higher frequency content will be between signal and ground. Since I'm bringing back thesignal and ground wires together, would this be considered a balanced circuit?

Thanks,
Doug

 

[Ecnerwal]

What kind of wire should I use for this? If I used two twisted-pair
wires for each sensor, what would I connect the extra wire to?

Ground, or Vcc, as you prefer. Then again, two coax with one carrying
Vcc on the inner and the other carrying signal on the inner, or one coax
and another wire for Vcc might be preferable.

 

[Tauno Voipio]

I'm having difficulty finding a suitable cable for a wired sensor application, and I was wondering if anyone could check my design approach. Basically, my control circuit powers a wired sensor located about 2 meters away. The sensor circuit uses a single-ended op-amp to amplify a 5 MHz transducer signal, and then sends it back to the control circuit where it is digitized. So, I need 3 wires between the control circuit and the sensor: Vcc, ground, and signal.

Since the cables from multiple sensors will be running next to each other, I thought I should use twisted-triad wire to reduce cross-talk. External EMI shouldn't be a problem, so I don't think shielded cables are necessary. However, I haven't been able to find twisted-triad wire without a jacket or shield. What kind of wire should I use for this? If I used two twisted-pair wires for each sensor, what would I connect the extra wire to?

The sensor has a Vcc bypass capacitor, so I think most of the higher frequency content will be between signal and ground. Since I'm bringing back the signal and ground wires together, would this be considered a balanced circuit?

Thanks,
Doug

Your frequency is in the HF range, so you should seriously consider
impedance matching to the cable.

If you want to use twisted pair, the signal must be symmetrical to
achieve advantage of the cable. For single-ended signals, consider
coaxial cable instead.

Remember that a RF signal travels in the insulator between the conductors.

 

[[email protected]]

Ecnerwal, thanks for your reply.

Ground, or Vcc, as you prefer. Then again, two coax with one carrying
Vcc on the inner and the other carrying signal on the inner, or one coax
and another wire for Vcc might be preferable.

Would twisted triad be better than two twisted pairs? If I used two twisted pairs with ground on one wire of each pair, which ground wire would carry the return current? Or would the return current be split between the two ground wires?

Doug

 

[[email protected]]

Tauno, thanks for your reply.

Your frequency is in the HF range, so you should seriously consider
impedance matching to the cable.

I was going to use a 100 ohm series resistor at the output of the op-amp. Is this the proper way to match the cable impedance?

If you want to use twisted pair, the signal must be symmetrical to
achieve advantage of the cable. For single-ended signals, consider
coaxial cable instead.

I was hoping to use twisted pair since it is much cheaper than coaxial cable. If I bring back the signal and ground wires in the same twisted pair, would that be considered symmetrical? More current would be flowing out to the sensor over the Vcc wire than would be returning through either the signal or the ground wires.

Thanks,
Doug

 

[[email protected]]

John, thanks very much for your reply.

A balanced circuit would be one where a differential signal was being
propagated down two floating conductors, like 300 ohm TV twin-lead.

If one of the wires was ground and the other signal, (like in the 75
ohm coaxial cable that feeds most modern TVs) that's called an
unbalanced, or single-ended signal.

I'm still a bit confused about single-ended signals. In this application, the sensor is floating--it's not connected to earth ground. If the signal is referenced to the sensor's ground wire, and I return the ground wire along with the signal wire, wouldn't those two wires be balanced?

You might also be able to use twisted shielded pair (or twinax) by
running Vcc down one conductor, signal down the other, and the shield
as a common ground.

I've also been confused by how the return currents would flow. Would they flow in both twinaxes, with the currents equal and opposite in each twistedpair? How do the ground currents know which return wire to take?

 

[tm]

What's the signal level? What sort of s/n ratio do you need?

Twisting single-ended signals probably doesn't matter.

One less-obvious EMI problem is RF (like from TV and radio stations, cell
phones, motor controllers, things like that) getting rectified by opamp
front
ends... or back ends. That will be site dependant. Shielding would help
with
that.

If the sensor signal is truly AC, you could use a coax cable and send the
power down the coax along with the signal.

 

[Tim Williams]

Besides the in-circuit properties, endedness refers to the field given off
by the cable (or, equivalently, how the wires see the field around). If the
construction is symmetrical (a plus and a minus wire, same size, evenly
spaced, etc.), then the signals should be symmetrical as well. If not, an
average field radiates (or an external field will be picked up, same thing).

Return current follows whatever wires are close to the signal in question.
If other signals are nearby, they act as grounds, too. Imagine ribbon
cable: if you put a bunch of signals in a row with one ground at the end,
all those signals are going to screw with each other. If you alternate
ground, signal, ground, signal.., it's so much better (like 20dB). There's
an appnote somewhere out there with scope pictures illustrating the
difference.

Note that supply and ground are, for the purposes of signal quality, both
grounds. This is especially true with bypasses at *both* ends.

Tim

--
Deep Friar: a very philosophical monk.
Website: http://www.seventransistorlabs.com/

John, thanks very much for your reply.

A balanced circuit would be one where a differential signal was being
propagated down two floating conductors, like 300 ohm TV twin-lead.

If one of the wires was ground and the other signal, (like in the 75
ohm coaxial cable that feeds most modern TVs) that's called an
unbalanced, or single-ended signal.

I'm still a bit confused about single-ended signals. In this application,
the sensor is floating--it's not connected to earth ground. If the signal
is referenced to the sensor's ground wire, and I return the ground wire
along with the signal wire, wouldn't those two wires be balanced?

You might also be able to use twisted shielded pair (or twinax) by
running Vcc down one conductor, signal down the other, and the shield
as a common ground.

I've also been confused by how the return currents would flow. Would they
flow in both twinaxes, with the currents equal and opposite in each twisted
pair? How do the ground currents know which return wire to take?

 

[[email protected]]

John, thank you very much for explaining this to me. After 25 years, I finally understand the difference between balanced and unbalanced cables. "Balanced" means balanced _current_ flow on two wires going in both directions.. That's why coax is used for unbalanced cables--the shield stays at ground potential while the signal _voltage_ varies in the center conductor.

As John Larkin pointed out above, the cable is physically short compared tothe signal's electrical wavelength, so transmission-line reflections will have time to dissipate and won't matter at my relatively slow signal frequency. So, I don't have to consider impedance matching, other than for the op-amp output. This means I don't need expensive coax with its controlled impedance--simple foil-shielded, unbalanced wire should work to protect against cross-talk and external EMI.

Thanks again!
Doug

 

[[email protected]]

John, thank you very much!

 

[Phil Allison]

"Doug D."

I'm having difficulty finding a suitable cable for a wired sensor
application, and I was wondering if anyone could check my design approach.
Basically, my control circuit powers a wired sensor located about 2 meters
away. The sensor circuit uses a single-ended op-amp to amplify a 5 MHz
transducer signal, and then sends it back to the control circuit where it is
digitized. So, I need 3 wires between the control circuit and the sensor:
Vcc, ground, and signal.


** No - you need only two, seeing as one of them is DC.

The technique is known as "phantom powering".

Small ( ie 100uH ) inductors at each end of the +5V line will easily allow
a 5MHz signal to be superimposed on the same wire.

BTW: "tm" has already alluded to this idea and JF posted a schem.

Twisted pair should work as well as co-ax for your app - long as no nasty
RF exists in the vicinity.


.... Phil

 

[[email protected]]

I was hoping to use twisted pair since it is much cheaper than coaxial cable. If I bring back the signal and ground wires in the same twisted pair,would that be considered symmetrical? More current would be flowing out to the sensor over the Vcc wire than would be returning through either the signal or the ground wires.



Thanks,

Doug

How cheap did you want it? You can't beat Ethernet Cat 5 SFTP patch cables for cheap, in addition to maximal EMI performance both from external sources and between pairs, and reduction of radiation from the cable signals. This one is $6.99 from Amazon, already connectorized:

http://www.intellinet-network.com/en-US/products/3655-network-cable-cat5e-sftp

Your signal may be 5MHz ( is that total BW?) but the potential interferencemay not be.

 

[Phil Allison]

After 25 years, I finally understand the difference
between balanced and unbalanced cables.
"Balanced" means balanced _current_ flow on
two wires going in both directions.

** Fraid that is not correct.

Practically all cables have equal current flow in opposite directions -
even AC power cable.

Balanced refers to the conditions at each end of the cable.

That's why coax is used for unbalanced cables--
the shield stays at ground potential while the
signal _voltage_ varies in the center conductor.

** Co-axial cable has near ideal properties.

1. There is equal and opposite current flow in the inner and outer
conductors.

2. There is no external magnetic or electric field.

3. External electric and magnetic fields do not affect the cable.

4. Single ended drive and receive is used.

However, a twisted pair cable can do virtually the same - but only if it is
ES shielded.

If the send and receive signals are both balanced, then you can lose the ES
shield in most applications.


.... Phil

 

[Tauno Voipio]

Tauno, thanks for your reply.

I was going to use a 100 ohm series resistor at the output of the op-amp. Is this the proper way to match the cable impedance?

The impedance of a twisted pair cable is quite close,
CAT5 is around 120 ohm.

I was hoping to use twisted pair since it is much cheaper than coaxial cable. If I bring back the signal and ground wires in the same twisted pair, would that be considered symmetrical? More current would be flowing out to the sensor over the Vcc wire than would be returning through either the signal or the ground wires.

No. You need to have +signal and -signal to prevent asymmetric ground
currents. The DC polarities do not need be opposite, but AC components need.

 

[RobertMacy]

I'm having difficulty finding a suitable cable for a wired sensor
application, and I was wondering if anyone could check my design
approach. Basically, my control circuit powers a wired sensor located
about 2 meters away. The sensor circuit uses a single-ended op-amp to
amplify a 5 MHz transducer signal, and then sends it back to the control
circuit where it is digitized. So, I need 3 wires between the control
circuit and the sensor: Vcc, ground, and signal.

Since the cables from multiple sensors will be running next to each
other, I thought I should use twisted-triad wire to reduce cross-talk.
External EMI shouldn't be a problem, so I don't think shielded cables
are necessary. However, I haven't been able to find twisted-triad wire
without a jacket or shield. What kind of wire should I use for this?
If I used two twisted-pair wires for each sensor, what would I connect
the extra wire to?

The sensor has a Vcc bypass capacitor, so I think most of the higher
frequency content will be between signal and ground. Since I'm bringing
back the signal and ground wires together, would this be considered a
balanced circuit?

Thanks,
Doug

I assume you have both the signal gnd (common) and the power gnd (common)
connected at your controller? It is NEVER a good idea to use single ended
to transfer energy around, should have used balanced receiver at the
controller and balanced driver at the sensor. However, you can probably
get by with your architecture.

Two tiwsted pair should be the cheapest. You can use uncontrolled like
telco phone line, or controlled cat 5 type cabling, your choice.

one pair - Vcc and GND
2nd pair - SIGNAL and GND

at the source, your OpAmp, place a common mode choke in series with the
the signal/gnd wires to make them 'float' relatively.

Why? because you actually have an 'extra' connection to Earth that you
have NOT taken into account - the parasitic capacitance of your sensor.
Assuming you put some type of shielding all around it, which is referenced
to its ground, that in combination with the unbalanced currents in your
cabling can lead to havoc everywhere, with NO easy solutions. You
mentioned EMI coming IN. Don't forget about EMI going OUT!

But at least with the above, you might squeak by, and worse case make that
twisted pair inside shield should be ok for sure.

 

[Phil Hobbs]

How cheap did you want it? You can't beat Ethernet Cat 5 SFTP patch cables for cheap, in addition to maximal EMI performance both from external sources and between pairs, and reduction of radiation from the cable signals. This one is $6.99 from Amazon, already connectorized:

http://www.intellinet-network.com/en-US/products/3655-network-cable-cat5e-sftp

Your signal may be 5MHz ( is that total BW?) but the potential interference may not be.

You won't get good EMI rejection on twisted pair unless both the driver
and the receiver are balanced.

If you drive one and ground the other, the fields in the neighbourhood
of the cable are approximately half differential and half common mode.
That's because

V_CM = (V+ + V-)/2 and V_DIFF = (V+ - V-)/2.

If you drive V+ and ground V-,

V_CM = V_DIFF = V+ /2.

In other words, you've put half your signal swing in the differential
(transmission line) mode, and half in the common (i.e. antenna) mode.

Antennas work both ways, so if you set it up to radiate like that,
you're also setting it up to receive whatever's around.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics

160 North State Road #203
Briarcliff Manor NY 10510

hobbs at electrooptical dot net
http://electrooptical.net

 

[[email protected]]

You won't get good EMI rejection on twisted pair unless both the driver

and the receiver are balanced.



If you drive one and ground the other, the fields in the neighbourhood

of the cable are approximately half differential and half common mode.

That's because



V_CM = (V+ + V-)/2 and V_DIFF = (V+ - V-)/2.



If you drive V+ and ground V-,



V_CM = V_DIFF = V+ /2.



In other words, you've put half your signal swing in the differential

(transmission line) mode, and half in the common (i.e. antenna) mode.



Antennas work both ways, so if you set it up to radiate like that,

you're also setting it up to receive whatever's around.



Cheers



Phil Hobbs



--

Dr Philip C D Hobbs

Principal Consultant

ElectroOptical Innovations LLC

Optics, Electro-optics, Photonics, Analog Electronics



160 North State Road #203

Briarcliff Manor NY 10510



hobbs at electrooptical dot net

http://electrooptical.net

Thanks for the refresher, but I already know quite a bit about driving and receiving controlled impedance transmission lines, and differential and common mode impedance. The OP needs to tell us the precision of the his A/D and his ENOB spec at 5MHz so we can suggest drvr/rcvr pairs of adequate precision.

 

[whit3rd]

On Sun, 06 Oct 2013 22:24:36 +0300, Tauno Voipio

The cable is 2 meters long. The signal wavelength is 60 meters. There will be no
significant transmission-line effects.

Not quite true: it just means that the 'transmission-line effects' are nearly
indistinguishable from stray capacitance (if Z is high) or stray inductance (if Z is low).

So, the wiring can be an excess load, and cause frequency-dependent distortions.
I had an audio cable that snaked through two or three rooms, and it performed
very poorly at conducting line-level signals until I lowered the source impedance
(to about 100 ohms); matching impedance gets rid of capacitive rolloff.

 

[Ecnerwal]

This means I don't need expensive coax with its
controlled impedance--simple foil-shielded, unbalanced wire should work to
protect against cross-talk and external EMI.

Sure, there's "expensive coax" ­ teflon-insulated silver-plated fancy
stuff; There's also a good deal of cheap coax in the world ­ RG-59, for
instance. Due to its use in cable TV, if you can't find a chunk for
free, it's all of 6.75 cents/foot in bulk (without even shopping hard ­
7.4 cents for dual-shield) If you want shielded twisted pair, Cat 6 is
over double the price (though it has more pairs than you need, so
possibly you can find some STP that's actually cheaper than RG-59 - then
again, there's at least a bulk market for Cat6, so perhaps not, too.)

You can even get RG-59 bonded with a pair of 18Ga power wires for 16.9
cents a foot (still not shopping all that hard, so you might be able to
do better.)

RG-6 is also a comparatively inexpensive coax.

 

[Phil Allison]

"whit3rd"

John Larkin


Not quite true: it just means that the 'transmission-line effects' are
nearly
indistinguishable from stray capacitance (if Z is high) or stray
inductance (if Z is low).

So, the wiring can be an excess load, and cause frequency-dependent
distortions.

** Correct - the cable is an electrically short transmission line, which
if not terminated becomes be a pure C load.

I had an audio cable that snaked through two or three rooms, and it
performed
very poorly at conducting line-level signals until I lowered the source
impedance
(to about 100 ohms); matching impedance gets rid of capacitive rolloff.

** Simply reducing source Z to 100 ohms did the trick for your case - but
it did not eliminate cable C.


.... Phil