Forums

RJ-45 for instrument interface

Started by Steve Wilson March 12, 2018
Jeroen Belleman <jeroen@nospam.please> wrote:

> Steve Wilson wrote: >> Clifford Heath <no.spam@please.net> wrote:
>>> If you want isolation, you need co-ax, not twisted pair.
>>> We're designing an RF prototyping platform that will use thin co-ax >>> jumper leads, with U/FL (IPEX) or MMCX connectors.
>>> Clifford Heath.
>> Thanks. I need bidirectional control, and would like to minimize RFI to >> external circuits. Ordinary unshielded CAT5 seems to have little >> radiation, and shielded and double shielded should be even better.
>> Coax is single-ended, so radiation from sharp transitions can be a >> problem.
> What's that supposed to mean?
Fast transitions are wideband. Coax shielding degrades at high frequencies.
>> Twisted pair is balanced, which reduces transition noise.
> Coax is balanced too, with the current on the central conductor > being accompanied by an equal and opposite current along the > shield. Equal to 0.03% or so, better than twisted pair.
Cite? I have a few links on coax shielding effectiveness. They range from 57dB to 115dB for solid coax. I was unable to find anything on shielded twisted pair effectiveness and would appreciate any links you can give. In any case, I'm worried about RJ-45 cable noise affecting nearby sensitive circuits. So far, I haven't seen any problems from the LAN even when unshielded, but local FM stations can be a problem.
>> The coax shield can lead to ground loop problems. RJ-45 is inherently >> isolated by the coupling transformers. In the case of ground loop >> problems, there is no need to ground the shields at both ends since >> they are not involved in data transmission. One end can connect to >> ground, and the other end to the case through a 100k resistor in >> parallel with a 0.1uF or 0.01uF bypass cap to reduce risk to ESD. > [...]
> Nothing prevents the use of coupling transformers at the end(s) of > a length of coax.
I need to tranfer power. That is done in RJ-45 using poe. Coax would require a separate cable and connector. I also need bidirectional data transfer. That means another coax. All these cables and connectors would take time to connect and disconnect. Two identical coax cables could be mistakenly swapped. RJ-45 puts it all in one connector.
> Jeroen Belleman
Jeroen Belleman <jeroen@nospam.please> wrote:

> Nothing prevents the use of coupling transformers at the end(s) of > a length of coax.
No, that doesn't work. It allows return currents to flow on the outside of the shield. The makes the cable radiate RFI. To prevent it, the coax must be grounded to the case. This can provoke ground loops.
> Jeroen Belleman
Steve Wilson wrote:
> Jeroen Belleman <jeroen@nospam.please> wrote: > >> Nothing prevents the use of coupling transformers at the end(s) of >> a length of coax. > > No, that doesn't work. It allows return currents to flow on the outside of > the shield. The makes the cable radiate RFI. To prevent it, the coax must be > grounded to the case. This can provoke ground loops. > >> Jeroen Belleman > >
Oh dear. To have any current on the outside of the shield, you need a closed loop. With a coupling transformer at the end, you've just opened that loop, ergo, no current. You're supposed to have a termination resistance at the other end of the transformer, which will absorb all incident energy. I'll admit that some current can still flow because of the inter- winding capacitance of the transformer. It may even resonate with the common-mode inductance of the cable, which would be unlikely, but bad. A few hundred Ohms from screen to GND across the transformer possible with a series cap to prevent DC and LF current, will damp that efficiently. Jeroen Belleman
Steve Wilson wrote:
> Jeroen Belleman <jeroen@nospam.please> wrote: > >> Steve Wilson wrote: >>> Clifford Heath <no.spam@please.net> wrote: > >>>> If you want isolation, you need co-ax, not twisted pair. > >>>> We're designing an RF prototyping platform that will use thin co-ax >>>> jumper leads, with U/FL (IPEX) or MMCX connectors. > >>>> Clifford Heath. > >>> Thanks. I need bidirectional control, and would like to minimize RFI to >>> external circuits. Ordinary unshielded CAT5 seems to have little >>> radiation, and shielded and double shielded should be even better. > >>> Coax is single-ended, so radiation from sharp transitions can be a >>> problem. > >> What's that supposed to mean? > > Fast transitions are wideband. Coax shielding degrades at high frequencies. > >>> Twisted pair is balanced, which reduces transition noise. > >> Coax is balanced too, with the current on the central conductor >> being accompanied by an equal and opposite current along the >> shield. Equal to 0.03% or so, better than twisted pair. > > Cite? I have a few links on coax shielding effectiveness. They range from > 57dB to 115dB for solid coax. I was unable to find anything on shielded > twisted pair effectiveness and would appreciate any links you can give. > > In any case, I'm worried about RJ-45 cable noise affecting nearby sensitive > circuits. So far, I haven't seen any problems from the LAN even when > unshielded, but local FM stations can be a problem. > >>> The coax shield can lead to ground loop problems. RJ-45 is inherently >>> isolated by the coupling transformers. In the case of ground loop >>> problems, there is no need to ground the shields at both ends since >>> they are not involved in data transmission. One end can connect to >>> ground, and the other end to the case through a 100k resistor in >>> parallel with a 0.1uF or 0.01uF bypass cap to reduce risk to ESD. >> [...] > >> Nothing prevents the use of coupling transformers at the end(s) of >> a length of coax. > > I need to tranfer power. That is done in RJ-45 using poe. Coax would > require a separate cable and connector. > > I also need bidirectional data transfer. That means another coax. All these > cables and connectors would take time to connect and disconnect. Two > identical coax cables could be mistakenly swapped. RJ-45 puts it all in one > connector. > >> Jeroen Belleman >
You're never going to get the same shielding effectiveness from a twisted pair than you can get from coax. Twisted pairs radiate, because the fields of the two wires do not cancel even in idealized theory. At least in theory, coax has no external field at all. I've made some measurements on actual coax shielding effectiveness, see <http://cern.ch/jeroen/coaxleakage/leakage.shtml>. I did not do so for twisted pairs, alas. It's not hard to send bi-directional data and DC power all over one single coax. Jeroen Belleman
Jeroen Belleman <jeroen@nospam.please> wrote:

> Steve Wilson wrote: >> Jeroen Belleman <jeroen@nospam.please> wrote: >> >>> Nothing prevents the use of coupling transformers at the end(s) of a >>> length of coax.
>> No, that doesn't work. It allows return currents to flow on the outside >> of the shield. The makes the cable radiate RFI. To prevent it, the coax >> must be grounded to the case. This can provoke ground loops.
>>> Jeroen Belleman
> Oh dear.
> To have any current on the outside of the shield, you need a closed > loop. With a coupling transformer at the end, you've just opened > that loop, ergo, no current. You're supposed to have a termination > resistance at the other end of the transformer, which will absorb > all incident energy.
> I'll admit that some current can still flow because of the inter- > winding capacitance of the transformer. It may even resonate with > the common-mode inductance of the cable, which would be unlikely, > but bad. A few hundred Ohms from screen to GND across the transformer > possible with a series cap to prevent DC and LF current, will > damp that efficiently.
> Jeroen Belleman
You do have a loop. The transfomer must match the 50 ohm cable. The center conductor has voltage applied. The return path is through the 50 ohms to the coax shield. Current can then flow on the inside and the outside of the coax. Current that flows on the outside can radiate RFI. The coax shield must be grounded with a circumferential ground, such as a BNC connector. This keeps the return current on the inside of the shield. For example, see Analog Devices, "EMI, RFI, and Shielding Concepts" http://www.analog.com/media/en/training-seminars/tutorials/MT-095.pdf See Fig. 8, page 11 See page 12: "For high-frequency interference (>1 MHz), the preferred method is grounding the shield at both ends, using 360&#2013266096; circumferential bonds between the shield and the connector, and maintaining metal-to-metal continuity between the connectors and the enclosure."
Jeroen Belleman <jeroen@nospam.please> wrote:

> You're never going to get the same shielding effectiveness from > a twisted pair than you can get from coax. Twisted pairs radiate, > because the fields of the two wires do not cancel even in idealized > theory. At least in theory, coax has no external field at all. I've > made some measurements on actual coax shielding effectiveness, see > <http://cern.ch/jeroen/coaxleakage/leakage.shtml>. I did not do > so for twisted pairs, alas.
There are about as many ways to measure shielding effectiveness as there are people making the measurements! Nice curves, but notice the results at 100MHz. This is the frequency where the transition risetimes start, and they can extend through the GHz region. You show UT141 with a solid copper shield has the same effectiveness as RG-58 at 100MHz. I think a solid copper shield, or hardline, is much better than that. Perhaps down to 121dB. If you take a HP network analyzer of spectrum analyzer, you find most of the internal coax connections are hardline. These instruments can go down to -130dB and below. If they used plain coax, or didn't ground both ends, they could never do as well. To measure the sheilding effectiveness, you must ground both ends with a circumferential ground, as shown in Fig. 8 of http://www.analog.com/media/en/training-seminars/tutorials/MT-095.pdf
Steve Wilson <no@spam.com> wrote:

> There are about as many ways to measure shielding effectiveness as there > are people making the measurements!
Here's how I measured shielding effectiveness. 1. get two female 50 ohm BNC terminators, such as https://www.pasternack.com/0.5-watts-bnc-female-rf-load-up-to-1-ghz-pe6027- 50-p.aspx 2. get two identical 50 ohm coax cables of a suitable length. 3. connect one coax to the signal generator and terminate in 50 ohms. 4. connect the other coax to a spectrum analyzer. Terminate the free end with a 50 ohm terminator. 5. run the two coax cables together side by side. I tried various ways to change the coupling. Twisting the cables together with a slow twist seemed to produce the best and most repeatable results. 6. The spectrum analyzer usually showed the crosstalk about 80 db down from the signal generator output using ordinary cheap coax. This method simulates the problem of running coax cables side-by-side, such as a signal generator driving an amplifier, with the output going through another coax to a spectrum analyzer or network analyzer. The coax should be high quality double shielded or hardline.
On Wednesday, March 14, 2018 at 3:36:03 AM UTC-7, Steve Wilson wrote:
> Jeroen Belleman <jeroen@nospam.please> wrote: > In any case, I'm worried about RJ-45 cable noise affecting nearby sensitive > circuits. So far, I haven't seen any problems from the LAN even when > unshielded, but local FM stations can be a problem.
> > Nothing prevents the use of coupling transformers at the end(s) of > > a length of coax. > > I need to tranfer power. That is done in RJ-45 using poe. Coax would > require a separate cable and connector. > > I also need bidirectional data transfer. That means another coax. All these > cables and connectors would take time to connect and disconnect. Two > identical coax cables could be mistakenly swapped. RJ-45 puts it all in one > connector.
There's plenty of connectors out there: the usual S-video connector is dual coax with an optional overall shield. You can do power POE-style over the two braids, and high bandwidth signals on the center conductors, using the usual Ethernet-like magnetics. It ain't high performance coax, but it's attractively off-the-shelf at low cost.
Jeff Liebermann wrote:
> On Tue, 13 Mar 2018 02:45:04 GMT, Steve Wilson <no@spam.com> wrote: >
<snip>
> > Besides users that cram a USB connector into the ethernet jack, > there's also the problem of the gold wire connections being mangled > inside the RJ45 jack. Three ways that I know of to do that: > 1. Cram something into the RJ45 that doesn't belong and wiggle. > 2. Use an RJ11 or RJ14 plug in an RJ45 jack. (More on this below) > 3. Cram an RJ45 plug into the jack which was NOT crimped. > The result of all of these is to have the gold wires in the jack move > over a notch, thus shorting two adjacent pins together. If the > adjacent pin carries power, it might blow something up. Internally, > the jack has a plastic "comb" that acts as a separator and spreader > for the gold wires. It doesn't take much to lift a gold wire from the > comb, move it to the adjacent slot, and produce a short. > > The mangled RJ45 has become common enough that designers are defending > themselves by making the connector an easily replaceable part. Most > common is using a dual RJ45 "barrel" connector. Something like this: > <https://www.amazon.com/Waterproof-CAT5e-RJ45-Feed-Thru-Coupler/dp/B00CULLAWE> > If the connector gets mangled, just unplug, remove and replace. >
There's also something to be said for M12 on the outside of boxes. They're harder to kill and make positive contact. You can have a throughhole pigtail like you'd have using the above object.
> Anyway, good luck. > >
-- Les Cargill
On Wed, 14 Mar 2018 18:20:45 -0500, Les Cargill
<lcargill99@comcast.com> wrote:

>There's also something to be said for M12 on the outside of boxes.
Yep: <http://www.iebmedia.com/index.php?id=5873&parentid=63&themeid=255&showdetail=true>
>They're harder to kill and make positive contact. You can have a >throughhole pigtail like you'd have using the above object.
They're also substanitally more expensive than RJ45. -- Jeff Liebermann jeffl@cruzio.com 150 Felker St #D http://www.LearnByDestroying.com Santa Cruz CA 95060 http://802.11junk.com Skype: JeffLiebermann AE6KS 831-336-2558