RS-485 and switched AC in same connector - interference?

[RyreInc]

Hello,

For a product we make we are looking into a single connector that provides RS-485 communications, 24VDC control power, and 120VAC heater power. The connector would be a Harting or Epic style connector (if you search "epic connector" the top images are representative). I am concerned with EMI between the RS-485 and 120VAC lines. The subject of EMI is largely out of my depth, but I have a basic understanding.

The 24VDC is fairly low and steady current of at most 1A, so I am not concerned with EMI on these lines. But the 120VAC line will be switched - right now it's half-wave zero-cross switched (low harmonics), but in the future it could be phase angle (high harmonics). Up to 15A can flow through this line, so EMI is a concern here. As for the RS-485, the current baud rate is 460800 but I would like to provision for up to 2M.

Regarding the connector, we would prefer to use a standard mono block type due to cost and ease of assembly. (There is also a modular design that can include a fully shielded portion for the communications and here I would not be concerned with EMI). But for the mono block type, their would be no shielding and no twisted pair across the connector, a length of about 4 inches. We would place the RS-485 lines at the opposite end of the connector pinout as the 120VAC, separated by about 2 inches. The overall connector housing would be shielded, but the AC line is within that shield.

How can I verify interference will be of an acceptable level?

I did some calculations that show that the electromagnetic frequency in copper with a wavelength of 4 inches is about 2GHz, which is well in excess of the RS-485 baud or slew rate.

I looked at it geometrically and with equal but opposite currents on the AC line it should produce a relatively symmetric magnetic field that should cancel out at the RS-485 differential receiver.

I am looking into testing. I could produced a phase-angle switched load of 15Apk and measure the impact on the RS-485 lines with an oscilloscope. In that case I could possibly use a transformer with a low voltage secondary to reduce the load power needed for that much current. That is, only if the 120 VAC electric field is not important, and only the 15A magnetic field is. If the electric field is important then I need a much higher powered load resistor.

I have also contacted an EMC test lab who could perform a similar test, but at a relatively higher cost.

So is testing needed, or are the analytical methods I used sufficient? If testing is needed, do I need 120V/15A on the AC line or just 15A at any voltage?

Thanks!

 

[FreddyNurk]

Never mind the EMI, what about the insulation rating on the RS485 lines when that close to the AC Power?

EDMS Australia

 

[waross]

If you have a ground fault the current may be several times the normal 15 Amps.
Some thermal magnetic breakers trip instantaneously on a fault current of 150 Amps.
Currents less than 150 Amps trip on an inverse time curve.
If a ground fault returns through the machine frame rather than through the connector there may be little cancelling of the magnetic field.
Maybe it doesn't matter in your application but there are often unintended consequences.

Bill
--------------------
"Why not the best?"
Jimmy Carter

 

[itsmoked]

You've not said how long this cable will be.

Around 10 feet you'd not have any problems.
50 feet maybe not, a hundred - probably.

You must use 150V rated insulation or higher for all conductors. The suggested connectors have to be rated 150V or higher.

Whatever is spewing across the 485 has to be adequately vetted and you need a sufficient protocol to handle communications having botched check sums.

You haven't said what you're doing. It could be that the comm over fiber makes more sense.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[VEBill]

It's not uncommon for connectors to contain both power and signal pins. Plenty of military avionics are boxes with a single connector. On the other hand, many use separate connectors for power.

If you need to predict the EMI coupling, then either calculate it from first principles, or use modelling software. Or set up a bench test.

Four inches seems excessive. We've specified wiring details inside connector shells, to bring the twisted pairs right up to the pins.

 

[RyreInc]

Thank you for the responses! To address each of your concerns:

FreddyNurk and itsmoked: All cable outer insulations are rated 300V, as well as the connector. For the small area where the outer insulation is stripped back there will be a physical gap of about 2 inches.

waross: Good point! Fortunately ground faults are not a concern with respect to comms integrity.

itsmoked: Overall cable lengths are up to 10 m or so (but in most applications 5 m max), but the 485 and AC cables will not run parallel to each other for some or even most of that length, depending on the application. Also the 485 cable is shielded, so I am not too concerned about that. Its just the portion inside this connector where it will be unshielded and untwisted, a length of about 4 inches. Regarding the physical medium, we are locked into copper wires for a number of reasons. The 485 protocol does include checksums and message retries, but of course I don't want to rely on those.

VE1BLL: My concern regarding calculating from first principles is that I am a bit rusty and don't have a ton of confidence that I know the right things to check. As I mentioned I did calculate that the EM frequency corresponding to 4 inches is about 2 GHz in copper, and I am confident about that figure - but is that meaningful? Regarding the 4 inches, this does account for bringing the twisted pair right up to the terminals, but the connector length is still about 3 inches (I did round everything to be conservative).

 

[RyreInc]

I think I answered my question regarding the need for 120V on the AC line for a test, versus using a low voltage: while the current will be pulsed and harmonically rich, the voltage will be a pretty consistent 50/60 Hz sinusoidal wave with very little harmonics, so the electric field should not be a big contributor to noise.

 

[waross]

Electro-static field is a function of voltage.
Electro-magnetic field is a function of current.


Bill
--------------------
"Why not the best?"
Jimmy Carter

 

[VEBill]

"...the EM frequency corresponding to 4 inches is about 2 GHz ..."

For EMI between a power line (even switched and noisy) and serial communications, you really don't need to worry about GHz at all. The peak noise coupling would probably be in the kHz, maybe low MHz worst case. There are other loss factors that diminish the GHz vastly more effectively than the optimum length (4 inches) can overcome.

If you're worried, then experimental bench test.

Another project management approach is to take as many countermeasures as possible (extending the twisting and shielding inside the connector housing), and then accept the risk. If, after doing everything you could have done, it still fails, then you can think about the outcome of that now (major redesign).

 

[itsmoked]

The shield is only good for capacitive coupling.. The twists are for inductive coupling.
With shielded twisted 485 and that distance I agree you shouldn't have a problem. The untwisted aspect in the connectors is not a problem. If possible I would try to have the 120V run in a twisted pair also as the twisting works both ways.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[RyreInc]

Thank you for the followups!

The shield is only good for capacitive coupling.. The twists are for inductive coupling.

I'm not sure I realized this... makes sense though! waross's definitions are helpful in this regard.

I do have a test devised, and a BOM picked out to do so, but I am increasingly convinced this is probably not necessary. I think part of my paranoia is early on we had intermittent communication drops that seemed to be related to having a short untwisted/unshielded length adjacent to a 120V line. But with software fixes and a baud rate change these have been eliminated in those installations.