AoE x-Chapters - 1x.1 Wire-&-Connectors

You probably use tabs instead of spaces too. Blech.

Cheers

Phil Hobbs

Stranded wire: no mention of extra-flex or the main reason unstranded is used: lower cost.

NT

T.pdf?dl=1

used: lower cost.

also tinsel wire. PVC is frequently rejected due to its vulnerability to chemicals as well as heat Magnet wire comes in various temperature/lifetime grades, single & double c oated, solderable and gotta-scrape. And colours. Higher temp insulations: glass fibre, ceramic beads. Asbestos historically.

Lots of domestic kit has been made on ss & ds boards. They haven't disappea red. FR-4... FR2 is still popular for domestic grade goods. Some use other varia nts too. Robot stuffing? Masses of stuff is hand soldered in the 3rd world. If the soldering scorches anything, it went wrong. Soldering temp profiles. What's strange about ounces per sqft?

"If there?s more than one wire, you call it a cable." or flex if stranded, short for flexible cable. Multicore.

"flows through a conductor, the current is not uniform throughout the bulk ? it is concen-trated in an outer layer " sometimes

NT

Very kind of you, but I'd never use them. (The ones that come out of the cake though...nah, I'd never use those either.) ;)

Cheers

Phil Hobbs

pdf?dl=1

"Now for the connector types to avoid (reasons withheld ? just trus t us!)"

that won't cut any ice, and the advice is wrong. Try selling consumer grade audio gear with odd connectors, see how that works out. Edison mains plugs are easy to criticise too, should we put something else on our mains produ cts? Aluminium Belling-Lee UHFs are particularly awful, but without explana tion it tells the reader nothing.

NT

But then you get induced [electric] currents corrupting your signal.

No free lunch. Use differential regardless. :-)

Tim

That's... huge?

I may have to get you a dictionary.

That's board level. Ground plane takes care of that. Do the same usual thing on the board, keep loops local and avoid passing signals across it, and you're fine.

Tim

Does a single-phase power plug have four prongs? Or five?

With standard 20 mA current loop driven from a 24 Vdc supply will carry nearly 0.5 W signal power, thus a quite strong interference is needed to disturb the signal.

Using twisted pairs .e.g. telephone cables also limits the induced interference.

Current loops have been done with optoisolators for nearly half a century, so you also get galvanic isolation for 'free'. Powering the loop always from the transmitter side, the receiver can be passive (an IR LED) so no ground potential issues between boxes. .

Properly terminated RS-422 line is essentially a bipolar current loop.

On a sunny day (Tue, 6 Aug 2019 18:24:32 -0400) it happened Phil Hobbs wrote in :

~ # locate 3055 | grep -i pdf /root/download/html/2N3055-D.pdf

What is a big VFD ? I have only experience with a few dozen VFDs, each in the 10-100 kW range. Control signals with 20 mA current loops.

Unless it's quite precise. Recently had a customer ask for 24-bit inputs. Ended up with a precision burden resistor, channel mux, in-amp and a TI SPI ADC.

Noise floor looks close to what the datasheet claims, so that's good news about my front end. CMRR isn't bad, is visible but of a similar magnitude (~100 counts over the input range). And hey, resistor matching, what do you expect. (I added functionality so they can measure VCM and calibrate it out in software if they want. And yes, customer's handing software.)

Since the inputs are filtered and essentially unloaded (only the filter caps, mux switch, and in-amp input leakages), the inputs float very nicely, and can be biased (in the common mode) by a touch of the finger. There is some sequential charge transfer between channels, due the mux.

They couldn't give me a hard commitment on what ground reference or polarity* the 4-20mA channels might be, so I had to do this (your traditional +/-15V AFE). It's a 24V system and they really would've rather had a 30V range, but that gets harder and harder to pull off (not so hard regarding precision amps, but analog switches?).

Yup, pretty nice. Unless there's a nick in the insulation and now you have current leakage from one or the other line...

Optos are terrible! Unstable gain, poor distortion, slow speed. Fine for serial. Basically useless above say 8 ENOB, and a few megbits.

It's also a bipolar voltage pair. "Properly terminated" being the key word there. It is necessarily and simultaneously both! Concentrating on just one is setting yourself up for edge cases in the other. :)

Tim

I would dare say any size VFD is "big" enough, if it's badly filtered. Hundreds of volts swing, unshielded cables? Going from 100W to 100kW is only a few dB more volts; sure it's many more dB of amps, but the amps don't radiate quite as much.

It depends, of course. The normal case is, the amps are mostly filtered by winding inductance. In that case, the offending radiation is mostly due to switching voltage times cable capacitance, and not so dependent on load current.

You can filter out that kind of noise, if you don't mind that all your signals become ~ms scale status signals. Good luck getting real data through there; shielded cables and isolated transceivers are basically mandatory.

Tim

Have you looked at the current waveform of a simple old style six pulse rectifier. ? It contains quite a lot of high harmonics. Current requirements usually require some kind of PFC, for instance a 12 pulse rectifier.

Then there are the PWM drive side, which should use shielded power cables to the motor to reduce radiation, even if there are some filter in at VFD output.

That 24 bit is quite special requirement, since it is more than 7 decimal digits. That might be needed when measuring the weight of a car and then something 1 gram on the same scale. In such cases the sampling period can be quite long, such as 100 ms, which cancels some of 50 Hz as well as 60 Hz interference.

Some high quality audio use 24 bit ADCs, but the real accuracy is about 20 bits (120 dB SNR).

Going digital at the signal source helps a lot.

Using dual optos at the transmitting side helps, if the other opto is used for feedback.

It is quite typical to use 12 bit ADCs with current loops, which gives better than 0.1 % resolution. Few transducers are better than this, at least repeatedly.

Doesn't even matter -- the harmonics don't extend that high. Well, they do, but not because of rectification as such, but because of diode recovery. That's usually in the low MHz range, and some volts in magnitude. A pain for EMC, but nowhere near as violent as the other side...

(A phase controlled rectifier can get up there in amplitude. SCRs don't switch all that fast, fractional microseconds maybe, so again, low MHz harmonics; but the worst case magnitude is a step change of the full peak line voltage.)

This. Or conduit more likely. But that can be hard to use. For which they might opt for flexible conduit e.g. SealTite, which is supposed to be grounded through the end fittings so it should be okay. But, maybe it doesn't make such a good connection? (As it is, regular rigid conduit may only make a few points of contact, with those setscrew-locked fittings. Compression fittings should be excellent though.) Or someone forgets to use metal and goes with plastic instead?

Or if it's not even grounded correctly, so, presumably the VFD has some local RF grounding inside it (or all the same, the EMI just goes up the mains conduit instead), but that's not been well-grounded to the output conduit, and the loop between grounds ends up huge.

And that can happen easily in a normal industrial panel -- you install all the components: fuses, terminal blocks, DIN rail, contactor, PSU, VFD, PLC, on a common plate. You lift the plate into the panel and bolt it down. It's only grounded to the panel itself through those bolts, and whatever wires have been added. Maybe the paint wasn't properly scraped off the panel and none of its mounts make ground? Then you'd be relying on the galvanic ground -- some long piece of 12AWG spaghetti -- for RF ground.

If nothing else, all the wiring on such a panel is neatly tucked together into raceways. It's not like you can run a conduit right up to the DIN-mounted VFD. (How would you get the plate in or out?) So all your, say, 24V or 4-20mA control signals get intimately mingled with those 360V peak square waves.

It could still be routed in a responsible way, absolutely; you could add a ground strap between the plate and the target conduit, keep the VFD output cables independent from the other raceways, and routed along the plate and ground strap. If mains is badly filtered, you can put a line filter right beside the VFD (you can get DIN-mounted ones that ground through the DIN clamp, not exactly perfect for RFI, but a damn good start on EMI), and then you aren't so worried about the mains being routed wherever. And you could use R+C or MOV or other snubbers across all the contacts: relays, contactors, solenoids, whatever. But that's a degree of knowledge most technicians don't have, so unless the panel was designed carefully by an engineer first, it's unlikely your average industrial panel is built quite so neatly.

So in short, it's no surprise that industrial environments can get god-awful noisy.

And it's a good thing that stand-alone VFDs exist, with standard electrical breakouts on the bottom, and hopefully good filtering and grounding, all together in one package. Run conduits in and out, connect up your control signals, and you're golden.

Still likely to be a few volts of noise floating around, but that's at least tolerable to RS-485 and such (7V CM range). Now, that still sucks if you're trying to do anything much quieter (say, scoping those serial lines), but such is life.

Tim

Try doing ultrasensitive measurements near one, and you'll see.

Cheers

Phil Hobbs

OK, just trying to help you find the right tool. After a while pounding nails with the crescent* hammer gets old.

George h.

Great filesystem organization you have there, Jan. ;)

Cheers

Phil Hobbs