RS-485 Full duplex.
About the ground issue, I feed GND from another PCB which is the circuit I am trying to protect the most, so this is why I did not want to dump a surge into it. I was going to use mains neutral as it seemed to be recommended vs safety ground to avoid frying sensitive electronic equipment that can deal better with a surge on neutral than on safety ground. I understand your point about neutral crossed wiring and surges though. I would not want an issue with neutral to back feed in my circuit that is for sure. So would you use safety ground then?
Huh. I haven't seen anyone using those split transceivers before, and, I can't really think of a use-case for one at the moment. Why would the pairs not simply be connected immediately together?
Note that putting separate pulls and TVS on them increases loading (resistance and capacitance), doubling the effect if they're wired to the same bus in the end anyway. Maybe that doesn't matter here, dunno.
Grounding is always a locale dependent thing, check code in your area. There is some legacy equipment I think which receives shared ground and neutral, or should be grounded with a separately wired lug (i.e. to the receptacle housing, or a water pipe, etc.). Examples that come to mind are things like stoves, welders, etc.; years ago I had an AC welder box with a NEMA 6-50P, which is H1-N-H2 here (homes are wired with 240V CT). Preferably these should just be upgraded to a safer combination, and obviously you can't design new equipment to such a standard. Maybe it still matters if you're wiring to such equipment.
Yes I wanted to use forced PWM for the buck converter. Fly-Buck is simply the trademark name from TI for Buck + Fly-Back. TI uses constant ON time converters. I was going to use PWM in forced mode (MP4576), is that an issue? Here is my SMPS circuit. If you could tell me if it makes sense or not that would be great, you seem very knowledgeable about this kind of supply, you are the first member that mentions the need for forced CCM. Many had told me to use whatever buck converter I can find.
Yeah, that'll be fine then.
Regarding all the capacitors... what's 12V coming from? Do you have any particular EMI concerns about it (in or out)?
Note that all those caps do nothing about the common mode, which is likely the biggest concern here, having an isolated channel.
The caps don't do much because the larger (1, 10 uF) dominate the response, and the smaller values either don't have much if any less ESL, but their values are smaller so they have higher Q resonances with the larger ones, generally making response worse (peaks in the impedance / transmission at high frequencies). Like, consider this case:
Note the capacitors (body, pad, trace and via) and trace lengths are represented by small inductors.
There are more valleys of low impedance which is nice, but the worst-case maximum is what matters, and you've put in several new peaks in so doing.
There are some old appnotes that recommend staggered or tapered values in parallel, which should be considered erroneous. That might've been appropriate back in the days of THT capacitors, but it's sadly harmful most of the time with ceramic chips.
Better to use one or two large caps for low ESR/ESL at high frequency, then R plus a few in parallel for damping at middle frequency:
Here, a supply is at a modest distance (several cm, say) to the load. R1 has been added (notice the value), or changed to a lossy bulk type (electrolytic, tantalum, polymer*) several times the value of the parallel low-ESR capacitor(s).
*Polymer are available in a range of ESRs, tants as well; they cluster lower than tant, but shop around and choose appropriate values.
You're kind of doing this already (e.g. C1-R1), but the values are inappropriate -- 0.1uF is far too little to do anything at all, and 10R is much too high compared to the surroundings. If C31/C34 are ~adjacent, their loop inductance will be say 5-10nH so sqrt(10nH/5uF) ~= 44mR is what should be targeted, and then you'd need more like 20uF with the ESR, for which a chip electrolytic would be fine I suppose. (ESR can be within say a factor of 2 of Zo = sqrt(L/C) and it's fine.)
Also keep in mind, the transceivers won't care very much about supply noise; maybe just a 10uF would be enough filtering there, and no chokes. (So, you'll be fine in this regard.) If it is found problematic, maybe a couple more caps, or adding a ferrite bead or small choke (in the <200mA range, a 1206 size ferrite bead in the 30-100R range should be adequate without too much saturation).
As for common mode, a 'Y' cap (in the sense of being ground-to-ground; potentially the safety rating is relevant here) might be used to return current coupled across the transformer, i.e. the switch node has some capacitance to the isolated side so the switching edge drives into the common mode. And maybe some ferrite beads or whatever on the RS-485 connectors, to dampen the resonance created between the cable and that capacitance. Which can be coupled (use a data line type CMC) if you need to keep differential bandwidth up.
Tim
Home
Help
Search
Login
Register
