Killed prop RAM with 12vdc to 120vac inverters. Suggest prop protectors. Zeners?

Clock Loop

Aside from using opto isolators, the 12vdc inverter and the 3.3v prop regulator are on the same ground and supply line. emi, spikes, etc i think is what killed the prop. how about some suggestions to attach to the 3.3v line to help bypass all that emi etc into the ground? a TVS would be ideal.... say around 3.6v...

or would a 3.6v ZD be a good solution here?

I just want to start attaching something to the 3.3v supplies of all my prop projects because the prop is quite sensitive to anything that does spikes and noise above 5v..
*when i powered my circuit to test it using 5v, everything worked fine, but immediately when putting the full 12v into the circuit, it killed the prop ram.

jmg

It is smarter to not get spikes in the first place.

If you have high and erratic loop currents, then keep the Prop on a separate Vcc and keep the areas very small, and maximize the distance from your Prop, to the high current area.
If you do need a communal/wide spread 3v3, use a separate regulator(s).

If you do want to clamp pins, the TVS clamps used for USB lines are effective and made in high volumes.
These have a Vcc zener, and 4 fast low C clamp diodes

whicker

clock loop, you can try using a DC-DC converter module to power the prop electronics. most of these do some filtering.

Pay attention to your ground return path to make sure the current from the 12V device doesn't flow back through any part of the 3.3V ground... one 0V (-) wire goes back from the 3.3V devices to the battery or whatever you have, then another 0V (-) ground wire goes back from the 12V device to the battery. both wires are tied together permanently at the - of the battery so one is not connected before the other.

evanh

Good ground plane layout is number one, and by extension, as Whicker said, avoid ground loops as much as possible. The last thing you want is unexpected voltages forming along the copper tracks. Treat it as part of the shielding design. It means you have to think about it as part of the board layout but it doesn't cost any extra in components.

At least some filtering/snubbing/isolating might have to be used on the digital inputs. Isolating is the most robust option but a simple RC filter can do wonders when you aren't too worried about fast response time. Resistor-zenor combo is a good fast acting input and also doubles as a voltage shifter without directly loading the input of the chip. I have been known to combine capacitor, zenor and resistor all together.

Extra wrap-around shielding can be wise. I've even seen just a small conductor jumpered above the sensitive area as a fix. All depends on how much RF is being generated by the motor/inverter. Good idea to block the source of the noisy RF instead.

Design of digital outputs deal mostly with ground-bounce and fly-back which are self generated. Power supplies have a level of inherent filtering but, as Whicker mentioned again, isolated the supply can be an excellent solution. Analogue is whole other level of precision of design again.

Clock Loop

I replaced my transistor based IRF510 mosfet with a HCPL-314J gate driver optoisolator chip, this chip drives the gate.
Circuit works great now. I thought I might be able to get away with not using optoisolation on prop I/O, but I was wrong.
Thanks for all the great ideas...

Gonna havta order some and test their high voltage, high current, and emi capabilities.
http://www.digikey.com/product-detail/en/TPD2EUSB30ADRTR/296-28153-1-ND/2520840
http://www.ti.com/lit/ds/symlink/tpd2eusb30a.pdf
$5.30 for 10.

Fernand

FWIW, I had the great educational experience of designing controls for aerial tramways and ski lifts in the mid 1970s. It's hard to create a more hostile test environment. The temperatures are totally mil-spec. The control lines ran several kilometers, as wireless was illegal for allegedly lacking sufficient fail-safe capability. Relay logic was the industry standard. The wiring had to be "closed to run", with all manner of brute force fail-safe features required. The crews used high-powered VHF walkie-talkies right next to wiring. The ground potential between the bottom and top of a mountain is often over 100 volts. The motors I had to smoothly ramp ranged from 200 up to 700 HP, 480 volts. I got to design SCR style DC controls, handling several hundred amps. The power cabling used something like half-inch conductors. Once a bug in my code kicked the SCRs full-on from full off. All the breakers in the plant popped, and I got to see the wiring jump right off the floor. Talk about some noise and ground loop issues! In addition, there are frequent storms with lightning strikes.

So, as this was the first microprocessor-based design in that setting, and I was just a kid, I got to seat o' the pants try some isolation and protection circuitry. When a lightning strike took out all the relay logic in the same cabinet, my prototype system survived 100%, which, in concert with averting a catastrophe I'll tell about some other time, so impressed the electricians in the ultra-conservative "relays only" Colorado Safety Board they stopped throwing up roadblocks to its adoption.

So let me tell you what worked, in practice. Some actuators had to run a loop over the whole line (2 km?) into CMOS into the (6800) MCU. On such inputs Optos were used. But even if you use optos, the wires run into your boards, and without good layout, supreme attention to grounding and heroic suppression, transients will find a way to couple and cause failure. On the front end I used little neon bulbs with a low value 5 watt resistor and little fuses as a first line of defense, and resistor-zener clamps behind them as secondary. That fed in current mode into optos with ceramic bypass caps for noise suppression. Debouncing was done in software, and multiple hardware watchdog timers, on both IRQ routines and the main loops, would quickly force a safe state if the MCU failed or jumped the tracks.

But on some less critical lines without the giant ground potential differences, I tested using the resistor-zeners alone, without neon bulbs, fuses or optos. I was amazed at their effectiveness. If I had to build a highly transient-resistant system today, I'd basically use the same approach, probably using Transzorbs (or those TI packages) instead of zeners. Grounding, clamping and filtering are absolutely essential, much more important than optos. Of course in my tram case they were inevitable (because of the ground potential differences), and cost was not a factor. But IMHO, optos are only necessary if you cannot control the grounding and keep the signaling levels within spec. In something like an automotive app, they add to cost and make the design and layout more complex without necessarily solving anything.