DIYEFI.org Forum

So I just read yet another article that tells me we are doing this grounding thing all wrong. So I figured I'd post about it.

Does anyone have an article or something that notes we do in fact need multiple ground planes? Right now, I see our needs as power, and non-power.

As noted here http://opencircuits.com/Printed_Circuit_Boards analog and digital grounds should all converge at one point of the board. This is a close correlation of the NFPA70, which is why all house and branch circuits in the US are required to converge to one ground reference.

As noted in Circuitcellar mag 253 article "RF and High-Speed PCBs" by Robert Lacoste, that above reference is wrong in suggesting splitting of the ground planes. He claims that for EMC purposes/signal integrity issues, you want one ground plane and you steer your currents not physical, but electrically. Basically, the ground currents will follow the trace on the other layer. He claims if you split the grounds into planes, you will likely force the ground current to flow in a path that isn't the shortest possible path. Which will act like antenna, coupling energy where you don't want it. If done correctly, you can split your ground currents, while keeping one copper plane. He agrees there should be one ground wire returning to the supply.

I feel these articles don't fully apply here, as we have power, digital, and analog grounds. The power signals need to be on their own, as the voltage drops will cause signal integrity issues. So I feel we do need at least two grounds, but I can't seem to find an article that claims we need any more.

I guess I can also see the need for the infamous Guard Ring as discussed here. http://www.ce-mag.com/ce-mag.com/archiv ... E_028.html or the infamous safety ground, for protection against hazardous voltages. So I can see a need for A third ground wire, however with a metalic case, that's screwed to the sheet metal of a vehicle, which is then grounded, I can see how we won't need that. Any how, I don't think the layout folks here are concerned with EMC, and our electrical energy potentials aren't typically life threatening. So this third isn't really that important.

I see PUMA has a power ground, and other ground. However it does have semi split planes on the other ground side of things. So far I see PUMA as having the most reference documents to back it's design theory.

So does anyone have an article or something that notes we do in fact need multiple ground planes?

One ground plane is fine, however what I mean by that is that we do not need a ground plane as such for the driver stuff.

Grounds should be pin-counted and sized appropriately for the current involved. If we connect the other grounds to (protected) ADC inputs we can monitor the performance of them and pump out errors if they are floating too much. I was about to suggest that for ravage too as it's a very good idea indeed.

The power signals need to be on their own, as the voltage drops will cause signal integrity issues. So I feel we do need at least two grounds, but I can't seem to find an article that claims we need any more.

You won't find one. The simple truth is that for many setups, we don't need more than two. However in the case of two banks of injectors being run from one board, it would be advantageous to have both sets have one each. Likewise for on board ignition drivers, they deserve their own ground. If the board is supplied with GP low side drivers that could be expected to do all sorts of tasks, some heavy duty, they should get one shared ground too, so as not to interfere with the injectors and/or coils. One ground can share many pins, if required. So the situation, purely for example could be "3 grounds, 2 pins for bank 1, 2 pins for bank 2, 1 pin for reference".

What will happen if you allow them to be shared is that a current on other circuits will cause V differences on the circuits that matter (especially injectors, but also coils, where applicable). You can get the situation where your idle is fucked up by a fan being switched on by a direct drive FET as it drags the ground voltage high and causes your injectors to see less total voltage and your dead time curve to be wrong. This is exactly what happened to me in California a few months back, the ignitors would not dwell the coils because insufficient voltage (and therefore current) was getting to them due to the injector ground (wrongly shared by the low level ignition circuits) was being dragged around by the injector load. All the sensor readings were perfect, and the fueling was going fine, but no spark, and that was why.

On a simple high z + ignitors in engine bay setup with a standard fuel pump relay and little else, you could reasonably expect 2 to be enough. The relay current (even if large) is steady state and wont dynamically affect the injector voltage, and the injector load moving the relay voltage around is extremely unlikely to cause the relay to disengage. So they can be shared without hassle. If GP accessory low side stuff is provided, an extra ground should be included and the fuel pump drive should probably go in with that. Not doing so could cause a headache tuning when that fan switching on messes with your tune more than it should by alternator/engine load alone.

I hope that is clearly written and well articulated enough to get the message across.

The summary is, ground count 2+, power count 2+, exact configuration entirely dependent on the features included on any particular design.

Fred.

I say neah, on 2 pins for one ground path. The ground current should be able to be handled by one harness pin alone. If a design has a redundant pin, that's OK, as long as each pin can handle the max current, with sufficiently low ohms, ect.

I may agree with separate gnd for separate banks of injectors. Depends on the layout. For an 8 cyl, highZ, the max current will be something under 10amps. If your pin can handle 10 amps, then I don't see it as a requirement. However I certainly don't mind putting only 4 on one pwr-gnd, and the other 4 on a different pwr-gnd, if one has pins in the harness connector for it.

However taking a DB37 with a rating at 1 amp, and trying to use 10 pins to make an equivalent of a 10 amp connection, that's where I say neah.

I'm also tempted to suggest an opto for galvanic isolation of the power grounds. Such that the small ohms we are likely to see on that path, won't cause the injectors to go lean. I haven't fully decided. It's a bunch of parts, and space is a bugger. I think to do it, we would also need to split the 5V ref to that same ground, such that the isolated outputs can float with the gnd ref.

jharvey wrote:I say neah, on 2 pins for one ground path. The ground current should be able to be handled by one harness pin alone. If a design has a redundant pin, that's OK, as long as each pin can handle the max current, with sufficiently low ohms, ect.

Fair enough! Good call.

However I certainly don't mind putting only 4 on one pwr-gnd, and the other 4 on a different pwr-gnd, if one has pins in the harness connector for it.

Right, exactly. And, if you were smart, and routing wasn't a bitch, you'd put 0,2,4,6 on one and 1,3,5,7 on the other such taht if only 0 - 5 were used, two didn't out perform the other four.

However taking a DB37 with a rating at 1 amp, and trying to use 10 pins to make an equivalent of a 10 amp connection, that's where I say neah.

:-)

I'm also tempted to suggest an opto for galvanic isolation of the power grounds. Such that the small ohms we are likely to see on that path, won't cause the injectors to go lean. I haven't fully decided. It's a bunch of parts, and space is a bugger. I think to do it, we would also need to split the 5V ref to that same ground, such that the isolated outputs can float with the gnd ref.

Stellar idea for ultimate robustness. I have to say, though, the only time you'll get that issue is with shit wiring, bad connector/loom to load ratio, or a fault condition, and under all of these circumstances it will just keep working badly instead of stopping working. Both could result in lean conditions, and lean conditions from both should never happen. So overall, it's probably much of a muchness? It's certainly worth considering, though.

Fred.

Hmm, I wonder if the OR/XOR could be included in the opto some how....

The typical approach is to have multiple ground planes indeed, and join them at one point which may be inside or outside the ECU.
The OEMs differ on how they do it, sometimes they join the power and logic grounds at the engine and sometimes they do it in the ECU. The safe thing is to join them in the ECU, that way the ECU doesn't fry if one ground wire gets disconnected but that's not as good for EMI performance.
However, the analog ground is pretty much always joined to the logic ground inside the ECU because for best signal integrity the sensor current must return the same route as the signal goes and there's not much tolerance for a voltage difference between the ADC chips and the logic that interfaces with them.

baldur wrote:The safe thing is to join them in the ECU, that way the ECU doesn't fry if one ground wire gets disconnected but that's not as good for EMI performance.

Nor is it as good for any sensor that isn't grounded at your EMS, such as an analogue feed wideband and the battery itself. I understand that there are other solutions to both, however the other solutions still work if you provide for the external analogues and the reverse is not true.