Snubbing Parallel MOSFETs

I have 5 Half-bridges that I am paralleling in my motor drive. THe problem I am running into is that the MOSFETs themselves are big and on the PCB they are going to be lined up in 10cm rows daisy chained together by a wide, but long straight traces. I have rail clamp diodes and RC snubbers to try and shield the "MOSFET bunch" from the inductive flyback of the motor and it's wires. BUt I foresee problems from the parasitic inductances that connect the MOSFETs to each other within the bunch. It's not practical for me to add rail clamp diodes and/or RC snubbers for every single MOSFET.

Is there a way to go about this? RIght now, my best...but somewhat precarious bet is to rely on my one big discrete snubber to absorb the motor's flyback, and rely on each FET's parasitic diode to deal with the remaining inductive kick that will occur from the traces paralleing the FETs together. I hate relying on that instrinisc diode though, and adding individual snubbers to each FET just lengthens these paralleling traces as well as destroying how tight and clean they already are.

I'm unsure if the inductance from the traces alone (assuming the snubber has completely dealt with the motor inductance) is enough to cause the problems I am afraid of. The totally length of the trace daisy chaining the FETs together would be about 10~12cm divided into ~2cm of trace connecting each FET together.

To give you an idea, it's going to look something like this:
**broken link removed**

Are you sure you need snubbers? I would check some of the high-power model controllers. I don't believe they have snubbing. Also, here are two pictures of paralleled half-bridge controllers I built. The left is an early version using copper strips. The right is on 4 oz PCB. Neither has snubbers. Both have a "main" capacitor as we discussed previously, which is not shown in the image of the older design. I have no idea of the amperage capacity of the left controller. It was probably 200 to 300 A at 12V. The right controller is for a 1.5 HP PM motor at 12V. John

WHen I say snubber in this case, I am referring to any kind of discrete device like diodes or R-C. DId you only have the main capacitor? YOu didn't have any diodes or anything?

I am not certain if I need snubbers, but I am uber paranoid about it because it's something that's very dependent on the layout and can't be predicted very well ahead of time. If it ends up not working, then I'm out $1k or so.

The best I can do are some realy crude calculations:
20nH/inch of PCB traces, 10nH source inductance, with 50A, and my rise times of 200ns with V=L*di/dt just tell me a spike of 25V is possible assuming the motor inductances are completely supressed by the main snubber. My thermal calculations tell me each MOSFET can only handle about 10A (since I'm trying not to use heatsinks and it suits my purposes anyways), and since the MOSFETs are daisy chained along the same trace I went and added up all the inductive spikes from each trace all the way to the MOSFET at the end of the daisy chain (ie. V = Ldi/dt with 50A, then 40A, then 30A...all the way to 10A, and then added all the Vs up to get the largest voltage drop).

I think you want a seperate groundplane for your FETS. In doing so you can snub at the ground point of entry. Breadboard work may prove this.

The point of entry from the batteries you mean?

Currently only the gate drivers and Power FETs run off the main battery and everything else runs off of a smaller separate battery and is completely optically isolated from the gate drivers and power FETs. THese will be on a separate 4-layer PCB with a ground plane and all the snubbers sit at where the main battery power or the motor wires enter the board but right next to the FET groups)

RIght now I don't have a ground plane for the power-side (because I can't get 4 layer boards with 6oz copper) and any heavy copper 4-layer board I can get is really expensive. This "power PCB" will eseentially be nothing but +V and high-side FETs on one side and GND with low-side FETs on the other side. THe ground and +V traces basically mirror each other on both sides of the board though, as do the motor power wires. I'm not sure if this would count as a ground plane, however.

Check the avalanche rating of the MOSFETs and compare it to how much inductive energy is stored in the motor. It's possible that all those MOSFETs will be happy to take the spikes with no problem.

Ah, I've never take that approach to evaluating it before. The energy stored in the PCB traces is two orders of magnitude less than the avalanche energy the MOSFETs can withstand. So I guess I do not need individual snubbers, assuming the motor snubber and diodes do their job.

THanks.

I would err on the side of caution. At the very least with real estate permitting, I would add MOV's and a snubber, along with a diode in some area of your chain. You can always leave the pads unpopulated. Inductive spikes can be real nasty, and end the life of an expensive FET.

Real-estate is not a problem, the problem is layers. It's difficult to jump snubbers over large high current traces (right now I have the gate drive trace running in between a the power rail trace and the motor drive output trace since that was the cleanest way I could get it. And jumping starts making things...tricky, unless I go multi-layer that is, but I don't like vias because of the extra inductance and multi-layer is expensive. What I need are some LONG SMD packages that can jump over traces.

Unless I can find another way to route the gate traces.