Isolating ground planes to prevent noise

[UnaClocker]

I'm laying out a motor controller board that's got a uC onboard. I've got GND1 and GND2, GND2 is a high current ground for the FET's driving the motors, GND1 is a low current ground plane for the logic. I've got all the sensors, logic ground, etc coming in on the left side of my board. And all the noisy high current stuff is on the right, with the high current GND2 on the right (this is a unidirectional motor driver, so it's only got high current ground) and all the various outputs are coming out on the right as well.
My question is, I know the two ground planes need to be connected, otherwise the uC wouldn't be able to control the outputs (at least not with tons of optocouplers, which I don't want to do, they'd be way overkill for this). All of the grounds will eventually be going to the same chassis, so they'll be joined outside the unit. How should I join the two ground planes on the board? I was thinking a decoupling cap might work, but intuitively, I feel that'd be a mistake. Maybe a really low value resistor, .001 ohms or something? Or simply a jumper, be it a .1" two pin header that I solder a jumper across, or maybe a single trace connecting the two.
Having no formal training in this, I may not even have my terms right.
And I'm sorry, I can't post the board up to share, my reasons are complicated.

[bobcat]

Actually, you want just 1 ground connected to the board. Run a jumper from the digital side directly to where the high power ground connects to the board. Put a filter cap (220uf) across the digital power planes (Vcc, Gnd) to prevent ground bounce. Don't make another connection from the digital side to the chassis. That will create a ground loop and possibly lots of other problems.

[tszaboo]

If you ask about a the ground planes on a forum, you will eventually end up with the following answers:
high value resistor
low value resistor
inductor
back to back schottky
capacitor
tie them together below the analog stuff
tie them together next to the power supply
dont use separate ground
only connect them via the chassis
And Hitler because of Godwin's law

[Neilm]

I'm with the "don't use separate 0Vs" just carefully layout the circuit so the high currents would be naturally confined away from everything else.

[UnaClocker]

Ok, well at least I don't feel like I'm making a massive blunder then. The way I set it up in the end, I can try just about all of the options. I have a .1" jumper between the two ground planes, so I can jumper that and only ground the high current side, or I can ground both sides. Or I can toss a component in that jumper (cap, inductor, resistor, etc). Good to hear this is a tough decision, I think I've got all my bases covered.

[DanielS]

Connecting grounds through the chassis alone would be no good if the microcontroller drives the MOSFETs without electrical isolation of some sort since the gate drive current would have to take the long inductive path through the chassis. If the gate driver IC is on the 'noisy ground" side then you still have the lesser control signal's return current to worry about.

[Marco]

Just to prove NANDBlog right : Don't use separate grounds (just put the battery/power connections in the middle).

If the signal is PWM you will need at least one relatively direct low impedance connection between the ground planes if you do split them ... doing the ground return path for a PWM signal through the chassis is not a good idea.

[UnaClocker]

Yeah, honestly I knew they needed to be joined in some way on the board, just wasn't sure what way. Direct, decoupled, resistor, etc.. And I wasn't sure if I should run only the high current ground or both separately.