I've printed up some shoulder mounts for four shoulder roll motors to give Nybble some extra agility, but I'm running into problems on a Nyboard v0_2. The motors I've added are a pair of MG90D's for the front legs, and a pair of MG16R's for the hips. They're just plugged in to the vacant pins on the nyboard. With just the front two installed, everything's fine: Nybble boots up, I can calibrate, everything seems to be working as expected. With the back two installed, I run into problems, but not where I expect: the MPU6050 seems to run into difficulties. Either it fails to initialise, and the setup() sequence hangs, or I get a hang during operation as it tries to either reset or read from the FIFO. This happens whether I've got the battery jumper on V+ or BATT. Am I right in assuming that what I'm seeing is likely due to the extra drain from the motors? Or is noise more likely? In either case, what's the best thing to be doing here? I'm not above soldering an additional cap on, or similar, if that's what it takes; another option would possibly be to add a second P9685 board to double up the 5A limit and possibly improve the isolation, but that's obviously more work to make the code aware of (and use) the new i2c address. For reference, the batteries I'm using are 7.4V 1300mAh 2S li-po's. Turns out that "airsoft batteries" is what you need to search for to get long, thin packages. Has anyone else successfully done this?  some more details on this: I couldn't find MG92D's in stock when I was buying. If you try to fit an MG90D at the shoulder, it's not got enough torque to hold still and it shakes the leg uncontrollably, so I swapped out the knee motor and used that as a shoulder roller instead. The MG16R's were a bit of a guess: given the experience with the front legs, I wanted something that had more than enough torque.