High DC current makes strong magenetic fields, thus obviously interfering with the magnetometer. You must keep power wiring way from the APM and further, never make any loops. When you added the battery you moved the wiring and it's magenetic fields.
It makes sense. I will try to reduce magnetic field.
I was mistake. My setup is parallel, not serial.
Today i test magnetic field and clearly i have some noise.
I can't move my apm board further. Is there any other way to reduce magfield?
Yes, clearly from your wiring, you would have a problem. You can move the wires and common sense should have said to keep the main power distribition and ESCS on the lower level, far below the APM. Magnetic fields drop strength by distance. Thus 1/2 inch or more makes a huge difference. There is NO reason to have main power wires carring current from the batteries to the ESCs near the APM (any big fat wires near the APM tells me you are doing it wrong). I can see the ESCs are on the lowest level by the arms and so should the power distro board. The only wires going to the APM from that power section are PWM and 5 volt to power the APM (nothing larger than 22-24 ga wiring). Maybe if you have a current shunt, it too should also be on the power level and then the low current leads go back up to the APM from the sensor but any and all large wires should be more than 1/2 inch or more away from the APM.
As far as shielding, this is the ONLY way a very special metal (AKA MuShield http://www.mushield.com/ ) Moving the wires gives the same effect and is less costly and probably works better anyway.
I know it sucks to take it apart but build it right and then you won't have problems, You should be mounting your power distro board below the deck plate, between the upper and lower arm plates. Then, no high current wiring would even be near the APM.