@godel9 has a good suggested solution, although, in my own testing, the explanation given for the unexpected behaviour is not correct.
From the SKPhysicsBody Class Reference:
The force is applied for a single simulation step (one frame).
Referring back to the SKScene Class Reference's section on the -update method:
...it is called exactly once per frame, so long as the scene is presented in a view and is not paused.
So we can assume that calling -applyForce: in SKScene's -update method should not cause a problem. But as observed, the force does not exceed gravity, despite applying an upward force much greater than gravity (400 newtons vs 9.81).
I created a test project that would create two nodes, one that falls naturally, setting affectedByGravity to TRUE, and another that calls -applyForce with the same expected gravity vector (0 newtons in the x direction, and -9.81 in the y direction). I then calculated the difference in velocity of each node in one time step, and the length of time step. From this, I then logged the acceleration (change in velocity / change in time).
Here is a snippet from my SKScene subclass:
- (id)initWithSize:(CGSize)size
{
if (self = [super initWithSize:size])
{
self.backgroundColor = [UIColor purpleColor];
SKShapeNode *node = [[SKShapeNode alloc] init];
node.path = CGPathCreateWithEllipseInRect(CGRectMake(0, 0, 10, 10), nil);
node.name = @"n";
node.physicsBody = [SKPhysicsBody bodyWithCircleOfRadius:5];
node.position = CGPointMake(0, 450);
node.physicsBody.linearDamping = 0;
node.physicsBody.affectedByGravity = NO;
[self addChild:node];
node = [[SKShapeNode alloc] init];
node.path = CGPathCreateWithEllipseInRect(CGRectMake(0, 0, 10, 10), nil);
node.name = @"n2";
node.physicsBody = [SKPhysicsBody bodyWithCircleOfRadius:5];
node.position = CGPointMake(20, 450);
node.physicsBody.linearDamping = 0;
[self addChild:node];
}
return self;
}
- (void)update:(NSTimeInterval)currentTime
{
SKNode *node = [self childNodeWithName:@"n"];
SKNode *node2 = [self childNodeWithName:@"n2"];
CGFloat acc1 = (node.physicsBody.velocity.dy - self.previousVelocity) / (currentTime - self.previousTime);
CGFloat acc2 = (node2.physicsBody.velocity.dy - self.previousVelocity2) / (currentTime - self.previousTime);
[node2.physicsBody applyForce:CGVectorMake(0, node.physicsBody.mass * -150 * self.physicsWorld.gravity.dy)];
NSLog(@"x:%f, y:%f, acc1:%f, acc2:%f", node.position.x, node.position.y, acc1, acc2);
self.previousVelocity = node.physicsBody.velocity.dy;
self.previousTime = currentTime;
self.previousVelocity2 = node2.physicsBody.velocity.dy;
}
The results are unusual. The node that is affected by gravity in the simulation has an acceleration that is consistently multiplied by a factor of 150 when compared to the node whose force was manually applied. I have attempted this with nodes of varying size and density, but the same scalar multiplier exists.
From this I must deduce that SpriteKit internally has a default 'pixel-to-meter' ratio. That is to say that each 'meter' is equal to exactly 150 pixels. This is sometimes useful, as otherwise the scene is often too large, meaning forces react slowly (think watching an airplane from the ground, it is travelling very fast but seemingly moving very slowly).
Sprite Kit documentation frequently suggests that exact physics calculations are not recommended (seen specifically in the section 'Fudging the Numbers'), but this inconsistency took me a long time to pin down. Hope this helps!
restitutionproperty of the physics body controls the "bounciness" of the node. – Diggins