No it is tan(phaseerror/2) not sin
It comes from the 10*LOG10{[1-cos(phaseerror)]/[1+cos(phaseerror)]} which is equal to 20*LOG10[tan(IQphaseerror/2)]
Look at Razavi RF microelectronics for derivation or
http://ieeexplore.ieee.org/iel5/4138799/4138800/04138864.pdfA key thing is that in a SSB IQ modulator error vectors which cause the "image" sideband, i.e quadrature phase error and quadrature amplitude error, are orthogonal. Therefore you can null both phase and amplitude error independantly.