There is no formal requirement to test a measurement model first/separately but it can sometimes be helpful to figure out the proper measurement structure first before adding a complex structural (latent variable) model. Otherwise, it may be difficult to figure out sources of potential model misfit (i.e., misfit could be due to misfit in the measurement model, the structural model, or both).
Uday Arun Bhale In SEM, it is generally recommended that you first test the measurement model before proceeding to the full SEM. The measurement model is the part of the model that specifies how the indicators are related to the factors. Testing for it confirms whether model is well-specified or not and that the results are reliable or not.