“…In initial design scheme, an aluminum girder is connected to back beam by only one guide rail and eighteen sliders, a two-period test piece has been manufactured, we have found that the clearance between arrays fluctuates from 1 mm to 1.4 mm in different polarization model, the clearance has changed 0.4 mm, which is unacceptable. In sight of this, a two guide rails scheme has been put forward (figure 8), each aluminum girder is connected to back beam by two guide rails (such as guide rail 2 and 3) and thirty-six sliders, the preloading class also increased by one grade, then a new one-period test piece has been manufactured, the clearance fluctuated 0.12 mm when changed from horizontal polarization model to vertical polarization model [3], the improvement is evident, and the two guide rails scheme has been adopted, on the full four-row APPLE-Knot undulator this clearance between arrays is found to fluctuate within 0.1 mm. The fluctuation of the clearance between arrays is influenced by three factors.…”