Why would you collimate a newtonian and not adjust the primary mirror,
tl;dr: in a normal Newt your reference is the focuser axis and you collimate every element from the focuser inward to it, but you could also take the fixed optical axis of the primary as a reference and work _the other way around_, i.e. "catch" the optical axis be moving the secondary, then tilt it to point it to the focuser, then tilt the focuser to make its axis parallel to the reflected optical axis.
That's sort of what you have to do on a small Skyscanner, since it doesn't have a moveable primary. What's critical is not "to adjust the primary mirror", it is to make the focuser axis and optical axis cross at the focal plane. _Normally_ we do that by tilting the primary, since we also want a focal plane perpendicular to the focuser axis and we've set the tilt of the secondary for doing that (which prevents us from using _that_ in subsequent steps).
But if the primary mirror is fixed, then you have to use a Cheshire and try to make things work using the tilt of the secondary, and live with the focal plane tilt that you end up with (which is doable for visual use, since the eye can focus at distances other than infinity and can accommodate for a slightly tilted focal plane.)
Both moving the secondary away/from the focuser and to/from the primary will also translate the optical axis seen at the focuser along one dimension, and rotating the secondary will move it in the other direction (but also change its angle with the focuser axis).
Since usually you want a centred fully illuminated field and you have too many choices, just as when you have no sight tube, on scopes like that you first do 'axial collimation' using tilt of the secondary --which in this case is a misnomer, since the axes will cross at the focal plane but will not be guaranteed to be parallel-- and then you evaluate whether the secondary is concentric with the primary's reflection and appears round. If not, rotate it and you move it towards where the primary's reflection appeared and start again.
In theory you could play with the length of the spider vanes and the centring of the secondary in the tube to get good axial collimation (the optical axis crosses at the focal plane _and_ the focal plane is perpendiculair to the focuser axis) but then you lose control over vignetting by the front of the tube.
If you can also move the tube opening...well, moving the tube opening (with the secondary attached to it) with respect to the primary is exactly the same thing as tilting the primary, so you could chase the primary with the tube rather than chase the tube (and the secondary centred in it) using the primary. On a truss scope, instead of tilting the primary, you could in theory work on the length of the trusses to collimate!
Edited by sixela, 25 October 2018 - 07:08 PM.