I find Lens Align difficult to use. Without a proper laboratory test bench it very hard to be sure that the camera and target are correctly aligned.
Setting up the LensAlign perfectly is a bit tricky, but there is a limit on how accurate one needs to be. One of the benefits of the LensAlign kit is that one also gets a sense of how much of a DOF tolerance one has at the chosen shooting distance.
Recently I came across the method advocated by Keith Cooper of Northlight images using moire interference patterns on a computer display. This is very simple. One focuses very accurately using live view then drops the mirror down and presses the AF button to re-focus whilst watching the lens. Any movement will indicate the error.
I'm the 'designer' of
the moiré technique to improve ones AF Micro Adjustment. It is very accurate, it is based on minimizing optical blur that functions as a low-pass filter that prevents such moiré. Minimize the blur, maximize the moiré, by optimal focusing.
Clearly this assumes that the pentaprism live view will give an identical image to that on the sensor. Is this the case?
The pentaprism doesn't play a role. The Live View image comes straight from the sensor that will capture the image itself.
Will contrast and phase detection always give the same result.
Not necessarily, but the goal is to have good focus in either case. When contrast is low or has a confusing pattern, contrast detect can fail. Contrast detect should be more accurate than phase detect, but it is also slower. Phase detect can fail if the optical system produces different results with the aperture wide open, and when stopped down for the shot itself.
What distance should one use? Canon say at least 50 x the focal length. Others say x 30 is adequate and some advocate using a short distance. What is the rationale behind this? Your comments would be appreciated.
The best distance is the one you use most for your actual shooting. However, that may be unpractical for testing. Since many lens designs are optimized for a certain distance (somewhere between infinity and close by), one can use that distance, but we don't know what design distance that is. Some say it's 10 metres, some say it is at 50x the focal length, but we simply don't know. It is also a bit irrelevant if in practice we normally shoot at a different distance, making that distance more relevant to get correct focus at, IMHO.
Also, traditionally one tested resolution and focus by shooting a test chart on a flat wall. In order to find the limiting resolution on film, that required shooting at distances like 100x the focal length to avoid getting everything in focus. Digital sensors usually have lower resolution limits than very fine grain high contrast film, so we can shoot at shorter distances and still have some of the finest detail unresolved. The shorter shooting distances like 25-50x the focal length also make it easier because we don't need as much space/distance. The shorter the shooting distance, the more accurate our test target needs to be, because the magnification factor of the optical projection of that target on our sensor is larger.
To achieve a manifestation of clearly visible moiré, we may need to vary the distance a bit, so I recommend to choose a distance that produces an obvious moiré pattern when in focus, and which is close to your common shooting distance. When that is impractible, (e.g. telelens and your room is not long enough) I would test at a couple of metres distance and later verify with shots taken at common actual shooting distances. There will always be some front or back focus, but we want to remove the front/back bias so that on average the focus is closest to the intended position.
Cheers,
Bart