It's not at all important at low NA, using objectives at say 10X and below. But the scaling goes as NA^4, so it gets more important quickly for higher magnification objectives.mawyatt wrote:Is it that important to get this distance just right from an IQ standpoint, rather than an exact magnification standpoint?
My standard reference for this is the set of graphs at http://www.science-info.net/docs/etc/Tube-Length-na.gif , which I'll show as an inline display here for convenience:
Long ago I shelled out for the full text of an original article that contains this figure and explains how it was constructed .
Quick summary is that the curves come from experimental observation, using data from a couple of observers who were very strict. Quoting their words: "The star test is so critical that data obtained with it represent the least possible image degradation, one that probably could not be detected in most images."
The observers noted that at NA 0.25, their nominal 160 TL objective could tolerate over 200 mm of additional tube length before even their star test showed a problem. Their NA 0.50 tolerated almost 15 mm, and even at NA 0.65, tolerance was 5 mm. Interpolating their results, it would require NA 0.55 to barely detect a difference with +-3.5 mm of tube length, which is the accuracy that be achieved with standard extension tube sets.
The experiments reported in the article were concentrating on spherical aberration near the center of the frame. Other aberrations can kick in farther from center, and they won't necessarily scale the same way. But even so, the rule of thumb is that small changes are fine, especially at low NA.
Perhaps the biggest problem I face in trying to provide guidance in this area is that what's OK depends so strongly on NA. The tolerance goes inversely as NA^4, so a delta that has no significant effect at NA 0.25 can become a big deal at NA 0.50, where NA^4 is 16 times larger. Even a delta of +-3.5 mm could become relevant if somebody mounts up an NA 0.90 or 0.95 objective.