Originally Posted by rader
For the black measurement the xy values are not used in the spreadsheet, only Y. If you cannot measure Y for black then it is impossible to calculate gamma (see Umr's post a few pages back).
You are referring to Post #7 in this? If it is, then you are misinterpreting Jeff's statements. The issue Jeff was describing is that your calculations had previously assumed a true black (no light) black level, and that was materially impacting your overall calculated gamma. In reality, there are several viable alternatives for calculating gamma. Here are two:
1) Take a curve fit. In this case, you can ignore the linear tail when trying to match the measured curve to the ideal. What you do want here, though, is a sufficient population of measurement points to make your fit result more statistically significant. (This is what Gary was doing originally)
2) Measure point gammas. Given that you only have one unknown (gamma) for each measurement point, you can actually solve for gamma for each measurement point beyond the linear tail (there is no gamma factor inside the linear tail). This is essentially what you are doing now.
Neither of these methods equire there to be a black level (0%) reading to be valid. All that the black level provides is another data point. One would note, though, that using a regression esitmation technique, the error in white (high light output) will dominate the calculated gamma value given the absolute numbers involved (i.e., there is limited variation in absolute terms at low stimulus levels).