Originally Posted by gregr
Intra-image contrast has traditionally been measured using a 4x4 checkerboard pattern and an ANSI specified process that could have only been designed by a committee. I've explained this in other threads in the past so I won't detail it again here, but the ANSI process leads to measurement errors, poor repeatability, and produces higher contrast values in the least important parts of the picture (especially for non-CRT projectors), which are then weighted equally with the lower contrast values in the more important parts of the picture. So in my opinion, although the ANSI contrast ratio is just a figure of merit, it is about as bad a figure of merit as anyone could imagine (especially for non-CRT projectors). So for the last few years I have been using my own modified-ANSI contrast ratio measurement process (which I've detailed here previously), which I believe substantially corrects these issues. But the result is still just a figure of merit. i.e. It allows you to compare projector A vs projector B, but there is still no way to numerically compute the resulting intra-image contrast for any real image content. That is way beyond anything that could be done with a few test patterns and without very complex computations.
When we refer to intra-image contrast we are usually talking about contrast at fairly low frequencies. i.e. not the smallest image details. MTF is a way of measuring contrast vs spatial frequency isolated from the factors that affect contrast at low frequency, so it tells us about the sharpness and resolution of the projector independent of other factors. It takes some rather sophisticated equipment to measure MTF at high spatial frequencies.
And i agree, but still in the practical sense you are reporting numbers based on ONOFF and your modified ANSI board which gives the reader a hint for contrast performance.
It is the ANSI board as such that is being questioned by people since the numbers often doesn't give a fair hint of contrast performance, combine that with that people think that ANSI is the real measurement and you have a problem.
Often people shout out for ANSI contrast and often refer all the depth of DLP to the ANSI contrast, so people have a wrong picture of what the ANSI number really is.
People think that just because you have black and white boxes in a pattern means that the measurement is for "mixed scenes".
Often you here "The DLP had much greater depth in mixed scenes due to its high ANSI contrast"
Mixed? what is that, a mixed scene in a typical movie isn't near what the ANSI board measures in terms of APL.
And as have been said, ONOFF and ANSI are measuring the "end" points of the scale, however IMO ONOFF lies closer to movie material than ANSI.
If we where to measure computer applications which tends to be close to 40 in APL then the ANSI measurement would be a better indicator. And a LCD monitor is perceived as more "contrasy" compared to a CRT monitor when using Word for example, even though that depends on many factors the ANSI number in this situation gives a hint of what we see. Put on a movie on the same screens and the LCD is blown away.
All of this is numbers, all of it is a simplification of the "real world" none is completely accurate. But you are supposed to make the models as close to the thing your are trying to represent with the model.
If you are representing movie material in numbers than measuring on a standard 50APL ANSI board doesn't make any sense?
Even ONOFF may not make any sense but still i believe that it lies closer to the reality you are trying to model.
There has to be a balance between measurements and practical applicability, you could probably make the measurements and patterns very complicated but then you wouldn't have to many sources of numbers. Just as you can add a parameter to an equation in physics to get closer to the reality.
So we still need these "basic" patterns and measurements like the ANSI board, but that doesn't mean that we cant change them, bring them closer to the thing we are trying to measure.
Even though you and many here may know how to interpret ANSI numbers and other figures, MANY people dont, and therefore it is good to give them numbers that make sense on their own to have a parameter which you could compare.
Often people don't see the numbers in a context, you may see ANSI numbers and think MTF, sharpness and many other things etc. But people usually look for a measurement to use simply comparing products.
Lets say breaking distance for a car, usually measured from about 60mph, why do you do that, because often you face situations where you have to break from those kinds of speeds. Now the ANSI board is like measuring the breaking distance from 200mph.
Or even horse power of a car, usually horse power is good if you are racing and constantly lying at top rmp where you usually get max horse power, it still effects lower rmp though. However in normal driving situation you would like to know the power you get from about 2000-3000 rmp, for example disel engines have less horse power but much more torque earlier, and they usually feel more powerful in normal driving situations but they are useless if you are only going to be at high rpm.
So having the torque at 2000-3000rpm usually gives a better performance indicator for normal use, that is why turbos feel so powerful, they give you the torque early on but they may not have much more horsepower.
Car turbos is almost like an iris, which gives you a boost in ONOFF, and thus creates more depth in film material, the iris doesn't do any good for computer applications just as many turbos are not as useful if you are constantly in higher rpms. But since we are at lower APL and since we are driving at lower rpms they make a difference in performance.
Example from BMW
530i at 190BHP
0-60 = 6,5 sek, (here you rev out all gears until you reach 60mph)
50-75mph = 6,6 sek (in 4/5 gear which is a usual overtake)
530d at 170BHP
Tourqe 500 @ 1700-3000rpm
0-60 = 6,8
50-75 = 5,1
Now you tell me which performance parameter torque or BHP gives the best indicator of how these cars feel at normal use.
I would argue that BHP is like ANSI and Torque here is like ONOFF in this example.
BHP gives you potential power at the high rpm and these torque figures gives you potential power at the lower rpms, bot are important and both affect the rpm scale, however normal driving is closer to the torque numbers in this example. Thus that number is better for comparing the above cars for normal use.
I'm sorry for the incoherent babbling and these terrible examples