Theres an italian in the house! - Page 14

Yes exactly. I think that the notion of MTF correlating with ANSI has been made mostly because it's been happenstance that some technologies like DLP that have high ANSI contrast also have high MTF. Both measures are probably independent from each other. In fact if we look at a plot of MTF % vs frequency (say line pair/mm) it's hard to equate a point on that curve with the ANSI pattern because contrast measurements from that pattern are unique to that pattern. But if we make the leap that it may be roughly similar to a low frequency MTF pattern, the corresponding MTF anywhere at those low frequencies will be very close to 1 for every type of native 1080p display. In other words there isn't a way to infer ANSI contrast from the MTF curve. Working it from the opposite direction (trying to infer MTF from ANSI) is also difficult because there are so many 3-panel and lens issues (mainly convergence, CA and focus) which don't play much of a role in ANSI contrast but play huge roles in MTF.

As you've pointed out it's easy to mess up the MTF of a 3-panel by unfocusing or misconverging the lens but the ANSI cr is unchanged. As another example of this, my old RS1 had about 1/2 of a pixel of misconvergence and some red CA, but my RS20 has excellent convergence, low CA and is strikingly sharp compared to my RS1. The combination of sharpness and improved on/off gives the image a sharper and crisper image which is noticeable even with mixed scenes. It was mainly because of the crispness in mixed scenes that I was expecting ANSI to be significantly improved, but when I measured ANSI it is down on the RS20 compared to the RS1. Mlang a retired optics engineer and forum member, also made this same assumption at Cedia (that the RS20 had higher ANSI) and was also surpised to see that this wasn't the case.

At any rate, I think that what's happened is that there has been a dearth of MTF data and as a community we've maybe made some seemingly sound assumptions in the absence of this data that probably doesn't really hold water. It's too bad that reviewers don't perform MTF measurements as part of their reviews, but given the variance of 3-chippers that can cause huge differences in MTF, I'll bet that there is some reluctance by reviewers to publish these sorts of tests and probably a reluctance by manufacturers to allow the reviewers to make the measurements in the first place. Even ignoring variances between units, there's probably a large variance of MTF within the field of a given unit.

Some movie caps. Camera is a cheap Sony DSC-T100.

A closeup of John Goodman's ring. from Speed racer. Superb detail at 14 ft



The City, Speed Racer. Fabulous flamboyant color rendition.



Sin City. A couple of interesting "points" worth noting.



Sin City. This film is a real ANSI torture test in places, though not as tough as Renaissance.



Sin City. As above.



Sin City.



James Boned, License to Drill.



Beowulf

The reds look gorgeous. Sin City is a torture test for panel aignment and is an ANSI contrast wet dream.

Thanks. I also like to run Renaissance as an ANSI test. I will be posting screens of that soon.

Some more comments about ANSI CR and MTF relating to the posts by Darin and Mark.

The correlation between ANSI CR and MTF is not perfect, but it's quite good. There are confounding factors that affect specific frequencies, like panel convergence and the problem LCOS has at the limiting resolution. These can result in good ANSI CR but poor MTF at a few frequencies. For most of the frequencies, the correlation will be good. Also, there are no exceptions to the reverse that I can think of. Poor ANSI CR will always correlate well with poor MTF at all frequencies.

The MTF of very low frequencies, like the ANSI CR pattern, are only meaningful if you look at the data before the lowest frequency is normalized to 1. If you do that, they are measuring the same thing, which is the contrast between the rectangles on that pattern.

The lens quality I was referring to is not CA but that of the optical coatings and lack of glass imperfections. ANSI CR is degraded by reflections and so is MTF.

We are not going to get MTF measurements. It is too difficult and expensive to do. We should get MTF data on the lenses used. The manufacturers have these but it's unlikely that they will release them.

I would love to do some more testing with this Joe Kane Samsung A800B with relatively low ANSI CR (or another one if it is the same way), but I will not be surprised if it doesn't have poor MTF at all frequencies.

But can be to very different degrees. Much like a light colored room is likely to have a much bigger percentage effect on ANSI CR than MTF at full resolution. Just as an example in case that isn't obvious, consider a projector with 500:1 ANSI CR in a room and with a screen that is resulting in a washout of the black boxes of about 1%, giving an ANSI CR off the screen of around 84:1. If that projector was doing 3:1 for a 1x1 checkerboard then that same room would hardly lower that at all (only down to around 2.9:1).

I don't know whether this issue with the A800B is the type that would be like the room or would cause significantly more washout closest to the white objects, but I could see internal projector issues that could cause either way.

--Darin

Darin,

I was thinking more of internal reflections in the projector optical components than room effects. These would cause both a general diffuse effect and local effect close to the light source. In any case, we are all guessing about this since we don't have MTF data.

I was using the room effects as an example of the kind of reflections that can hurt ANSI CR and have little effect on MTF. I suspect that there are things that can happen inside a projector that have way more effect on ANSI CR than MTF, just like the room reflection issue. A 1% reflection of the image that was flipped and offset by say 1/6th of the image would be such a thing.

You are right that I am guessing (especially about this particular A800B), but we do know that whatever the cause of a reflection is, it would have to have way more local effect percentage-wise in order to negatively affect MTF to the same degree that it affected ANSI CR. Just having some local effect wouldn't be enough. I have no doubt that there are some internal projector effects that will result in both poor ANSI CR and poor MTF at full resolution, but I would be surprised if there are no internal projector issues that would result in poor ANSI CR that would not result in poor MTF at full resolution, just as we know there can be internal projector issues that can cause poor MTF at full resolution and not cause low ANSI CR. A projector with bad focus is an example of the latter. I was just measuring a Sharp 20k (actually the South American model) where I got about 700:1 ANSI CR with the irises closed (about 460:1 with the irises open and measuring in a white room with a black velvet board to minimize the room effects), but it has manual focus and I was having trouble getting the pixels sharp at the screen from 20' back. As setup I suspect that would have had poor MTF at full resolution.

BTW: Just changing the lens offset for the RS20 between having the projector centered to the screen and more like out at the screen edge changes the ANSI CR a significant amount from measurements I've done. Maybe ~200:1 ANSI CR to ~300:1 ANSI CR. I doubt it has close to the same percentage affect on MTF at full resolution or even half resolution (a 2x2 checkerboard). Unfortunately, I don't think I have any way to verify that.

EDIT: I should have mentioned that the 700:1 for the Sharp was just for the 4 center boxes.

--Darin

Some shots from Renaissance. This is a fantastic test of ANSI, not just because its black and white, but there are frequently tiny tiny black details encapsulated in white. Its also a brutal test of alignment. It also shows up motion issues quite nicely. These are at just over 30fL.

Because of the extreme contrast of the image, the camera simply cannot beging to convey how much visual impact this film has

As requested,Images 5,7 & 8 have top, bottom and left screen border in view.

1.


2.


3.


4.


5.


6.


7.

8.


9.


10. This scene is unbelievable in real life, the delicate mist in the distance contrasts nicely with the absolute sharpness.

hi Cold
superb
how do you do those shots ? you use a stand i"m sure ?
to make them so precise.
Alan , screenshots

Hi

Would a lumis be overkill for a 120" 2:35 curved screen
To bright or other problems

Coldmachine,

You are doing a top job with the captures. As has been already noted, it would be quite easy to mistake them for straight still images and the fact that these are from movie content makes it all the more impressive.

Sean

Yes, I'm anxious to see more numbers from produciton units on brightness. Worrying about something being TOO bright on a 128 2.35 (110" diag 16x9), is a new problem for me . Makes that DNP 0.8 screen look attractive for my semi-multipurpose room. Too bad it's not AT, I could fit a bigger screen if it were. But a larger curved MP firehawk might still work well. If these brightness figures are right, I could do without the ~1.3 gain. Ash, I think you said you had a curved MP firehawk, is that right?

Never fear viktor!

I am planning on pointing one at a 110" wide 2.40 SMX ProMaskCurv! From what has been said already, there are apparently good options for dealing with excessive lumens.

Sean

Get Gray,
What a world we live in? From not enough light output to too much light output within few months.
Yes, I have a Stewart 2:35 Firehawk (1.25 Gain) 12 foot wide Microperf screen.
The Lumis is perfect for my setup and with a lamp with 25% decay in light I will get 22 to 26 foot lamberts with low lamp and closed Iris.
Remeber, there are ND filters for those who want to dial down the light!
Ash

Hey ash. Yes, too much is good. I've been there and already own the ND to prove it . But my preferece is to go sans glass. It's a good problem for sure.

P.S., if you emailed me, it didn't arrive.

Cheers,
Scott

I'm sorry if I missed it (I scanned through all the pages rapidly), but has there been any information provided relating to the specific Dynamic Black methodology and lamp modulation technology being employed in the Lumis? I've heard that it is not a conventional DI. ...something about a tear-drop shaped opening on a movable reed that slides up and down in front of the aperture just after the lamp? Supposedly this prevents back-scatter in the light engine. There's also something about the lamp driver being coupled to the reed driver increasing or decreasing the amount of lamp output based on the content of each video frame. The claim is that this system is infinitely faster than an analog dynamic iris. I'd like to see some sort of white paper on all this.

It's a pear shaped opening as I understand it, and you have the basic idea. I could not tell there was any DI working - it ws pretty much invisible. Except for the excellent blacks it produced on the screen.

Alan / Coldmachine, thanks for the effort...Even though it is hard to judge from the pictures, but it looks like that the flesh-tone-quality is nearly equal to what you would expect from a xenon-lamp, can you confirm that ?

I had a couple of pjs here lately, but they all had trouble to achieve this very natural "look" of my ruby (even though they where in mayn other discplines much better than the ruby) regarding flesh tones and many have commented on that already regarding the quality of Xenon-Bulbs. So, do you miss Xenon-Light on this one ?

"The claim is that this system is infinitely faster than an analog dynamic iris."

I don't see what the benefit would be over existing iris' that work on a per-frame basis.

A couple of other benefits:
- There is no reduction in ANSI as there is with a DI because there is no backscatter back into the light engine as the iris aperture is reduced.
- There have also been comments that there is no BC and dynamic gamma is not being used (at least those were the comments regarding other DB implementations). This seems really interesting and I hope more information is forthcoming.

I think I see your point that low ANSI will always act to reduce MTF and I agree with it, but not to the exclusion of other factors. At the nyquist limit of alternating single pixel rows the ANSI component will sometimes play a lesser role than other effects such as convergence, CA, etc. This is what I mean:

If we look at MTF as a contrast measurement:
MTF = I(max)-I(min) / [I(max)+I(min)]

What's really of most interest is the I(min) reading which is affected by ANSI contrast, convergence, CA, etc. If for example the convergence of say green is off by 1/2 of a pixel, the MTF will be seriously affected and will likely be more of a first order effect while ANSI contrast will be a second order effect. If on the other hand ANSI is really bad and the other factors are good, the MTF will still be bad.

My guess is this is probably another way of saying that MTF is a cascade of other factors. In other words something like this:
MTF (system) = MTF(ANSI)+MTF(lens sharpness)+MTF(CA)+MTF(convergence)

So if ANSI is bad, MTF will be bad but if ANSI is good it doesn't necessarily mean that MTF will be good.


Unfortunately I think you're right.

Alan Gouger you alluded to this in an earlier comment comparing the Lumis to the RS20 I believe, but how do you like the Lumis using the zoom method for 2.35:1 AR movies? Particularly since any reduced light ouput using the zoom method should not be a problem for you with your "small" screen.

CM those shots you are posting appear to be 2.35:1 as well. How do you like the Lumis using the zoom method?

Also for you both, how is the motorized zoom and focus on these units. Do you have to hunt back and forth to hit the mark or can you dial it right in with the remote?

Finally, in a related topic, does the Lumis have zoom/focus/lens shift memories like the new inexpensive Panasonic 3000 and the not inexpensive Christie HDK-M series incorporate that allow folks to more easily use the zoom method for 16:9 and 2.35:1 AR switching.

What about motorized lens offset?

I'm not convinced of that partially because the CRs that define good and bad ANSI CR and MTF at full resolution are going to be so different. For example, if 100:1 ANSI CR is where somebody says it goes over to being bad, then 2:1 MTF at full resolution is more like where it goes over to being bad. That is, 1% washout effect vs 50% washout effect in order to get to being "bad" in this example. A 100:1 CR during a full resolution MTF test would be incredible, but 100:1 ANSI CR is nothing to be very proud about.

Just consider a reflection inside the projector that results in an inverted image that is exactly the same as the normal projected image except 1/100th as bright. That would limit ANSI CR to 100:1, but would likely have little effect on MTF at full resolution, since 1% washout isn't going to have a huge effect on something that is already at 20% or higher. I have my doubts that the issues that cause the RS20 to have less ANSI CR when centered on the screen than when the lens shift is setup to a screen edge have a large local effect. It wouldn't surprise me if the ANSI CR is lower when the lens shift is all centered, but the MTF would be higher because of less chromatic aberration, better convergence, or something similar. I wonder how MTF is affected by shooting light through a lens at an angle compared to straight on. Straight on seems like it has more potential for reflections to go back into the light engine and make their way out again, but I wouldn't expect that to have the relatively large local effect it would take to hurt MTF by a similar percentage as ANSI CR.

I should say that I am guessing somewhat at the MTF figures here, but when I got to play with a system for measuring MTF with a camera and a scope in Florida with a CRT I think the white to black CR for the MTF would have been about 1.5:1 in some of the cases we measured. That is, the black lines about 60% as bright as the white lines. There may have also been some where they were 30% as bright, but it depended on how well we focused the tube and this was a few years ago, so my memory could be off as far as specific levels. This was for alternating lines of black and "white" at 1080p resolution (I don't recall whether they were vertical or horizontal lines), but I believe just for the green tube (so if convergence was off it would have been lower when using all 3 colors).

--Darin

The Lumis will be no different to any other machine in this respect. If you are asking my opinion of the zoom method, let me know and I'll answer you.


You dont have to hunt back and forth, but that depends on your control system too. I've seen a ton of BS about slow responding motors and overshooting. Frequently the issue is the response time of the IR system. If you use a powerful repeater, or especially if you use RS232, there is no issue. Focus, by its very nature requires a lot of back and forth ( for me anyway), but the motor stops when you ask it to. Simplu using the supplied remote presented no problems other that the normal cone limitations.


No, thankfully, it doesn't offer this facility.

In a unit like the Christie it makes perfect sense and is done to perfection . Its application profile make it a requirement. The cost of doing this correctly is not inconsiderable and requires a very sophisticated focusing arrangement. Offering this facility on a mid priced unit like the Lumis, let alone a budget unit is, in my opinion, a mistake.


Its manual. Similar to above, its mandatory to motorize this feature with a certain type of machine. The typical Lumis install would not require it. Ive have seen it on entry level units, but I believe at that end of the spectrum the money would be better spent on the image.

We know that the Lumis has high ANSI CR and high sequential CR. I would predict high MTF as well, except at some frequencies where other factors dominate. Darin and Mark have strong arguments that this may not be true.

This is a very interesting article that shows how MTF is used to evaluate image quality in a movie application:
http://www.jts2004.org/english/proceedings/Galt.html

I would like to know if you don't mind.