Real time Tube Burn Compensation ?

[Alaric]

Hi There,

I'm looking for a little help in finding / doing something i mentioned on the Theatertek forum and thought the idea may be of interest to people here.

I have 4:3 light image burn on my green tube, it manifests itself as two patches either side on my 16:9 screen that are slightly brighter/lighter than the middle. Two trapizoids that are 6" at the top slanting to 12" at the bottom.

eg :

**** ****
*** ***
** **
* *

A full white/green field shows the fault quite disinctly, but for a dark film you may not even notice.

The above is caused as the inner part of the picture is less bright than the edges, its the centre that is slightly worn and the bright new phosper that is causing the problems.

I have heard of running an inverse bright field and trying to create wear only in the fresh areas of phospher but i haven't heard any major sucesses and its also touted as risky.

Now with all the postprocessing capabilities of the HTPC and a bit of set-up, i think it should be able to tweak the brightness of an image in the Green proportion of an RGB signal by use of some form of static mask (tube burn doesn't change!) and hence compensate for the tube burn.

This method is only like adjusting the brightness a point or two, and it would actualy work best by adjusting the brightness down in the outer edges, which would allow facillitate normal colour calibration of the middle.

The mask would be reasonably simple to make, but would have to be made on a tube by tube basis, but its only a case of getting a full screen image editing package and drawing two lines, filling the area and bluring the line slightly (you need a featherd edge or i guess you'd get a bright or dull line instead of a area).

So, does anyone know of a directshow filter that can do this, or can be modded, or something similar ?

Oh and what do people think ?

Cya,
Lee

[cosaw]

Hi Lee,

I've often toyed with similar ideas in my mind but although the principles may seem simple I really think this is a really complex issue.

I don't beleive (but to be fair I don't know) that "on the fly compensation" will work. By that I mean that even with contrast modulation some considerable colour difference will still be visable. The reason I think this is to do with whats happening to the phosphor. Generally you'll notice that as a tube becomes worn whites will start to become yellow. So you'd think you could compensate on one of the tubes in order to get back to full white in the particular region of the tube that is affected (at the obvious expese of messing up the rest of the tube colours). Essentially you'd be redoing the greyscale for one particular point of the screen just to see if the theory works.

I have a slight ware pattern - so I tried it! This was about 6 months ago now. I played with g2 settings all day to see if a particular setting could return the slightly yellowy worn areas back to full white just to see if the theory worked. There is no way I could get a really clean white back. On the same basis I'm thinking that contrast modulation (on the fly) will not get rid of a visible wear pattern in the immediate term. I think when the phosphor wears there must be a shift so that the phosphor outputs more yellow (overall for the combination of tubes) than virgin phosphor.

Thats not to say that you can't equalise the wear so that the projected colours either become equalised over time or in a short time by driving the tubes harder. I presume you saw Lex's attempt at this? - it was the most succesful I've seen.

I was thinking of a real time tube wear prog via pc. You'd need some sort of
feedback system with a video camera or lens monitoring the current onscreen colours and modulating the contrast to compensate and achieve an even usable wear pattern in a short period of time. So it would be like a mega advanced IRIS style system that could do contrast modulation. My idea would be to base my software round artificial intelligence and something called "fuzzy logic" which can be used as a self compensating self monitoring decision system for altering contrast parametters on the feedback of what the camera sees. Thats the idea on paper. In practice it could be very complicated. However if this could be made to work the power of fuzzy logic is the sort of thing that would make this achievable. My background is Computer Science you see. A long time since I've played with artifical inteligence though - the theory is sound but I don't have the time to invest currently and the system would really be of limited use as far as I can see.

Its an iteresting debate though!

[Alaric]

Hi There,

Maybe this should be re-titled wear compensation as to be honest, burn is far too extream for what i have got. It isn't that bad, just mildly irritating and hence i think a little brightness adjustment would do it.

My shadowing is a slight green push, not yellow, and bringing up a full field green shows why as the outer edges are a touch brighter.

After fiddling with cuts and gains (CalMan+Spyder) my white point is bang on in the centre so there isn't a problem there.

The question is finding some software to do what i want, i don't think the processing would be that bad as its a fixed matrix once its set up.

The webcam idea is curious and i agree tweaking via AI has potential (Also ex-comp grad), but i'd have thought that the web cam resolution over a large screen to be problematic - I have a tracker pod at work with some tracking software that works quite well on a near focus item (2-10M) but for outside (10-30M) it struggles to track even large objects - I think that you'd have problems acuratly watching an 7-10' screen at the detail you'd need.

Cya,
Lee

[cosaw]

The webcam idea is curious and i agree tweaking via AI has potential (Also ex-comp grad), but i'd have thought that the web cam resolution over a large screen to be problematic - I have a tracker pod at work with some tracking software that works quite well on a near focus item (2-10M) but for outside (10-30M) it struggles to track even large objects - I think that you'd have problems acuratly watching an 7-10' screen at the detail you'd need.

Yes this is the real crux of the issue it would seem - like you say getting a camera to track the developments would probably be a tricky and expnsive task. I have know idea what technologies exist. That being said if you could get something to accurately "see" then I'd think you're more than half way there - the software would then come more naturally and theoretically could be far more effective than a static mask. We'll probably never know unless one of us decides to develop the idea further. Funds, time and inclination make this not an option at the moment But its interersting to debate nontheless.

Si

P.S. The other way would be to take a lens off and monitor at tube level. You'd have to find out which one was more feasible - screen level or tube level. Somehow I think tube level could be just as difficult if not moreso.

[Alaric]

Hi There,

Just been thinking, there is a reasonably easy way to test my idea of brightness compensation. Create a mask and then adjust the brightness in that area (on the mask image), if i can get a reasonable match then there is a good chance, otherwise its a failure straight off.

Time to see if the gimp or Paintshop Pro has a full screen version...

Cya,
Lee

[Barcoing Mad]

I've always wanted a decent filter to contrast modulate onscreen logos before they burn the tubes - imagine a water mark to subtract the bloody things. Dscaler logo killer came close (and it took me a couple of years to realise I could run multiple incidences of it).

[Boy Lex]

This is a cop out - but the real solution is obviously to move the pj backwards a touch and just adjust the H and V to within the wear area.

The results of my burn evening brought the sides to the same level as the middle (more or less) but it was impossible to remove the line where the previous raster had ended.

I think you might end up with the same issue - you'll even up the difference with a filter but still see the line.

[Alaric]

Hi There,

Barcoing - FFDshow has something called Logoaway. Not used it as i don't have/use a TV feed. My V is Nov29 2005.

Lex - My Pj is about 50mm off the back wall. I origionaly set the room up for an 801s but came across a BG808. I think its officialy an 1" or so forwards on the Lens progs ideal, though i've read varying theories on PJ distances, with some saying a touch forward is better as you use more phospher, i guess probably not in my case !

Though its all wimping out as a new green tube would be the best option.

Cya,
Lee

[Barcoing Mad]

Going from memory, Dscaler logo killer is better than Logoaway.

Annoyingly, even decent scalers don't seem to have a logo killer facility as an option. Given that various blanking controls and with the C2, pip available sometime, it shouldn't be an issue to implement. Anywat, getting OT....

[cosaw]

Just been thinking, there is a reasonably easy way to test my idea of brightness compensation. Create a mask and then adjust the brightness in that area (on the mask image), if i can get a reasonable match then there is a good chance, otherwise its a failure straight off.

I tried something similar with little success - thats not to say you can't make it work though - good luck in your attempt and let us know if you have success.

The results of my burn evening brought the sides to the same level as the middle (more or less) but it was impossible to remove the line where the previous raster had ended.

I think you might end up with the same issue - you'll even up the difference with a filter but still see the line.

This is my fear also. In theory this is where the computer software and camera feedback system would come in - and it should be far more powerful than what a human can do. As stated before its too much time, effort and expense for me to attempt and seems to have limited use in the real world.

Luckily I'm the only one that sees my wear and I'm happy to live with it as it is rarely visible.

[Boy Lex]

Yeah - I'm very grateful that my 801s has zero wear. It makes it really difficult to find fault with the end result... just my screen warble on pans that winds me up.

Honestly though, I don't think it matters how hard you try and what money you spend (short of a fifth mortgage), you are always going to see something that you know isn't perfect!!!

[cosaw]

just my screen warble on pans that winds me up

Not heard warble used in this context before. Are you talking about the effect caused by a ripple in an uneven screen?

One of the reasons I built a frame canvas style screen was to get it as flat as possible. I was worried a pull down or roll down screen may curl in at the sides distorting the pic. To this end the screen appears perfectly flat on pans. Disadvantage - unless you're gonna shell out on "Screen Goo" then you're talking no gain. Suppose brilliant white matt emmulsion may not even be unity gain but you get a real natural pic viewable from any angle and no hotspotting. Swings and roundabouts as they say.

I do have PC related issues with pans though! Like you say there is always something but I think I've got the best out of a lowly 650Mhz pc.

[Boy Lex]

Originally Posted by cosaw

Not heard warble used in this context before. Are you talking about the effect caused by a ripple in an uneven screen?

Yep exactly.

Originally Posted by cosaw

I do have PC related issues with pans though! Like you say there is always something but I think I've got the best out of a lowly 650Mhz pc.

It's a wind up, 650 MHz should be more than enough to cope with anything you want to do, but software is written so badly that is needs 3.6 Ghz to run properly. Let me guess, you are running windows on it. We still have macs running avid to edit video that do real time effects running on less than 300 Mhz processors!

If you have the time and the patience, you could walk down a linux route. I've just ditched sky plus in favour of mythtv:

http://en.wikipedia.org/wiki/MythTV

It took me ages to get it running properly, but now I can 'set the video' over the net and when the new series of top gear started up, it remembered that I recorded the last series and so automatically recorded it for me even though I was on holiday and didn't know the new series had started!

SORRY! WELL OFF TOPIC!

[Barcoing Mad]

I still have Acorn Risc machines at work....well off topic.

[cosaw]

It's a wind up, 650 MHz should be more than enough to cope with anything you want to do, but software is written so badly that is needs 3.6 Ghz to run properly. Let me guess, you are running windows on it. We still have macs running avid to edit video that do real time effects running on less than 300 Mhz processors!

I couldn't agree more. Its ridiculous. Yep windows! I remember having a 350Mhz Pentium II that would drop frames quite regularly. I thought well if I swap in a 400Mhz version that should clear the frame drops up......oh how wrong I was! Like you say windows is the big problem. That Myth TV looks interesting. Might think about the Linux route its not the first time I've been advised.

I still have Acorn Risc machines at work....well off topic.

Reduced instruction Set Computer?? What the hell is one of those? Nothing so elegant these days! Who'd have thought it would be software that makes things increasingly more difficult and complex! Argghh Bill Gates!!!! Rant continues...etc. etc.