JVC RS1/HD1 CC filter tweak - Page 4

This surprises me a little, not that you feel that this is not a good idea but because you, more than most, understand the concept of compromise to achieve the greater good. All else being equal, sure, sacrificing color makes little sense. Sacrificing color to obtain appropriate lumens on the screen to within standards which can help reduce the visibility of noise many see with this PJ to increase depth and 3D effect and things aren't quite so cut and dry, add in a couple thousand free points to the CR and things become even more less cut and dry should there be color or other compromise. Conversely would you calibrate a customers display to achieve that last few % for perfect color if that meant possible eye-burning lumens and a 25% cut in CR? Some would find that acceptable and some not so much. At this point I don't see where "recalibration of the gray scale" is an issue as it appears anyone experimenting has equipment.

Assuming my equipment and methods reasonably reflect reality I believe I have shown that at least with the filter I experimented with there was no sacrifice in color to obtain increased CR and a decrease in lumens to better levels in my setup. There was one negative which may be associated fully to the cheap filter I used caused increased obvious optical distortion with increased use. Again, assuming my measurements reflects reality, using this filter appears to little different than using a ND filter to cut brightness and seldom have I've read of any negative side-effects using a good quality filter for this purpose. I can only surmise that using a decent quality CC filter will reduce the negative result of using a filter. I've also learned that there indeed can be optical artifacts which at this point I can only attribute the specific (not glass) filter I used. Each end user has his/her own set of "meaningful" priorities for which they feel will offer to most life-like image factored on many variables, one being color accuracy.

Quite honestly I normally balk at putting anything additional in front of the lens except the screen and began this experiment primarily to provide myself with another project to keep myself occupied with the possible off-shoot that something could be learned (+ or -) from it. What I've learned so far is that, if you're a little lucky, using a CC filter tool to calibrate RGB balance and bring down lumens might yield surprisingly good results by maintaining CR close to torch mode levels and increase absolute black levels while doing so, and I neglected to mention that with this PJ, the reduction in absolute black level also decreases the visibility of bright corners should that be an issue with present setup. How much a PJ owner is willing to compromise to trans-morph to a seemingly better set of problems is anyone guess.

I look forward to trying the Hoya 85a filter I have coming in but it's on back-order till May 11. I also look forward to anbjornks color fidelity measurements.

Worth the Trouble?? Most here spend tens if not hundreds of hours tapping on keyboards often about nothing at all, what's an hour setting up a filter.

The 25% lumen drop is with using an 85A filter which is very aggressive, the 85C is a much paler orange which should have far less lumen drop.

I've been reading the same thing about a reduction with ANSI CR which makes perfect sense and I've been meaning to measure it but feel I don't have the gear to do it properly (seems like spot meter would be ideal for this measurement). However I just put on a 16 square ANSI test pattern and measured with the meter facing the PJ lens about 1/4 distance from screen to projector with sensor centered in one of the squares. Typcial ansi tests require several measurements from different screen locations but this will have to suffice for now. I used the PJ's vertical flip option to switch the square from black or white.

ANSI CR no filter = 299
ANSI with filter= 276

An approx 8% reduction in ANSI CR with filter installed.

Just for conversion sake my first ANSI measurement with filter installed spit out a CR of 232 but I've doing a little work in the HT which I anticipated would create a little dust so put on the lens cap on the PJ but left the filter uncovered. After noting the lowish ANSI I checked the filter and there was a slight coating of dust on it which I cleaned off most with some air after which I attained the 271 measurement. Keep your room and optics as dust free as possible, even a small amount of dust is an ANSI killer!

I ran the test again after giving both the lens and filter a blast of air which increased the numbers slightly which I've edited my previous post with. I'm a little surprised, I thought the ANSI with filter would be less than that. Yes these measurements are with the Cokin 85A filter which I believe is the most aggressive orange filter of this series, I have a large Hoya 85C on order which I anticipate will have less than half the lumens cut than the 85A but with a slightly different green blue filtering ratio which may or may not work as well for my situation which could alone impact CR measurements. I should have the Hoya middle of May and will post some measurements when I do.

You are correct the reduction in light output with this particular filter in place is quite noticeable but there are many instances where this can be beneficial. You are implying that maximizing or maintaining lumen output is the only option for all installations, any methods employed to reduce lumen output therefore is undesirable. Using filters to cut lumen output is fairly standard and accepted practice, the benefits can outweight the negatives.

I'll state again that I've found a few positives using the filter:

1- Increase CR of 20 %. whether it's visible is debatable but I'd could never imagine an instance where an owner or calibrator would throw away CR by 20% if it could be avoided and just say you can't see it anyway.

2- Drop in lumens, same effect as employing an ND filter. Some situations could welcome this.

3-Decrease visibility of bright corners.

This thread was started to provide a few measurements in an attempt to quantify the results of using the filters as opposed to relying on intuition of their effects. In the process I have noted some subjective positives and a couple negatives in my setup which may or may not be Cokin filter specific. If you wish to dismiss what I, at this point, subjectively feel may be an improvement that's fine, hopefully whatever measurements are taken by myself and others which may be helpful to those interested. I note from your site that you are ISF certified, have you first-hand witnessed and/or measured the negative effects of using filters?

Tweaking for tweaking's sake....never!

1) I'm in Canada and the Canadian online dealers supplier was out of the large 77mm filters. I suspect USA purchases would be quickly supplied. As usual I paid too much, good filter price of about $51 CAD plus crazy shipping and crazier tax. A fact of life living in Canada.

2) I posted a couple CIE charts on the first page of this thread of what you're looking for. Are those what you're looking for?

3) Anything which reduces clarity or sense of focus drives me nuts so I was expecting this to be a problem. Looking at the on screen menus I see no changes whatsoever, the text is still razor sharp. Watching movies I at times feel there may be loss of optical resolution, other times I'm convinced there is not. Without matching brightness levels and eliminating personal bias it may be difficult to ascertain for sure but I'm leaning towards the "no loss" camp. (this is when the filter is not suffering the effects of heat)

4) In post #14 I have a photo of the temporary filter and holder, this hangs in front of the lens via a piece of tape. The filter holders circumferance is the same size as the PJ lens. The new Hoya filteron order is 77MM and I measured the PJ shroud at about 78MM, my plan is to use a very small dab of tac "n" stick on the filter threads and push in in place.

5) I haven't noticed any drift from the Cokin filter but there does seem to be a amount of color variation across the screen and in the corners. This could be fault of the filter but more likely non-uniformity of the panels.

I apologize to MikeSRC, I sorta forgot about posting that, I ran the test last week but there's nothing of value to report, no compromise could be found. The color control does not function in a linear manner or equally for saturation and brightness. The primary color locations move relatively little for the first -14 or so clicks then really begin moving between that and -30. On the other hand color brightness is so sucked out by -15 where the primary movement gets ramped up that any sort of middle ground is impossible.

If there were more primary movement between -4 and -8 then perhaps a good compromise could have been reached but such was not the case.

Just wish to clarify that when I introduced the filter and took measurements the color control was not manipulated in any way from the pre-filter measures. The only adjustments were to the RGB controls to get gray back to D65 or thereabouts. Any changes in color lightness is do to either the filter, RBG controls, equipment tolerances or UBD error, but definitely not the color control.

Here are my most recent numbers again, posted in the previous page.

Filtered first:

___dE(u'v')__dL____dC____dH

R__15.6____1.5___15.3___2.7
G__28.0____-1.3___28.0___0.0
B__4.2_____-1.7___2.3___-3.0
Y__20.4____0.0____20.4___0.7
C__15.6____-1.0___15.5___0.6
M__6.3_____0.5_____5.8__-2.4

No Filter

R__19.3____2.5____18.9____2.7
G__30.6____-1.3____30.5____2.5
B__4.3_____-3.3____-2.3____-1.6
Y__20.7____0.3____20.7____-0.2
C__16.5____-1.7___16.4_____0.6
M__8.9______1.0____8.1_____3.7

No problem. Positive numbers denote impromement with filter, negative numbers a worsening with filter.

___ddE(u'v')_ddL__ddC___ddH

R__3.7______1.0__3.6_____0
G__2.6______0____2.5____2.5
B__0.1______1.6___0____-1.4
Y__0.3______0.3___0.3___-0.5
C__0.9______0.7___0.9____0
M__2.6______0.5___2.3___1.3

I've been reluctant to make any claims or proclamations with respect to my measurement result without corroborating measures so in an effort to appease my own doubts of accuracy and relevance or, at the very least, consistency of my measures I ran a complete new set of and ran them through Krasmuziks spreadsheet. Keep in mind that it's been two full weeks since I ran the measures posted just above, my tripod/sensor is in a slightly different location.

I ran 100% windows and calibrated the 100% white window pre-filter to dE .26 to D65, and managed to get 100% white with filter to dE .14 to D65. I have not touched the color control and didn't even check to see if all was still ok.


Filtered first

__dE (u'v')_dL____dC_____dH
R___17.6___2.1___17.3___2.7
G___28.5___-1.1___28.4___0.4
B___5.8___-1.0____4.6____-3.4
Y___20.8___0.2____20.2___-4.8
C___16.4 ___0.8___16.0___3.1
M___7.7____1.4___7.5____-0.9

No filter

R__20.7___2.9___20.4____2.7
G__31.1___-1.2___30.9___2.9
B___3.7___-3.0___ -0.7___-2.1
Y___20.5___0.4___ 20.5___-1.4
C___17.3___-1.6___17.1___1.8
M___10.6 ___.6___ 8.6_____6.0

And the "delta dE/dC/dL/dH to show the difference the filter created"

___ddE(u'v')_ddL__ddC___ddH

R__3.1______.8___3.1_____0
G__2.6______0.1__2.5____2.5
B__-2.1_____2.0__-3.9____-1.3
Y__-.3______0.2___0.3___-3.4
C__.9______0.8___1.1____1.3
M__2.9____-0.8___1.1____5.1

With the exception the barely perceptible changes (about 3 dE) of yellow/magenta the relative dE changes from the filter are very consistent compared to my previous results Essentially, the end result of using the Cokin 85A filter with my RS1 has measureably improved on average color chroma, lightness and hue. Any measured change ultimately is under delta of 3 which for most intent and purposes is imperceptable to the "Hughman" experience even in side by side comparison.

I think I found my new moniker.

At some point during this thread I stated I'd attempt to find a compromise using the color control which yields the least average delta error between color brightness, color saturation and color hue. I did some quick testing last week focussing primarily on the chroma movement of gree and reported no comprmise could be found but this was kind of a knee jerk reaction to just viewing the numbers in Krasmuziks spreadsheet accompanied with viewing the shifts within the CIE charts but I did not actually average all the results.

I felt I short changed a few including myself so I again ran the numbers tonight but calculated the averge delta error to determine what I originally indicated I would. I measured the xyY values while reducing the color control from -2, -4, -6, -8, -10, -15, and calculated the average error of lightness, chroma and hue for each position to determine which offered the best compromise. I re-established 100% white level before each run as lumens increased slightly throughout the test. To calculate average I performed root mean square (RMS) calcs from the dEu'v' which has accounted for the weighting assigned to LCH as they combined relate to perception.

Here are the RMS results, the first at -2 which is the proper color calibation using DVE filter and Getgray disk. on my setup which does give the best overall lightness error.

-2, RMS dE= 17.9
-4, RMS dE= 16.4
-6, RMS dE= 15.4
-8, RMS dE= 14.8
-10 RMS dE= 15.4
-15 RMS dE=16.0

I'll also submit a straight average of dEu'v' for each color position:

-2 average dE=16.1
-4 average dE=14.2
-6 average dE=12.4
-8 average dE=12.9
-10 average dE=11.2
-15 average dE=15.3

Utilizing the LCH model which takes into account color brightness and it's perceptual effect on saturation it's stands to reason that the goal would be to minimize the average delta error for primary and secondary colors lightness, chroma and hue. With that goal in mind setting the color control between -8 and -10 appears to the sweet spot in my setup and accounts for the lowest RMS and average delta error taking all the above factors into account.

I'm quite surprised at these results, assuming the measurements and calculations are accurate then using the LCH model to calibrate overall color accuracy shows the best perceived result could be anywhere between -4 and -15 on the color control. The presumably filter calibrated result of color control at -2 would appear to very setting not to use if the intent is to maximize overall color accuracy.