or Connect
AVS › AVS Forum › Display Devices › Digital Hi-End Projectors - \$3,000+ USD MSRP › Thoughts about Projector convergence and effect on resolution
New Posts  All Forums:Forum Nav:

# Thoughts about Projector convergence and effect on resolution

I was thinking about which projector I should upgrade to and thinking about convergence. There is all sorts of opinion about convergence and what is acceptable. I've looked on Sony, JVC and Epson sites to see if they have any published standard, but didn't see it.

Here are my semi-unorganized thoughts. It's using info from LCD monitors and "projecting" it into the projector space (LCoS or otherwise).

An LCD monitor has subpixels. If you have a 1920x1080 LCD monitor, you have 1920*3 by 1080 sub pixels. For each regular pixel, you have 1 Red, 1 Green and 1 Blue sub pixel. They are INHERENTLY not converged. Rather they are arranged next to each other. So an LCD monitor that was only 2x2 pixels (silly monitor, using just for example purpose) would have the following

(R G B) (R G B)
(R G B) (R G B)

Note the parenthesis indicate a pixel and the R,G,B inside represents the sub pixels.

Now let's move to a projector with bad convergence. Say a projector has poor horizontal convergence. A vertical line, when viewed up close, shows 3 vertical lines. One red line next to one green line which is next to one blue line. Red is offset by 1 pixel to the left of green and Blue is offset by 1 pixel to the right of green.

This would give you the behavior that a typical LCD monitor's has with regard to subpixels. The key item here, however, is that the 1080p projector only has 1920 horizontal pixels, not 1920*3 = 5760 like the monitor.

One might conclude that a 1080p projector with convergence as described has only 1920/3 = 640 pixels of effective horizontal resolution!

(I'd venture a guess that can make the exact same argument for poor vertical vertical convergence if arrayed as stated above). Just imagine you rotate a projector with poor vertical convergence 90 degrees and then expect 1080 pixels rather than 1920.)

Just something to discuss... I realize any 3-LCD/LCos projector is prone to some convergence issue, but that doesn't mean we shouldn't understand how it effects the resolution of the image.

Any thoughts on this? I welcome agreement or disagreement.

... Altan

### AVS Top Picks

The difficulty with the LCD display analogy is that the red, green and blue subpixels are spaced appropriately to allow for their inherent misconvergence, which isn't the case with projectors.

If a projector is misaligned by say one pixel in red, then that red pixel is overlapping what should be the adjascent blue and green pixel. Consequently, if the red values are different between the pixel that SHOULD be overlapping and the one that IS overlapping, that pixel colour will be wrong, and the same for each and every instance where that value is different.

So in the extreme example you posted above, the resolution in effect remains, but many of the colours would be incorrect.

Now, if we only turned on every third row in each colour of a projector (alternating) so that R and G and B were lined up side by side, then you'd get a situation similar to an LCD monitor, and the resolution would be as you described.
Video has effectively no detail at the single pixel level as MTF is normally less than 10% which means very blurred. Any detail you see on screen will be made up of several pixels never just one so getting worked up over minor convergence errors is not constructive unless itâ€™s significant enough to be seen from the normal viewing position with video source not a PC generated test pattern. PC text and graphics have nothing in common with video as the PC created content can have 100% MTF at the pixel level compared to 10% or less for video.

Projectors with convergence errors of a full pixel should be considered defective IMHO.
Quote:
Originally Posted by Owen

Projectors with convergence errors of a full pixel should be considered defective IMHO.

Then I should be throwing away all of my past 4 JVCs then! I have never had a projector that has not hit about 1 pixel error at least in one part of the screen. But as discussed before, I think different people measure this in different ways.

If you take a colour as a frame of reference, say green, and red is half a pixel above green and blue is half a pixel below....is that as defective, as green and blue being perfectly conveged and red being one pixel above both of them? The answer is that they are both the same error of convergence but the resulting effect on the image is totally different. The typical tolerance of most manufacturers is 2 pixels. No manufacturer at the price points being discussed here can yet consistently produce 3 panel projectors with errors less than a pixel across the entire image. If you get one, you are lucky!!

Also blue misconvergence doesn't seem to matter anywhere near as much as green and red because our brains seem to pull in detail primarily from the green and red spectrum.
Quote:
Originally Posted by Fat Dave

The difficulty with the LCD display analogy is that the red, green and blue subpixels are spaced appropriately to allow for their inherent misconvergence, which isn't the case with projectors.

If a projector is misaligned by say one pixel in red, then that red pixel is overlapping what should be the adjascent blue and green pixel. Consequently, if the red values are different between the pixel that SHOULD be overlapping and the one that IS overlapping, that pixel colour will be wrong, and the same for each and every instance where that value is different.

So in the extreme example you posted above, the resolution in effect remains, but many of the colours would be incorrect.

Now, if we only turned on every third row in each colour of a projector (alternating) so that R and G and B were lined up side by side, then you'd get a situation similar to an LCD monitor, and the resolution would be as you described.

Indeed. If you display a checkerboard pattern with white and black on every alternating pixel, and intentionally push the convergence out, you will find that it ALWAYS looks like a checkerboard pattern no matter what you do. The colours may be wrong such that you could end up with a red/green pattern instead of a white/black one in parts of the screen, but the resolution is not affected and you still see the pattern no matter what. This of course does have an effect on visible fringing, colour accuracy, and effective contrast if black areas are filled with colours. The latter can effect perceived resolution or make an image look softer in specific instances. But your underlying point that resolution is not affected by misconvergence seems to hold true from my own research into this.

That being said, good convergence is highly desirable and I am not seeing the consistency from projector to projector that I believe should be there by now. I know we are dealing with microns, but we don't seem to have really improved at all in the last 5+ years.
Quote:
Originally Posted by JonStatt

Then I should be throwing away all of my past 4 JVCs then! I have never had a projector that has not hit about 1 pixel error at least in one part of the screen. But as discussed before, I think different people measure this in different ways.

If you take a colour as a frame of reference, say green, and red is half a pixel above green and blue is half a pixel below....is that as defective, as green and blue being perfectly conveged and red being one pixel above both of them? The answer is that they are both the same error of convergence but the resulting effect on the image is totally different. The typical tolerance of most manufacturers is 2 pixels. No manufacturer at the price points being discussed here can yet consistently produce 3 panel projectors with errors less than a pixel across the entire image. If you get one, you are lucky!!

Also blue misconvergence doesn't seem to matter anywhere near as much as green and red because our brains seem to pull in detail primarily from the green and red spectrum.

The eye is most sensitive to green and red so green-red convergence is more important than green-blue or red-blue.

I choose a color as referance, normally green and relate red and blue to it. My JVC is +/- 1/2 a pixel for all colors over the entire screen and my 5 year old Sony is not much worse. I am obviously very lucky.
I also have near-perfect convergence on my RS-45.

My JVC is 0.2 to 0.5 pixels across the screen for the most part, but it does drift a bit depending on temperature. Mainly RED is off 0.1 to 0.3, and blue is off occasionally 0.1 to 0.2. To see the real convergence on a JVC, you need to get the zoom position and the lens shift reset, let the projector warm up for at least 20-30 minutes, and then check various points of the screen. The first day I had my JVC I thought it was off 1+ pixels because of warmup time + lens shift.

The JVC is the first projector I've ever had with this good of convergence.
There are a number of factors other than pixel convergence that's going to make an image have resolution. LCOS designs are inherently less sharp as DLP. It's just in the nature of the panels used. This is a great read and I highly suggest some of you read it. The whole article breaks down the sharpness difference between DILA and DLP on more than just a convergence and optics perspective.

http://www.videovantage.com/?p=819
Yah that is a somewhat older study, it is a good one, and I've read it before.

My JVC RS-45 has just as good convergence or better than the RS-35 they used in that study if judging by the pics, and I feel that for video my JVC is just as sharp as DLP's for the most part. The lower-end JVC's have really gotten improved optics since that study was performed.

However, DLP has a more definitive looking pixel grid that you can make out easier which adds a sharpening effect, and distinct black lines may be slightly quicker to go from all black to all white, but that is not visible to me. The text on my RS-45 looks very DLP-like as well, although some DLP's will look a tiny better on text, but not by much at all. Keep in mind I've owned several DLP's and have one here.
Quote:
Originally Posted by coderguy

I also have near-perfect convergence on my RS-45.

My JVC is 0.2 to 0.5 pixels across the screen for the most part, but it does drift a bit depending on temperature. Mainly RED is off 0.1 to 0.3, and blue is off occasionally 0.1 to 0.2.

To see the real convergence on a JVC, you need to get the zoom position and the lens shift reset, let the projector warm up for at least 20-30 minutes, and then check various points of the screen.

The first day I had my JVC I thought it was off 1+ pixels because of warmup time + lens shift.

The JVC is the first projector I've ever had with this good of convergence.

coderguy,

What type of grid or pattern do you use to test your convergence with?

How do you reset the Zoom Position and Lens Shift?

Can you provide detail as to how I should best test an RS45's convergence?

Thanks!
There is a pattern in the service menu. Also, you can burn the free AVS-HD downloadable test disc (look in the calibration topic, there is a posting about it), you'll see on their web site I think it tells you which ones are for convergence. The Spears and Munsil and DVE disc also have a test for convergence. You can also use a basic grid pattern on an HTPC to see basic convergence issues, although it might not be as exact as some of the more detailed patterns.

The lens shift can be reset in the service menu, but be careful as there are 2 options, one resets the permanent default zero'd posiiton (and you don't want to do this). Without going into the service menu, you can use a piece of paper method or white cardboard box on a table in front of the lens to line the lens back up to have 0 lens shift.

Zoom position, well that's easy to do and doesn't have to be as perfect as lens shift, just zoom all the way in, take note of convergence, zoom halfway and all the way out. Generally around mid-zoom to farthest throw (somewhere in there probably has the best convergence, but it varies on every unit by a tad). The zoom isn't as much a factor on the JVC as the lens shift.

Also make sure the projector is focused very nearly perfectly, because an out-of-focus image can invalidate the convergence test.
Quote:
Originally Posted by coderguy

There is a pattern in the service menu. Also, you can burn the free AVS-HD downloadable test disc (look in the calibration topic, there is a posting about it), you'll see on their web site I think it tells you which ones are for convergence. The Spears and Munsil and DVE disc also have a test for convergence. You can also use a basic grid pattern on an HTPC to see basic convergence issues, although it might not be as exact as some of the more detailed patterns.

The lens shift can be reset in the service menu, but be careful as there are 2 options, one resets the permanent default zero'd posiiton (and you don't want to do this). Without going into the service menu, you can use a piece of paper method or white cardboard box on a table in front of the lens to line the lens back up to have 0 lens shift.

Zoom position, well that's easy to do and doesn't have to be as perfect as lens shift, just zoom all the way in, take note of convergence, zoom halfway and all the way out. Generally around mid-zoom to farthest throw (somewhere in there probably has the best convergence, but it varies on every unit by a tad). The zoom isn't as much a factor on the JVC as the lens shift.

Also make sure the projector is focused very nearly perfectly, because an out-of-focus image can invalidate the convergence test.

Thanks for the info and hope you have a GREAT New Year!

...Glenn
Quote:
Originally Posted by JonStatt

Also blue misconvergence doesn't seem to matter anywhere near as much as green and red because our brains seem to pull in detail primarily from the green and red spectrum.

Quote:
Originally Posted by Owen

The eye is most sensitive to green and red so green-red convergence is more important than green-blue or red-blue.

Human eyes have around 30 green and red receptors per 1 blue receptor in linear dimensions, 1 blue to several hundred green and red in area terms. Typically blue is out of focus in the eye by anything up to 1.5 diopter. Human vision 'in-fills' perception of blue.
For comparison if red and green are out of focus by 0.25 diopter or more you would probably benefit from glasses. Glasses are typically prescribed in 1/4 diopter increments, occasionally 1/8 diopter increments.

You also have the issue of chromatic abberation. Equal distant red/green objects and blue objects can not be simultaneously in focus on the retina due to the chromatic abberation introduced by the eyes lens. This is one possilble explaination for the phenomenon of warmer colors appearing nearer in depth to viewers than cooler colors when looking at a flat picure.

Virtually the entire incoming data to the eyes are immediately transformed into sets of local difference signals, within the retina prior to transmission along the optic nerve to the central cortex. These are created by a variety of concentric 2D groupings nested together. It is believed there are essentially four sizes of these concentric nested R.F.'s, with the surrounds of each effectively providing the centers of the next scale up

1) Highest resolution R.F.s involve the inhibition of individual R or G cone outputs by a ring of mixed (R+G) cone outputs.
2) Larger R.F.s are then known to be created via amacrine cells & diffuse bipolars from 2D groups of R, G or (R+G) .
3) The groups of R & G lead to lower resolution opponencies directly between R & G
4) The grouped Y outputs are believed to interact with the known very sparse B receptors which intercept a very out-of-focus blue incident light. This latter leads to a very low resolution B vs Y opponency
Quote:
Originally Posted by Owen

Video has effectively no detail at the single pixel level as MTF is normally less than 10% which means very blurred. Any detail you see on screen will be made up of several pixels never just one so getting worked up over minor convergence errors is not constructive unless it's significant enough to be seen from the normal viewing position with video source not a PC generated test pattern. PC text and graphics have nothing in common with video as the PC created content can have 100% MTF at the pixel level compared to 10% or less for video.

I think the Kell factor also limits picture detail size to larger than one pixel, as having picture detail size down to one pixel would reduce picture quality.
Quote:
Originally Posted by Owen

Video has effectively no detail at the single pixel level as MTF is normally less than 10% which means very blurred. Any detail you see on screen will be made up of several pixels never just one so getting worked up over minor convergence errors is not constructive unless it's significant enough to be seen from the normal viewing position with video source not a PC generated test pattern. PC text and graphics have nothing in common with video as the PC created content can have 100% MTF at the pixel level compared to 10% or less for video.

Projectors with convergence errors of a full pixel should be considered defective IMHO.

Agreed.

Quote:
Originally Posted by JonStatt

Then I should be throwing away all of my past 4 JVCs then! I have never had a projector that has not hit about 1 pixel error at least in one part of the screen. But as discussed before, I think different people measure this in different ways.

If you take a colour as a frame of reference, say green, and red is half a pixel above green and blue is half a pixel below....is that as defective, as green and blue being perfectly conveged and red being one pixel above both of them? The answer is that they are both the same error of convergence but the resulting effect on the image is totally different. The typical tolerance of most manufacturers is 2 pixels. No manufacturer at the price points being discussed here can yet consistently produce 3 panel projectors with errors less than a pixel across the entire image. If you get one, you are lucky!!

Also blue misconvergence doesn't seem to matter anywhere near as much as green and red because our brains seem to pull in detail primarily from the green and red spectrum.

Having a CRT I have seen over one pixel misconverge and perfect converge. The difference is definitely noticeable. If over one pixel is deemed acceptable, then I may never buy a three chip pj.
Quote:
Originally Posted by Ericglo

Agreed.

Having a CRT I have seen over one pixel misconverge and perfect converge. The difference is definitely noticeable. If over one pixel is deemed acceptable, then I may never buy a three chip pj.

Is that a CRT that can actually display 1080 truly indepedent pixel lines? I have seen a badly converged CRT but it wasnt truly an HD projector and so the dots were much larger to start with. As 1 pixel was effectively a much a larger delta in physical terms, it was more noticeable. If we imagine how many microns of error results in 1 pixel error on an LCD or LCoS projector with true 1080p resolution...once we get to 4k projectors, the "effective" pixel error will double if the manufacturing tolerances don't improve. But whether you could actually "see" the error from seated once the pixels are so small is another matter altogether.
Yes, I have seen multiple CRTs.

As for 4k, I could see it not being as big of a deal for 1080p, but could soften native 4k.
Quote:
Originally Posted by dovercat

I think the Kell factor also limits picture detail size to larger than one pixel, as having picture detail size down to one pixel would reduce picture quality.

Kell and digital sampling limitations have much in common, effective resolution is limited to about 75% of the pixel/line count.

Digital video is low pass filtered to remove single pixel detail as much as possible to avoid aliasing and interline shimmer.

What also seems to be forgotten in discussions like this is that video is encoded with 4:2:0 color which means color is presented in averaged 4 pixel blocks.

At the pixel level video is heavily distorted, very blurred (low MTF) and has color averaged into blocks, why get all worked up about convergence errors that are not visible from the normal viewing distance when the source is such a mess?
Quote:
Originally Posted by coderguy

Yah that is a somewhat older study, it is a good one, and I've read it before.

My JVC RS-45 has just as good convergence or better than the RS-35 they used in that study if judging by the pics, and I feel that for video my JVC is just as sharp as DLP's for the most part. The lower-end JVC's have really gotten improved optics since that study was performed.

However, DLP has a more definitive looking pixel grid that you can make out easier which adds a sharpening effect, and distinct black lines may be slightly quicker to go from all black to all white, but that is not visible to me. The text on my RS-45 looks very DLP-like as well, although some DLP's will look a tiny better on text, but not by much at all. Keep in mind I've owned several DLP's and have one here.

Would it be fare to assume the RS55 would have the same ability for the good convergence your RS45 has, or are there differences in the RS55 (e-shift...) that could have a negative, may-be positive, impact?
Haven't seen e-shift myself, so I cannot say, but you can turn e-shift off, so I would think a well-converged RS-55 should look just as good or better.
Quote:
Originally Posted by coderguy

Haven't seen e-shift myself, so I cannot say, but you can turn e-shift off, so I would think a well-converged RS-55 should look just as good or better.

e-shift effectively thickens the pixels when you review a one pixel grid, so the convergence actually appears to improve.
An old discussion on convergence I thought I'd insert in here to make finding it much easier for someone else.