BUG REPORT: How much overscan on the 921?

[John Quaglino]

I DVR'd the test patterns off HDNet the other day. When I view the "resolution" pattern, the commentator states that I should be able to see the number 3 on top and 4 on the sides. With the 921 set to normal and adjusting in my service manual on the Toshiba, I cannot get the horizontal below 10 or so. Vertical overscan can be brought down to an acceptable limit. Does anybody else have this problem? Is it the overscan setting of the 921? By the way this is via component output.

[Mark Lamutt]

I also have taken a look at the HDNet test pattern for overscan, and I don't understand what their numbers and lines mean. I'm almost completely certain that the numbers on their pattern are not percentage overscan like on the Avia pattern.

My television is set to just over 3% overscan on all sides, and the Avia pattern confirms that in 1080i mode. On my set, the HDNet pattern shows either 12 or 14 on the sides, and 10 on the top and bottom in 1080i mode.

Doing direct comparisons of recorded 1080i OTA material from CBS between my 921 and one of my hipix cards, it appears that the 921 introduces a very slight overscan, but not a huge amount - maybe 1% more. Just enough to barely notice it on my 51" screen.

[CAL7]

Mark - Do you have Avia on DVHS? Just wondering if there was a way to use the DVD version, which is all I have, to calibrate 1080i.

My television is a bit old (Mits WS55907, CRT RPTV, ca. late 2000), so my results are most likely limited by its capability. But the overscan I register is:Left 14
Right 13
Top 8
Bottom 6
Also, on the IF YOU CAN READ THIS... test, I can make out Line 8 pretty well and a bit on Line 9.

Just curious, does overscan go away on a DLP or LCD?

[Mark Lamutt]

I have Avia on dvd, but use it at 1080i by playing it on my HTPC which scales it to 1080i. The television then sees a 1080i signal, which puts it into 1080i mode.

BTW, nice kitty picture - my cats also love the top of my monitor. Nice and warm.

[928gt]

I know the 921 doesn't overscan much if any on the DVI Digital output as my plasma showed 0 on top, almost 2 on the bottom, 1-1.5 on the left and 2 on the right (see photo) using the HD Net pattern. Actually I wish it did overscan just a hair more as this would make the aspect modes a little better in some respects.

Dave

[John Quaglino]

I know the 921 doesn't overscan much if any on the DVI Digital output as my plasma showed 0 on top, almost 2 on the bottom, 1-1.5 on the left and 2 on the right (see photo) using the HD Net pattern. Actually I wish it did overscan just a hair more as this would make the aspect modes a little better in some respects.

Dave

928gt,

That is what I was afraid of. What you took a picture of is what the pattern should look like. I think the 921 has major overscan on the horizontal plane.

Mark,

I calibrated my television via Avia and HTPC the same as you. I'm running sub 5% vertical and horizontal. When I display the HD Net pattern I get 13 horizontal and 10 vertical. It is nowhere near what 928gt gets via DVI. I've has a toshiba 4*3 with the vertical squeeze. Just for kicks I turned the squeeze off (now using the full raster) and changed the 921 to 4*3 #1. This yeiled a 16*9 image with significantly less horizontal overscan and virtually no vertical overscan. To me, this demonstrates that the picture material is there but the current normal mode is stretched on all axis. I beleive the normal mode for 16*9 on the 921 has significant overscan issues. It is not using the full 1920*1080 of the picture. I would guess that it introduces an additional 7-10% horizontal overscan. We are probably only getting 1750*950 of the picture via component in normal mode.

Is there anyway you can confirm the overscan setting for component output from the developers? Since most televisions have too much overscan, the 921 should have near zero.

[928gt]

Hi John,

The gross overscan on component is probably my fault ! I have been calling and complaining for a month that some SD material doesn't quite fill my screen (Letterbox programs on "Zoom" aspect mode) and leave about a half inch on each side. I have also complained that some 4x3 programs on CBS-HD (commericals mostly) have a "fuzzy/funky" edge at the very, very top. I think I forgot to mention I was using DVI-D to them and they have been increasingly overscanning the component output in an effort to stop me from calling......

JUST KIDDING !!

Seriously since I have seen those test patterns I now understand why I have the issues I mentioned above. I feel if they increase the DVI-D overscan so I had a 3-4 at all sides I would have less "underscan" issues coming from the slightly varied outputs of the content providers.

Now maybe we can get the 921 guys to reduce the overscan on component and add just a little on the DVI-D output and we will all be happy(er).

Dave

[John Quaglino]

Amen Dave!

[Mark Lamutt]

John, I'll see what I can come up with from the developers. It may be a few days, because they are working right now just about 24x7 getting the next software version ready to go to meet the next Friday deadline. (It was said on the Tech Forum on Monday night that the next 921 public software release would be coming next Friday.)

[jsanders]

Can you reduce the overscan on your TV such that the whole picture appears? Can you make it so you can actually see the zero lines on the horizontal and vertical parts of the screen?

If you do it such that black bars appear around the picture, and you can only see the lines to go three or four, then the 921 is adding overscan. If you can see the whole picture, then the TV is adding the overscan.

Has anyone tried this??

Maybe I'm not totallly understanding the thread yet.

My TV is just as bad as the rest, with way too much horizontal overscan and a bit too much on the verticle with the HDNet test pattern.

[Slordak]

I don't think the DVI output needs any additional overscan either; televisions already seem to have enough, so adding further overscan is probably not helpful (and may in fact be a further hindrance).

I'm using a Sony KF-50WE610 over DVI and there does not appear to be a need for any additional overscan on any axis, at least with the factory default service menu settings. I do agree that the overscan levels in general, for both outputs, should be revisited, but I would not be in favor of just a blanket modification without detailed analysis performed by the development team and beta testers.

[John Quaglino]

Can you reduce the overscan on your TV such that the whole picture appears? Can you make it so you can actually see the zero lines on the horizontal and vertical parts of the screen?

If you do it such that black bars appear around the picture, and you can only see the lines to go three or four, then the 921 is adding overscan. If you can see the whole picture, then the TV is adding the overscan.

Has anyone tried this??

Maybe I'm not totallly understanding the thread yet.

My TV is just as bad as the rest, with way too much horizontal overscan and a bit too much on the verticle with the HDNet test pattern.

I have tried this. I reduced the width and height in the service menu until I had black all around. It still had overscan. That is why I concluded it was the 921 adding overscan. I'll try and post some pictures tonight.

[jsanders]

I have tried this. I reduced the width and height in the service menu until I had black all around. It still had overscan. That is why I concluded it was the 921 adding overscan. I'll try and post some pictures tonight.

ha HA! That is the definitive answer! ( Given that HDNet actually transmits a complete picture :rolleyes: ) We will look forward to seeing those pictures!

[John Quaglino]

ha HA! That is the definitive answer! ( Given that HDNet actually transmits a complete picture :rolleyes: ) We will look forward to seeing those pictures!

Here are some pictures. 1 is 16*9 mode (normal) on my 4*3 unsqueezed. It has major overscan. The total image can be viewed by selecting the very flawed gray bars which compress the picture horizontally and add gray bars as viewed in 2 and 2. This mode should just overlay gray bars over the existing image like the 6000. Notice there is no horizontal overscan. Picture 4 is with the 921 set in 4*3 letterbox mode. Notice there is almost no vertical overscan. Picture 5 is the 921 back in 16*9 Normal with my televisions vertical compression on. Once again there is horrible overscan both vertical and horizontal.

[John Quaglino]

Here are some pictures. 1 is 16*9 mode (normal) on my 4*3 unsqueezed. It has major overscan. The total image can be viewed by selecting the very flawed gray bars which compress the picture horizontally and add gray bars as viewed in 1 and 3. This mode should just overlay gray bars over the existing image like the 6000. Notice there is no horizontal overscan. Picture 4 is with the 921 set in 4*3 letterbox mode. Notice there is almost no vertical overscan. Picture 5 is the 921 back in 16*9 Normal with my televisions vertical compression on. Once again there is horrible overscan both vertical and horizontal.

Here are some more pictures. 1 and 2 are close ups of the 921 in 16*9 Normal. It appears to be running about 8% overscan vertically and even more horizontally.

For picture 3, I went into the service menu on the Toshiba and reduced the width until I was getting no image on both the left and the right side. This wacked the convergence but you can still see with I mean. Here my television is displaying severe underscan revealing the 921 16*9 Normal true overscan setting. The image is hard to read because of the convergence problem but it is running about 8-9% horizontal overscan. For pictures 4 and 5 I reduced my height until the raster was visible. Here you can see that the 921 16*9 mode is running about 4% vertical overscan.

Look again at 1 and 4 from my first post. Notice in 1 how much more of image there is on outer portion of the left and right circles horizontally! Notice in 4 how much more of the circle is visible vertically!

I imagine this could be fixed in a software update. The 921 Normal modes should run 1-2% overscan if any on the component output. All televisions have some degree of overscan. Adding additional overscan serves no purpose. All it does is deprive the viewer of the true image! It like watching HBO's cropovision times 10.

[Mark Lamutt]

Thanks a bunch for the pictures, John. They are going to be very helpful for the developers. Don't expect a fix for next week's software release - there won't be time to get it in. But, for the one after that hopefully.

[John Quaglino]

Thanks a bunch for the pictures, John. They are going to be very helpful for the developers. Don't expect a fix for next week's software release - there won't be time to get it in. But, for the one after that hopefully.

No problem Mark. By the way I was editing when you posted this to add the notes on pictures 1 and 4 from my first post to my second post.

[Richard Chalk]

When I looked at the test pattern on a Samsung 42" Projection set, I thought that the set was way overscanned in Horizontal. I went into the setup, and reduced the width until I could see the raster edges, and the cutoff is about 8-10 on the chart. I have the receiver set now to show the 12-14 area at the edges of the screen, which would be about 3-4% overscan. Oddly enough, the main circle in the center looks pretty good - not stretched much. Makes me think that the Horizontal scan is not very linear, causing stretch at the edges more than the center. The geometry of a projector would do this, but I thought there was compensation in the scanning circuits to correct that. Maybe they don't do enough "slowing" of the scan at the edges.

Vertical scan is fine.

Richard

P. S. My RCA 61" is much better. I don't know how to change the width, but the display shows more than the Sammy...

[SJ HART]

When does HDNET broadcast the test patterns?

[Mark Lamutt]

Either every Tuesday morning or every other Tuesday morning, usually at 8:00am EST.

[jsanders]

This really seems weird to me that the hardware would add overscan to the picture! I've never heard of a video card ever doing that, why would the 921 video do it?

The tests done are conclusive, but not completely. There are a couple things that could be done just to make sure. The first is to verify that the problem isn't with HDNet. A way to do this is to set negative overscan on the TV, such that there is a black region all around the picture. Then, use a receiver other than the 921 to actually watch the HDNet test patterns live, and verify that the picture itself is transmitted with no overscan. That means getting up at 8am on the east coast, or 5am on the west coast and watching with an upgraded 6000 receiver (I don't think the 5000 could do 8PSK). If anyone is game, it would be worthwhile just to verify that the bits being recorded on the 921 have the full picture. I wouldn't put it past a broadcaster to add a bit of overscan themselves, stranger things have happened.

The second one could be to use Video Essentials, run that into the 921 composite input, and let the 921 upconvert the overscan picture to 1080i. This would tell us if the 921 is adding the overscan, but unfortunately, this only tells us if it does this when up-converting and stretching the picture. We know it stretches too much already, right :nono: ..... Probably a moot point, but it might be worth trying..

There is a neat conclusion if all of this is TRUE. If the picture is already a little stretched because more horizontal overscan is being introduced, then this would explain why the "stretch" mode looks pretty bad. It is stretching an already stretched picture, right??

If they are going to have a fix for the aspect mode problems in the next software release, then they may well have it fixed by addressing *this* problem!

We can always hope!

[jcd4878]

I think I can help to disprove that the video broadcasts are missing information. I have a DVI-D connection to a DLP front projector and on the HDNet test pattern with 921 set to 720p/16:9/Normal mode my numbers are 2/2/1/3 left/right/top/bottom. I doubt the broadcast is missing any information as I see the video on my 921 setup.

The most likely problem is that the component outputs of the 921 are adding the additional overscan. The software on the 921 generating the image is obviously correct as the DVI-D output shows me almost edge to edge on all four sides. The RAMDAC chip in the 921 that converts the digital video into the analog outputs is obviously adding a bunch of overscan, but the digital image in the 921's video ram actually does contain the whole image. Since DVI-D bypasses the RAMDAC processing, I am seeing exactlly what is sitting in the 921's video card's output buffers.

I will post detailed pictures of my HDNet patterns in about a week, need to get my camera back from a friend. There is definitely problems with Dishes' idea of what the aspect modes should do. They are totally messed up!

-Jerry

[jsanders]

I think I can help to disprove that the video broadcasts are missing information. I have a DVI-D connection to a DLP front projector and on the HDNet test pattern with 921 set to 720p/16:9/Normal mode my numbers are 2/2/1/3 left/right/top/bottom. I doubt the broadcast is missing any information as I see the video on my 921 setup.


Very interesting theory. It seems to explain why some of the people here don't see as much overscan as others do.

What would be really cool is if you could post your pictures with the DVI, and then with component, to see if we can see the difference! That would explain a lot! :)

[jcd4878]

Very interesting theory. It seems to explain why some of the people here don't see as much overscan as others do.

What would be really cool is if you could post your pictures with the DVI, and then with component, to see if we can see the difference! That would explain a lot! :)

Okay, I finally got around to doing this. There is definitely a problem with the 921's component outputs. The overscan is all messed up. I also noticed a wierd color problem, on the top color bar I get a magenta color bleed all the way across the image to the left and right of the color bars? Anyone else have seen this? Pictures are attached as 1280x960. I apologize for some blurring of the images, couldn't find my tripod and I took the pictures with a somewhat steady hand. Hard to take flashless pictures with a slow shutter and no tripod. You can still make out the overscan pattern allright.

Here is the first group 720p Component

-Jerry

[jcd4878]

720p DVI-D

[jcd4878]

1080i Component

[jcd4878]

1080i DVI-D

[jcd4878]

Oh, one final thing. While doing this experiment I found that indeed on my system that the DVI-D AND the analog componet outputs do actually output a signal at the SAME EXACT time. I do not know why some cannot do this, but I can.

-Jerry

[Mark Lamutt]

Jerry - don't you have those 2 sets of pictures (the 720p ones) labled backwards? The DVI should be showing much less overscan than the component set, at least I would think based on some things that I've been hearing...

[jcd4878]

Jerry - don't you have those 2 sets of pictures (the 720p ones) labled backwards? The DVI should be showing much less overscan than the component set, at least I would think based on some things that I've been hearing...

The board software flipped the order around. I edited the post and kept uploading until it came out right, had to do it like 15 times before it stuck it in the right order. If you look at the filenames, they are correct, its just that the board software wanted to show the stretched one first.

-Jerry