CCD question

[Robert_Niemann]

May somebody explain to me, what the differences are between an interlaced and a progressive working CCD? I know, what interlaced and progressive means, but not according to a CCD. Can an interlaced working CCD not also work as a progressive one and the other way round?

[Anhar Miah]

Hm... read this website, may answer your Question


http://www.shortcourses.com/how/sensors/sensors.htm


Anhar

[Antoine_Fabi]

Wow !

very instructive article.

thanks Anhar

[Loki]

In Reality.. all CCD's are "technically" and inherently progressive by nature. I don't think it is physically possible to make something interlaced... unless my info on "matter" and "3D spaces" is lacking. Interlacing is a matter of interpretation.

The interlaced vs. progressive thing happens when the CCD is read and interpreted. Interpreted being the keyword here. If the CCD is designed to be read in alternating lines then it is an interlaced system. If it reads all the photosites in order then it is progressive.

All CCD's have photosites arranged in an Array. Since there is no such thing as "Interlacing" in the "real world" you can technically call all CCD's progressive.

Canon's 24F solution seems like it could make some very nice pictures from the interlaced chip. I would wager that since they are interpolating at such high resolution that the distinction between 24p and 24F will be small.. but thats another topic..

[jaegersing]

Canon's 24F solution seems like it could make some very nice pictures from the interlaced chip. I would wager that since they are interpolating at such high resolution that the distinction between 24p and 24F will be small.. but thats another topic..

Interpolating? Loki, where did you see that 24F involves interpolation? (Can you provide a link please?) This would mean that all the claims about it being "identical" to 24p are junk.

Richard Hunter

[braw]

Here's what I don't get. If an interlaced sensor scans one field (every other line) and then the next, and a progressive one scans every line one after the other, how are they achieving a progressive look?

If they scan each field and then combine them, what's the difference between that and deinterlacing? If they have a software method of sampling each line like a progressive chip but are using an interlaced chip, why can't they just call it progressive? So what if the chip was designed for interlaced cameras. If it is sampling the CCD in the same way, just using a chip that wasn't designed for it, it should be progressive.

What concerns me is that if it is taking one field at a time and combining them, there should be a slight difference in the image it captures...even if it's incredibly subtle. Like when you deinterlace footage, any major movement shows slight differences in each field even though it's a high frame rate.

Anyone?

[Neil Rowe]

..someone.. somewhere here posted something about a canon rep saying that the fields are captured at a very very short time interval differance so you cant really tell after their software does its magic.. or something like that.

sorry to be so vague.. but the actuall post is here somewhere if anyone is so inclined to find it and quote it for us.

[braw]

Well, I read that it samples at 24fps, is that correct? Like the DVX, just adds pulldown to embed it in the 60i timeline.

Maybe that's not accurate, but even at 60fps, each field will have subtle time differences and with big or fast movements are pretty noticeable when looking at each field. If it's true that it's sampling any slower than 60fps, then fast movements will have even greater differences in each field.

[Loki]

Interpolating? Loki, where did you see that 24F involves interpolation? (Can you provide a link please?) This would mean that all the claims about it being "identical" to 24p are junk.

Richard Hunter

This has been mentioned in articles.. and is really just common sense... you cannot combine two interlaced frames together without some form of interpolation. Unless of course you are merely just de-interlacing and throwing away half the data. But even then there is interpolation to re-create the missing data.

2 fields combined into one progressive frame are always going to be interpolated.

According to Canon, they are using their 8 years of experience in converting video from an interlaced chip to progressive-looking video to achieve the effect
- www.camcorderinfo.com

[mgalvan]

..someone.. somewhere here posted something about a canon rep saying that the fields are captured at a very very short time interval differance so you cant really tell after their software does its magic.. or something like that.

sorry to be so vague.. but the actuall post is here somewhere if anyone is so inclined to find it and quote it for us.


That'd be me :laugh:

http://www.dvxuser.com/V3/showthread.php?t=34056

[Anhar Miah]

Loki, your are half correct but you forgot the fourth dimension which is time (relative of course) thus an interlace camera when capturing an image does capture (when the subject is in motion) at two different "moments in time", where as in a progressive CCD all the information is captured at the same "moment in time" time BUT may be read out line by line, therein lies the difference, anyway I'm no CCD expert but thats my understading.

@ iamaloser:

The F900 has its own method to obtain the 24P, But for now we do not know if the Canon does the same if so then it will be true progressive.

[mgalvan]

I spoke with Canon at the ResFest event in NYC (which they also had a full working model of the XLH1 which I played around with for a while). Their explanation of 24F is indeed it should be identical to 24P ... all the way down to the workflow. The CCDs are still interlaced and initially capture the fooatge that way, but are then outputted at 1/24 per second. The fields are combined before output at the 24 frame rate. The Canon rep said that the fields are so close together, that when combined and outputted at 24fps, it looks exactly like 24p. They said that the process is very much like the frame mode from the XL1s, only "8 years in design later. It has basically been perfected." They said they will be releasing footage in the near future to prove this.

They say 24F is so identical to 24P that even in the SD camera mode, it is layed down to tape exactly like the XL2 with options of 2:3 and 2:3:3:2. Hope this helps.

[Walter_Graff]

Take a look at Sony's 24PsF and you will get an idea of how something can be progressive or interlace sort of at the same time. In Sony's case a progressive CCD captures the picture then it is broken into two separate segments after capture (effectively interlaced) so that later yo have the choice of recombining them or keeping them interlaced. Canon does it slightly different. They take the segments out at the chip, but down the line combine them. The key for Canon is their DSP which is a second generation of a version they use with their high end still cameras that is able to seamlessly combine elements from a CCD to make as good a picture as if you started out making the picture another way. It's all just different ways to achieve the same effect. All called different things, 24F, 24PsF, and 24P but the end result identical.

[Loki]

Loki, your are half correct but you forgot the fourth dimension which is time (relative of course) thus an interlace camera when capturing an image does capture (when the subject is in motion) at two different "moments in time", where as in a progressive CCD all the information is captured at the same "moment in time" time BUT may be read out line by line, therein lies the difference, anyway I'm no CCD expert but thats my understading.

@ iamaloser:

The F900 has its own method to obtain the 24P, But for now we do not know if the Canon does the same if so then it will be true progressive.


This is correct indeed. Time is a factor, but time is a factor in the interpretation of the CCD. So while you are correct my original statement can still stand on it's own.

[braw]

So I'm not sure this was addressed. A progressive CCD captures the whole image at one point in time. Progressive captures two fields at two different moments in time. Granted they are very close together, but not the same instant. So as Loki says, there has to be some sort of interpolation that "motion blurs" the difference in the two fields, or you would get interlaced stepping.

So I do get the sales pitch that it looks exactly like 24p, but what I'm looking for is technically how it's doing this. I can only imagine that if it does have to do some sort of "magic" to blend fields that have big differences, I would imagine you might lose some perceived resolution. I'm sure not a very large amount, but some. Also, if it's true that you loose some resolution in HDV when there's a lot of movement, there's a double whammy.

I'm seriously not trying to spread FUD here or dog this camera, I just want someone to either acknowledge that some interpolation has to occur, or tell me how it doesn't.

[Anhar Miah]

@ Loki, basically we sayin the same thing in a round about way lol :)

[Robert_Niemann]

1. Why has Canon not built in progressively working CCDs? Are these CCDs more expensive, or what?

2. I hope, that the XL H1 does not interpolate. Maybe the two interlaced pictures are recorded so close to each other, that you would not see any edges. Then there could be the full resolution of 1080 lines for the frame modes. I hope so!

[mgalvan]

I asked that at Resfest, and it has something to do with them wanting 1080 resolution but needing the highest pixel count on the CCDs for HD-SDI output. They said 1080p chips were out of the question because of the sheer price for them.

[jdsmith19]

This is from the DV Magazine website, DV.com--if you have a membership (it's free), you can check out their frontpage article on the XL H1. But here is the answer to the interpolation question.

Quoted from DV.com:

The XL H1 shoots 60i and two Frame modes of 30f and 24f. The Frame mode is similar to early Canon digital camcorders--it's not technically a progressive frame because the CCDs aren't progressive, but fields are interpolated and reinterlaced in Canon's proprietary method to allow the "f" frames to behave just like "p" frames during postproduction. Canon representatives said their method would work much better than other manufacturers' attempts at 24p-like HDV (read Sony).

J.D.

[Robert_Niemann]

Really sad news, but thanks! Now I am hoping of Panasonic and its HVX200 (1080p), but I guess, that there will be no native 1080 because of the price. You just cannot have everything...

[braw]

Thanks jdsmith19. That's what I thought was going on.

[Robert_Niemann]

I have just found this at www.hdforindies.com: "[...] This is even more interesting after I talked to the canon rep, who said that canon really wanted to go 1080p, but a chip that could actually do 1080p would have made the camera cost $100,000, and they had no choice but to make the compromise. [...]"

[Barry_Green]

I know curiousity is a powerful motivator, and everyone's curious about how the Canon will perform, but please, let's keep something in context here: nobody, outside of a couple of Canon engineers, knows how 24F actually works. And until someone like Adam Wilt etc. gets a camera for testing, nobody is going to be able to answer the questions being asked.

Within five minutes of testing we'll be able to answer definitively as to a) whether it retains full resolution, and b) whether it delivers filmlike motion. But until that testing can take place, there's nothing people can learn definitively about it. So, I guess I'm saying "patience, please", but nobody should rule out anything about the cam until it's been given a chance to prove how it works.

There was a lot of similar energy expended about CineFrame 24 on the Sony, with some people convinced (absolutely convinced) that the Z1 would offer a new way to do it, with its CCD clocked at 48hz, etc. In the end it was much ado about nothing, as the camera offered absolutely nothing different than the existing FX1 when it came to CF24 mode. So engaging in the discussions may be fun, but don't go basing any opinions or making any decisions until actual facts are revealed.

The truth will be learned about the XL H1's 24F mode, but not before the camera is in the hands of people who know how to test it.

[Antoine_Fabi]

I've always asked myself how the DVX's CCDs were able to output progessive AND interlaced signals. The DVX and the XL2 do that, no ?

I always felt that the DVX CCDs were in fact native interlaced, and progressive was made using intensive processing.
If you carefully watch the color noise in progressive mode, you'll see that with a shutter speed of 48, you'll see that there the red color do not cover the whole surface of a small objet, there is a lot of noise.
With the shutter speed at 24, the color seems less noisy, in a static shot.
...In interlaced mode, no such a problem, the noise level is much better.

Am i right ?

[Robert_Niemann]

Antoines message leads me to the question: Why can progressive CCDs output progressive AND interlaced signals, while interlaced CCDs cannot output progressive signals?

[Walter_Graff]

It's an inherent problem to get more noise with progressive outputs more than interlaced.

"Why can progressive CCDs output progressive AND interlaced signals, while interlaced CCDs cannot output progressive signals?"

Easy to make an interlace output on a chip than the other way around.

[Antoine_Fabi]

but...why ?

[Walter_Graff]

CCDs are very difficult to manufacturer. In Fact the yield is not 100%. The CCD in your camera is not perfect either. Depending on the market (consumer, prosumer, professional) allowable defects are acceptable. Inside a CCD are lots of little switches (gates) and all sorts of electronics that gets the signals out, making them require a 100% clean room to make and even then problems arise both due to contamination and the gates not working properly, etc. make for a very difficult device to produce. And the bigger the device the harder it is to make. That is one reason why yo will not se 2/3 inch CCDs in prosumer cameras. So now imagine that little device with a size of your pinky nail. You can only do so much with it. ANd its better to design a CCD for an application that just to make one that does everything. So CCDs with registers that are designed specifically for interlace are made just for that. But one of the easier things to do is to use a progressive chip and change the output timings and you can more easily make interlace. But starting out with a register designed for interlace and having to try to combine it for progressive output is much more difficult. Take into account how hard it is to make the chip in the first place and you see that it's easier to start out with more and simply change output timings than to start out with less and have to rely on circuitry and chip to combine the signals. That's how it was told to me.

[Antoine_Fabi]

That makes sense to me, so the only question left is why progressive CCDs would cost more than interlaced CCDs ?

[jdsmith19]

Why use interlaced CCDs at all? They can't be that much more expensive, because the DVX and HVX are both cheaper than the XL2 and XL H1. Does it provide better low light performance because of the faster sampling rate? Maybe I just don't understand.

J.D.

[Walter_Graff]

It's not just the CCD here but everything else. Progressive output contains more information and that info needs a pipeline to process it. Frankly all CCDs are really progressive, it's just the way they output. Using a line pair trick a CCD is designated as interlaced. Sorry to say in the prosumer world, and even on a professional camera it takes a lot more electronics and a lot more money to make it all happen in progressive. And frankly you would not be able to really see the difference so it matters little. And in addition we still live in an interlaced world so progressive isn't found a place yet, if ever.

[mmm]

And in addition we still live in an interlaced world so progressive isn't found a place yet, if ever.

That looks like tasty bait...

In Europe, it looks like interlaced days will be numbered once HD becomes common place...whenever that is!

[Barry_Green]

The existing base of televisions are overwhelmingly interlaced, yes. But interlace is dead technology; nobody's building interlace anymore, nobody's designing the next generation of interlace, etc. It's over. Television manufacturing going forward is overwhelmingly progressive. All LCD, DLP, OLED, SED, computer monitors, plasma, they're all progressive. And the EBU specifically rejected interlace, basically declaring that there was no reason to saddle European consumers with interlace going into the future, they definitely and clearly preferred progressive.

Again, even in the US, the ATSC defined six formats for HDTV broadcast. Of those six, five are progressive; only 1080/60i is interlaced. And the EBU in Europe won't endorse interlaced (although they do acknowledge that interlace broadcasts will probably happen); they endorsed 720/50p as the best for today, with an eye towards an eventual move to 1080/50p, completely skipping 1080/50i.

So there's the view from here going forward, which is overwhelmingly progressive-scan. And then there's the "what do we do with the existing base" issue, which is extremely overwhelmingly interlaced. As long as those existing sets need to be supported, interlaced will cling on, but it won't grow, it will inexorably continue to shrink.