Ignorant 16x9 Question?
 

This is me exhibiting my ignorance.

Panasonic have this truly very covertable new camera. Seen the footage, marvellous! OK, it has an almost "square" sensor plate. I would be videoing in 16x9. Does this mean that a central slice is the only part of the sensor that is uilized? If this is the case how much of the 4/3 is ignored?

Apologies for uninformed question, but I can't get my head around this one at all.

Grazie

I wouldn't worry, quite a lot of the standard 35mm motion picture frame isn't used either when it's being used for most productions since they stopped shooting 4x3.

That's why 3 perforation pull downs became more popular now that digital post production has become common. It saves money on film stock.

Thank you Brian.

Ok, then how much of the HD reso, as this a central slice, is being made use of? Or putting it another way, how much reso is available on the whole sensor?

So, as this is be triumphed as a full 4/3 this has more to do with it coming from a stills dynasty. Is that correct?

Grazie

Graham, what makes you think the sensor is "almost square"?

There is nothing in the 4/3rds standard which stipulates the sensor must have a 4:3 aspect ratio (although most early ones did). The only measurement it has to adhere to is a diagonal of 22.5mm.

With the GH1, Panasonic actually made a chip that was larger than it needed to be in order to accommodate 16:9, 4:3 & 3:2 aspect ratios. So even though the chip was being cropped at every aspect ratio, you still always ended up with a diagonal of 22.5mm.

It seems Panasonic are being tight lipped about the actual sensor they've used but I hardly think they'd take a step backwards by going to a 4:3 sensor and cropping to a diagonal which is less than 22.5mm. Most likely they use a similar oversized chip to the GH1/GH2 so your diagonal is going to be 22.5mm.

A lot of people overlooked this fact when comparing crop factors between the GH1 and 7d - they failed to note that the GH1 was generating wider 16:9 images by adding pixels on the side, while the 7d was cropping it's 3:2 sensor. The result was that the crop factors of the two sensors in video mode was closer than the usal 2x vs 1.6x comparison.

Looking at the IBC video in new thread, they're using a 16:9 "slice" on the sensor, that 4x3 won't be an option on this camera and a guesstimate is 11 mega pixels.

Thanks John.

What I was wanting to see is a 16x9 area clearly, diagrammatically laying on top of this 4/3 sensor. Tight lipped? OK.

Here is the video I saw and at 2:02 the 16x9 Graphic is slotted-in, but over what? I'm confused. Is this just to show the comparisons TO 16x9? But just how would that 16x9 look if it was placed over this new 4/3?

Taking careful note of the numbers, it looks like 1.15mm top and bottom (13mm>10.7mm). But then the width of 19mm (this for the 16x9) how does this then fit into the 17mm width?

YouTube - 4/3" Sensor vs 35mm comparison

Cheers

Grazie

You'll find a multi aspect diagram here:

Panasonic Lumix DMC-GH1 Review: 1. Introduction: Digital Photography Review

Graham, it is a little confusing, especially as Panasonic are being tight lipped on the sensor being used. John Wiley has given quite a good rundown of this. I think it is probably likely that they have used an oversize sensor, as in the GH1/2 and the diagram Brian has posted should explain how it fills more area horizontally.

Looking at the diagram, I take it that more or less 16:9 sits in the centre of the 4x3 with bands top and bottom that are no longer used. Which is basically how the standard 35mm motion picture frame is used for various aspect ratios.

Panasonic are being tight lipped about the actual sensor, but Jan Crittenden's description of it a "slice" would seem to indicate this is how they're doing it.

Although, if the sensor is rounded either side, i suppose they could have more width wiggle room with 16:9 compared to using a straight edged 4 x 3 sensor.

To be honest, I'm not really sure what the question is -- but maybe this link to Panasonic's description of the sensor in the GH2 will help. Look down the page for a flash comparison of Lumix aspect ratios and how the same imaging chip can handle multiple formats. Note too that the pixel count -- both horizontal and vertical -- changes for each of the three choices, i.e. the 16:9 does not simply drop lines top and bottom.

Aspect ratio:
DMC-GH2 | PRODUCTS | LUMIX | Digital Camera | Panasonic Global

Pixel count by aspect ratio (File size - still image):
https://panasonic.ca/english/audiovi...ries_specs.asp

HTH,
GB

That makes sense, basically the sensor is oversized for 4 x 3 in width, which is used as the aspect ratio changes to a wider format, so that the image diagonal remaining pretty constant. A combination of cropping top & bottom and widening.

Seemingly this camera is only available as 16:9 anyway, so it could be an academic exercise in this case.

Thanks Geoff, I hadn't seen that diagram yet.

It's still difficult to find any info on what sensor is used in the AF100, but what you can see from that diagram of the GH2 is that even though every aspect ratio takes only a "slice" of the sensor, they all still maintain the maximum possible area allowed by the 4/3rds spec which dictates a 22.5mm diagonal.

I feel it's necassary, just to clear things up, to state that 4/3 is very different to 4x3 or 4:3. 4/3rds is a measurement of the sensors size diagonally. It is a measurement directly comparable to those of other sensors such as 1/3", 1/4", 2/3" etc. 4x3 and 4:3, however, are aspect ratios. They compare the relative width and height of an individual sensor or image. A 4/3rds sensor is not necessarily 4:3 or 4x3 - many early ones were, though that changed as the format grew and developed. So a 4/3rds sensor can be natively 16x9 or 3:2.

When first announced at NAB, I seem to remember it was definitely announced as a 12.1 megapixel four-thirds sensor at the Panasonic press conference, so generally assumed to be adapting one of the sensors from the stills range - which seems highly logical. It seems to have been further confirmed that the adaptation includes an optical low-pass filter - so expect it to have overcome the worst of the problems that plague vDSLRs, even though still using a similar type of sensor.

What hasn't been commented on is how the sensor is being read, which tends to make me think the worst - that it's still having to pixel skip, rather than read out the entire sensor each frame and downconvert. That will mean that at very least the sensitivity is unlikely to be as good as may be expected from that size of chip, even if the OLPF gets over the worst of the DSLR problems.

However good it turns out to be, it won't be as good as a custom designed for video sensor with the "right" number of pixels. For 1080 video, "right" is likely to be around 4 megapixels - enough to give full resolution (4 megapixel = 2x1920x1080 total, so 1920x1080 green), but few enough to be easily processed, and large enough to give optimum sensitivity.

Exactly - they gave a number (12.1 megapixles) and a format (micro 4/3rds) but not much else to go off. Is it 12.1 megapixels total? Or effective? If it's the effective number of pixels, then which pixels are thrown away? ie how is it cropped? What is the aspect ratio of the sensor? Is it the same as the multi-aspect ratio in the GH1?

As for how the image is read off the sensor - well, that's beyond my level of undrestanding but so far the footage looks really impressive - less of the aliasing and moire which troubles the other DSLR's/hybrids. And in the end, all that really matters is that the picture looks great!

Yes John, quite. And I'm still around and still reading.

Sooooo, boys, what is/how many pixies=reso is the 16x9 slice finally, actually, in reality, reading on this 4/3 sensor? A 4/3" sensor is big, and placed against a 35mm plate, it is jolly impressive.

And yes, now I've seen much footage, I do want this camera - so far. Knowing something through the physics always guides and tones down my expectations. Having a good enough tool can go to satisfying my creative urges. Knowing the difference between these two tensions is for my bank manager to know!

When I know something, I know it. When I don't, I ask.

Interesting thread, huh?

Grazie

Hope I'm reading this right and happy to be corrected as the camera sounds brilliant but the sensor should be just under 22mm along its diagonal applied to any aspect ratio and may line skip. Oh dear that takes out the possibilty of using my 16mm zeiss superspeed lenses and makes me think this is much inferior to my Sony EX1?

Hmmm? Should I know consider the SONY route?

Grazie

I don't think the precise no of pixels on the chip really matters a great deal, it's the ball park figure that's most important. It seems pretty certain that there will be "enough", and whether it's 12, 12.1 or 11 megapixels is pretty academic - they'd all potentially give better than 1080 HD resolution.

But it's the fact that the no is "around 12" that IS important, since it brings up the far bigger question of read out, and how it's being done. The best way would be to read all 12 million off every frame, then downconvert - but that is likely to cost in terms off complexity and power heat issues.

Hence it's far more likely to pixel skip as DSLRs do - only read a percentage of the pixels each frame. There should still be more than enough for 1080 resolution not to be compromised - but it will impact performance in other ways.
Well, yes, and I'm fairly sure it will. But the question may be if there's better still?

As said in the previous post, a purpose designed chip is far more likely to have about 4 megapixels - but read them all out every frame. The big advantage that is likely to have over the AF100 is sensitivity, either being able to be used in lower light, or far better s/n ratio.

The indications seem to be that is what Sony have done for the F3 and it's as yet unnamed little brother. And that's why a lot of people are waiting to see how they turn out rather than jumping in with an AF100 order now.

How so?

Because it is too big to use certain lenses, it makes it inferior to a camera with much smaller chips? That's like saying becuase you can't use APS-C lenses on your 5DmkII, it's inferior to your compact camera!

The truth is they are made for different purposes. The EX1 will be a much better choice for certain scenarios while the AF100 will outshine it in others.

As I explained inferior if the sensor is line skipping.

The needed bigger lenses is only an issue for me wanting to use my Zeiss 16mm primes My thinking has up till now been that if the sensor is 17mm x 13.5mm in a 4:3 shape then my Zeiss primes with diameters of 20mm for the 25mm 19mm for the 16mm and 17mm for the 12mm would be more or less covered especially if shooting for 2.35 But I can see now.

The sensor can be diferent sizes as it follows just under 22mm diagonal line for the aspect ratio used So undoubtably the sensor will be 16/9 and just under 22mm diagonal so you will need a minimum 22mm rear lens to fully cover this.

Also the Sony EX1 offers 10 bit out and a superb lens. Factor in the lens costs and the camera is significantly lots more money for a sensor that line skips.
Not trying to promote any camera as I really like the things the Panny can do Just that line skipping is a deal breaker for me.

I read somewhere that 16mm film lenses are ok if longer than 30mm, shorter will definitely vignette.

Olaf thats a generalisation. Many 16mm lenses are larger than they need to be.

If this is the case, would you think that Panasonic would care to comment?

Grazie

An easy way to find out is to use a Lumix gf1 (or similar M43 still camera) with an adapter to M43 take a photo with your lens and check for vignetting.

A good rental house may be able to help you test this w/o purchasing.

The problem appears to be the sensor size may be different for a stills camera to this one. At this point who knows.

The lack of a reply from Panasonic might indicate that line skipping is indeed used in the sensor.

I believe the only difference from the still version is that it is cropped to 16:9. The lumix can even be set to shoot stills in this aspect.

So this would give you a very good idea if your lens will cover the sensor. It certainly is not a larger sensor than the standard 4/3.

I personally doubt its the same sensor cropped. Just my opinion but I think it most likely sticks to the four thirds spec with a diagonal of just under 22mm that's usable this would make the camcorders sensor wider than the still camera sensor.

Best

Mark

I just meant that the M43 spec sensor is the same size.

I am real sure the it is not the same sensor as well.

Personally I am using the Novoflex Canon FL/FD adapter to test 7 old lenses I have kicking around in preparation for the arrival of my AF100.

The 35mm f2 and 24mm f2.8 are very nice as is the old 50 f1.4. I did also order the Voigtlander 25mm f.95, it should be here soon. It is nice to compare the different lenses at 12 Mpx RAW in Aperture and PSD.

My understanding is that the AF-100 sensor does not perform line-skipping, though I've never been able to find technical details about how the sensor functions.

Recent production comments from cinematographers shooting with the AF-100 indicate that the camera does not have any noticeable moire or aliasing visible, and that resolution appears to match the Sony EX-1/3 cameras. ( the cinematographer who made this comment owns an EX camera )

AF-100 does not line-skip, and image quality appears to be comparable to the Sony EX-1/3 cameras. ( Yes it's 8-bit, but almost all prosumer cameras under $10K are 8-bit )

All that said, the AF-100 is not an ENG camera, and would likely not be a great match for run'n'gun ENG style work. ( Sony EX-1/3, Canon XF300/305, Panaonic HPX 370/170 cameras would be a much better solution )

Can I ask why you don't think it pixel skips? Do you have any evidence to back that up?

It's important to realise that whilst pixelskipping has got a very bad name, a lot of that is down more to the absence of any optical low-pass filter in DSLRs than pixelskipping itself. Add an OLPF, (which the AF100 does) and the worst of the moire etc issues will go away, pixelskipping or not. Absence of moire or aliasing is not by itself an indication of no pixel skipping - it's an indication of an OLPF.

Given an OLPF, the ADVANTAGE to pixel skipping in such a camera is reduced complexity, power consumption, heat problems etc if a chip of over 10 megapixel is being used. The DISADVANTAGE will be (mainly) lower sensitivity or a higher noise level, compared to reading the whole chip and downconverting.

For optimal results, you really want a chip of more like 4-6 megapixel, same physical size, but the lower pixel count meaning pixel skipping is unnecessary. In many ways, you may not be able to tell much immediate difference between such and what the AF100 is doing - but it is likely to be inherently much more sensitive.

Two reasons:

1- Because it's precursor the Panasonic GH-1 / GH-2 DLSRs do not use line-skipping to reduce resolution. ( they down res via pixel binning )

2- Every camera that I've ever heard of that uses line-skipping to down res always has aliasing and moire issues, and it's been confirmed by many shooters that the AF-100 has no visible aliasing or moire, even when shooting highly detailed environments with wide angle lenses.

My understanding is that the current crop of Canon DSLR cameras all have OLPF, but because these filters are designed for full resolution from the sensor, they do not prevent aliasing or moire when shooting at much lower HD video resolutions. ( i.e. 1080P, 720P, etc... )

Pixel binning (as I'm sure you're aware) is combining adjacent pixels to form a single effective pixel. But used with a Bayer array, the problem is that adjacent pixels are filtered to different colours, which makes pixel binning far more problematic. In this case, and doing it in real time, my suspicion would be that it will be too coarse an effect.
But up until now, every camera that has used pixel skipping to down res has not had an OLPF matched to HD video. If the AF100 has no visible aliasing or moire, it's an indication of the presence of a suitable OLPF - not necessarily that it's not pixel skipping.

An OLPF in a DSLR suitable for stills will be of no use in preventing video moire/aliasing - if it cut off at a low enough frequency, the video would be fine - but the stills would be unacceptably soft! (So don't expect the AF100 to ever give stills a fraction as good as a GH-1, even if it were to have the same sensor.) You can't have your cake and eat it.

Practically, when comparing it to such as an EX1, the differences are far less about quality as usability. The AF100 (and Sonys F3) will give far shallower depth of field - but will have significant disadvantages as a "general purpose" camera, most notably in the effective absence of a cost effective servo zoom lens of decent zoom range, manual operation and aperture.

My preferred way of filming till now has been an EX1 for mid and wide shots using a letus adapter with 35mm still camera lenses for closeups and some mids and using the 10 bit out with a Ki Pro.

The panasonic seemed at first to offer a good solution but now with possible line skipping the use of four third lens and the expense of decent lenses for the format. The 8 bit out. The crop factor issues and Super 16mm lenses not covering the sensor, combined with a lack of information leaves me AMAZINGLY wondering if the EX1 is still the best deal for quality flexibility and cost effectiveness.

I agree with Mark that the EX1/3's are an incredible value, and I have no intention in selling mine in the next few years. I use the NanoFlash on them, I know that the NF is only 8 bit. But I am not sure the EXcams at 10 bit really are justified, the signal is not 950 quality. The NF 422 at 100mb/s is really amazing looking even after some heavy grading.

What I think is really cool about the AF100 is the SDOF properties in a compact inexpensive camera. And as an added plus I can use my old FD and FL lenses and all my newer L glass. And some new exciting glass like the Voightlander 25mm f.95. And others will be able to use their cine quality glass and Nikon glass.

This is not an ENG cam it is really a great 7D class solution. A real video camera. I am still keeping my 5DmkII as well it is also a great tool.

So for me it is my new SDOF adapter that will be easy to use and light.

Stills with the AF-100 are at HD video resolution only, thus at 1080P they are roughly equivalent to a 2 Mpx still camera. Also this camera can't do the quick punch-in for focus, which makes me think that the 1080P optimized OLPF might be the reason why.

The EX-1/3 cameras can resolve 1000 lines, which make them one of the sharper 1080P cameras on the market. The AF-100 resolves in the same range, though the images are supposed to have a more "organic" look to them. Hopefully we will have some direct side-by-side comparisons in January, to see how these new cameras hold up.

Shallow DOF is the main reason for buying these cameras, and I doubt that servo-zooms will ever make these cameras suitable for ENG work. The lens alone would just be too big, too heavy, and too expensive.

Myself, I'm keeping my Panasonic HMC-150 for ENG style work. ( one small bonus is that these cameras share the same batteries )

Yes it is.

Or it could be, depending on whether or not they used the multi-aspect sensor as they have in the GH1 & GH2.

But at any given aspect ratio, it crops to a section of the chip which has the sama area as a 4/3rds chip natively designed for that particular aspect ratio. Each aspect still adheres to the 4/3rds standard which is a 21.6mm diagonal imaging area.

Software has an easy time piecing together an image from a fixed chart. But I wonder about real world resolution. Both compression and noise reduction can reduce real resolution. Subjectively the AF100 images don't look particularly hi res to me. But perhaps the files are just not sharpened to my preference.

Of course I want the perfect camera for $5000, and that's not going to happen.

According to the brochure then a cropped 16/9 frame from the 17.3mm x 13mm sensor which means I will be able to use my 16mm superspeeds and if I use an aspect ratio of 2.35 Maybe a little bit of vignetting and maybe a little bit of zooming in on the frame with the 12mm but all in all with a nanoflash or ki pro this should be outstanding.

One thing I noticed is that the pics in the brochure seem to be lower in quality than my Sony EX1? Softer?

I would read too much into an on-line pdf, chances are it's not a full print quality resolution.

Panasonic claim the frame size is almost the same as movie 35mm and could be if it was just under 22mm on the diagonal for 16.9 But ti isn't The brochure states it's cropped and so therefore is less than halfway between 16mm and Super 35mm film Unless I'm missing something.

Super 16mm aspect ratio 16.9 Frame size 12.52 by 07.41 mm
Four thirds aspect ratio 4.3 Frame size 17.30 by 13.00 mm
Movie Super 35mm aspect ration 2.35 Frame size 24.00 by 12.97mm

However this is a fantastic size for me!