PDA

View Full Version : Confusion about resolution



FelixGER
07-05-2011, 06:15 AM
At first, this is not meant to bash the FS100. I already bought it and will get it next week, so I´m a fan.

I found several statements about real resolution of the FS100.

Nigel Cooper wrote it resolved about 780 lines

http://www.dvuser.co.uk/content.php?CID=268

http://www.dvuser.co.uk/images/img/reviews/camcorders/sony-fs100/chart.jpg

In this Philip Bloom video, they talked about 770 lines.


Yesterday I bought the new "Videoaktiv Digital" magazine ( http://www.videoaktiv.de/ ) and they come up with some quite disturbing numbers.

They meassured 671 line pairs vertical and 660 lp horizontal resolution!

They write, that´s about 300 lp less than the Sony F3. Some pages later they test a bunch of cheap new consumer camcorder which all resolve about 800-960 lp (960 seems to be maximum for 1080p).

AND, that´s a real shock, they test the new Canon 600D which is, I quote: "on pair with the 550D and resolves 709 line pairs vertically". That´s an "excellent" while the FS100 gets a "very good".


Can this be? The FS100 has poorer resolution than a Canon DSLR?

That´s quite the opposite I thougt. What do you guys think?

Jean-Philippe
07-05-2011, 07:19 AM
Personnaly I don't care about numbers. Last weeks I had a job where I used both my 7D and my FS100. While the two cutted together pretty well, The FS100 appeared way better to my eyes in every aspect: overal details, shadow details, highlight handling, color reproduction. And obviously aliassing. Magazines can say what they wants.

morgan_moore
07-05-2011, 07:45 AM
Its got 'twice' the res of the canons to my eye and a bit less than the EX1

http://dslr4real.tv/index.php?option=com_zoo&task=item&item_id=80&Itemid=1

S

Postmaster
07-05-2011, 08:01 AM
Yeah, something went super wrong in this test.
I did a ton of tests since I have mine and had many problems (and a long wish list), but resolution was non of them.

If this is really only 709 line pairs (which I doubt), Sony found some magic sauce, that makes it look 5 times as much to the eye.

Also from a technical standpoint, we all know the dirty tricks Canon employs to get video of a huge still sensor right?
And we also know that the FS100 grossly oversamples - so there is way more information, than just 1920 x 1080 to begin with.
I can see no way the FS100 should have less resolution, than a DSLR.

Frank

mandrean
07-05-2011, 02:04 PM
Something must have went bonkers in the videoaktiv test.

I shot some NYC skyline stuff, plus some Times Sq stuff, with the FS100, 5D and T2i. We used some vintage Zeiss lenses (Flektogon 35, Pancolar 50.)

The footage from the Canons was visually less sharp/resolving. Clearly visible with the naked eye. Everyone in the room were we viewed the raw footage awed at the FS100's clean image, no moiré/aliasing etc. while the 5D had so much moiré in every brick wall that the owner of the 5D started to feel bad that his $2.5K camera gave him footage that looked like crap.

Razz16mm
07-05-2011, 04:11 PM
You have to question how some of these tests are conducted or if people even understand the terminology they use. TV line vertical, now just called lines vertical, counts every distinguishable line on the chart, black and white. Line pairs per image height is exactly half of that figure since it counts a black/white line pair as one.
I can see an FS100 coming in at 670 LINES vertical, that actually is pretty good mid range performance for a 1080p camera. About what one should expect from an under $10k HD camera. But line PAIRs? That would suggest it is resolving higher than its pixel count since the absolute maximum vertical line pair resolution any 1080p output camera can possibly do is half of 1080p, or 540 line pairs vertical. And that is only if the sensor is absolutely perfectly aligned so a white line falls completely on one row of pixels and a black line on the next. change the alignment so half of a white line and half of a black fall on one row of pixels and you get a gray field, 0 resolution. So higher than 540 line pairs vertical from 1080p is physically impossible.
DSLR's come nowhere close to 600 lines vertical, best I have seen off of a chart is around 540 lines, less if you don't count aliasing.

TV lines and LPPH ( line pairs per image height) both refer only to vertical resolution, which is the limiting factor on overall spatial resolution of the image. You get more total lines or line pairs horizontally since there are more pixels, but a resolution chart measures spatial resolution, which for square pixels is the same vertically and horizontally and slightly less diagonally. you have to multiply by the aspect ratio to derive total horizontal resolution.

TheDingo
07-05-2011, 08:09 PM
Taking a look at the NAB 2011 - SCCE Charts : Resolution, the Canon DSLRs show lower resolution ( 428 line-pairs ) compared to either the Panasonic AF-100 ( 500 line-pairs ) or the Sony F3 ( 540 line-pairs ). Since the FS-100 is supposed to use the same sensor as the F3 ( different electronics ), it should be producing something close to the F3 results.

NAB 2011 - SCCE Charts : Resolution (http://provideocoalition.com/index.php/awilt/story/nab_2011_-_scce_charts/)

Razz16mm
07-05-2011, 09:50 PM
The SCCE tests are very unusual in their results and they have not published their methodology. I think in some cases they were interpreting aliasing and false detail as resolution.

Barry_Green
07-05-2011, 11:13 PM
Since the FS-100 is supposed to use the same sensor as the F3 ( different electronics ), it should be producing something close to the F3 results.
Not necessarily. There's nine thousand dollars of hardware backing up the sensor in the F3, that the FS100 doesn't have. You shouldn't attribute comparable results to them just because one element is the same.

Barry_Green
07-05-2011, 11:14 PM
The SCCE tests are very unusual in their results and they have not published their methodology. I think in some cases they were interpreting aliasing and false detail as resolution.
I'll be polite -- please, please don't go disparaging or calling into question the methodology and competence of the folks that produced that test, because they are and were the best of the best. I would be shocked beyond belief if the combined efforts of that team somehow managed to let slip such an amateur mistake as not checking for actual resolution and banking on fake aliasing detail.

morgan_moore
07-06-2011, 12:31 AM
There is a possibility that the FS has 'low' resolution

Compared to my I EX I would judge by eye that is does

However

Once the camera is attempting to reslove something it cant resolve what we see is blur

Which is far superior to DSLRs

Because with a DSLR if you have any tight resolution items in frame you get a shot you probably need to throw away, with the FS you just cant resolve but get a shot you dont need to thow out

IMO

(informal test).. http://dslr4real.tv/index.php?option=com_zoo&task=item&item_id=80&Itemid=1

TheDingo
07-06-2011, 07:42 AM
Not necessarily. There's nine thousand dollars of hardware backing up the sensor in the F3, that the FS100 doesn't have. You shouldn't attribute comparable results to them just because one element is the same.

So better electronics would improve the resolution of the sensor ?

I could understand it would make a difference for things like: Dynamic range, Color-space, Gamma-controls, Motion-blur, Noise-reduction, Image-sharpening, Encoding, etc...

But isn't resolution more of a sensor thing ? ( unless image data is being thrown away in post-sensor processing )

Dustin Moore
07-06-2011, 08:02 AM
But isn't resolution more of a sensor thing ? ( unless image data is being thrown away in post-sensor processing )

One can always throw away image data in post-sensor processing and there's always a trade between electricity consumption/heat
generation and processing quality.

A computationally cheap debayering algorithm like a bilinear resample will trash massive amounts of resolution compared to a computationally
expensive debayering algorithm based on gradients or other edge detection.

Rick Burnett
07-06-2011, 08:48 AM
One can always throw away image data in post-sensor processing and there's always a trade between electricity consumption/heat
generation and processing quality.

A computationally cheap debayering algorithm like a bilinear resample will trash massive amounts of resolution compared to a computationally
expensive debayering algorithm based on gradients or other edge detection.

Exactly. The DeBayer process can be different for both of them, which can change the achievable resolution. Just think about the fact of what a still image and a video frame looks like off of a Canon DSLR. Same sensor, VASTLY difference resolution performance. Canon claims they don't skip but that they pixel bin, regardless, different methods of handling the same data can produce different results. We have no clue really what they are doing with the data.

And pertaining to resolution, when I shot with my 7D and FS100, the difference in sharpness was clear as day. It's not subjective. This magazine sounds like something I would avoid.

cheezweezl
07-06-2011, 01:22 PM
agreed. if you want clear cut, non-nonsense, proof. throw a really wide lens on the fs100, like the tokina 11-16, and shoot a landscape with lots of detail. the difference between my 7d and the fs100 is night and day. before the fs100, i would occasionally have to shoot a still with the tokina to remind myself that it's a good piece of glass because the video looks like death. the fs100 resolves it beautifully.

sinapps
07-06-2011, 04:23 PM
agreed. if you want clear cut, non-nonsense, proof. throw a really wide lens on the fs100, like the tokina 11-16, and shoot a landscape with lots of detail. the difference between my 7d and the fs100 is night and day. before the fs100, i would occasionally have to shoot a still with the tokina to remind myself that it's a good piece of glass because the video looks like death. the fs100 resolves it beautifully.Agree 100%. exactly same experience with D7000 and 11-16.