2.5K S35 Please

[Stephen Gentle]

Those who have said 2.5 is illogical, have you read the first post of this thread?

It's logical to use a non-standard resolution based on a rough estimate of the level of detail produced by a different camera's larger sized sensor?

 

[Katty Galloo]

Sorry people, but 2.5 does not make sence at all, than I can aswell keep shooting on my HVX200 at 1080p. There would be not a big progress in here.

Euro Girl

 

[Stephen Williams]

Sorry people, but 2.5 does not make sence at all, than I can aswell keep shooting on my HVX200 at 1080p. There would be not a big progress in here.

Euro Girl

Hi,

You are joking! HVX is a very long way from native 1080 cameras such as the Sony EX1. I would expect Scarlet even at 2.5k to be way better.

Stephen

 

[Charles Angus]

I don't see why it makes no sense. To me it makes as much sense as 3k - you get reduced resolution but also reduced storage and/or compression, plus bigger sensor sites. Whether it fits at all with Red's plans is doubtful but if I was building my dream camera it may well be a s35 2.5k handycam. I really don't need more resolution in the finished product than good 1080p.

It makes no sense because it is strictly worse in every area except one - file size. And, (although it hasn't been pointed out in this thread, the point has been made elsewhere) that higher res at the same bandwidth looks better.

More pixel density = lower/same noise, more resolution, more DR

 

[Katty Galloo]

Hi,

You are joking! HVX is a very long way from native 1080 cameras such as the Sony EX1. I would expect Scarlet even at 2.5k to be way better.

Stephen

I dont know pal, Sony EX1 has a mutch higher compression like mpg 2 or 4 to what I have heard, so surely not native and a lot higher compression then the HD DVCPRO100 compression.
Sony is also more the cam for Tele or reporter interlaced shoots. I know that 2k RED would be a lot better then both of them, but I wana have all the options for the future when I'am buying a RED, if I ever wana shoot one for the big screen I can then. So my option is at least 3K big sensor or the s35.
I'am tyred of buying a new cam almost every year you know. Whats out next year? got it.

Euro girl :waaa:

 

[Stephen Gentle]

I dont know pal, Sony EX1 has a mutch higher compression like mpg 2 or 4 to what I have heard, so surely not native and a lot higher compression then the HD DVCPRO100 compression.
Sony is also more the cam for Tele or reporter interlaced shoots.

The EX series cameras are really not bad at all - they are native 1080p, with 1/2 inch sensors (the HVX has smaller 1/3 inch sensors). Their codec is essentially a higher-resolution, higher bandwidth HDV, which is unfortunately long GOP, but still, it produces really nice footage. I've shot with an EX3, and I was really impressed. If I wasn't planning on getting a Scarlet, I would probably be looking at buying an EX1 or and EX3.

 

[Katty Galloo]

Hey,

No hart feelings :wink: , the Sony is a good cam. I had one before 170PD and was pretty satisfy at that time for standard dvcam resolution. I have heard indeed from a collegue that the EX has a nice collor and great image.
But as I said before the competion is going on every year and again others like Canon, Pana, jvc will reply on the 2/3 of Sony. Thats why i hope RED will be far ahead with his new collection.

Euro girl

 

[Stephen Williams]

It makes no sense because it is strictly worse in every area except one - file size. And, (although it hasn't been pointed out in this thread, the point has been made elsewhere) that higher res at the same bandwidth looks better.

More pixel density = lower/same noise, more resolution, more DR

Do a back to back comparison the EX1 looks many times better than an HVX200, it also has a faster sensor. Had to use a HVX 200 recently very disappointed after using an EX1.

 

[Stephen Williams]

I dont know pal, Sony EX1 has a mutch higher compression like mpg 2 or 4 to what I have heard, so surely not native and a lot higher compression then the HD DVCPRO100 compression. .

Euro girl :waaa:

Hi,

Pal has nothing to do with it, we are discussing an full HD camera with the EX! not an up resed SD camera like the HVX.
Remember the 100 data rate is based on 60fps, so at 24/25 it's nearer 40.
Looking at the final pictures is the best way to judge equipment IMO,

Stephen

 

[John Caballero]

Pal has nothing to do with it, we are discussing an full HD camera with the EX! not an up resed SD camera like the HVX.
Remember the 100 data rate is based on 60fps, so at 24/25 it's nearer 40.
Looking at the final pictures is the best way to judge equipment IMO,

LOL! You should read and learn a bit more before making statements like these.

 

[Tom Lowe]

I don't see why it makes no sense. To me it makes as much sense as 3k - you get reduced resolution but also reduced storage and/or compression, plus bigger sensor sites. Whether it fits at all with Red's plans is doubtful but if I was building my dream camera it may well be a s35 2.5k handycam. I really don't need more resolution in the finished product than good 1080p.

A 3K>2K or 1080p downconversion will look great. A 2.5K>2K downsample will be questionable, IMO. Plus you have zero headroom for reframing or image stabilization. Why ask for 2.5K when 3K is already the stated resolution? It seems like nitpicking, to me.

Why not save these "request" threads for meaningful changes we all can really get behind?

This is not directed at Stu. It's just a general statement.

 

[Cedric Akins]

Asking for less resolution seems..., well a bad idea. It makes no sense. Less resolution means a poorer quality image. Here is a retorical question. WHY?!?!?!?!

 

[Stu Mannion]

Why ask for 2.5K when 3K is already the stated resolution? It seems like nitpicking, to me.

You're right. I guess the real request would be for a s35 sensor at one of these smaller resolutions.

 

[combatentropy]

Pixels don't come for free.

For a sensor of a given size, more pixels = smaller pixels = less fill factor = less low-light sensitivity and less dynamic range.

Therefore if the sensor has more pixels than you will use, you are throwing away not only pixels but also sensitivity and dynamic range.

 

[Daniel Browning]

Pixels don't come for free.

No, they don't. There is an increased demand on in-camera computer power, file sizes, storage transfer times, raw conversion CPU time, post-processing time, backup systems.

Not to mention the incredible amount of research and development that goes into designing the sensor and manufacturing process.

For a sensor of a given size, more pixels = smaller pixels = less fill factor.

Therefore if the sensor has more pixels than you will use, you are throwing away not only pixels but also sensitivity and dynamic range.

I know that 99% of web sites, magazines, and photographers believe that. It has been repeated for so long that no one questions the authority of it.

But it's not true. It's urban legend. Objective, scientific experiments (e.g. below) prove it is incorrect. I think the reason this myth has persisted so long is that people make flawed comparisons using 100% crops, converted raws instead of raw data itself, and comparing two different sensor sizes incorrectly.

For example, the cheap little $400 LX3 digicam, with 2 micron pixels, has higher QE per area than the $5,000 Nikon D3 and its lumbering 8 micron pixels after normalization of ISO:

http://forums.dpreview.com/forums/read.asp?forum=1033&message=29866728

That and other tests have proved that smaller pixels have had superior QE, FWC, and base-ISO read noise for at least the last five years. The only area where larger pixels have had an advantage is a slight reduction in high-ISO read noise, perhaps due to using on-sensor transistors for read noise reduction.

Of course, most photographers can't use these superior pixels because they *only* come in the very tiniest sensor sizes. Even if they were 10 times better per area, the sensor is 25 times smaller, so in the end we must use the larger-but-worse-per-area sensors we have.

I'll re-post some of the information I linked above.

Eric Fossum, inventor of the CMOS APS, the technology behind almost all new DSLR cameras, coined [1] Agranov's Law: "More pixels improve image quality even in the presence of noise". This is based on a paper [2] presented by G. Agranov at 2007 International Image Sensor Workshop, where it was demonstrated that pixels sizes between 5.6 and 1.7 microns all give the same low light performance, but improved performance with more light.

Eric also said that FWC tends to increase with smaller pixels [3]: "What we really want to know is storage capacity per unit area, that is, electrons per um^2. Generally, as technology advances to smaller dimensions, this number also increases. So, in your terms, smaller pixels have greater depth (per unit area) and saturate 'later in time'".

[1] Eric Fossum coining Agranov's Law: http://forums.dpreview.com/forums/read.asp?forum=1000&message=24828114

[2] G. Agranov's paper: http://www.imagesensors.org/Past Workshops/2007 Workshop/2007 Papers/079 Agranov et al.pdf

[3] Eric fossum http://forums.dpreview.com/forums/read.asp?forum=1000&message=30017021

 

[combatentropy]

smaller pixels have had superior QE, FWC, and base-ISO read noise

Daniel, you say the smaller the pixel, the better the quantum efficiency (QE), full well capacity (FWC), and noise. I may be misreading your posts, but are you not comparing unfairly?

First, you compare a large sensor to a smaller one, but only the smaller one has microlenses. Compare the smaller one to a larger one that also has microlenses.

Then you compare two sensors of different ages. One of your references even explicitly says that the reason the smaller sensors are better is because technology has improved.

Daniel, if you have two sensors of exactly the same technology, and one has larger pixels than the other, which one will then be better?

 

[number6]

Graehme said a long time ago that the sensors all resolve only about 75 (or 78 or or 73 or somesuch) percent of the actual pixels. So, if you do the math, you already have something in the neighborhood of 2.5k from a 3k sensor.

 

[Daniel Browning]

First, you compare a large sensor to a smaller one,

I removed sensor size from the equation by basing the analysis on performance per area. In other words, I'm comparing an LX30-sized crop from the D3 against the LX30. Same amount of sensor area. Obviously, this is only useful for drawing conclusions about pixel size (the subject of this discussion), not the performance of any camera as a whole (some other discussion).

but only the smaller one has microlenses. Compare the smaller one to a larger one that also has microlenses.

The D3, like most DSLR, most certainly do have microlenses. In fact, it's the first and only full frame DSLR with gapless microlenses. That is a big part of the reason why it has a very respectable 15% gain in QE over all other FF35 DSLR.

Then you compare two sensors of different ages.

You got me there. Let's see... so many examples to choose from... how about a comparison of the 5D and FZ50? Those came out at the same time.

For the same exposure at base ISO, the FZ50 QE is 1200 photons per square micron, and 5D is 830. Read noise is 0.20 e-/um^2 for the FZ50 and 5D is 0.46 e-/um^2. These are based on raw files and measurements by Emil Martinec and John Sheehy.

One of your references even explicitly says that the reason the smaller sensors are better is because technology has improved.

Daniel, if you have two sensors of exactly the same technology, and one has larger pixels than the other, which one will then be better?

Performance would be the same except one has higher resolution. Obviously, right now we have a situation on the market where smaller pixels have higher technology, so they are *better* per area, but if things were even then the performance would be the same per area.

The current trend has been proven out to 1.7 microns, but perhaps somewhere smaller than that the trend will start to reverse, and 1.7 micron pixels will be better per area than, say, 1.2 micron pixels. Or maybe not until 0.5 microns.

Strictly speaking, smaller pixels are higher technology just by definition, in the same way that 45 nm process is higher technology than 90 nm process. It's not like we start with 1 micron pixels and spend billions on R&D to find ways to make the pixels bigger and make the CMOS process bigger.

 

[combatentropy]

Performance would be the same except one has higher resolution. Obviously, right now we have a situation on the market where smaller pixels have higher technology, so they are *better* per area, but if things were even then the performance would be the same per area.

Since we are talking about what RED will put in its next line of cameras, let's assume that technology will be the same no matter which pixel pitch they pick.

RED has made many smart choices --- Bayer sensors, wavelet compression, nice customer service, just to name a few. If our constructive criticism seems ingratuitous, I apologize. RED has made a better camera than I have ever.

Cameras RED has made: 1
Cameras I have made: 0

I would not even bring up this stuff about a camera after it is already in production, like the RED One. But, emboldened by Red's willingness to shuffle plans if a better idea shows itself, and since Scarlet looks like it is still on the drawing board, I thought I might give this idea a chance. It seems like one of the cheaper changes to make.

Great as the RED One may be, people still pine for:

- more dynamic range
- more low light sensitivity
- higher frame rates
- easier data wrangling

Most are at least satisfied with the current number of pixels. Many others say, HD resolution will last them for a long time. Therefore, it seems to me (and I am not a sensor fabricator) that one of the simpler ways to address the things that people still ask for is simply to reduce pixel count to the minimum needed.

 

[Tom Lowe]

Graehme said a long time ago that the sensors all resolve only about 75 (or 78 or or 73 or somesuch) percent of the actual pixels. So, if you do the math, you already have something in the neighborhood of 2.5k from a 3k sensor.

Some people put that number closer to 60% or 70%, but it depends on the camera, and the particular software. It also very much depends on how well exposed the shots are. Low-light Bayer-pattern imagery suffers greatly and can require larger downsamples to look clean, IMO. Well-exposed shots might very well reach 80% efficiency.