Foveon Sensor

[Kayaker72]

So, like several here, I occasionally follow canon rumors. The hot topic right now is that a new Canon 1D/Ds may (or may not) be annouced tomorrow (or some other time in the future). The current rumored spec is 18 MP, but as recently as a few weeks ago the rumor was 54 MP. This has some thinking that the new sensor may be Foveon-like, in that each pixel, instead of reading a single color (as our current sensors do), will record data for three colors. So, an 18 MP Foveon has the same resolving power as a current 18 MP CMOS/CCDsensor, but has 3 times (54 MB) worth of color data.


So, first of all, please feel free to clarify my above understanding. But my question is about memory usage. Wouldn't RAW files from a Foveon require 54+ MB as they record three colors for every pixel whereas our current senorsrequire 18+ MB (the "+" being things like the internal jpg, etc). So a Foveon senor might be a memory hog when shooting in RAW but when the images are converted to jpg they'll be very similar to the current jpgs from the current 18 MP sensors? If so, the burst rate may be faster for jpg than for RAW in the new camera?


Not that I put much stock in any of this, but I am mostly just curious and didn't find the answer in a brief internet search.


Thanks,


Brant

[Sean Setters]

According to CanonRumors, you may only have to wait 24 hours or so to get your answer...


;-)

[HDNitehawk]

So, like several here, I occasionally follow canon rumors. The hot topic right now is that a new Canon 1D/Ds may (or may not) be annouced tomorrow (or some other time in the future). The current rumored spec is 18 MP, but as recently as a few weeks ago the rumor was 54 MP. This has some thinking that the new sensor may be Foveon-like, in that each pixel, instead of reading a single color (as our current sensors do), will record data for three colors. So, an 18 MP Foveon has the same resolving power as a current 18 MP CMOS/CCDsensor, but has 3 times (54 MB) worth of color data.

This information is very definitive on its release. I might add that it may be released, or possibly will be released, or maybe not released at all. All of which will happen or not happen in the near future. I think that covers my prediction which will be 100% correct depending on which part you read.

If so, the burst rate may be faster for jpg than for RAW in the new camera?

Do you mean a larger burst rate, not faster. Meaning that if you shoot jpg now you can shoot many more shots before you get the nasty"busy" signal whenyour processor gets backlogged. Other than larger files from the "rumored sensor", this will depend more on the processors they use in the new camera. In theory larger files that are the same type in the same processor should take longer to process, but then again you do not know what kind of upgrades the firmware will have. Possibly better firmware and better file management will make it faster. So....it is all speculation that really does not have an answer unless you are in the know at Canon.





If a 1Ds IV is getting released tomorrow, we should be making bets for the price point it will come in at....I say $12,499.00....

[neuroanatomist]

Wouldn't RAW files from a Foveon require 54+ MB as they record three colors for every pixel whereas our current senorsrequire 18+ MB (the "+" being things like the internal jpg, etc)

Correct - the RAW file size from an 18 MP Foveon-type sensor would be three times larger than a RAW file from an 18 MP sensor with a Bayer filter. Note than after demosaicing the file from the Bayer sensor, the size would be similar to that of the Foveon-type, e.g. if you output a 16-bit TIFF file from both formats.

[Sheiky]

Hmm...it was also rumoured that the new camera would be capable of doing 10-12fps. With a 54MB+ data amount I would assume that this would be quite a big issue to tackle?


I'll see what happens tomorrow. I'm curious though.

[Kayaker72]

Hmm...it was also rumoured that the new camera would be capable of doing 10-12fps. With a 54MB+ data amount I would assume that this would be quite a big issue to tackle?

That is one of the things that got me thinking....12 fps in RAW (54+ MB-edit: if Foveon)is 648+ MB/sec of processing, whereas if the 12 fps is for <span style="text-decoration: line-through;"]jpgRAW if Bayer/current only (18+ MB)is 216+ MB/sec of processing speed. A big (3x ) difference in either the processing speed or buffersize.....if the sensor is Foveon-like, jpg would be even smaller than the RAW output from a Bayer/current sensor.


Rick...good point in that they may have the same burst rate but the camera will maintain that rate for 3x fewer <span style="text-decoration: line-through;"]RAW pictures than jpg RAW from a Foveon Sensor compared to RAW filesfrom a Bayer sensor. Even fewer compared to jpg.


Anyway, I hope that they are right in that a new flagship is released tomorrow. I may never own one, but it will be great to see what technologies/capabilities may filter down(conspiracy theoriesaside) to cameras I will buy.


Edit: It hit me that I was labeling jpg but thinking RAW files from the current Bayer sensors.

[Daniel Browning]

This has some thinking that the new sensor may be Foveon-like, in that each pixel, instead of reading a single color (as our current sensors do), will record data for three colors.

Foveon technology has a host of permanent disadvantages compared to Bayer, most notably: terribly poor color accuracy and bad chroma noise. They can never be fixed or improved because they are integral to the technology itself. The only possible advantage is increased color resolution, but is that really any benefit? Most users are going to immediately throw away that color resolution, such as by converting their raw files to JPEG (with usually uses chroma subsampling AKA compression). Even if they don't throw it away, can they tell it's there? Humans are very insensitive to minute variations color details.


It would take a *major* breakthrough, such as sub-pixel prisms or dichroic mirrors, to achieve a non-Foveon method of 3-color co-sited sampling within each pixel. Even if someone were to achieve that, it would only be at the cost of color accuracy. At least until an even *bigger* breakthrough comes on some way to achieve accurate color without filters.

[neuroanatomist]

It would take a *major* breakthrough...to achieve a non-Foveon method of 3-color co-sited sampling within each pixel.

Nah. All it takes is subjects that don't move.


The Zeiss AxioCams on several of my microscopes use co-site sampling to eliminate the need for color interpolation - they move the Bayer mask twice during acquisition, so each pixel captures a separate R G B image. Furthermore, they use 'microscanning' - translating the sensor itself in a 2x2 or 3x3 sub-pixel array to increase resolution by moving the photosensitive portion of each pixel around within the area of that pixel (no microlenses on the sensor). So, that simple 1.4 MP 2/3" Sony CCD sensor can generate a 13 MP, non color-interpolated 14-bit image (of course, it takes 27 separate exposures to do so...).

[Daniel Browning]

The Zeiss AxioCams on several of my microscopes use co-site sampling to eliminate the need for color interpolation - they move the Bayer mask twice during acquisition, so each pixel captures a separate R G B image.

That is really cool!

All it takes is subjects that don't move.