Some updates from Foveon in a new video interview posted on YouTube:
- Sigma is “still working on the development of the sensor” [17:00].
- Current status: The project is still in the “technology development” stage [17:11]. They have not yet started the design of the actual, final sensor [17:11].
- Focus: The team is currently working on the “design of the pixel architecture” [17:20].
- Delays: The project has been “a little bit delayed” [17:30] because as they test prototype wafers, they encounter “technical issues” [17:53].
- Development team: The sensor development is now being handled primarily by the Sigma Japan engineering team [18:02].
- Path forward: Mr. Yamaki mentions that the technical problems “have been narrowing down” [18:12]. Once the team is confident that the technology is ready, they will start the final sensor design and move toward production [18:23].
When this sensor will be available I plan to buy the camera and go on a Safari vacation. When I'll rest in my room in the evenings I'll read the final book in the "Song of Ice and Fire" series.
ReplyDeleteFor the record, I sold Foveon sensors into non-photographic markets for about 10 years and I wrote a lot of documents about them. The most interesting is probably the F7 sensor on the ESA CLUPI camera that will go to Mars if they ever launch it.
ReplyDeleteHere are the shortcomings of the existing Foveon design, limited specifically by the capabilities of semiconductor processing at the time. These will give readers some idea of the complexity of the task now before Sigma.
1. Pixel size - The Foveon device is made from a sequence of alternating n and p dopings at various depths, each doping inside the previous one. This takes a lot of space so the minimum pixel size is around 10 microns.
2. Dark Current - There is little opportunity to provide any of the dark current reducing features used in other sensor types so the dark current is high. The shape of the implants required to produce the stacked photodiodes actually increases the junction area available to leak.
3. Non-linearity - Except for the top (blue) layer, the photodioes are unpinned so the response in non-linear - somewhere between linear and square root. This necessitates very accurate post-processing corrections so the color tracks as the intensity changes.
4 - Shuttering - The device's architecture leaves no spot for global shuttering so only a rolling shutter is available.
5 - Noise - Optimizing for other performance factors, produces a device that has high noise. Since there is no storage available, CDS is not possible. The high dark current contributes to the high noise floor at low light levels.
6 - Color Separation - Silicon provides enough variation in penetration depth vs. wavelength to produce quality color pictures but the correction matrix requires high off-axis values and all channels are infrared sensitive so the noise in the corrected signals is substantially higher.
Some other limitations were design decisions made because the old designs were intended only for still photography. The readout speed was slow, the power and signal requirements were complex and the packaging was difficult to accommodate.
These item descriptions are, of course, greatly condensed. If anyone is interested in more detail or in a newer method of producing stacked pixels offering improved performance in all listed factors, feel free to reply or e-mail me - gilblomdl@engineer.com.