43 Rumors

I agree. I hoped this MPX-racing-BS stopped in 2007. The normal compact-user got used to digital cameras and knows what the problems are. Auto White-balance, low light performance, fast focus, this are all the points people are asking me about when they search for a new camera. Nobody wants to print in A2 (metric, very big paper format
a lot of people put the settings down to 4 or 6 MP to save space on the Card and on the Drive

You might be surprised if you took Sony’s 16mp APS sensor and compared it correctly to your 6mp RD1. The 16mp sensor is better. Visibly better, which is as it should be because sensors are running on a improvement curve that’s been pretty much unchanged back into the late 1980′s.

I suspect you’re looking at neighboring pixels at 100% view. Don’t. Try comparing same area, same output (which in this case would be APS crop and 6mp, which requires downsizing the 16mp sensor).

Fuji’s early F-series model like F30/F31 had bigger sensors than the competitors and that combined with fujis processing was the main reason for good high Iso performance.

mFT has a fixed (usable) sensor size and any improvement in high Iso will have to come from better sensor technology and/or noise reduction. The number of pixels don’t matter much for Iso performance when you look at a certain image size. Per pixel performance is likely to be better with bigger pixels though.

dummy00001, Panasonic and Olympus have had lossless compressed RAW files for years. If you didn’t, your 12MP files would be around 28MB’s each. I think the last Panasonic camera I owned the L1 had uncompressed RAW files, as did the Panny FZ50 and LX2 (I believe). Those were a nuisance, and got me used to converting to DNG to get some compression.

Yeah I think it’s mainly marketing, IIRC panna said in a interview that 12mp was more than enough but marketing needs drove the GH2 up to 18mp. I can see where they’re coming from as well, if a consumer has two cameras in there hand, one 24mp, one 12mp, many would go for the 24mp IMHO.
I think alot of manufactors look at fuji, who steeped out of mp race with DSLR’s and concentrated on DR, and although the camera preformed well, it was the last DSLR fuji made and until the X100 they had pretty much given up on more serious cameras.

> It is fairly well established that as the density of photoreceptors on a chip increase, issues with cross talk and thermal noise increase.

Yes and no. It’s much more complex than that. But note that NO company has yet done much of anything to deal with thermal aspects internally in cameras. We don’t even have heat sinks on many critical components yet. There’s plenty that can, and will, be done in the future.

> This problem increase as the sensor sensitivity is increased.

I’m not aware of any real breakthroughs in “sensor sensitivity.” The underlying photo diode that’s doing the photon/electron conversion hasn’t changed much over the years. “Sensor sensitivity” in the sense you probably mean it is much more likely to be changed by Bayer filtration and other factors. What’s changed is not the sensitivity, but the noise propagation (or lack of it, actually). We simply are getting better at keeping electrons where they belong and counting them correctly. Has nothing to do with a more sensitive sensor.

> This would lead one to believe that once a certain density is reached the problems would negate the increased resolution possible.

I’ve been running linear regression on sensor data since the late 1980′s. Sensors are still on the same basic improvement line that they’ve always been on. Every now and again we get a sensor that’s above (D3s) or below (GH2) the projected curve, but the curve is a pretty good predictor overall. So I’d say no to this notion.

> Also, since only about 12 mega pixels are required for a good resolution photo,

Funny thing is, if you had perfect 24mp and 12mp sensors and all else were treated the same (same final print size, for example), most people would pick the 24mp output. Why? Visual acuity would look higher.

> that would lead one to conclude that this would force manufactures to conform to a fixed sensor density.

Nothing of the sort. I don’t know of an industry or example where once it was clear that you could make constant, incremental progress, that wasn’t pursued until the end of the line. Those who aren’t competitive (Olympus doesn’t have a sensor design of their own) are usually the only ones that say “let’s get off the improvement curve.”

> If greater ISO performance or greater dynamic range is needed, then a larger sensor would be required.

Again, no. The current 16mp APS/DX Sony sensor is better than the old 6mp APS/DX sensor, and in so many ways that it’s difficult to describe. Again, it’s that constant improvement line thing.

> Marketing is a factor, but I do not believe this is significant. When holding two cameras in my hand, one with 12 mega pixels and the other with 16 would I necessarily choose the 16 because I think it is better? I also would think, that I do not want to clog up my computer with more data than I need.

The smart salesman sells you a new computer, too ;~). Good thing you’re buying at Best Buy ;~). Marketing is more of a factor than you think. When a competitor like Olympus announces that 12mp is enough and everyone else is still improving sensors, you’ve got a real marketing advantage (assuming you actually can execute marketing well, which isn’t a given with the Japanese companies). It’s like GM saying 30mpg is enough because they don’t have an engine that will go better. How much do you think Ford, Toyota, et. al would jump on that with a different marketing message?

> What I believe is happening is that image quality is improving by increasing the pixel density, depite the issues with noise.

This is true, except for the last phrase. Noise needs to be measured over a constant area. I can show you 16mp Sony DX sensor images that blow away the old 6mp Sony DX sensor at ISO 3200. Now how can that be? If you do a Standard Deviation on a gray field, the 16mp sensor is “noisier.” Problem is, the TEST is wrong ;~).

>Look at the results of the GH2 which has the performance of an APS-C sensor.

Uh, no. The GH2 sensor does not approach the current APS/DX state-of-the-art sensor. Indeed, the GH2 sensor seems to be a slight step backwards from the GH1 sensor overall (remember, I measured it as below the expected improvement curve).

It’s actually difficult for me to figure out what’s driving Panasonic’s sensor design decisions at the moment. I suspect that the GH2 was driven by video parameters that they wanted to meet, not still parameters.

> It appears they do this by selectively driving some of the photoreceptors at a slightly higher voltage to gather image detail in dark areas.

That’s not how you get a sensor to have less noise ;~). Remember, the photo diode technology hasn’t changed a lot. What’s changing is the stuff AROUND the photo diode. Essentially, we’re getting better at keeping electrons where they should be and counting them. Simple as that.

> Segmenting the Sensor

Simply put: no.

> If I am right, we are going to see the mega pixel density to continue to go up.

Well, you’re right about the conclusion, but your logic doesn’t support your conclusion ;~).

> I would not be surprised to find that in 5 years the 4/3 sized sensor will be taking 30 mega pixel images, (outputting only half of that) and have better performance than the full frame sensors of today.

I think that slightly doubtful, actually. There’s more bang for the buck in getting rid of Bayer. Bayer takes a sensor that would be, oh, ISO 400 or 640 or something like that, and robs light before it gets to the light gathering component. Foveon’s approach has its flaws, but Nikon’s approach is highly promising if it can be manufactured.

Driving sensors too high in resolution becomes a road of declining returns. Diffraction and lens performance suddenly starts to be taking away almost as much as you gain. In APS/DX, I calculate the breakover point as about 24mp. Below that and pixel gains show up readily. Above that, and there’s so little gain (especially with existing lenses) that most people would see no difference. (It takes about a 14% increase in resolution for most people to see it).

Delete my comment as I just noticed the caption under the sensor pictures does say sensor for GH2 and not G3 as I assumed all too readily.

Re: 10Mp’s would be more than enough ….

-Actually, look at the compact cameras, using CCD or BSI CMOS. They are already at 16 MPix today.

However, the larger sensors for the system cameras (eg. PEN & DSLR) will also be using more modern semiconductor processes and lithography. I think that is what we are seeing with the upcoming Panasonic sensor.

I anticipate that Pana has been working with this for say 3-5 years now, hopefully with a rearchitecture in order to move the digital interface closer to the light gathering pixel.

Also the less space the supporting electronics will consume, the larger the fill factor becomes, further enhanced with micro lenses.

And how about the trade off regarding sensor optimization for digital photo versus digital video?

Expect the Olympus as well as the DMC-G3 to prioritize digital still photo, but as the market faces Full HD, the 1920x1080p with European as well as American Framrates are definitely needed!

I think the trend is towards more pixels, and at the same time better electronics for low light performance. The dynamic range issues will be resolved with advanced image processing along the evolutionary way.