[Pentaxor]

First, color separation done the Foveon way is extremely lousy regarding accuracy. The sensor relies on the photon penetration depth probabilities and by nature creates very low color separation. This causes increased noise. And this lowers the DR. Also, I specifically mentioned DR in high quality imaging. The massive sensor size advantage allows for far larger full well capacity for the MF, allowing far superior image quality over the areas where photon shot noise dominates sensor generated noise.

Second, resolution is a function of luminance, not chrominance. Without the anti-alias filter Bayer sensors lose only something like 10% resolution against a black-and white sensor (or Foveon). Demosaicig does not lose much resolution at all. If the MF cameras had the low pass filter, the differencre would naturally be greater.

Interistingly, if the patent 20100155576 describes the Foveon sensor used in this camera, it does not have 3*15 million photosites as advertised, but 15 million blue photosites and 3.8 million green and red photosites. Of course it is possible that this patent does not describe this sensor, but I've not seen any recent other (sensor) patents from Sigma either.



Plenty of the technology is known and how it influenced the DR also. Like that the poor color separation demands agressive color matrixing and this leads to increased noise.

Also, I did not say it has poor dynamic range. Just inferior to MF under certain conditions. Against the best APS-C (the Sony sensor in the K5 and D7000) it will be inferior for several reasons. If you insist, I'll go through them.

just one question, do you have the SD1 on hand that draw such conclusion?

from my understanding, you based this on the assumption that the Foveon Sensor used on the SD1 is the same Foveon Sensor used on it's smaller MP predecessors. this is based on your assumption that Sigma didn't or hasn't made any improvements on the their past sensor design due to lack of advertised or publicly shown patent. I think Pentax did the same thing as well. honestly, I would had agreed with you if and only if this was the same previous Foveon sensor design that is only increased with MP count. besides, it wouldn't make any sense that Sigma would increase sensitivity by 2 stops if they already had some problems with regards to noise with their previous cameras which I'm not refuting. I'm saying this if the SD1's Foveon sensor is the same sensor on the SD14/15, Sigma is making a big and stupid mistake to increase an already poor High ISO performance. and even yet more stupid thing if such camera is priced at $9k with a performance of an improved SD15 with the very same sensor.

Luminous Landscape said it best that it is wise to wait and see if such camera does indeed live up to the hype and worth the price. besides, I think it's better to be optimistic rather than not. who knows, they hit a goldmine rather than an old relic.

[JohnBee]

I think that photo is a bad example, the focus is on the hair with some whack lighting.

I think these images does a better job showcasing the sensor:

The second image certainly shows more eye detail than with the initial image.
Though, the level of detail on the skin is about the same(when in focus).

One interesting aspect of these samples are with the shadow details within the iris.
And I don't know if this is due to the lighting or setting, but given the perceived level of detail within the image, I'd be tempted to say there is no detail where there should be detail(DR?)

Anyways, the samples do look good, but at 3000px, I'm really not feeling the 45mp output in these images to date.
Anyone know if there are any RAW files out there yet?

[ogl]

I think that photo is a bad example, the focus is on the hair with some whack lighting.

It's NORMAL 15.3 MP photos from APS-C camera. Nothing else. Rather soft for me. But with good resolution.

I don't see MF cameras IQ or even FF. And I don't see for what to pay $10000.
M9 makes RAZOR sharp images and costs much lower.

have a look at specs....a bit funny....
WR body without WR lenses.
LCD - 460 000 dots...He-he...
5 fps and 7 RAW buffer - he-he...K-r can 6 fps with 12-13 RAW...
Camera can use only CF Type I cards...Why?

[Aku Ankka]

just one question, do you have the SD1 on hand that draw such conclusion?

No. Nor do I need one as I do have basic understanding on sensor technology and the laws of physics. They are all that are needed.

from my understanding, you based this on the assumption that the Foveon Sensor used on the SD1 is the same Foveon Sensor used on it's smaller MP predecessors.

You have misunderstood. I am basing my estimates on how silicon and light interact, how demosaicing works (for Bayer), how sensors work. I've not mentioned the old sensor at all, and if I recall right, I mentioned the quite recent Sigma patent describing another sensor, with 15/3.8/3.8 Mp layers.

I am puzzled on how you could come to the idea that I somehow have guessed what the SD1 sensor if from the performance of some other sensors.

this is based on your assumption that Sigma didn't or hasn't made any improvements on the their past sensor design due to lack of advertised or publicly shown patent. I think Pentax did the same thing as well. honestly, I would had

First, I did mention certain patent (either in reply to you or to another person) whch describes an improved sensor. The biggest single improvement is the use of (a modified) 4T design which allows for correlated double sampling to take care of one of the two biggest sources of Foveon noise, the reset noise. This should lower the read noise from what it is now (it's somewhere between 30 and 60e-, can't remember for sure) to somewhere between 5 and 10e- (or a bit more or less) at the base ISO. This is a massive improvement. The other main source of noise with Foveon is unfortunately not fixable with a three-layer silicon design - it is the color separation problem. Due to the probabilistic nature of photon absorbion in medium, the system which relies on photon penetration depth for color separation is by nature of very weakly signal separating. This leads to massive noise for color imagery.

agreed with you if and only if this was the same previous Foveon sensor design that is only increased with MP count. besides, it wouldn't make any sense that Sigma would increase sensitivity by 2 stops if they already had some problems with regards to noise with their previous cameras which I'm not refuting.

Well, their pricing makes no sense either, lots of other stuff in the world also make no sense.

Anyhow, you evidently quoted the ISO 6400 specification compared to ISO 1600 or the earlier Sigma cameras. You should make a note that this is a camera specification, not sensor. Assuming that there is a variable signal apmlifier in the sensor, it is likely to be able to boost the signal at most to ISO 1600 (this is pretty much standard for all the sensors as it makes little sense going beyond - ideally of course there would be no need for VGA, and that day is getting closer) and the higher ISOs are just software post processing. For example the sensor inside the Pentax K-5 has a maximum ISO of 1600 as well. The rest is just camera pushing in software.

I'm saying this if the SD1's Foveon sensor is the same sensor on the SD14/15,

The sensor has different dimensions, different pixel pitch, different resolution - how on earth it could be the same sensor?

Sigma is making a big and stupid mistake to increase an already poor High ISO performance. and even yet more stupid thing if such camera is priced at $9k with a performance of an improved SD15 with the very same sensor.

The ISO numbers of the camera are absolutely irrelevant for this topic, nor do they have anything to do with the noise performance of the sensor compared it's predecessor. It would have made sensor for the Sigma marketing department t demand ISO 100.000 just to make the camera specs look better...

Luminous Landscape said it best that it is wise to wait and see if such camera does indeed live up to the hype and worth the price. besides, I think it's better to be optimistic rather than not. who knows, they hit a goldmine rather than an old relic.

LL is one of the worst sources for technology information in the internet. And yes, before you ask, even I am a better source

There is no reason to be optimistic or pessimistic. Enough facts are known - about sensor technology in general, about the Foveon sensor technology principles, about how resolution is measured and how our eys see it, about how our eyes see color compared to how the color is created by different technologies, incluging the silicon separation of Foveon, of photon behavior and so on. There is no magic, nor many parameters we don't know.

This new sensor will have quite a bit better signa/noise ratio than it's predecessor, especially for black and white photography, it'll also have quite a bit more resolution than it's predecessor, though far less than Sigma fans and marketing would want us to know. Borrowing your terminology - it won't be a goldmine, nor a relic, just a logical evolution of the previous generation.

Anyhow, at that price point it won't sell at all. It simply is a weak product for professionals of pretty much any field, apart from someone shooting landscapes and travelling light. It won't be good for sports and such for obvious reason, it hardly is the most sensible tool for photo journalism at that price, and typically what is shot with medium format requires accurate colors (for human color vision), something that is impossible to to achieve with the Foveon tehnology (it is not someting that can be fixed in that techonology) and also MF has far higher resolution, far better tonality and so on.

Still, if/when the price comes down to under 1k, I might buy one as it is an interesting product, though probably not as by that time the world has moved on.

[ogl]

The sensor has different dimensions, different pixel pitch, different resolution - how on earth it could be the same sensor?

LL is one of the worst sources for technology information in the internet. And yes, before you ask, even I am a better source
.

It seems to me you are troll. Sensor could has different dimension, but it's the same TECHNOLOGY. Have a look at official samples - the same problem with colours.

LL is MUCH clever than you and more professional

[Pentaxor]

No. Nor do I need one as I do have basic understanding on sensor technology and the laws of physics. They are all that are needed.

then your findings are flawed. understanding the laws of physics and sensor technology that you only know isn't enough if you don't have the actual sample that would support such claim as such. textbook knowledge is only as good as a mere theory without proper experimentation that would support the as evident fact.

You have misunderstood. I am basing my estimates on how silicon and light interact, how demosaicing works (for Bayer), how sensors work. I've not mentioned the old sensor at all, and if I recall right, I mentioned the quite recent Sigma patent describing another sensor, with 15/3.8/3.8 Mp layers.

I am puzzled on how you could come to the idea that I somehow have guessed what the SD1 sensor if from the performance of some other sensors.

thank you for clarifying this as I had assumed that you had any experience with any Sigma cameras. but since it is evident that you haven't, this makes your findings highly suspect.

First, I did mention certain patent (either in reply to you or to another person) whch describes an improved sensor. The biggest single improvement is the use of (a modified) 4T design which allows for correlated double sampling to take care of one of the two biggest sources of Foveon noise, the reset noise. This should lower the read noise from what it is now (it's somewhere between 30 and 60e-, can't remember for sure) to somewhere between 5 and 10e- (or a bit more or less) at the base ISO. This is a massive improvement. The other main source of noise with Foveon is unfortunately not fixable with a three-layer silicon design - it is the color separation problem. Due to the probabilistic nature of photon absorbion in medium, the system which relies on photon penetration depth for color separation is by nature of very weakly signal separating. This leads to massive noise for color imagery.

again, this is still highly questionable due to absence of real experimentation and experience of the SD1.

Anyhow, you evidently quoted the ISO 6400 specification compared to ISO 1600 or the earlier Sigma cameras. You should make a note that this is a camera specification, not sensor. Assuming that there is a variable signal apmlifier in the sensor, it is likely to be able to boost the signal at most to ISO 1600 (this is pretty much standard for all the sensors as it makes little sense going beyond - ideally of course there would be no need for VGA, and that day is getting closer) and the higher ISOs are just software post processing. For example the sensor inside the Pentax K-5 has a maximum ISO of 1600 as well. The rest is just camera pushing in software.

the Pentax K-5 real maximum ISO is 1600? where did you get that?

The sensor has different dimensions, different pixel pitch, different resolution - how on earth it could be the same sensor?

I wasn't talking about APS-C sensor size versus FF sensor size versus MF sensor size. I'm talking about the sensor technology used on each system. you got lost.

The ISO numbers of the camera are absolutely irrelevant for this topic, nor do they have anything to do with the noise performance of the sensor compared it's predecessor. It would have made sensor for the Sigma marketing department t demand ISO 100.000 just to make the camera specs look better...

of course it is relevant. it would show how Sigma improved the so-called noise performance at ISO levels that you are worried about. of course, anyone can add a few more stops on the camera. but I was referring to the performance at High ISO levels. as I said previously, it wouldn't make sense for Sigma to boost up the ISO if the performance is not even up to it. the only possible reason why they would boost up the ISO is if they made some real significant improvement of the Foveon sensor with respect to performance.

LL is one of the worst sources for technology information in the internet. And yes, before you ask, even I am a better source

thanks for the laugh. but the difference between you and LL is that LL actually do some tests and experimentations, aside from textbook knowledge inorder to see the consistency to their findings if it is such or not. in other words, they have more bearing than what you could only offer.

Still, if/when the price comes down to under 1k, I might buy one as it is an interesting product, though probably not as by that time the world has moved on.

this ultimately defeats your argument considering that there are better cameras out there as you have generalized. also why the interest if you already knew more about the SD1 and I quote "you knew already about sensor technology and their limitations and the laws of physics" even if you don't have the SD1 nor have any experience of it yet. it's funny how you the devalue a camera yet interested to get one for a price that would be able to afford it. very ironic, not to mention pointless as well.

[ogl]

K-5 ISO800


SD1 ISO100


good answer of johnbee
645D & Sigma SD1 are in the same boat: Pentax SLR Talk Forum: Digital Photography Review

[thibs]

Yep, K5 max sensor iso is 1600 which why firmware only allowed (fixed since then) to use 1600 max for B mode, which is somehow quite logical.

[Digitalis]

Sigma cameras are ok, they have good skin tones - not as nice as the out-of-the-camera skin tones fuji cameras have. I used one in a studio in sydney for a bit,several years ago. With flash at 5500K the sensor delivers quite good results - I never saw anything resembling the 3D rendering sigma photographers debate to the death over.

The biggest issue I have seen from the sigma X3 sensor is when you mix lighting, say flash and tungsten - it handles this rather poorly because of the noise from the other channels and the way the sensor separates colours - flourescent lighting plays merry hell with X3 sensors too, from what I have seen X3 sensors also have weak response at the blue end of the spectrum.

[Pentaxor]

Sigma cameras are ok, they have good skin tones - not as nice as the out-of-the-camera skin tones fuji cameras have. I used one in a studio in sydney for a bit,several years ago. With flash at 5500K the sensor delivers quite good results - I never saw anything resembling the 3D rendering sigma photographers debate to the death over.

The biggest issue I have seen from the sigma X3 sensor is when you mix lighting, say flash and tungsten - it handles this rather poorly because of the noise from the other channels and the way the sensor separates colours - flourescent lighting plays merry hell with X3 sensors too, from what I have seen X3 sensors also have weak response at the blue end of the spectrum.

yeah, the blue end seems to be the major issue in the Foveon. it interests me to know how Sigma would be able to overcome that on the SD1.

[gazonk]

I'll save my verdict until at least we can compare samples at dpr or ir - but I'll be really surprised if this camera is even close to the 645d in detail.

[Pentaxor]

price has been reposted again and back at $9,700. what sounds strange is that buying the SD1 kit is significantly less expensive than buying the camera only. but anyways, I guess that's it. any hopes of attracting consumers into buying a Sigma is up in smokes. I guess I have to look up elsewhere where a newly released camera would cost around $1,500.

[macTak]

what sounds strange is that buying the SD1 kit is significantly less expensive than buying the camera only. but anyways, I guess that's it.

I agree that is very strange. While I can understand a camera company wanting to attract new customers with attractive bundle pricing, this is (in my opinion) something of a slap in the face to those loyalists who have already bought into the system and own a number of SA lenses. I still can hardly believe all of this pricing stuff...

[Pentaxor]

I agree that is very strange. While I can understand a camera company wanting to attract new customers with attractive bundle pricing, this is (in my opinion) something of a slap in the face to those loyalists who have already bought into the system and own a number of SA lenses. I still can hardly believe all of this pricing stuff...

imagine how many shares from Sigma I could had bought for $9,700. atleast I wouldn't mind or hesitate to put out that kind of money if those were stocks that I'm buying.

anyway, I think Sigma lost more rather than gain more followers with their announcement. most of their loyal followers had already made up their mind of buying into other systems. if there was something positive that came out of this price, it did help boost the sales of other camera manufacturers in an instant.

I'm not sure if Sigma is going to do something about this or if someone's head is on the chopping block. I would hate to be in Sigma's position right now. talking about epic fail in the massive scale.

[uccemebug]

I'm not sure if Sigma is going to do something about this or if someone's head is on the chopping block. I would hate to be in Sigma's position right now. talking about epic fail in the massive scale.

I don't get it at all (the LL article comes to the same conclusion you do). It strikes such an odd note that it's disturbing. Perhaps we're going to see substantial price increases from all of the manufacturers in Japan? Maybe not on this scale, but maybe we're in for some surprises.