Examples DP1

[sdai]

Advertisement (gone after registration)

X3 sensors are actually slightly better at resolving black and white patterns (according to the figures you quoted, the 4.6 MP SD14 is on a par with the 6 MP D70s, but see below), but the X3’s real forte is its resolution in the colour channel, where conventional Bayer sensors fare much worse. This is never taken into account in resolution tests which is why lp/ph figures typically don’t square with visual results when comparing Foveon’s and Bayer sensors.

 

As far as I know, there's no industrial standard to measure "color resolution" and the terminology usually refers to the bit depth of an image, i.e. when you say the picture has 16 bit color resolution then it means it has 2^16 (16.7 million) colors.

 

The Japanese standard organization for digital camera industry CIPA only has a standard to "measure" the resolution and it doesn't take individual color channels into consideration.

 

The SD14 sensor still only has 4.6 MP in each "base color" channel, full color for each pixel (theoretically) is reproduced by stacking R, G and B pixels, right? the result is still at the 4.6MP level.

 

If somebody is going to measure the resolution of each single color channel, all he is getting is a 4.6MP resolution.

 

As many have agreed, the less than stellar results from higher pixel count "conventional" cameras are mostly due to artifacts in software algorithm, the outcome varies based on the software you use.

 

IMO, with current technology, bayer pattern has sufficient advantage to offset its shortcomings. This is especially true when you also take costs into consideration.

 

There could be something else to take the place of Bayer desmosaicing, but I suspect it could be Foveon, its slow and grinding to halt development says a lot about itself.

 

There can be a lot of great ideas on paper, turning it into (profitable) product is a whole different matter.

[mjh]

The SD14 sensor still only has 4.6 MP in each "base color" channel, full color for each pixel (theoretically) is reproduced by stacking R, G and B pixels, right? the result is still at the 4.6MP level.

No, it isn’t. With Bayer-type sensors, one has grown to expect a certain resolution from a given number of pixels, but with Foveon sensors, these expectations are way off. The quality of the images from the SD9 and 10 (3.4 MP) compared favourably to that of conventional 6 MP cameras, and the SD14 (4.6 MP) can hold its own against 8 or 10 MP models. The Foveon sensors have their share of problems, most noticeably noise, but their resolution is nothing to be ashamed of. That this doesn’t show up in standard tests using B&W charts shouldn’t deter anyone except those obsessively photographing test charts.

[sdai]

IMO, what really matters to (most of) the end customers, photographers is cost/performance ratio, right? to achieve comparable results, people will choose to pay less.

 

Let's see, suppose the SD14 works at 100% efficiency, we're not using resolution charts here, the SD14 should go for 1760 lines per picture height (according to manufacturer's spec.).

 

While the Nikon D80 works at 86% efficiency (based on Color Foto 10/2006 test results) it resolves 2254 lines per picture height, that's a 494 lines (or 28%) advantage.

 

At the same source I quoted previously in this thread (Cameta Camera on eBay), the SD14 goes for 750 US, D80 goes for about 680 US. The choice is easy.

[zapp]

The DP1 is available for pre-order from Amazon for $799. Some got it for $750 when it first appeared on the website. Expected to ship in mid to late March.

Let's call the 1000 dollar offer the suggested retail price. 800 sounds more reasonable.

 

I am using the SD14 and the Ricoh GR digital. SD14 images are by far better than the Ricoh images even with just half the physical pixels. The DP1 with the same sensor as the SD14 should be able to do the trick. For a copmpact camera the pixel size/sensor size of the DP1 is unique. You should not compare a 200g compact camera with a DSLR.

[sdai]

The Foveon sensors have their share of problems, most noticeably noise, but their resolution is nothing to be ashamed of. That this doesn’t show up in standard tests using B&W charts shouldn’t deter anyone except those obsessively photographing test charts.

 

Please bear with me for my repetitive posts.

 

Actually, I just thought that the Color Foto standard test has done a huge favor to SD14 because they're saying the SD14 can resolve 2112 lines per picture height while there are only 1760 pixel rows on board.

 

I've done some home work and compare all the past Color Foto test results and found the only two digital SLRs that can do this are all without anti-aliasing filters. SD14 and the DMR.

 

I haven't dug into this carefully but my guess is the lack of an anti-aliasing filter can produce a certain amount of false details. If you have the answers please let me know. Thanks.

[Riley]

what you mean like bayer principle sensors with interpolation and a whole lot of false details. thats gotta be rich

 

you only have to look at X3's resolution to see that it has something going for it beyond the trashing it gets and the lowest estimate 4.60Mp. Foveon has problems, but acuity of its images isnt one of them

[giordano]

Advertisement (gone after registration)

You should not compare a 200g compact camera with a DSLR.

Why ever not? With 35mm film, good compact cameras stood comparison with SLRs (Rollei 35, Konica Hexar...) and the change to digital shouldn't be an excuse.

 

More to the point, the DP1 doesn't so much invite the comparison as clamour for it. Its only reason for existence is to deliver a compact camera with a DSLR sensor.

[carstenw]

Simon, you are completely ignoring a really good point that Michael made: the resolution numbers you are basing your cost/performance argument on are bogus. If you want to look at a deeper test than Color Foto (which is a rag, IMO):

 

Sigma SD14 Resolution: Can it Hang with the Big Dogs?

 

The SD14 is up there, competing with the 5D, at least at low ISO. I am not going to say it beats the 5D, but clearly it is putting on a good show. Now, if you want to bash the SD14/Foveon for high ISO noise, go ahead, but your resolution argument is no good.

[mjh]

I haven't dug into this carefully but my guess is the lack of an anti-aliasing filter can produce a certain amount of false details.

With a Bayer-type sensor: yes. For a Foveon X3 sensor, you don’t really need an antialiasing filter. It doesn’t suffer from the issue the antialiasing filter should solve.

[sdai]

Sigma SD14 Resolution: Can it Hang with the Big Dogs?

 

The different software approach this guy adopted in his comparison pretty much invalids his test, to say the least ... so, he took a walk in his backyard, grabbed some snapshots, came back and declare Foveon is better? LOL sorry, he has no methodology at all, reading that stuff is purely wasting everybody's time IMO.

 

I repeat, what everybody sees on his web page is the difference between different processing algorithms and parameter settings, ok?

 

Also, I wouldn't rush to play down Color Foto. Can you tell me anything better (available publicly)?

 

Please, don't mention about Chasseur d"Image ... they once published a rumor made up by my friend on dpreview as gospel.

[sdai]

With a Bayer-type sensor: yes. For a Foveon X3 sensor, you don’t really need an antialiasing filter. It doesn’t suffer from the issue the antialiasing filter should solve.

 

But, you didn't answer my question, and I mean - the lack of a proper anti-aliasing filter is adding false details to the SD14 picture and test results.

 

The extra "detail" is not there in reality, where does it come from?

[sdai]

what you mean like bayer principle sensors with interpolation and a whole lot of false details. thats gotta be rich

 

AA filter will only smooth out details, fake or real ... the lack of AA will add false details.

[philipotto]

Sigma DP1 Las Vegas Photo Gallery

[neelin]

Sigma DP1 Las Vegas Photo Gallery

good to see some photos from the camera. sad to see a shy photographer afraid to get up close IN FRONT of people.

 

r

[carstenw]

Am I the only one who thinks that there might be something odd with the colours in those photos? Like they are laminated postcards from a couple of years ago or something. They don't pop at all.

[mjh]

But, you didn't answer my question, and I mean - the lack of a proper anti-aliasing filter is adding false details to the SD14 picture and test results.

In theory, there could be aliasing artifacts due to spatial frequencies beyond the Nyquist frequency, but in practice, I’ve never found this to be a problem. The main issue the antialiasing filter is supposed to solve is colour moiré, and a camera with a Foveon X3 sensor isn’t suffering from this.

[sdai]

I'm sorry if I haven't made my point clearly, Michael.

 

I agree that the SD14 doesn't need a AA filter, but this is exactly the reason why false details at a higher frequency than the highest real details are being recorded, so the Sigma shots would appear to be sharper than normal and richer in detail, but this is technically not right. It's an artifact.

 

What everyone is seeing from the casual flower sample shots falls exactly into the illusion category.

 

Any test using the conventional test procedure is actually doing Sigma a big favor. I'm not an expert in this area but I suspect a good RAW programmer could possibly develop an algorithm to remove the false details above the Nymquist frequency, then we shall the stuff in real.

 

My best bet? it's like the Hollywood celebrities without makeup.

[carstenw]

Simon, if you what see appears to be detail, and you don't see anything which is clearly wrong, why do you care? With a Canon you see a very clear and unambiguous blur, which you then need to post-process to clean up, and get some semblance of detail back into the image, minus the detail that is irrevocably lost.

 

If you see artifacts in Foveon images which are objectionable and clearly wrong, please point them out.

[mjh]

Any test using the conventional test procedure is actually doing Sigma a big favor.

If you take a look at resolution test charts (and I did), you see how contrast decreases as the spatial frequency increases, until at some point it gets all mushy and greyish. This is similar to conventional cameras, only without the colour moiré adorning the lines of the test chart.

 

Again: One might think that there could be a problem, but in practice, there isn’t.

[Advertisement]

Hallo,

Look here: Interesting Leica gear

[sdai]

I haven't seen a SD14 test chart myself but, color moire has no play in the test here as I recall that I've read somewhere that starting from 250 lp/ph desaturation kicks in and above 500 lp/ph all turns into black and white, so unlike many Foveon advocates claim, a "color resolution" test is meaningless, the black and white charts should do its justice.

 

The reason why I think that Foveon is favorably compared to Bayer pattern sensors in a conventional test is because of its capability recording higher spatial frequencies above Nymquist, the result is invalid because false details are also counted in measurement.