Is Sigma SD 14 really that good?

[Dakotah]

In reading and looking at the information sent me by Sigma the camera looks pretty good. I am only interested in the ultimate image quality and don't need fast frame advance and some of the other features I can get with Canon or Nikon in comparison.

Is the Sigma with the three layers on top of each other really 'that good'?

It sounds as if it is but all I see is their information, no third party testing.

What is reality for this body?

I come from years of shooting 11x14 and contact printing...... that is the comparison I am using while trying to decide which digital system to get into. Am looking for the sharpest image with the cleanest transitions in color and gradation.

[ARD]

Review

[Dakotah]

Yes, I saw that. It is pretty much what Sigma sends out. It will be nice to see a real comparison to see if the camera actually is 'better' in terms of ultimate image quality than the competition. I will mainly be using the digital camera I do get on a tripod and want the highest quality possible. So far the Canon 16 megapixel model is what I am looking at. If this is an improvement over that I will give it a go.

[JLK]

Yes, I saw that. It is pretty much what Sigma sends out. It will be nice to see a real comparison to see if the camera actually is 'better' in terms of ultimate image quality than the competition. I will mainly be using the digital camera I do get on a tripod and want the highest quality possible. So far the Canon 16 megapixel model is what I am looking at. If this is an improvement over that I will give it a go.
[a href=\"index.php?act=findpost&pid=107035\"][{POST_SNAPBACK}][/a]

I'm a dyed-in-the-wool Sigma shooter. The SD14 is an extremely nice camera, but I do not think it is in the same category as the Canon 1D series for most stuff (if you're only doing macro work, that might be another story).

Having said that, it won't hurt you to go to www.pbase.com/sigmadslr to see some full sized images for yourself. What kind of shooting do you do? If price is no object and your style is appropriate, a scanning back is a good choice...

[pss]

In reading and looking at the information sent me by Sigma the camera looks pretty good. I am only interested in the ultimate image quality and don't need fast frame advance and some of the other features I can get with Canon or Nikon in comparison.

Is the Sigma with the three layers on top of each other really 'that good'?

It sounds as if it is but all I see is their information, no third party testing.

What is reality for this body?

I come from years of shooting 11x14 and contact printing...... that is the comparison I am using while trying to decide which digital system to get into. Am looking for the sharpest image with the cleanest transitions in color and gradation.
[a href=\"index.php?act=findpost&pid=107022\"][{POST_SNAPBACK}][/a]

if you are really used to a 11x14! camera (where do you get the film?!) not even a P45 will make you happy....and any canon, nikon or DSLR will just make you laugh....
the sigma 3layer sensors are very nice, crisp...i used to have a sd9..very nice...colors had some problems, high iso wasn't there, but all in all a very capable camera....i believe the sd14 will be a major improvement.....but nothing compared to even a P20.....

[Eric Myrvaagnes]

If (and it's a big if) you are never going to print bigger than 11x14", then you might be happy with a Canon 1D series or an MF digital back.

I shot 35mm, 6x7, 4x5, and 8x10 in my film days (never 11x14, but I have two friends who did), and my Canon 5D does everything I want now. Does it give the same quality as a good contact print? No.

Digital is still catching up, but film is getting taken away faster than digital can fill the gap, IMHO.

Good luck finding the quality you want.

[Let Biogons be Biogons]

That's not a "review".  That's a manufacturer's press release!

Review
[a href=\"index.php?act=findpost&pid=107033\"][{POST_SNAPBACK}][/a]

[John Clifford]

I know I'm late to this thread, but...

I find that the Sigma dSLRs will outresolve any Bayer sensor-equipped dSLR at the pixel level. However, as Bismark said (about armies), 'mass has a quality all its own.' Because Bayer sensors will never be able to resolve down to the pixel level, the manufacturers keep cramming more and more pixels in. Once they can get pixel density to a level so that the lens becomes the limiting factor and no AA filter is required, that's as good as it gets. My back-of-the-envelope calculations tell me that, on an APS-C sized sensor, you'd need about 50 MP (12 MP for Foveon).

To compare Foveon sensors with Bayer sensors, realize that Bayer sensors resolve to about 66% of theoretical resolution, e.g., a 12 MP Bayer sensor will be equivalent to a 4 MP Foveon sensor. Now, Bayer sensors will do much better than this on a black and white resolution chart, but with full-color subjects this is about what you get.

The big advantage of Foveon sensor-based images is that they can be up-rezzed to much higher levels than comparable Bayer sensor-based images.

[BernardLanguillier]

I come from years of shooting 11x14 and contact printing...... that is the comparison I am using while trying to decide which digital system to get into. Am looking for the sharpest image with the cleanest transitions in color and gradation.

Only one word comes to mind: stitching.

Today, you could get a used Phaseone P30 and use it with a Mamiya and 100+ mm sharp lens on a RRS pano head, buy PTgui and Autopano Pro licenses. That should mostly do the trick.

It might be better to wait a few more weeks if Nikon indeed releases something interesting in the high end of not.

Cheers,
Bernard

[petermarrek]

If you are used to 11x14 contact prints, you might as well get a nice mug of hot chocolate, turn on a tv and bore yourself to death.

[Graeme Nattress]

There are a number of aspects to the Sigma camera and  it's sensor that need to be taken into account.

1) it lacks an optical low pass filter, and hence it aliases quite badly. That means a significant part of the resolution you see is either false, or corrupted. Now, I know people like the look of this, but....

2) although the stacked foveon doesn't "need" interpolation  to make an RGB image (whereas a bayer pattern CFA sensor image does), the "decode" pipeline for a raw Foveon image to RGB includes conversion to a luma / chroma colour  space, and chroma noise reduction (ie, interpolation) so if you're looking for strictly "no interpolation" the Foveon does not deliver it for you.

3) Why the chroma noise reduction? Because silicon is a pretty poor colour filter, and hence although the luma you get is relatively noise free, the chroma is not, and hence needs a bit  of work to make it acceptable. This also limits colour accuracy.

4) Resolution. Resolution needs to be measured, not inferred from  the pixel count, no matter how you define the word "pixel". Some would say that the Foveon resolution is the same as  it's pixel count, and sure enough, it will probably measure that, but only because of it being allowed to alias. Real world measurements of bayer sensors are numerous, but the one I'm familiar with measures 78% of the linear resolution. That would give your 12mp sensor equivalent to a Foveon at 7.2, but you'd have less aliasing (good) and better colour accuracy (good) and probably much better high ISO (very good). Because of the interpolation on the chroma needed on Foveon, you cannot strictly declare it's chroma resolution to be equal to it's luma resolution - again, you'd have to measure it, but this is very tricky due to the adaptive nature of whatever noise reduction is being applied. I've not seen a good chroma resolution test that doesn't also involve a luma edge...

5) if you get get the pixel density high enough, and still keep noise low enough, then yes, you can do without an OLPF and the lens does then become the limiting factor. However, we're quite a way off being there yet.....

I know I'm late to this thread, but...

I find that the Sigma dSLRs will outresolve any Bayer sensor-equipped dSLR at the pixel level. However, as Bismark said (about armies), 'mass has a quality all its own.' Because Bayer sensors will never be able to resolve down to the pixel level, the manufacturers keep cramming more and more pixels in. Once they can get pixel density to a level so that the lens becomes the limiting factor and no AA filter is required, that's as good as it gets. My back-of-the-envelope calculations tell me that, on an APS-C sized sensor, you'd need about 50 MP (12 MP for Foveon).

To compare Foveon sensors with Bayer sensors, realize that Bayer sensors resolve to about 66% of theoretical resolution, e.g., a 12 MP Bayer sensor will be equivalent to a 4 MP Foveon sensor. Now, Bayer sensors will do much better than this on a black and white resolution chart, but with full-color subjects this is about what you get.

The big advantage of Foveon sensor-based images is that they can be up-rezzed to much higher levels than comparable Bayer sensor-based images.

[Plekto]

Been away for a few weeks.  Busy in RL...    

The technology of the Sigma is currently the closest to film that has been made with digital formats.  It has by far the best shoulder and dynamic range that I have seen(one except, which I'll get to below), and truly does look like film.

But the resolution is poor.  Literally like APS format if you're looking at film.

Film at a standard 2400DPI scan is ~94 full color dots per mm.  This yields roughly 2250X3400 resolution, or about 7.6MP.  Note, though, that this is pure resolution.  32 bit color data at each location with no moires or filters or other junk in the way.    The best Bayer type patterns barely get 0.7 X in each dimension, which is pretty good, actually, considering what a kludge the original pattern was.  This yields roughly 3200*4800 with a Bayer sensor to make 35mm film moot.   15.3MP, give or take.  Let's just say 16MP.  There are many cameras out now that do this and really DO make film an exercise in futility.  Other than the problems with the Bayer pattern and CCD/CMOS/etc sensors.

The best are at about(just under) 0.8, which are a couple that use non Bayer type alignment patterns.  The Fuji S5 is a good example.  By utilizing a hexagonal pattern.  This lowers the number of locations to about 2800*4200.  (11.7MP) - this is why the Fuji cameras at 12MP can equal a typical 16MP Canon.  

Anyways - the problem is that the Foveon/Sigma sensor has a huge problem.  The resolution is like film at 1.0X factor(yay!), but the darn thing only has a 1,768 X 2,652 pixels.  Not close to the ~2250*3400 that would be required.  So while it looks fantastic, it lacks detail, because it's just not high enough resolution.  As a result it creates very "soft" photos.

* I mentioned one other camera earlier that deals with shoulder and dynamic range properly.  It's the Fuji S series.  It has less moire due to a hexagonal pattern(our eyes don't see the aliasing and moires and defects as easily, much like how a lens with a 9 leaf aperture produces less noticeable starring than say, a 5 or 6 leaf design.

It also has a special mode where it lowers the resolution to about Sigma's but essentially shoots two bracketed shots at the same time and blends them together.  This is similar to the "Zero Noise Technique" that is discussed on this forum.  It effectively boosts the ratio for the camera to about 0.9X as near as I can tell and adds huge dynamic shoulder as well.  The results look stunning to me.  As much as I'm a die-hard Sigma fan, the S5 just does better in real life shots where there is difficult lighting or less than storybook outdoor material.(cold rainy day in a forest - no contest between the two)

Note - the S5 isn't *quite* full resolution, just like the Sigma.  I'd say it's close in this mode to 90% of 35mm film per pixel.   It's sharper than the Sigma by a little bit and most people honestly don't care.  

Also, the Fuji software is a LOT easier to use than the Sigma nonsense.  And it uses Nikon lenses(huge bonus).  

***
Though - let's be honest, neither is a 20MP+ SLR.  They both need a higher resolution version of the same design ASAP.  Both effectively are like shooting 35mm film with the last 5-6mm on each side cropped off.  It's SO close to what we want, but not good enough.  And it's been a could of years for both of them.  While they've stood still, 16-20MP cameras with the poorer Bayer technology are breeding like rabbits.

[Graeme Nattress]

I've not seen any test charts showing an enhanced dynamic range or shoulder non-linearity in the Foveon sensor. Also, it only achieves it's 1.0x resolution factor (in luma, not chroma, as I mention above due to the NR) at the expense of masses of aliasing artifacts, which film, due to the random sampling of the grains, avoids.

Graeme

[Plekto]

True.  The sensor itself has no such issues with quality - it appears to be Sigma's crappy hardware/implementation that's responsible for the problems and artifacts in the image.  I found that the Sigma basically works as a Velvia 50/100 camera and attempts to move it from its lowest ISO setting result in a disaster, much like pushing 50 speed film does(except since it's digital, it goes into massive amounts of noise and other ills).  Way faster and harsher than a typical DSLR as well.  Even ISO 400 is unusable for anything other than *maybe* Ebay photos.

Great for portraits and outdoor scenery where there's tons of light and a tripod or the lighting is carefully controlled, so that you can use it at the slowest speed and manually push the exposure a bit.

It just got to be too much for me to deal with.  From the software to the idiocy of the controls to the fact that you basically were limited to the lowest ISO all of the time...  Gorgeous technology.  That they should have given to Sony or someone else who had a clue how to properly implement it.

Sigh.  Such a disappointment given that when it works right, it really makes your jaw drop.  But worrying about whether each shot would come out correctly every time - I might as well have been back to my old fully manual Rollei.

***
The Fuji is a joy to use by comparison from even the half hour or so I've messed around with it at a show.  Sure, it has its foibles, but basing the thing off of a Nikon body was a huge plus(much like how Sony has basically left Minolta's designs intact).

Think of it as a Bayer 2.0 design.  An appropriate evolution of the original.

It has good dynamic range, and looks to be very forgiving of tweaking with software to recover overexposed shots as well, plus you don't have to manually bracket and blend, which is a godsend for those really low light and nighttime shots - since you get high dynamic range AND the darker areas don't get all full of noise nearly as easily.  

Where with the Sigma, you were toast - ISO 400+ was a disaster/unusable and pushing it via software resulted in an even worse mess.  I found myself with the Sigma, honestly, reverting to my 20 year old Minolta film SLR and a fast lens instead for city and night time shots.  It was a nice month or two that I owned the Sigma(earlier model, but basically the exact same sensor technology), but it was just not the right choice for me.   Thankfully I bought it used, so didn't really lose any money in the deal.

I'm just wishing Fuji would jump to 16-20MP...

[Graeme Nattress]

The fuji does a simple trade of resolution for dynamic range. If you  don't want to trade, then shoot bracketted locked off shots and post-process.

I wouldn't lay all the blame of the image at Sigma. Foveon themselves say you still need an OLPF, but Sigma don't use one. Foveon themselves outline the image processing pathway and how the chroma NR is needed. If silicon were a better colour filter, then all would  work a lot better, but it's not... Think of it as a great luma sensor, with a noisy chroma overlay.

[Plekto]

True.  Fuji *does* give you the option to go to high resolution for those more composed shots, but being able to essentially bracket with action shots on the fly as well is a really nice feature.  It kind of spoils you, since the images are really quite free of moires and other junk.

Also, evidently, you can take the raw file which is really two images and DIY blend them if you want - that's why the raw file size is so huge.  Given how massive memory cards are today, it's a real non-issue, IMO, other than the slow cycle time.  Though, it's no worse than bracketing.  That also causes a wait for most cameras.

Yeah, I was really hoping that Kodak or one of their other partners would have implemented the technology correctly.  True, silicon isn't the best/optimal option, but I'm sure there are other materials as well that might work better.  But Foveon isn't really looking like it's doing anything recently.   We wait and wait and... nothing.  Same with Sigma.  

So my money is on Fuji between the two, for now, or at least until we see what Nikon and Sony are doing with their next generation cameras.  Lots of speculation but not much to actually hold in our hands...

[Graeme Nattress]

Well, I think the big sensor push is now coming from  the video side of things. Once you can be writing your raw to your CF in real time, it opens up a lot more possibilities for photographers, even if you're  not capturing action, I think.

[John Clifford]

True.  The sensor itself has no such issues with quality - it appears to be Sigma's crappy hardware/implementation that's responsible for the problems and artifacts in the image.  I found that the Sigma basically works as a Velvia 50/100 camera and attempts to move it from its lowest ISO setting result in a disaster, much like pushing 50 speed film does(except since it's digital, it goes into massive amounts of noise and other ills).  Way faster and harsher than a typical DSLR as well.  Even ISO 400 is unusable for anything other than *maybe* Ebay photos.

[...] It was a nice month or two that I owned the Sigma(earlier model, but basically the exact same sensor technology), but it was just not the right choice for me.

You can't compare the SD9/SD10 with the SD14, especially when the SD14 is using the latest firmware (1.08). The difference is enormous. The earlier models, as you said, didn't like low light. The SD14 will tolerate low light without excessive grain or green casts in the shadows. Even so, the SD10 will also take beautiful landscape images.

Again, the SD14 is not the do-all, be-all, end-all of dSLRs. If you want to shoot action, choose something else. If you want to shoot high speed under low light conditions, choose something else. But, if you are into landscape photography, then know that the SD14 will hold it's own against the 5D in terms of per-pixel image quality, and that a 4.6 MP Foveon image that is up-rezzed during post-processing (not in-camera or using Sigma's SPP raw software) will print as nice or nicer than any 12 MP image from a Bayer sensor-equipped camera. Hard to believe, I know... but true.

This is why I think even a 7 MP Foveon sensor, perhaps in a 1.5x size, would totally kick even a 1DSMKII's butt... for landscape photography.

[Plekto]

You can't compare the SD9/SD10 with the SD14, especially when the SD14 is using the latest firmware (1.08). The difference is enormous. The earlier models, as you said, didn't like low light. The SD14 will tolerate low light without excessive grain or green casts in the shadows. Even so, the SD10 will also take beautiful landscape images.

The difference between the two isn't THAT great, really.  It's great for landscapes as I already mentioned.  But I expect a camera to be able to do more than one trick.  It's just too limited as it currently stands.  And Foveon seems to be taking its sweet time in doing anything to improve the situation.

And removing hot pixels(noise) from the sensor via software at higher ISOs isn't a real fix - the more you push the SD14, the more it looks like an oil painting at best, and crap at worst.  ISO400 is garbage on the SD14 - and ISO200 was the limit on the SD10.

Again, the SD14 is not the do-all, be-all, end-all of dSLRs. If you want to shoot action, choose something else. If you want to shoot high speed under low light conditions, choose something else. But, if you are into landscape photography, then know that the SD14 will hold it's own against the 5D in terms of per-pixel image quality, and that a 4.6 MP Foveon image that is up-rezzed during post-processing (not in-camera or using Sigma's SPP raw software) will print as nice or nicer than any 12 MP image from a Bayer sensor-equipped camera. Hard to believe, I know... but true.

No, not hard to believe at all   BT, DT, owned the camera...  heh.

But it's just not in the league of the Fuji.  The two look almost identical in quality, but the Fuji sensor - at least the older one in the S5 that had the dual size sensors - did a great job with low light and action as well.  Much fewer compromises.  I'm waiting though to see what the jury says about Fuji's new version of the sensor.  

This is why I think even a 7 MP Foveon sensor, perhaps in a 1.5x size, would totally kick even a 1DSMKII's butt... for landscape photography.

Maybe... but where the hell IS it?  I want a 1:1 film like sensor.  NOW.  This would also have a huge advantage with upscaling and printing.  It would remove a couple of variables from the software and produce cleaner results, since it would be exactly like a scanned 2400DPI film's result.  Sigh.  They really wouldn't even have to do much else other than make a larger version of the same exact thing.  Though, I'd personally be tempted to add a fourth color channel to create a smoother blend.(ie - CYMK vs RGB).  Hey, I can dream, no?