A better way of phrasing the color differences of the Quattro vs. the Merrill

In another thread Kendall came up with a good term "true color" that I thought might help clear up the differences we've been having over our initial impressions of the Quattro. So using his term this is how I've rephrased the difference:

Because of the averaging the Quattro does at the 5mp level it can only know "true color" at that level, whereas the Merrill can know "true color" at the 14.7mp level. Now due to the top level having 4 separate pixels it can show some color variation at the 19mp level but those variations aren't "true color" themselves, just minor variations of the larger "true color" from the 5mp level.

It is still a Foveon chip and is obtaining its colors in the traditional Foveon way, but only at the 5mp level. So the Quattro will have higher color discrimination that the SD15 but less than the Merrill.

Personally I would have preferred, strictly from an artistic/fine art point of view, for Sigma to have retained the higher level of color discrimination of the previous generation Merrill in a new improved body, but I fully understand Sigma's choice.

If the only option for them to address the many criticisms against their cameras, especially poor low light capability, slow processing times and limited battery life, which definitely limited their marketability and appeal to a larger audience (and hence limited their income--they are a for profit company after all), then I can see why they did it.

The new Quattro should be better than the Merrill in those three areas along with (probably) improved DR and less shadow noise, then the improvements may indeed outweigh the higher color discrimination losses.

We'll see for sure when the first samples arrive.

Interesting stuff. I think Sigma knew the hardcore Sigmatites would have some trouble with Quattro. Like you said, it seems to be the best way for Sigma to go at this point in view of the competition.

Here's the biggie for me : Looks like the Merrill sensor will not appear in future SDx cameras or anywhere else.

The notion of increased luminance resolution being the better part of color perception goes back to the development of NTSC color TV in the 1940s and early 1950s, or maybe earlier I don't know. The performance of this NTSC system was good under the best possible conditions, but you could seldom see it in any TV sets from the period. Vision studies bore out the fact that below a certain threshold color perception is lost, leaving only luminance perception, and this was highly dependent on viewing distance. Since the impression of full color was working with very low real resolution, it was possible to pack the extra data into the existing 4.5MHz video channel in a compatible way.

This trick worked well in 25" screens viewed from 8-10 feet away on the other side of the living room, but when we take our SD1 photos, print them at 24"x36", and put our noses to the print, we can see all the color resolution right up to pixel level. Prints do not offer you any control over viewing distance but TV sets sort of did, back then.

We now end up with a Sigma Quattro image made of pixel quads containing 4 different luminance values at a single color and possibly a little more color info but not much. On the other hand, color film delivered full resolution color without limitation (except for grain). I think Sigma has traded itself away from film equivalence.

Test images requiring full color resolution should show the differences plainly.

The larger pixel areas below the top layer will definitely increase performance at higher ISOs, but for me the tradeoff seems to be not desirable. If I want high ISO, I'll use Bayer.

So far the buyers of Sigmas have been mostly high end artsy phtotographer types (or hobbyist wannabes like me) and there are only so many of us. Sigma must be realizing this now, and moving on. They leave us behind in hopes that they will find a larger market, and I wish them well.

Maybe this new sensor development will increase sales of SD1Ms and DPxMs!

To me, it looks like the highest development of full-res Foveon is the DP3M (one step beyond the SD1), and none better will ever be produced. Better get them before they are gone forever! I'd order one today but since I already have a SD1 I can't seem to justify it financially.

From now on, I wear black...

I think DP Quattro is made for similar or close resolution output with DP Merrill. because DP Merrill resolving output to Bayer equivalent is around 24MP, thanks to 15x3 color resolution. DP Quattro is somehow the same but with lower color resolution and offset by high luminance resolution. overall resolution of DP Quattro may be lower than DP Merrills and possibly only at 20MP but still maintain Foveon colors. my opinion is Quattro can go up in resolution applying same principle of 4:1:1 ratio. if Sigma do that, it can easily outresolve Merrills. e.g. of 24MP Quattro with 6MP colors will be better than Merrill at similar pixel level. I think Sigma reserve this for future Quattro cameras or higher end body, possible 30MP SD Quattro of course more expensive.

my belief is Quattro is more of refined Foveon technology because of overall improvement and possibly solve weakness of old Foveon sensors.

in theory, Quattro benefits is:

-retain Foveon colors and resolution

- solve sensor flare/glare

- solve funky colors

- solve problematic skin tones in some conditions

- solve blown reds

- high dr

- high resolution

-faster processing

-small file size and more space for more pictures

- better high iso performance.

nightmare is:

-SD1 pricing again

-Bayer-like output (loss of Foveon resolution)

-theory did not materialize

Why be discouraged, when we don't even have any images yet?

And Sigma is as I recall sort of touting a 39MP Bayer equivalence. If that is true, and clearly that is not based on simply adding up pixels, then it should blow the doors off the Best Bayer cameras out there.

The "look" might be slightly different, but if it works like an older Foveon on demand, where's the problem?

Maybe you should wait to see some pictures?

Richard

Tom Schum said:

The larger pixel areas below the top layer will definitely increase performance at higher ISOs, but for me the tradeoff seems to be not desirable. If I want high ISO, I'll use Bayer.

Click to expand...

I also don't care about high ISO, but i'm sure, better high ISO performance also means better (cleaner) low ISO performance.
I miss the "soft" and clean low ISO image quality of the SD15 generation.
Now you have to carefully over expose ISO 100 to get clean shadows.

Maybe the quattro will be the perfect combination of the 4.7 MP Foveon with its nice, smooth output and the much higher resolution of the Merrills.

Maceo

I think with Quattro, Foveon Colors are better dispersed. result to accurate and cleaner color also. 20MP is more than enough my opinion. if not, Sigma can increase resolution for future Quattro models.

difficult for me to decide now whether to buy 30mm Quattro or 50mm Quattro. I have DP3M also.

while possible is 39MP output, optimal result is 20MP per pixel comparison.

Member said:

I also don't care about high ISO, but i'm sure, better high ISO performance also means better (cleaner) low ISO performance.
I miss the "soft" and clean low ISO image quality of the SD15 generation.
Now you have to carefully over expose ISO 100 to get clean shadows.

Maybe the quattro will be the perfect combination of the 4.7 MP Foveon with its nice, smooth output and the much higher resolution of the Merrills.
Maceo

Click to expand...

As stated earlier in another thread, i for 90% of the time use ISO100 and only sometimes ISO200 and i am happy with that.

But still i have some/a lot of noise in shadow/low light areas in the picture - i look forward to see the noiselevl of this new sensor.

Good appraisal. Only time will tell if less is more. But I do feel the DP Merrills are a watershed in APS-C imaging never to be equaled.

Tom Schum said:

Interesting stuff. I think Sigma knew the hardcore Sigmatites would have some trouble with Quattro.

Click to expand...

Who has trouble with it? It seems like nearly everyone likes it quite a lot, especially most of the "hardcore Sigmatites" (are they related to the Hittites?).

Indeed to me it seems the more causally attached people are the ones with the greatest doubt. In fact they are the ONLY ones as some of the greatest Foveon detractors as saying "this looks great".

Tom Schum said:

<...>

Test images requiring full color resolution should show the differences plainly.

Click to expand...

I am thinking they will show that with advanced processing of the data you can get close enough to single pixel equivalence so as not to matter.

Making statements as to audience for the new camera is meaningless without understanding real performance. But even just based on the very rough examples people have mocked together, I as a "purist" am very very happy with the new sensor.

mike earussi said:

In another thread Kendall came up with a good term "true color" that I thought might help clear up the differences we've been having over our initial impressions of the Quattro. So using his term this is how I've rephrased the difference:

Because of the averaging the Quattro does at the 5mp level it can only know "true color" at that level,

Click to expand...

Because of the top layer varying luminance but also affected by layers below it can know the "true color" at a single pixel, or so close that there's essentially no difference.

mike earussi said:

whereas the Merrill can know "true color" at the 14.7mp level.

Click to expand...

Most of the time, unless there is noise in the lower levels which negates that understanding.

mike earussi said:

Now due to the top level having 4 separate pixels it can show some color variation at the 19mp level but those variations aren't "true color" themselves

Click to expand...

They may not be but they CAN be.

mike earussi said:

It is still a Foveon chip and is obtaining its colors in the traditional Foveon way, but only at the 5mp level. So the Quattro will have higher color discrimination that the SD15 but less than the Merrill.

Click to expand...

Since the Quattro has more top level sensors, when you factor in the greater nose in the Merrill sensor's lower layers it could easily be a wash.

mike earussi said:

Personally I would have preferred, strictly from an artistic/fine art point of view, for Sigma to have retained the higher level of color discrimination of the previous generation Merrill in a new improved body, but I fully understand Sigma's choice.

Click to expand...

We don't know they have not retained a high level of color discrimination.

mike earussi said:

If the only option for them to address the many criticisms against their cameras, especially poor low light capability, slow processing times and limited battery life, which definitely limited their marketability and appeal to a larger audience (and hence limited their income--they are a for profit company after all), then I can see why they did it.

Click to expand...

Or, they just did it because it's plainly better all around? Many changes in the body already addressed some of those things (like battery life and processing times), so the new sensor was not exactly needed. The fact that it comes so soon after the Merrill to me says something, like "this is so much better we could not wait".

mike earussi said:

<..>

We'll see for sure when the first samples arrive.

Click to expand...

Yes, but in the meantime we should not be saying it CAN'T do this or that either.

Kendall Helmstetter Gelner said:

Because of the top layer varying luminance but also affected by layers below it can know the "true color" at a single pixel, or so close that there's essentially no difference.

Click to expand...

That's more than DickLyon claims for this sensor. He describes it as "imagine adding 2X resolution in each dimension, but with extra luma detail only." The reason there is essentially no visible difference is the same human perceptual concepts as with JPEG chroma subsampling.

Kendall Helmstetter Gelner said:

Kendall Helmstetter Gelner said:

whereas the Merrill can know "true color" at the 14.7mp level.

Click to expand...

Most of the time, unless there is noise in the lower levels which negates that understanding.

Click to expand...

There is always noise. And other things.

Kendall Helmstetter Gelner said:

They may not be but they CAN be.

Click to expand...

It will depend on how fast the hue changes w.r.t. the luma. Realistically, it will impact some edges that might be different when pixel peeped. You would likely only notice the difference with a contrived test or a side-by-side with just the right alignment of edge to pixels.

Kendall Helmstetter Gelner said:

Or, they just did it because it's plainly better all around?

Click to expand...

It's not better for marketing and internet forum talking points about sensor superiority ;-)

Kendall Helmstetter Gelner said:

Yes, but in the meantime we should not be saying it CAN'T do this or that either.

Click to expand...

At a minimum, you need to first prove the Merrill sensor actually CAN do this or that. Then you have a test case for the Quattro sensor.

maceoQ said:

Tom Schum said:

The larger pixel areas below the top layer will definitely increase performance at higher ISOs, but for me the tradeoff seems to be not desirable. If I want high ISO, I'll use Bayer.

Click to expand...

I also don't care about high ISO, but i'm sure, better high ISO performance also means better (cleaner) low ISO performance.
I miss the "soft" and clean low ISO image quality of the SD15 generation.
Now you have to carefully over expose ISO 100 to get clean shadows.

Maybe the quattro will be the perfect combination of the 4.7 MP Foveon with its nice, smooth output and the much higher resolution of the Merrills.
Maceo

Click to expand...

That is exactly what I also am hoping will happen. The skies should be much cleaner, and I don't think we'll lose much if any color detail over the Merrill sensors.