Would X3 catch up with Sony 6MP CCD... in time?

Peter, 70% efficiency in horizontal and vertical axis means 0.7 * 0.7 = 0.49 total efficiency = 49% efficiency.

90% horizontal + vertical is:
0.9 * 0.9 = 0.81 = 81% efficiency.

So even if your numbers are right, 81/49 = 1.653

1.653 * 3.4mp = 5.6236mp equivalence.

Good enough for me...

Rgrds,
Moshe

Peter G said:

Steven Noyes said:

It is true that you can not triple it but I think you underestimate
its sharpness as well. It is at least 2X sharper than the best
Interpolation methods. That is assuming the scene is of average
color density.

Click to expand...

Do you have a source for this info? Current decent prosumer cameras
are getting 70 -80% of the theretical max resolution on a res
chart. A good X3 based camera will probably get 90%. Even at 100%
it won't be 2x sharper. It can't be greater than 100%.

Peter G said:

In short, I would take a 1.2 MP Foveon over any 2 MP and most 3MP
P&S cameras any day. This assumes that Noise is under control. If

Click to expand...

How can you know that when you haven't seen the camera and have no
idea if the lens is any good etc....

Peter G said:

stunning sharp enlargements. Think. No Anti-Aliasing filters, no

Click to expand...

Again is there a source for this information? Generally all
sampling systems need AA filters.

Peter

Click to expand...

I am aware that it works in both dimensions, but when rating the sharpness of lens or the ability of film capture resolution, the standard is lp/mm.

If one lens does 70 lp/mm and another 90 lp/mm I don't think too many people would insist on saying the second is twice as sharp. Or insist on squaring the value to exaggerate the impact, especially when the Foveon number are hypothetical right now. The lens plays a very important part in the system and with current cameras capturing consumer cameras capturing up to 80% of max, there is currently no proof that Foveon cameras will do any beter (on a monochrome res chart). So saying they will be Twice as sharp right now seems to lack any foundation.

Even if it works out that X3 is worth a doubling of the pixels (and this is by no means a given), Bayer cameras are available with double or more. I can buy a 5mp consumer camera today. From Foveons information, It will be Xmas before I can get a 1.3 mp consumer camera. If they had gone at least to 2MP I might be interested.

In few years I hope to own a very nice X3 DSLR when the price is right, but for the next year, Fovean is essentially a non-player in the consumer space.

I originally joined this argument when the foveon bandwagon got rolling, stating that they would purchase no Mosiac camera every again and they were waiting for foveon X3 tech.

Well my view is the exact opposite. I will purchase a good mature Mosiac camera now while waiting for the truly great Foveon X3 camera that are probably a couple of years away. I let other people be "early adopters" these days.

Peter

Moshe Vainer said:

90% horizontal + vertical is:
0.9 * 0.9 = 0.81 = 81% efficiency.

So even if your numbers are right, 81/49 = 1.653

1.653 * 3.4mp = 5.6236mp equivalence.

Good enough for me...

Rgrds,
Moshe

Peter G said:

Steven Noyes said:

It is true that you can not triple it but I think you underestimate
its sharpness as well. It is at least 2X sharper than the best
Interpolation methods. That is assuming the scene is of average
color density.

Click to expand...

Do you have a source for this info? Current decent prosumer cameras
are getting 70 -80% of the theretical max resolution on a res
chart. A good X3 based camera will probably get 90%. Even at 100%
it won't be 2x sharper. It can't be greater than 100%.

Moshe Vainer said:

In short, I would take a 1.2 MP Foveon over any 2 MP and most 3MP
P&S cameras any day. This assumes that Noise is under control. If

Click to expand...

How can you know that when you haven't seen the camera and have no
idea if the lens is any good etc....

Moshe Vainer said:

stunning sharp enlargements. Think. No Anti-Aliasing filters, no

Click to expand...

Again is there a source for this information? Generally all
sampling systems need AA filters.

Peter

Click to expand...

Click to expand...

Much of the reactions and results with this chip in the market
will depend on how the public reacts. If people don't buy
the cameras with Sony chips, or there is a big dip in sales,
then the Foveon will be in everyones line up by the end of
the year. If everyone just keeps buying the same product
Sony has been producing, that needs noise reduction and
smoothing tech to make a usable image in even their own
cameras, then Foveon will have a long adoption cycle, and
will have to fight for every step. Since color accuracy is the
idea when shooting images, and not having to smooth a lot
of the detail out of the image is also optimum, I want a 3mp
Foveon, not a 6mp smoothed noise reduced Sony. They said
AMD was just a flash in the pan also.

bob d said:

Chris Brown said:

Sorry, but this is nonsense, as the included blow-up from a Bayer
caputred (D30) image shows. The image is out of the camera,
unsharpened. It's a power line against the sky in the background of
a street scene. There is a little chromatic abberation, but the
power line (not quite horizontal, so look in the middle of the
"step") is clearly represented as 1 pixel thick.

http://www.narcissus.uklinux.net/detail2.tif

Antialiasing is required, but it doesn't cause anywhere near the
resolution degredation you suggest.

Click to expand...

The relationship of Bayer to X3 is probably around 2 to 1, so 6MP
Bayer = 3MP X3. So the new Foveon doesn't blow away the best Bayer
as people are suggesting.

And Sony has lots more experience building CCDs, and Sony has
marketing channels... don't expect the Foveon to completely alter
the landscape and make all non-Foveon cameras obsolete.

Click to expand...

Peter G said:

I am aware that it works in both dimensions, but when rating the
sharpness of lens or the ability of film capture resolution, the
standard is lp/mm.

If one lens does 70 lp/mm and another 90 lp/mm I don't think too
many people would insist on saying the second is twice as sharp. Or
insist on squaring the value to exaggerate the impact, especially
when the Foveon number are hypothetical right now. The lens plays a
very important part in the system and with current cameras
capturing consumer cameras capturing up to 80% of max, there is
currently no proof that Foveon cameras will do any beter (on a
monochrome res chart). So saying they will be Twice as sharp right
now seems to lack any foundation.

Even if it works out that X3 is worth a doubling of the pixels (and
this is by no means a given), Bayer cameras are available with
double or more. I can buy a 5mp consumer camera today. From Foveons
information, It will be Xmas before I can get a 1.3 mp consumer
camera. If they had gone at least to 2MP I might be interested.

Click to expand...

Peter, You contradict yourself...

If lens can not resolve more than 3mp mosaic cameras, then what is the point of going for 5mp bayer camera?. Rather have clean 3mp, after all, as you say there is no more info to resolve?

Peter G said:

In few years I hope to own a very nice X3 DSLR when the price is
right, but for the next year, Fovean is essentially a non-player in
the consumer space.

Click to expand...

That i can agree with...

Peter G said:

I originally joined this argument when the foveon bandwagon got
rolling, stating that they would purchase no Mosiac camera every
again and they were waiting for foveon X3 tech.

Click to expand...

Just usual excitement at the sight of new tech, don't worry, people will still buy good mosaic cameras.

Peter G said:

Well my view is the exact opposite. I will purchase a good mature
Mosiac camera now while waiting for the truly great Foveon X3
camera that are probably a couple of years away. I let other people
be "early adopters" these days.

Click to expand...

I'd be glad to be an early adopter, if the price would be half of what is announced. As it stands now, i simply can not afford one :-(

As for my math, i still think it is correct, and 3mp camera does not have half the resolution of 6mp camera as some might think (given everything else is equal in cameras). The real difference is sqrt(6) sqrt(3) = 1.414 times more resolution.

Rgrds,
Moshe

Peter G said:

Peter

Moshe Vainer said:

90% horizontal + vertical is:
0.9 * 0.9 = 0.81 = 81% efficiency.

So even if your numbers are right, 81/49 = 1.653

1.653 * 3.4mp = 5.6236mp equivalence.

Good enough for me...

Rgrds,
Moshe

Peter G said:

Steven Noyes said:

It is true that you can not triple it but I think you underestimate
its sharpness as well. It is at least 2X sharper than the best
Interpolation methods. That is assuming the scene is of average
color density.

Click to expand...

Do you have a source for this info? Current decent prosumer cameras
are getting 70 -80% of the theretical max resolution on a res
chart. A good X3 based camera will probably get 90%. Even at 100%
it won't be 2x sharper. It can't be greater than 100%.

Peter G said:

In short, I would take a 1.2 MP Foveon over any 2 MP and most 3MP
P&S cameras any day. This assumes that Noise is under control. If

Click to expand...

How can you know that when you haven't seen the camera and have no
idea if the lens is any good etc....

Peter G said:

stunning sharp enlargements. Think. No Anti-Aliasing filters, no

Click to expand...

Again is there a source for this information? Generally all
sampling systems need AA filters.

Peter

Click to expand...

Click to expand...

Click to expand...

Talking about mass market, people are buying Nikon, Canon, Fuji, Sony, Olympus, etc. If Foveon can convince some major camera company to use their technology, this could lead to the eventual market dominance. Otherwise the broad public wan't even notice Foveon chip, even if it's wastly superior.

Vladimir.

Bob Williams said:

Much of the reactions and results with this chip in the market
will depend on how the public reacts. If people don't buy
the cameras with Sony chips, or there is a big dip in sales,
then the Foveon will be in everyones line up by the end of
the year. If everyone just keeps buying the same product
Sony has been producing, that needs noise reduction and
smoothing tech to make a usable image in even their own
cameras, then Foveon will have a long adoption cycle, and
will have to fight for every step. Since color accuracy is the
idea when shooting images, and not having to smooth a lot
of the detail out of the image is also optimum, I want a 3mp
Foveon, not a 6mp smoothed noise reduced Sony. They said
AMD was just a flash in the pan also.
bob d wrote:

Click to expand...

--Vladimir.

Moshe Vainer said:

Peter, You contradict yourself...
If lens can not resolve more than 3mp mosaic cameras, then what is
the point of going for 5mp bayer camera?. Rather have clean 3mp,
after all, as you say there is no more info to resolve?

Click to expand...

I don't remember saying I would want a 5mp camera with a lens that was designed for a 3mp camera.

Moshe Vainer said:

As for my math, i still think it is correct, and 3mp camera does
not have half the resolution of 6mp camera as some might think
(given everything else is equal in cameras). The real difference is
sqrt(6) sqrt(3) = 1.414 times more resolution.

Click to expand...

I think most people here understand the relation between linear and area (square) measures.

It seems to me you just contradicted yourself. First when you claimed the 90% was approx twice as sharp as 70%, you wanted to use the area for the comparison (since the difference in your favor is greater).

Now you seem to want to downplay the difference between 3mp and 6mp you switch back to linear.

Could you pick one?

Peter

If you use that light sensor term the X3 has 9 million.

I'd like a D-30 user to comment on the low light capability and noise of their CMOS systems. Seems to me that the D-30 pics I've seen were very low noise. Is there an appreciable difference between the current D-30 CMOS chip and the current Fuji and Sony CCDs?

bob d said:

I am not convinced that there is going to be much difference in
image quality between the new Sony 6MP CCD and the new Foveon 3MP
X3 CCD. The Sony had twice as many distinct light sensors as the
X3, making up for the fact that each light sensor only samples one
of three colors.

Click to expand...

Peter G said:

Moshe Vainer said:

Peter, You contradict yourself...
If lens can not resolve more than 3mp mosaic cameras, then what is
the point of going for 5mp bayer camera?. Rather have clean 3mp,
after all, as you say there is no more info to resolve?

Click to expand...

I don't remember saying I would want a 5mp camera with a lens that
was designed for a 3mp camera.

Click to expand...

Of course you did:
You said:

The lens plays a very important part in the system and with current cameras capturing consumer cameras capturing up to 80% of max,

there is currently no proof that Foveon cameras will do any beter (on a monochrome res chart)

How can one understand that other than there is no way for foveon to get better resolution, since even now the lens is a limiting factor. And later you say:

Bayer cameras are available with double or more. I can buy a 5mp consumer camera today.

The question is why would you, if you think that lens is a limiting factor?

Peter G said:

I think most people here understand the relation between linear and
area (square) measures.

Click to expand...

I am not so sure... :-(, as you yourself spotted, i myself did a mistake here:

Peter G said:

It seems to me you just contradicted yourself. First when you
claimed the 90% was approx twice as sharp as 70%, you wanted to use
the area for the comparison (since the difference in your favor is
greater).

Now you seem to want to downplay the difference between 3mp and 6mp
you switch back to linear.

Could you pick one?

Click to expand...

Grrr, sorry, of course you are right... My mistake..

it's sqrt(1.653) * 3.4 = 4.37mp. I guess i rushed into conclusions without checking myself. (it's late here, and i am quite tired...)

Cheers.

Moshe Vainer said:

Peter G said:

Moshe Vainer said:

Peter, You contradict yourself...
If lens can not resolve more than 3mp mosaic cameras, then what is
the point of going for 5mp bayer camera?. Rather have clean 3mp,
after all, as you say there is no more info to resolve?

Click to expand...

I don't remember saying I would want a 5mp camera with a lens that
was designed for a 3mp camera.

Click to expand...

Of course you did:
You said:

The lens plays a very important part in the system and with
current cameras capturing consumer cameras capturing up to 80% of
max,
there is currently no proof that Foveon cameras will do any beter
(on a monochrome res chart)

How can one understand that other than there is no way for foveon
to get better resolution, since even now the lens is a limiting
factor. And later you say:

Bayer cameras are available with double or more. I can buy a 5mp consumer camera today.

The question is why would you, if you think that lens is a limiting
factor?

Click to expand...

Again, none of the above implies a 3mp quality lens on 5mp camera.

The lens is one of the factors limiting current consumer digicams to an observed 80% of their maximum theoretical linear resolution.

3MP = 768 (theor. max) line pairs per picture height x 80% = 614
5mp = 960 (theor. max) line pairs per picture height x 80% = 768

768 is still better than 614. I know english can be imprecise, but you always seem to want to twist what I say. This is getting away from the topic at hand.

My whole point is that until someone (hopefully Phil) tests a real camera, we really will have no idea of the performance benefit, certainly not enough to make claims like it will be Twice as sharp or have sqrt(2) times as much linear resolution. We currently have NO evidence supporting this.

But what you can do is is look at our current camera and see just how bad they are. If they are achieving 80% of theoretical resolution, then you know there is only the last 20% available for improvement. We still do not know how much of this will be exclusively available to X3 cameras. All? None? But regardless it can't be greater than 20%.

Peter

But what would it look like without interpolation. This is a
question we don't know the answer to yet but my guess (Based off
of the programs I have worked on to do the Bayer interpolation)
is that the X3 will make this a much more high contrast image.
The power line would have been darker because it has been lightened
as a result of the brightened sky. It is very true that the
interpolation causes a loss of sharpness and introduces artifacts
at the pixel level. There are many tricks, mostly in smoothing and
then providing sharping, that are done to minimize those artifacts.

Any sensor that gets way from this will have an enormous image
advantage if it can create accurate color, low noise images.

Steven

Chris Brown said:

Sorry, but this is nonsense, as the included blow-up from a Bayer
caputred (D30) image shows. The image is out of the camera,
unsharpened. It's a power line against the sky in the background of
a street scene. There is a little chromatic abberation, but the
power line (not quite horizontal, so look in the middle of the
"step") is clearly represented as 1 pixel thick.

http://www.narcissus.uklinux.net/detail2.tif

Antialiasing is required, but it doesn't cause anywhere near the
resolution degredation you suggest.

Click to expand...

Karlg said:

My point is that you need to compare at a Price point at a point in
time. X3 is an interesting technology, but it does not appear
today to be a drop down winner.

Click to expand...

I think X3-based cameras must be cheaper than their CCD-equivalents. There was a link to an interview page with X3 inventor who was saying that:

1. Image right from the sensor (X3) needs FAR LESS post-processing comparing to CCDs output

2. Thus, you need less power-hangry processor, less memory, less everything: camera becomes simpler (less components) -> cheaper.

Cost of X3 chip itself is not an issue: it's no more expensive than average CCD from Sony (that's how I understood that). The guy said by the next christmas there will be plenty of X3-based cameras for about $400 with better capabilities than modern $999 prosumer CCD-containing noise generators.

;-)
----Eugueny

Not sure if you are seious or being toung in cheek with the smily face at the end.

By Christmas there will just as likely be CCD based cameras at $400 with better capabilities than todays $999 prosumer cameras. It is also not clear if he is talking about the F10 which is a little over 1MP X3.

Points 1&2 are addressing the trivial differences. The processing of the image is neither power hungry or very expensive anymore relative to everything else in a digita camara.

Oh well,

Karl

softtower said:

Karlg said:

My point is that you need to compare at a Price point at a point in
time. X3 is an interesting technology, but it does not appear
today to be a drop down winner.

Click to expand...

I think X3-based cameras must be cheaper than their
CCD-equivalents. There was a link to an interview page with X3
inventor who was saying that:

1. Image right from the sensor (X3) needs FAR LESS post-processing
comparing to CCDs output

2. Thus, you need less power-hangry processor, less memory, less
everything: camera becomes simpler (less components) -> cheaper.

Cost of X3 chip itself is not an issue: it's no more expensive than
average CCD from Sony (that's how I understood that). The guy said
by the next christmas there will be plenty of X3-based cameras for
about $400 with better capabilities than modern $999 prosumer
CCD-containing noise generators.

;-)

--
--
Eugueny

Click to expand...

--Karl

Well, having seen a resolution sample now, it is clear that X3 allows the capture of the maximum amount of data. 100% of what it should be capable of sampling. Each pixel can resolve a line.

So with this new info, my opinions is that 3.5MP will capture resolution similar to 6MP mosaic cameras. There should be Hot competition this year with all the ~$3000 SLRs. Sigma, Canon, Nikon and Fuji.

The "affordable" DSLR remains a futuists dream at the moment though. :-(

Peter

Peter G said:

My whole point is that until someone (hopefully Phil) tests a real
camera, we really will have no idea of the performance benefit,
certainly not enough to make claims like it will be Twice as sharp
or have sqrt(2) times as much linear resolution. We currently have
NO evidence supporting this.

But what you can do is is look at our current camera and see just
how bad they are. If they are achieving 80% of theoretical
resolution, then you know there is only the last 20% available for
improvement. We still do not know how much of this will be
exclusively available to X3 cameras. All? None? But regardless it
can't be greater than 20%.

Peter

Click to expand...

How sadly true (re affordable) :-(...

After all, we don't want much, just X3 quality at 1600 iso, with 1d/F1 autofocus system at no more than 1500$. How hard is that? :-}

To be serius however, what really is sad is that all those new cameras will replace rather than supplement previous models (d30, Fuji S1). If that were not the case, we might have seen some affordable DSLRs much sooner.
Rgrds,
Moshe

Peter G said:

So with this new info, my opinions is that 3.5MP will capture
resolution similar to 6MP mosaic cameras. There should be Hot
competition this year with all the ~$3000 SLRs. Sigma, Canon, Nikon
and Fuji.

The "affordable" DSLR remains a futuists dream at the moment
though. :-(

Peter

Peter G said:

My whole point is that until someone (hopefully Phil) tests a real
camera, we really will have no idea of the performance benefit,
certainly not enough to make claims like it will be Twice as sharp
or have sqrt(2) times as much linear resolution. We currently have
NO evidence supporting this.

But what you can do is is look at our current camera and see just
how bad they are. If they are achieving 80% of theoretical
resolution, then you know there is only the last 20% available for
improvement. We still do not know how much of this will be
exclusively available to X3 cameras. All? None? But regardless it
can't be greater than 20%.

Peter

Click to expand...

Click to expand...

Yes, unfortunately. I probably won't get a DSLR until they reach 2000CDN$, which is about $1250 USD. This is MANY years away IMO.

Actually the new models are upgrades that cost no more to produce, so there would be no sense keeping the older model in production. In fact the newer models might even benefit from cost reductions and be cheaper to produce.

Unfortunately sensors of a set size are not really on any large price reduction curve, so it will a long time to get a big sensor camera.

Peter

Moshe Vainer said:

To be serius however, what really is sad is that all those new
cameras will replace rather than supplement previous models (d30,
Fuji S1). If that were not the case, we might have seen some
affordable DSLRs much sooner.
Rgrds,
Moshe

Peter G said:

So with this new info, my opinions is that 3.5MP will capture
resolution similar to 6MP mosaic cameras. There should be Hot
competition this year with all the ~$3000 SLRs. Sigma, Canon, Nikon
and Fuji.

The "affordable" DSLR remains a futuists dream at the moment
though. :-(

Peter

Peter G said:

My whole point is that until someone (hopefully Phil) tests a real
camera, we really will have no idea of the performance benefit,
certainly not enough to make claims like it will be Twice as sharp
or have sqrt(2) times as much linear resolution. We currently have
NO evidence supporting this.

But what you can do is is look at our current camera and see just
how bad they are. If they are achieving 80% of theoretical
resolution, then you know there is only the last 20% available for
improvement. We still do not know how much of this will be
exclusively available to X3 cameras. All? None? But regardless it
can't be greater than 20%.

Peter

Click to expand...

Click to expand...

Click to expand...

Karlg said:

Points 1&2 are addressing the trivial differences. The processing
of the image is neither power hungry or very expensive anymore
relative to everything else in a digita camara.

Click to expand...

"Everything else"... What else? Sensor+lens+card writer. I don't think G2 lens cost more than $70-100, card writer cost $15. So... why camera's MSRP is about $900? Are you saying it's sensor that worth all those $785?

What makes the difference between Canon's G2 and Sony's S85? Same lens, same sensor, but very noticable difference in image quality. Where this difference is coming from? What if both cams were equipped with X3?
----Eugueny

Suppose we have two LCD diplays, one is 800 x 600 and the other is 1600 x 1200; Suppose on both displays pixels are made of 3 rectangular R G and B sub-pixels (so that those pixels are square); Supose we are close enough from the display to cleary see those LCD pixels. If black and white lines, one pixel wide, are displayed on both screens, the 1600 x 1200 screen displays twice more lines pairs than the 800 x 600 screen.

But I think the 1600 x 1200 screen is four times "better" than the 800 x 600 ! Because it is able to display four times more informations (if you can read one page of text on the 800 x 600 little display, then you can read four pages of the same text (same font, etc.) on the large 1600 x 1200 one. (actually, there are few real 1600 x 1200 CRT displays, if any).

Those points must be clear before actual comparisons between X3 and other technologies are done, or ... yes, a quarrel on the comparisons criteria is inescapable !

GURL said:

Suppose we have two LCD diplays, one is 800 x 600 and the other is
1600 x 1200; Suppose on both displays pixels are made of 3
rectangular R G and B sub-pixels (so that those pixels are square);
Supose we are close enough from the display to cleary see those LCD
pixels. If black and white lines, one pixel wide, are displayed on
both screens, the 1600 x 1200 screen displays twice more lines
pairs than the 800 x 600 screen.
But I think the 1600 x 1200 screen is four times "better" than
the 800 x 600 ! Because it is able to display four times more
informations (if you can read one page of text on the 800 x 600
little display, then you can read four pages of the same text (same
font, etc.) on the large 1600 x 1200 one. (actually, there are few
real 1600 x 1200 CRT displays, if any).

Click to expand...

You mean LCD? CRT's goes up to above 2048 x 1536.
I have seen LCD's with 1600x1200, very nice!
Jake.

GURL said:

Those points must be clear before actual comparisons between X3 and
other technologies are done, or ... yes, a quarrel on the
comparisons criteria is inescapable !

Click to expand...

--Jake.