[Anvh]

Yes f/2.8 or f/4 instead of f/5.6 so out of my head but i'm missing the point why the light meter must be more sensitive if the AF becomes more sensitive.

I assume they use same system like for example Nikon.
Nikon D7000 Review: Digital Photography Review
- TTL exposure metering using 2016-pixel RGB sensor
- Auto (TTL white-balance with 2016 pixels RGB sensor)

About what Ray Pulley said.
He says the cause of the focus error in the K5 is caused by aberration off the AF objective while you say it's caused by the light meter not being sensitive enough.

[bwDraco]

The white balance is set through a combination of sensor output and data from the light wavelength sensor. From PENTAX - K-30 Black:

White Balance

Type: Image sensor detection w light wavelength sensor assist

--DragonLord

[Asha]

Just like a prism, CA is a refraction problem, so a optic that has very specifically designed diffraction properties can offset the refraction. This is not the only way to correct for CA, but there must be some other reason (cost, size, etc.) to choose this technology over the other methods. Perhaps it is also more accurate?

I don't know a huge amount about the AF system design. However, I'd like to add that any optical system can benefit by increasing the number of optical elements. MTF especially can improve, and being able to discern finer detail does seem like it should translate to better focusing. A diffractive optical element (DOE) is basically like a singlet lens, except it is flat (very similar to a Fresnel lens).

I agree that if the DOE is tailored to compensate for chromatic aberration, that could also be a big plus. However, I'd be a little bit concerned about the wavelength sensitivity of the DOE--it may work really well in most kinds of lighting situations, but could still choke.

[Class A]

About what Ray Pulley said.
He says the cause of the focus error in the K5 is caused by aberration off the AF objective while you say it's caused by the light meter not being sensitive enough.

I always said the cause is a combination of the AF module CA and the insufficient light meter sensitivity.

As long as the light meter "sees" something, its data can be used to compensate against the CA errors of the AF optics. As soon as it peters out (and that's much earlier than the AF module itself), the famous front-focus sets in.

@Asha: The AF chips sees an out-of-focus image. It doesn't evaluate sharpness directly but the phase shift of two out-of-focus images. There is no need for small details to be resolved.

[Asha]

Thanks for the reply, Class A! I thought I read somewhere that the AF system looks for an "edge"...is that not correct? An edge can be thought of as a "high frequency" detail, so higher MTF would resolve it better. If you're talking about phase shift in the MTF, depending how far out of focus the image is, the "zero crossings" in the MTF curve could be at a rather high frequency. Again, having a higher MTF resolving capability would probably produce better results in that case, too.

Interesting about phase shift, though--a DOF might have an advantage over a standard optic because its optical power is partially defined by phase. I would think it makes the AF easier to control by computer algorithm, especially if there is a way to correlate phase shift more precisely across the entire field of view.

[Anvh]

Class A how would the sensor be able to dedect the CA when it's caused by the optics only the AF sensor sees?

[Asha]

Anvh, I thought we were talking about CA in the AF system. It is not the imaging portion of the camera, but is an optical system in its own right, and therefore will have any number of optical aberrations.

[Anvh]

That's preciesly the problem Asha.
The AF sensor with the optics is located on the underside of the camera while the metering sesnsor for the light and most likely whitebalance is located in the viewfinder, so how can something that's in the viewfinder meter what's happening in the lower part of the camera?

If you want to meter the CA of the AF optics you need to have a sensor for that near the AF sensor.

[Asha]

Anvh...just thinking out loud here...there are some optics that are common to all three paths, ie, path to the meter/viewfinder, path to the AF and path to the imaging sensor. So, theoretically, if one were to "difference" the response from two of the paths, the CA from the fore-optics would cancel out. That could account for a good portion of the CA correction. The remainder may be easily correlated, depending on behavior of the metering and AF optics/sensors. I'm just theorizing, but think it is very doable.

[garyk]

Interesting.
I thought DO optics refers to a piece of the glass NOT spherical in design. I know for a fact that the use of spherical elements in fast glass cause problems. What happens is the focal point is different because of the element bending the light and focusing at different spots or locations on the focal plane. To correct for this a German man. Bernard Schmidt, the father of Schmidt optics. By use of a vacuum method to grind a thin piece of glass in a non spherical shape to focus the light rays properly. Ingenious really, and very difficult to do on a mass scale properly and accurately. What he did was figure the inverse spherical angle of the primary mirror and pull a vacuum on a piece of glass and grind away the access causing a non spherical shape to correct for focus inaccuracy. In other words the light was bent and hit the focal plane at the perfect angle. The DO in canon lenses made them very light because of doing away with all the glass if I am not mistaken. Correction is done by a non spherical glass element.

You might check with Faulk on this?

So my conclusion is they just bend the light on the AF system for more accurate focusing, Or what ever way they want to bend the light to get the AF module more accurate info.
My guess anyway. Without looking into it.

[Asha]

garyk, the Schmidt plate is an asphere, not a diffractive element. In the original design, the Schmidt corrector is mainly for spherical aberration in a spherical mirror (reflective telescope). A diffractive optical element is more along the lines of a Fresnel lens. It DOES have optical power, ie, it behaves much as a "normal" lens would, but with subtle differences. Plus, it can be made more flat and thus make for a more compact system.

[Class A]

Class A how would the sensor be able to dedect the CA when it's caused by the optics only the AF sensor sees?

The light temperature sensor, i.e., the metering chip, does not detect the CA in the AF module.

It is used to predict CA in the AF module. By measuring the light temperature of the active AF area (just pick the corresponding metering zones), you obtain a compensation value to be applied to the result of the AF module, since you know that certain wavelengths will throw off the AF module.

[Anvh]

Anvh...just thinking out loud here...there are some optics that are common to all three paths, ie, path to the meter/viewfinder, path to the AF and path to the imaging sensor. So, theoretically, if one were to "difference" the response from two of the paths, the CA from the fore-optics would cancel out. That could account for a good portion of the CA correction. The remainder may be easily correlated, depending on behavior of the metering and AF optics/sensors. I'm just theorizing, but think it is very doable.

The only optics that are shared are the lenses themselves.
For the rest, the big mirror you see inside the camera reflect the light to the viewfinder and the metering sensor there and behind the big mirror there is a small mirror that reflect light to the AF sensor.


And this how the AF looks like.

The light comes in from the top and is reflected to the right.

[Anvh]

The light temperature sensor, i.e., the metering chip, does not detect the CA in the AF module.

It is used to predict CA in the AF module. By measuring the light temperature of the active AF area (just pick the corresponding metering zones), you obtain a compensation value to be applied to the result of the AF module, since you know that certain wavelengths will throw off the AF module.

That could indeed be but CA does always happen right or not?

[Asha]

Anvh, great top pic--it makes the three paths much more clear to me than the model on the bottom (which had already been posted). As I was saying, the common optics would only account for a portion of the CA. However, thanks to your pic, I should point out that the main mirror is also common, except that I see it is a beam splitter rather than a mirror. Side note, if the "mirrors" have dielectric coatings, there could be wavelength shift at greater angles of incidence. Interesting tangential thought.

Class A's comments are in line with what I said above...that if the camera knows something about the behavior of the AF portion and the metering portion as separate optical systems, they can put an algorithm in the computer to make a correction.

ETA: To your question about CA always happening...classic CA (Seidell aberration) mainly occurs in refractive optics. If the mirrors are "nominal" mirrors, they should be nearly no CA, but there could be other spectral effects due to the coatings.