FANDOM


Get Focus BracketsEdit

Written for/on: G7X and G1X
Should also work on: any camera that runs CHDK
Requires CHDK build: Anything recent as of 2018

Latest (non M3) version of this script may be downloaded from: https://gist.github.com/pigeonhill/3fe9e52621db7149eb3505bd60b395cb

The M3 version (which is different to the general version) may be downloaded from here: https://gist.github.com/pigeonhill/cdf25239bae6a4343e6431fb4c10c849

Notes & Usage Unlike macro focus bracketing, where the focus step is the same each time and very small; in landscape focus bracketing we need to calculate the focus for each step. To get the full benefits from this script, it is useful to remember that, ignoring diffraction, at the near and far depth of fields the defocus blur is, by definition, the so-called Circle of Confusion. This blur is usually accepted as being 30 microns for a 35mm full frame camera, and scaled by the crop factor for other formats, eg 11 microns on the G7X.

Having said this, 30 microns, is not the 'best focus' quality but you will not see the differences on your Facebook images, only when you create high (focus) quality prints that will be scrutinised close up, eg in a competition. As you need a line pair (black and white pixels, to see a line, it is not sensible to consider CoCs less than, say, 2 sensor pixels.

Defocus blur is, of course, zero at the point of focus. Also, at the Hyperfocal distance (H), the blur at infinity will be the CoC. It is useful to note that any blur at infinity can be 'dialed in', by simply focusing at a distance either side of H. Thus, on my G7X, with a CoC of 11 microns, if I focus at twice H, then the infinity blur will be CoC/2. Likewise, at H/4, the infinity blur will be 4*CoC.

The script scales everything relative to H and the following chart illustrates how defocus blur varies with respect to H. Note the chart shows an estimate of the near and far DoFs that can be used in your head: the script uses a more ‘accurate’ estimate.

H

When focus bracketing we must at least ensure the previous image's far DoF is the same as the next image's near DoF: assuming we are focusing near to far as we are in this script. However, many wish to introduce some 'insurance' and have an overlap between the focus brackets. The script provides four options for this overlap: none, 2*CoC/3, CoC/2 and diffraction aware.

With diffraction aware overlap, the script calculates the defocus blur to use, that ensures that the defocus blur and diffraction blur, taken together in quadrature, are equal to the CHDK CoC. This overlap blur criterion will never be less than the CHDK set CoC/3.

The following illustrates the overlap between image n and n+1. Note that using the overlap feature comes at the cost of needing to take more brackets

Pict 2

Having achieved the perfect focus brackets, the script will then create an additional bracket beyond H, ie to cover the infinity focus quality. The options being 2H, 3H or 4H, giving infinity blurs of CoC/2, CoC/3 or CoC/4. If diffraction aware is being used for the overlap, the last image will be around CoC/3.

By default the script creates a dark image at the beginning and end of the focus bracket set, to help differentiate the sequence in post. You can turn this off via the menu.

Picture2

There are five focus bracketing strategies: current position to infinity (X2Inf), min camera focus to current position (Min2X), min camera focus to infinity (Min2INF), current position to delta X in front (X2DelX) and current position to delta X behind (-DelX2X).

At each focus point you can also ask the script to create additional exposure brackets at either -xEv & +xEv, -xEv & --xEv or at +xEv and ++xEv. X can be 1 or 2Ev or 3Ev from the base exposure. These options cover all the usual exposure bracketing options, eg if following an ETTR approach, then a +/++ logic would be the best choice.

In addition to the 'normal' bracketing option, the script offers two additional schemes. First, an ISO-less bracketing option (at ISO 800 or ISO 1600) that assumes (sic) the camera is ISO-less above ISO 800 or 1600 (check your camera to ensure you select the right ISO so the sensor read noise swamps the banding). Thus there is no point in taking brackets in the camera above this, ie just push in post. The ISO option is good for hand-holding situations and the script assumes your base ISO is 100.

G1X G7X

As an illustration, the chart on the right, from the DXOMark site, shows the G1X vs G7X. As can be seen the G7X can be assumed to be fully ISO invariant, which means doing ISO bracketing is pointless, ie simply push a single base exposure in post (as long as you are not too underexposed ;-))

On the other hand the G1X is only ISO-less above about 1600. Hence take a single additional bracket at, say, ISO1600.

The second additional scheme makes use of the 'Zero-Noise' approach, as suggested by Guillermo Lujik. In this mode a second time bracket is taken at +4Ev or +3Ev or +2EV from the first one. To use the ZN option, simply ETTR to capture the highlights and the script will create an additional image for the shadow areas. But beware potential flares from bright highlights if you use the +4Ev option

To use the script, simply focus on the nearest point of interest, unless you are using the Min2 mode.
Test image
The script is started in the normal way, eg by doing a shutter press.

The script doesn't require the camera to be any specific mode. Thus it should run OK from M or Av mode, as well as the camera being AF mode. Note, if in AF mode, obviously, focus at the required position before switching to Alt mode and running the script.

But watch out for the Canon Av restriction, ie Tv < 1s.

Vyne-2

The script will exit Alt mode when complete. Here are a couple of test images I took. The post processing was carried out in Lightroom, with a round trip to Helicon Focus.

Finally, you will find information about this script and other photography matters on my blog at photography.grayheron.net

Save to your /SCRIPTS/ folder as usual.

Community content is available under CC-BY-SA unless otherwise noted.