FoCal 2 is coming! It’s been a bit delayed because of some resourcing issues, but we’ve resolved that now and we’re full steam ahead on the FoCal 2 project. One of the new tests that we’ll be bringing in FoCal 2 is the Focus Shift test, and this post gives some idea of what it can do.
What is Focus Shift?
Focus Shift is the phenomenon where the point of focus can shift as a lens is stopped down from wide open. It’s the results of spherical chromatic abberation, which results in the focus point being shifted a little depending on where on the lens surface the light is coming from. Some lenses are renowned for having this property – e.g. the Canon EF 50mm f/1.2L – but during our development of this test we’ve discovered that quite a few other lenses exhibit focus shift.
To understand what impact this might have, consider taking a shot: when you half-press the shutter button on a camera and use the phase-detect autofocus system, the focusing operation is performed with the lens wide open. This gives the autofocus sensor the most light to work with and also the shallowest depth-of-field, together meaning the focus results should be easiest for the autofocus system to analyse and resulting in the most accurate focus.
When you then fully press the shutter button to take the shot, nd the camera stops down the lens to your selected aperture and fires the shutter. If you have a lens which exhibits a high degree of focus shift, the focus point will move and you’ll end up with a shot focused somewhere unexpected!
How does the Focus Shift test work?
Quantifying the focus shift of a lens involves taking quite a few shots across the aperture range at different focus points, but you have to be incredibly methodical in your testing strategy in order to get valid data.
The new Focus Shift test in FoCal is fully automatic on all cameras (supported by FoCal) and handles all of the shooting and analysing. On top of this, by using our mathematical lens behaviour model to fill in the gaps in testing across the focus range, and behaviour prediction across the aperture range we can get a full analysis of the lens characteristics with only 15% of the number of shots for full coverage. To put this in context, the images in this article are created from 70 shots, but give the effective data from nearly 500 shots!
Operation of the test is similar to any other FoCal test – with the standard target in the correct place, verified with the Target Setup utility, you just click the Start button and the test will run. The test takes a few minutes to run and the results are presented as the test runs.
What can you tell from the results?
There’s a lot of information contained in the results from all the shots, so the test is organised to be able to show you various views of interest. The images below are from a preliminary report and are likely to be a bit different in the final version but show roughly what you can expect from the test. For information, the results are all from a Canon EF 70-200mm f/4L IS lens.
The Quality Grid shows a lot of information in a single chart. The aperture runs along the bottom of the chart, and the focus shift up the side (negative numbers are towards the camera, positive numbers away from the camera).
The colour of the squares – from green through to red – indicate the quality of the capture at this point. The quality can be either compared to the overall maximum (i.e. the absolute best quality capture from the whole test), or compared to the best quality captured just for the specific aperture. The example above shows the values compared to the absolute maximum quality.
The dots and red curved line across the graph show the focus required to get the best quality for each aperture. This is the important measure to quantify focus shift. In simple terms, it means that your camera will focus wide open – i.e. along the thick green line in the middle of the chart – but to get the best quality from this lens as you stop down you need to manually refocus a little nearer to the camera, or use manual focus.
The white diamond shows the absolute best quality point from this lens. It’s a little difficult to see in the chart above (and we’ll make it more obvious in the final version) – you’ll notice it’s pretty much at the tested point at f/6.3.
The purple and green vertical bars show the points of best quality achievable with this lens. The purple bar indicates the best aperture to use when you shoot “normally” – i.e. let the camera do all the focusing wide open then stop down to shoot. And the green bar shows you the best aperture to use if you shift the focus to the best possible point (essentially if you’re manually focusing).
Quality and Shift
The Quality and Shift chart shows the same information as above but with more of an emphasis on detail about the relative quality levels. Again, the chart has the test points (in green) across the aperture and shift range, and shows the lens needs to be refocused a little towards the camera to get the best images as you stop down.
The coloured “islands” show the boundaries of 90% (green), 80% (yellow), 70% (orange) and 60% (red) of the overall image quality. This means that any point within the green area will give you at least 90% of the best possible quality achievable from this lens. You’ll notice that the green area doesn’t extend to wide open (f/4) in the example above, meaning that the best quality image you’ll get wide open (manually focused) is less than 90% of the maximum achievable quality from the lens.
What’s quite interested is that if you track a line horizontally across from a 0 shift – which represents the quality you’d get if you just shot with the camera’s normal autofocus system – at no point does the line hit the green area. This means using normal autofocus with this particular lens, you’ll never get better than 90% of the quality achievable from this lens.
The chart above shows the quality compared to the absolute maximum (which is why the green area doesn’t extend to f/4), but you can also select to show the quality compared to the aperture maximums rather than the overall maximum. In this latter mode, every aperture will have all 4 of the colours as there will be one maximum quality point (100%) for each aperture.
FoCal already offers an Aperture Sharpness test, but it has a few shortcomings compared to the results from the Focus Shift test.
For the Aperture Sharpness test it’s imperative that you focus as perfectly as possible (manually) before running the standard FoCal Aperture Sharpness test. If you don’t the results will be incorrect. And to compound this problem, you have no way of really knowing whether the focus was perfect or not, other than through stringent test methods. FoCal was born out of trying to keep testing of this nature simple and reliable by removing the need to be a testing expert!
The Focus Shift test has no such requirement – you simply point the camera at the target and the test takes care of all the focusing, including a calibration phase to ensure the lens is being driven the optimal amount for each focus step. The Focus Shift test results also give more information about the relative quality, and can show the effect of the focus shift across the aperture range.
The chart above shows two lines. The top line – with the square markers – shows the best possible quality achievable at each aperture (if you manually focused). The lower line – with diamond markers – shows the quality of the image if you let the camera do the focusing. This bottom line is also the result you’d get from running FoCal’s standard Aperture Sharpness test and can be a bit misleading. In the above example, it indicates that the quality appears to get a little better when stopped down, then drops, then rises again, but when you look at the refocused (top line) quality you see the more expected shape of the image quality improving as you stop down until you hit the diffraction limit.
The blue area is the boundary of 90% of the maximum quality for each aperture. You can see that up to f/5 the camera focusing will get you a quality of greater than 90% of a manually focused shot at the same aperture, but between f/5.6 and f/14 the quality drops off below this threshold.
Wide-to-Refocused Quality Change
The Wide-to-Refocused Quality Change chart shows the difference in quality you could achieve by manually focusing instead of using the camera focusing system at each aperture. It’s effectively a plot of the difference between the 2 lines in the Aperture Sharpness plot above. The label on the y-axis is wrong for this chart (it is still in development after all!) – it actually shows the percentage difference between the two lines.
The Horizontal-to-Vertical Differnence Grid can highlight issues with internal lens alignment by showing the difference between the image quality horizontally and vertically. If there is a problem with your lens, you may see large areas far away from green.
Generally, the important area is the area that will be in focus. Above, we’re seeing some blueish areas but they’re a long way from the ideal focus point (which runs approximately down the middle of the chart – see the line on the Quality Grid), so it’s not really anything to worry about. When the image becomes very blurred (which would be at the top and bottom of the chart above), the horizontal-to-vertical ratio is a less relevant indicator of image quality.
Finally, the Data Quality chart shows the quality of the captures and calculations at each point. You can see from above that the shots from f/22 onwards are suspicious and any results on the other charts above to the far right should be treated with caution.
What’s left to do?
The test works well and reliably with Canon USM and STM lenses at the moment (so all L-series lenses and the newer STM lenses). Simpler motors used in lower priced lenses can be quite unpredictable and we are still working on the algorithm adjustments to reliably calibrate for these situations. Nikon focus shift is handled quite differently from Canon and, again, we need to do a little work on getting consistent and reliable results from these cameras. We have both of these situations covered, but it takes longer than we’d like to run the test for cheaper lenses and Nikon cameras.
There’s also some further filtering of results. There’s some unnecessary or suspicious data shown on graphs where it really shouldn’t be (for instance if the Data Quality measure drops too low, the data should be ignored), and the layout of some of the charts will be made easier to interpret.
We also have to finish the user interface – it’s functional but definitely not complete yet!
All in all, there’s a few more weeks work to do before this test is ready, but we’re planning on having it in the first test versions of FoCal 2 which we’re aiming to release in July.
If you’ve got any comments or suggestions about this new Focus Shift test then do let us know. Now is the time that refinements can be incorporated, and if you like or dislike some aspects then we’ll take them on board as we finalise the test.
Let us know your thought in the comments below or through the Contact Form.