Ian, I used LR because it is convenient - I already owned an old stand-alone copy and I had already worked out a preferred downsizing technique from the days when I used it as my general software, so it was convenient to use solely for the downsizing. I use DxO pretty much exclusively for raw processing - I agree it is excellent and provides an all in one solution including excellent noise reduction and outstanding local adjustment tools meaning I don't have to pay any subscription to the Adobe behemoth. I've never warmed to the DPP/Adobe/Neat Image process.

As I said in the last post, DxO only allows you to specify the dimensions of the largest edge size when downsizing, and then downsizes the file retaining the aspect ratio of the original. This was a pain when we used 1400 by 1050 px on BLP because if you specify that the largest size was 1400, often the downsized photo had the second dimension more than 1050, depending on how it has been cropped. LR does this much better, allowing you to specify a maximum for both dimensions as well as overall file size, so the downsized file was always within the limits. Now, with 1800 by 1800 file size for BLP, I can just specify 1800 as the maximum size in DxO and the resulting dimensions will always be within the BLP requirements regardless of the aspect ratio of the cropped photo. As such, I hardly ever use LR any more for anything.

I used 8 bit tiff because I reckoned that it was OK to do the reduction to 8 bits prior to the downsizing and production of an (8bit?) JPEG.

Glenn I did what you had suggested with one of my files with DPP and it does not show any artefacts using the settings you proposed - but it gives me a very soft file if I then downsize to a JPEG using 'Convert and Save' in DPP (not sure if there is a better way with DPP?). This sharpens up nicely if I reload the JPEG into DPP and apply another +2 of Sharpening (not Unsharp Mask) and is halo free. So if ever I have a particularly difficult file this might be a solution.

Any how, its not a major issue for me as I can control what I am seeing with a balance of settings in DxO including for some files, reducing the input sharpness before exporting as a downsized file, as well as using the smoother bicubic option.

The purpose of the original post was because I was curious as to whether others saw similar edge artifacts when downsizing, and whether these were a product of the downsizing process per se. However, it seems that neither of you see similar effects, which perhaps has something to do with the way sharpening and downsizing are applied in our respective cases.

Thanks Simon,

No problem re LR, I just assumed you had a subscription which as I understand it, comes bundled with PS. Clearly you have LR only and if you are happy with your workflow then no need to get a copy of PS.

Regarding your experience with DPP, the sharpening settings Glenn suggested should give a result without over-sharpening artifacts. Whether this 'looks' sharp is a matter of perception; one can increase the DPP sharpness amount further and it will increase the acutance (looks sharper) at the expense of introducing artifacts, especially halos on hard edges.

Cheers, Ian

Thanks Simon for outlining your experience with DPP. However, I wasn't suggesting you process entirely in DPP. Instead, I was trying to understand where the halo artefacts you are seeing are coming in by simply doing the RAW processing in DPP then any subsequent steps in either DxO or LR to see whether the a sharpening halo emerged. It would then be more obvious where the problem arose.

I am just back from a walk, taking in some clean air under sunny blue sky after the wettest start to the year in about 3 decades! It occurred to me that some members who are following this discussion might appreciate a sketch of the bigger picture.

The sharpness of a digital image is determined by the the Modulation Transfer Function (MTF) of the lens and the MTF of the camera. The MTF of the camera has two main components, the MTF of the sampling pixels and the MTF of the AA-filter. The camera + lens system has an overall MTF that is a combination of these. The camera MTF is roughly the same for each position on the sensor array. The lens MTF varies with position across the field of view and is best in the centre (on axis). The lens MTF also varies with aperture and for high quality lenses a wider aperture is better than closing down the aperture. This means that the overall system MTF will also vary with aperture. The purpose of capture sharpening in RAW processing is to recover some of the sharpness which is lost due to the fall-off of the lens MTF, the sampling response and the AA-filter. By measuring the system MTF for various aperture settings it is possible to discover the optimal sharpening amount required without over-sharpening the image. The results of such a study are shown below for a Canon 5Ds + 400 mm f/4 DO II system when using DPP4 image editing software. While these results are for a particular system, it turns out that for high quality lenses like the Canon L-series, the results are generally applicable when using DPP4.

The main consequence of over-sharpening is the appearance of halos around hard edges. This can be investigated by photographing a knife-edge (razor blade) against a white background. The results from this kind of test are shown below for a Canon 90D + 300 mm f/2.8 II system. As the amount of DPP4 sharpness is increased, the digital image of the knife-edge gets steeper and 'ringing' artifacts begin to appear as a light and dark line on either side of the edge. The image of the edge appears to get sharper due to a physiological effect in the human eye and this is called perceived sharpness. The light and dark line on either side of the edge creates the perception of sharpness and a measure of this is called 'acutance'. A point is reached where the amount of sharpening causes an obvious halo and we usually judge this to be too much sharpening. The system MTF study shows that for a f/5.6 system the optimum sharpening amount is Sharp = 2 and this correlates well with the Edge Spread Function measurements shown below. Anyone who uses DPP4 will know that there are default values for sharpening which are provided as a stating point. The same default sharpening is provided across the entire range of Canon lenses from 4 mm wide angle lenses to 1200 mm super-telephotos and the full range of zooms. Furthermore, the default (Sharp = 3) does not take account of the aperture setting and is generally too much sharpening for the kind of lenses we use in bird photography. The situation is worse for Picture Style 'fine detail' which sounds good but the default sharpening is Sharp = 4 and will most certainly cause unacceptable halos around hard edges.