Tamron 90mm 2.8 CA Problem!

[Andrew Garrard]

<p>Tom - for what it's worth, my experience with LoCA removal in software (DxO, in my case) is that it can deal with only mild cases. It's also a little odd that the Tamron's behaviour seems quite so specific to high-contrast edges. For reference, the test I showed above with a piece of paper would have shown very significant LoCA-related colour fringing either side of the focal plane had I used my 135mm DC or (less so) 85mm Samyang wide open. Then again, Photozone's review of the 180mm f/4 APO-Lanthar contains an image with high contrast edges that lacks colour fringes, so maybe it <i>is</i> simply a LoCA feature.<br />

<br />

I've not heard of LoCA "coming back" at very small apertures - I would expect it to drop off as a lens is stopped down because the differing magnifications of the circles of confusion at the focal plane with different colours will be less (the cones for different colour frequencies which don't focus at the same depth relative to the focal plane become more acute cones as the lens is stopped down). Diffraction is dependent on wavelength, so I guess I could imagine this being an issue at very small apertures.<br />

<br />

Anyway; other than with extreme highlights, the Tamron's LoCA performance seems reasonable to me - maybe there's a higher-order aberration that's uncorrected and is being shown up by the contrast of the edge. If the consensus is that LoCA is the problem, I'd still like to know how the 150mm Sigmas behave, since they appear to be nearly apochromatic (depending on which photozone review you read).</p>

[Andrew Garrard]

<p>Tom - that image is there for me.</p>

[walterh]

<p>The shot presented by <a href="../photodb/user?user_id=4303235">Andrew Garrard</a> <a href="../member-status-icons"><img title="Frequent poster" src="../v3graphics/member-status-icons/3rolls.gif" alt="" /></a>, Aug 16, 2011; 09:35 a.m is what I expect from this lens.<br>

I am in hospital currently so I have no chance to take any test shots.Sorry.<br>

The "Uncrisp transition" is a good try but the letters seem blue on my monitor and the blue edges if these are CA are not well seen. Also I see no indication of fringes of different color on the other side. So hard to conclude much. I hope by now you do not feel like I am picking on you - just trying to nail down the effect.<br>

What happens if you try to correct CA in post processing?<br>

Cheers<br>

Walter</p>

<p> </p>

[tom_clune]

<p>Hi, Andrew.<br>

RE: "I've not heard of LoCA "coming back" at very small apertures - I would expect it to drop off as a lens is stopped down because the differing magnifications of the circles of confusion at the focal plane with different colours will be less (the cones for different colour frequencies which don't focus at the same depth relative to the focal plane become more acute cones as the lens is stopped down). Diffraction is dependent on wavelength, so I guess I could imagine this being an issue at very small apertures."<br>

I expressed myself poorly. In the graphs that I am familiar with, LoCA is not graphed separately from lateral CA. Rather, the graph shows CA, which will be a mixture of things. If the graph starts out high at the wide-open end and drops to a local minimum at somewhere around f/5.6, that is a signature for LoCA. Generally, for reasons that are unclear to me, the graph starts drifting back up after that. Maybe it is lateral CA (although my undestanding is that lateral CA is indifferent to aperture) or maybe it is related to diffraction somehow. For whatever reason, that is the typical shape of a CA graph that shows LoCA. The graphs of lenses (or zooms -- zoom lenses may have LoCA at some zoom and not at others IME) that do not show LoCA do not start out high and drop off to a minimum by about f/5.6. That was the shape that I was trying to describe verbally. Sorry for the confusion.</p>

 

[tom_clune]

<p>Hi, Walter.<br>

When you correct LoCA in PP, one of two things happens. If it is mild, it appears to just go away when you move the LoCA slider (the only interface with which I am familiar). If the LoCA is bad, you can make the color go away, but you end up with an out-of-focus region where the LoCA had been. Think of tree branches -- the branches will no longer have a purple haze, but they will be blurry. Usually, that is still preferable to the purple -- it draws less attention to itself if it is not at the center of attention. But it isn't pretty.</p>

[radicalgel]

<p>Thanks guys!<br>

<a href="../photodb/user?user_id=2403817">Rodeo Joe</a> <br>

Looks like you've got quite a good copy of the lens. Take a look at what mine does @3.2 (100% crop attached below). I'm pretty sure somethings wrong!</p>

<p><a href="../photodb/user?user_id=6218160">Tom Clune</a><br>

Thanks for that link and the explanation! Very helpful. This whole CA/blooming/fringing/coma etc.. is awfully confusing. lol.</p><div></div>

[Andrew Garrard]

<p>Walter - get well soon. Tom - understood; I'll seek out some graphs. For what it's worth, my attempt to fix extreme LoCA (from a 135 DC) by smearing the chroma channels of an LAB representation to introduce some softness, because LoCA contributes to the luma - but the result was still slightly less objectionable than colour fringing. Since the fringes we see here are also affecting luma, I'd expect it to be difficult to reconstruct a sharp edge.<br />

<br />

Incidentally, I looked up blooming to make sure my terminology was right - in computer graphics it's allegedly supposed to represent diffraction-induced enlargement of bright highlights, but whether it's actually diffraction-dominated on this occasion I don't know. I notice in Raghunath's latest image that the leaves have green highlights, possibly because the transition isn't as hard as the branch, or possibly because of a difference in position relative to the focal plane. Not that knowing the cause is necessarily much help. :-)</p>

[rodeo_joe1]

<p>Longitudinal CA, technically "Sphero-chromatism" is caused by variation of the lens's residual spherical aberration with wavelength of light, hence its other alternative name of "Bokeh colour". The colours usually shown are red and green. It only shows up around out-of-focus edges and changes colour for objects in front of, and behind, the plane of focus. It also reduces absolutely with aperture as the depth-of-field increases. This blue fringing appears to show along the outline of <em>in focus</em> high-contrast subjects and is always blue or purplish. I'd therefore suggest that it's something other than what's conventionally called LoCA. True LoCA is also not exposure or contrast dependent, but is simply made more visible by higher contrast edges or in brighter areas of the image.</p>

<p>I should also mention at this point that my Tamron AF 90mm macro is the older version with "screwdriver" AF and no "Di" designation (the old ugly fat rubbery one in other words). Now whether Tamron have changed the optical design or not I don't know, but it certainly looks as if we're dealing with completely different lenses between my sample and others that have been shown here. Cue stampede to *bay to find older lenses?</p>

[Andrew Garrard]

<p>RJ - I've always thought that spherochromatism is one specific cause of LoCA, but that there's no reason LoCA couldn't be caused by something other than spherical aberration (especially in a lens with aspherical elements!) so I've always been wary of using that terminology.<br />

<br />

In theory, given that the different wavelengths take different paths through the lens, it should be possible even for an apochromatic lens that brings all the wavelengths in the image [assume that the source of illumination is three coloured lasers...] to focus on the focal plane, to have colour fringing; in this case the fringes should be the same colour in front and behind the focal plane. I don't really know what to call that - it's not strictly a longitudinal chromatic aberration, it's certainly not spherochromatism, but it's also clearly a chromatic aberration of some sort. I guess I'll worry about it when I see it - which will probably be when I buy an expensive telescope that's not a reflector.<br />

<br />

I agree that the bloom effect seems to be something else, but it's hard to diagnose. If it was just diffraction, I'd expect it to be red fringes, not blue ones, since (unless I've forgotten some physics) longer wavelengths diffract more - unless the lens design attempts to compensate for this (which is possible, if it's trying to be sharp at very small apertures).<br />

<br />

For what it's worth, I think my Tamron is the one with the in-lens motor.</p>

[rodeo_joe1]

<p>Andrew, I agree that we shouldn't lump all our aberrations together under the collective titles of "CA" or "LoCA". For example, the link that was pointed to earlier oversimplifies considerably, and considers axial colour, which is a fundamental 1st order chromatic aberration, as the only cause of longitudinal CA. Most modern lens designs are considerably more sophisticated than to show axial colour to any pronounced degree, since axial colour is relatively easy to correct in a lens of short(ish) focal length, and is generally formulated to bring the red and blue focii together while letting the green/yellow take care of itself. If what we're seeing is simply axial colour then I would expect the common purple/green fringes, and for it to manifest at all degrees of contrast. Also, since axial colour in its simplest form is a change of focal length with wavelength, it causes a change of magnification with colour and the effect is more noticeable away from the image centre.</p>

<p>This blue/purple fringing seems in an entirely different league from common chromatic errors, but it's obviously strongly affected by the design of the lens. It would be interesting to try the same lens on different models of camera with different sensors and see if results were the same.</p>

[rodeo_joe1]

<p>P.S. "I don't really know what to call that - it's not strictly a longitudinal chromatic aberration" - try "secondary spectrum".</p>

[radicalgel]

<p>Thanks guys for all the help! BTW I jus figured out a way to compare my problem with someone else's sample shot. Ok so someone here, who owns a tamron 90mm (HINT: Andrew? Rodeo? ;) haha), could you please take a test shot for me and post a 100% crop. I would be very grateful!</p>

<p>So here's the shot I need to make a comparison. Hold the tamron lens cap (with the silver lettering) at arms length and take a photo of it, straight on, with the pop up flash on TTL, at the widest aperture possible (I could go to f3.2 at arms length). If you now zoom in you should be able to see fringing around the silver lettering. Could you guys pls post 100% crops of this shot. That way I can know for sure if my copy of this lens is bad.</p>

[Andrew Garrard]

<p>RJ - always an education. :-) I always wondered what "secondary spectrum" was about, but I think I assumed that it referred to the uncorrected wavelengths, rather than the wavelengths only being corrected in one plane. I really need to get around to buying a lens design book.<br />

<br />

Raghunath - I'll see what I can do when I get home from work, if nobody had got there first.</p>

[rodeo_joe1]

<p>Raghunath, you broke my lens! ;-)<br /> Here's the crop you asked for, along with 2 more showing the front and back focus fringing. They're a bit worse than I was expecting, but I can live with it since I don't use direct flash if I can help it. Anyway I still think it's a great lens for macro, which is what it was designed for after all. And I'm still not entirely convinced that the blue flares are the same thing as this.</p>

<p>Further info: The shots were taken at a bit more than arms length and bear in mind I'm using a full-frame camera.</p>

[rodeo_joe1]

<p>Oooops! Here's the crops.</p><div></div>

[randy_kayle]

<p>Just got my copy today here is 2 samples 4.5, and 2.8</p><div></div>

[randy_kayle]

<p>Here is 2.8</p>

<p>It lessen the CA, by 5.6</p><div></div>

[randy_kayle]

<p>Here is the 2.8 resized</p><div></div>

[radicalgel]

<p>Thanks Randy for those shots. Yes fringing is quite visible at 2.8 and it does get better at like 5.6. </p>

<p><a href="../photodb/user?user_id=2403817">Rodeo Joe</a> thanks so much for taking the time to do the test. It's just that your crop seems to have been downsized so the fringing is not that clear for comparison. Could you pls post a crop similar to mine so that it's easier to compare. <br>

I have attached a 500px by 500px 100% crop of the test shot. </p><div></div>

[radicalgel]

<p>Hey guys I would like to thank you'll, esp andrew, rodeo & tom, for all the awesome help. I went to the store yesterday and got the tamron exchanged for the Tokina 100mm 2.8. I just liked the build more and the 10mm extra is a nice bonus. Hope this serves my purpose well.</p>

<p>IMHO Photo.net once again proves to be THE most helpful photography forum on the net. Thanks again! :)</p>