I have a friend – yes, a strange concept I know, but I do have some – we’ll call him Steve.
Steve is a very talented photographer – when he’ll give himself half a chance; but impatience can sometimes get the better of him.
He’ll have a great scene in front of him but then he’ll forget things such as any focus or exposure considerations the scene demands, and the resulting image will be crap!
Quite often, a few of Steve’s character flaws begin to emerge at this juncture.
Firstly, Steve only remembers his successes; this leads to the unassailable ‘fact’ that he couldn’t possibly have ‘screwed up’.
So now we can all guess the conclusive outcome of that scenario can’t we……..that’s right; his camera gear has fallen short in the performance department.
Clairvoyance department would actually be more accurate!
So this ‘error in his camera system’ needs to be stamped on – hard and fast!
This leads to Steve embarking on a massive information-gathering exercise from various learned sources on ‘that there inter web’ – where another of Steve’s flaws shows up; that of disjointed speed reading…..
The terrifying outcome of these situations usually concludes with Steve’s confident affirmation that some piece of his equipment has let him down; not just by becoming faulty but sometimes, more worryingly by initial design.
These conclusions are always arrived at in the same manner – the various little snippets of truth and random dis-associated facts that Steve gathers, all get forcibly hammered into some hellish, bastardized ‘factual’ jigsaw in his head.
There was a time when Steve used to ask me first, but he gave up on that because my usual answer contravened the outcome of his first mentioned character flaw!
Lately one of Steve’s biggest peeves has been the performance of one or two of his various lenses.
Ostensibly you’ll perhaps think there’s nothing wrong in that – after all, the image generated by the camera is only as good as the lens used to gather the light in the scene – isn’t it?
But there’s a potential problem, and it lies in what evidence you base your conclusions on……………
For Steve, at present, it’s manufacturers MTF charts, and comparisons thereof, coupled with his own images as they appear in Lightroom or Photoshop ACR.
Again, this might sound like a logical methodology – but it isn’t.
It’s flawed on so many levels.
The Image Path from Lens to Sensor
We could think of the path that light travels along in order to get to our camera sensor as a sort of Grand National horse race – a steeplechase for photons!
“They’re under starters orders ladies and gentlemen………………and they’re off!”
As light enters the lens it comes across it’s first set of hurdles – the various lens elements and element groups that it has to pass through.
Then they arrive at Becher’s Brook – the aperture, where there are many fallers.
Carefully staying clear of the inside rail and being watchful of any lose photons that have unseated their riders at Becher’s we move on over Foinavon – the rear lens elements, and we then arrive at the infamous Canal Turn – the Optical Low Pass filter; also known as the Anti-alias filter.
Crashing on past the low pass filter and on over Valentines only the bravest photons are left to tackle the the last big fence on their journey – The Chair – our camera sensor itself.
Okay, I’ll behave myself now, but you get the general idea – any obstacle that lies in the path of light between the front surface of our lens and the photo-voltaic surface of our sensor is a BAD thing.
The problems are many, but let’s list a few:
- Every element reduces the level of transmitted light.
- Because the lens elements have curved surfaces, light is refracted or bent; the trick is to make all wavelengths of light refract to the same degree – failure results in either lateral or longitudinal chromatic aberration – or worse still, both.
- The aperture causes diffraction – already discussed HERE
We have already seen in that same previous post on Sensor Resolution that the number of megapixels can effect overall image quality in terms of overall perceived sharpness due to pixel-pitch, so all things considered, using photographs of any 3 dimensional scene is not always a wise method of judging lens performance.
And here is another reason why it’s not a good idea – the effect on image quality/perceived lens resolution of anti-alias, moire or optical low pass filter; and any other pre-filtering.
I’m not going to delve into the functional whys and wherefores of an AA filter, save to say that it’s deemed a necessary evil on most sensors, and that it can make your images take on a certain softness because it basically adds blur to every edge in the image projected by the lens onto your sensor.
The reasoning behind it is that it stops ‘moire patterning’ in areas of high frequency repeated detail. This it does, but what about the areas in the image where its effect is not required – TOUGH!
Many photographers have paid service suppliers for AA filter removal just to squeeze the last bit of sharpness out of their sensors, and Nikon of course offer the ‘sort of AA filter-less’ D800E.
Side bar note: I’ve always found that with Nikon cameras at least, the pro-body range seem to suffer a lot less from undesirable AA filtration softening than than their “amateur” and “semi pro” bodies – most notably the D2X compared to a D200, and the D3 compared to the D700 & D300. Perhaps this is due to a ‘thinner’ filter, or a higher quality filter – I don’t know, and to be honest I’ve never had the desire to ‘poke Nikon with a sharp stick’ in order to find out.
Back in the days of film things were really simple – image resolution was governed by just two things; lens resolution and film resolution:
1/image resolution = 1/lens resolution + 1/film resolution
Film resolution was a variable depending on the Ag Halide distribution and structure, dye coupler efficacy within the film emulsion, and the thickness of the emulsion or tri-pack itself.
But today things are far more complicated.
With digital photography we have all those extra hurdles to jump over that I mentioned earlier, so we end up with a situation whereby:
1/Image Resolution = 1/lens resolution + 1/AA filter resolution + 1/sensor resolution + 1/image processor/imaging ASIC resolution
Steve is chasing after lens resolution under the slightly misguided idea the resolution equates to sharpness, which is not strictly true; but he is basing his conception of lens sharpness based on the detail content and perceived detail ‘sharpness’ of his images; which are ‘polluted’ if you like by the effects of the AA filter, sensor and imaging ASIC.
What it boils down to, in very simplified terms, is this:
You can have one particular lens that, in combination with one camera sensor produces a superb image, but in combination with another sensor produces a not-quite-so-superb image!
On top of the “fixed system” hurdles I’ve outlined above, we must not forget the potential for errors introduced by lens-to-body mount flange inaccuracies, and of course, the big elephant-in-the-room – operator error – ehh Steve.
So attempting to quantify the pure ‘optical performance’ of a lens using your ‘taken images’ is something of a pointless exercise; you cannot see the pure lens sharpness or resolution unless you put the lens on a fully equipped optical test bench – and how many of us have got access to one of those?
The truth of the matter is that the average photographer has to trust the manufacturers to supply accurately put together equipment, and he or she has to assume that all is well inside the box they’ve just purchased from their photographic supplier.
But how can we judge a lens against an assumed standard of perfection before we part with our cash?
A lot of folk, including Steve – look at MTF charts.
The MTF Chart
Firstly, MTF stands for Modulation Transfer Function – modu-what I hear your ask!
OK – let’s deal with the modulation bit. Forget colour for a minute and consider yourself living in a black & white world. Dark objects in a scene reflect few photons of light – ’tis why the appear dark! Conversely, bright objects reflect loads of the little buggers, hence these objects appear bright.
Imagine now that we are in a sealed room totally impervious to the ingress of any light from outside, and that the room is painted matte white from floor to ceiling – what is the perceived colour of the room? Black is the answer you are looking for!
Now turn on that 2 million candle-power 6500k searchlight in the corner. The split second before your retinas melted, what was the perceived colour of the room?
Note the use of the word ‘perceived’ – the actual colour never changed!
The luminosity value of every surface in the room changed from black to white/dark to bright – the luminosity values MODULATED.
Now back in reality we can say that a set of alternating black and white lines of equal width and crisp clean edges represent a high degree of contrast, and therefore tonal modulation; and the finer the lines the higher is the modulation frequency – which we measure in lines per millimeter (lpmm).
A lens takes in a scene of these alternating black and white lines and, just like it does with any other scene, projects it into an image circle; in other words it takes what it sees in front of it and ‘transfers’ the scene to the image circle behind it.
With a bit of luck and a fair wind this image circle is being projected sharply into the focal plane of the lens, and hopefully the focal plane matches up perfectly with the plane of the sensor – what used to be refereed to as the film plane.
The efficacy with which the lens carries out this ‘transfer’ in terms of maintaining both the contrast ratio of the modulated tones and the spatial separation of the lines is its transfer function.
So now you know what MTF stands for and what it means – good this isn’t it!
Let’s look at an MTF chart:
Now what does all this mean?
Firstly, the vertical axis – this can be regarded as that ‘efficacy’ I mentioned above – the accuracy of tonal contrast and separation reproduction in the projected image; 1.0 would be perfect, and 0 would be crappier than the crappiest version of a crap thing!
The horizontal axis – this requires a bit of brain power! It is scaled in increments of 5 millimeters from the lens axis AT THE FOCAL PLANE.
The terminus value at the right hand end of the axis is unmarked, but equates to 21.63mm – half the opposing corner-to-corner dimension of a 35mm frame.
Now consider the diagram below:
These are the radial dimensions, in millimeters, of a 35mm format frame (solid black rectangle).
The lens axis passes through the center axis of the sensor, so the radii of the green, yellow and dashed circles correspond to values along the horizontal axis of an MTF chart.
Let’s simplify what we’ve learned about MTF axes:
Now we come to the information data plots; firstly the meaning of Sagittal & Meridional. From our perspective in this instance I find it easier for folk to think of them as ‘parallel to’ and ‘at right angles to’ the axis of measurement, though strictly speaking Meridional is circular and Sagittal is radial.
This axis of measurement is from the lens/film plane/sensor center to the corner of a 35mm frame – in other words, along that 21.63mm radius.
Separate measurements are taken for each modulation frequency along the entire measurement axis:
Let’s look at that MTF curve for the 500m f4 Nikon together with a legend of ‘sharpness’ – the 300 f2.8:
Nikon say on their website that they measure MTF at maximum aperture, that is, wide open; so the 300mm chart is for an aperture of f2.8 (though they don’t say so) and the 500mm is for an f4 aperture – which they do specify on the chart – don’t ask me why ‘cos I’ve no idea.
As we can see, the best transfer values for the two lenses (and all other lenses) is 10 lines per millimeter, and generally speaking sagittal orientation usually performs slightly better than meridional, but not always.
10 lpmm is always going to give a good transfer value because its very coarse and represents a lower frequency of detail than 30 lpmm.
Funny thing, 10 lines per millimeter is 5 line pairs per millimeter – and where have we heard that before? HERE – it’s the resolution of the human eye at 25 centimeters.
Another interesting thing to bare in mind is that, as the charts clearly show, better transfer values occur closer to the lens axis/sensor center, and that performance falls as you get closer to the frame corners.
This is simply down to the fact that your are getting closer to the inner edge of the image circle (the dotted line in the diagrams above). If manufacturers made lenses that threw a larger image circle then corner MTF performance would increase – it can be done – that’s the basis upon which PCE/TS lenses work.
One way to take advantage of center MTF performance is to use a cropped sensor – I still use my trusty D2Xs for a lot of macro work; not only do I get the benefit of center MTF performance across the majority of the frame but I also have the ability to increase the lens to subject distance and get the composition I want, so my depth of field increases slightly for any given aperture.
Back to the matter at hand, here’s my first problem with the likes of Nikon, Canon etc: they don’t specify the lens-to-target distance. A lens that gives a transfer value of 9o% plus on a target of 10 lpmm sagittal at 2 meters distance is one thing; one that did the same but at 25 meters would be something else again.
You might look at the MTF chart above and think that the 300mm f2.8 lens is poor on a target resolution of 30 lines per millimeter compared to the 500mm, but we need to temper that conclusion with a few facts:
- A 300mm lens is a lot wider in Field of View (FoV) than a 500mm so there is a lot more ‘scene width’ being pushed through the lens – detail is ‘less magnified’.
- How much ‘less magnified’ – 40% less than at 500mm, and yet the 30 lpmm transfer value is within 6% to 7% that of the 500mm – overall a seemingly much better lens in MTF terms.
- The lens is f2.8 – great for letting light in but rubbish for everything else!
Most conventional lenses have one thing in common – their best working aperture for overall image quality is around f8.
But we have to counter balance the above with the lack of aforementioned target distance information. The minimum focus distances for the two comparison lenses are 2.3 meters and 4.0 meters respectively so obviously we know that the targets are imaged and measured at vastly different distances – but without factual knowledge of the testing distances we cannot really say that one lens is better than the other.
My next problem with most manufacturers MTF charts is that the values are supplied ‘a la white light’.
I mentioned earlier – much earlier! – that lens elements refracted light, and the importance of all wavelengths being refracted to the same degree, otherwise we end up with either lateral or longitudinal chromatic aberration – or worse still – both!
Longitudinal CA will give us different focal planes for different colours contained within white light – NOT GOOD!
Lateral CA gives us the same plane of focus but this time we get lateral shifts in the red, green and blue components of the image, as if the 3 colour channels have come out of register – again NOT GOOD!
Both CA types are most commonly seen along defined edges of colour and/or tone, and as such they both effect transferred edge definition and detail.
So why do manufacturers NOT publish this information – there is to my knowledge only one that does – Schneider (read ‘proper lens’).
They produce some very meaningful MTF data for their lenses with modulation frequencies in excess of 90 to 150 lpmm; separate R,G & B curves; spectral weighting variations for different colour temperatures of light and all sorts of other ‘geeky goodies’ – I just love it all!
SHAME ON YOU NIKON – and that goes for Canon and Sigma just as much.
So you might now be asking WHY they don’t publish the data – they must have it – are they treating us like fools that wouldn’t be able to understand it; OR – are they trying to hide something?
You guys think what you will – I’m not accusing anyone of anything here.
But if they are trying to hide something then that ‘something’ might not be what you guys are thinking.
What would you think if I told you that if you were a lens designer you could produce an MTF plot with a calculator – ‘cos you can, and they do!
So, in a nutshell, most manufacturers MTF charts as published for us to see are worse than useless. We can’t effectively use them to compare one lens against another because of missing data; we can’t get an idea of CA performance because of missing red, green and blue MTF curves; and finally we can’t even trust that the bit of data they do impart is even bloody genuine.
Please don’t get taken in by them next time you fancy spending money on glass – take your time and ask around – better still try one; and try it on more than 1 camera body!
Please consider supporting this blog.
Please note, I’ve not written this blog article for any other reason than to make its readers aware of the facts – no one pays me in cash or kind for any products mentioned in this article; it is written purely for information purposes; you do with this information “what you will”!
If you have any questions please feel free to drop me a line via email at firstname.lastname@example.org