How do you read MTFs

greenspun.com : LUSENET : Large format photography : One Thread

Hello: I am interested in knowing how you in this forum read or interpret MTFs and what you know about MTF (summary). If you please, Your own official and expert versions and even the very private unofficial versions, including those you were afraid to ask about. What do you look for in MTFs? What to you is OK and not?

-- Julio Fernandez (gluemax@ora.auracom.com), January 07, 2001

Answers

Are you asking how to properly read an MTF chart?

Or are you asking how someone thinks it should be read?

It is a scientific chart and there is no guess work involved.

if you are in the US and want a detailed and correct explanation of lens curves that Rodenstock published then I would be happy to mail you one. Providing you are in the US.

-- Bob Salomon (bobsalomon@mindspring.com), January 07, 2001.


Read them with a grain of Salt if they're from the lens maker or importer.

-- Bill Mitchell (bmitch@home.com), January 07, 2001.

Importers/distributor do not do MTF, distortion or color testing.

Lens and camera manufacturers do.

Now why would you be so gilded to not trust MTF curves?

In the case of Linhof some curves are also done by Hasselblad for the Arc camera lenses from Rodenstock so it is very easy to compare the ones from zRodenstock to the ones from Hasselblad.

As for the accuracy Rodenstock has ISO 9001 status. this assures government and industrial purchasing agents of a given level of assured quality. This encompasses the performance characteristics as stated by the MTF cuves, distortion curves and color curves.

Now while few photographers know or understand ISO 9000/9001/9002 reqierments and fail to comprehend its meaning and people in organizations who don't opt to spend the large amount of money required to qualify or just ignore the benefits of ISO status that does not mean that specs are automatically suspect.

In the case of Rodenstock it represents the basic quality level you should receive.

-- Bob Salomon (bobsalomon@mindspring.com), January 07, 2001.


My question pertains to how people in the forum interpret MTF charts and what information they derive therefrom. It is your own personal views that I am interested in, whatever they are. ALso I would be interested in knowing what effort it takes to gain a solid technical understanding of what the MTF data conveys and what effor the respondent has made to that effect. To all, thanks.

-- Julio Fernandez (gluemax@ora.auracom.com), January 07, 2001.

As in my original response we would be happy to mail you a brochure on how to read the curves used by lens manufacturers.

-- Bob Salomon (bobsalomon@mindspring.com), January 07, 2001.


 Bob, Julio lives in CANADA.  So please stop saying you would 
send him anything.  The US postal services is so expensive especially 
with the weak American dollar that you can't mail up there.  
Julio, I do not understand MTFs.  I look at the specs and user 
experiences.  While its not very scientific, looking at the images is 
key for me.  I try to keep in mind I am making images not MTFs graphs. 
 I am not trying to be dogmatic, just sometimes we can take the 
technical side too far.  For example my Rodenstock Sironar S 150mm is 
suppose to be optimal at 1:5 to infinity.  However macro shots look 
fine. 


-- David Payumo (dpayumo@home.com), January 07, 2001.

Bob: thank you for your reply. I am asking how the people in this forum read MTF charts, the information they get from them and also how various people think these should be read. And you are right, it is a scientific chart and there should be no guess work involved. However, since the information available on how to read these charts is not detailed enough, there are holes left that leave room for personal interpretations, which I believe there are many and varied. That is what I am interested in finding out and perhaps my own have been wrong all along? Bob, thank you for your kind offer to provide the Rodesntock lens data, I do have it for all their lenses. If however Rodenstock has a white paper on MTF testing and interpretation, other than what is said about it in their lens brochures, I would appreciate having such and if so, I will contact you directly. Your statements on ISO certification are a valid point, besides, Rodenstock's reputation transcends ISO.

-- Julio Fernandez (gluemax@ora.auracom.com), January 07, 2001.

Julio:

I look at MTF charts all the time. I am a seismologist and my work involves many facets of signal processing, which, like MTFs are based on Fourier analysis of signals. In the case of MTF charts, the signal is the input image, and the Fourier decomposition is done as a function of spatial frequency. MTFs are like spatial spectra, except that they omit phase information (essentially they are spatial power spectra). If you need pointers to technical explanations of such analysis, I can provide them, but as I read your question, you are more interested in how I use them in my decision making, so I will comment on that a bit.

First, you need to know that many MTF charts provided by manufacturers (Schneider and Rodenstock in particular) are theoretical not measured. They are based on computer models of the lens design and glass characteristics, and they assume a default imput color spectrum of light. Thus, they cannot address manufacturing tolerances or variations. They are mainly useful in determining the priorities that the lens designers used in designing the lens.

Next, you need to decide what format film you are interested in. I use 6x9 roll film, so I am interested in getting the highest performance at the highest spatial frequencies possible. Most LF lens manufacturers only show MTF curves up to 20 lp/mm. For MF work, I would love to have curves at 40 lp/mm, and I have some for some Rodenstock lenses off of the Hasselblad web site for the ArcBody lenses. Since I use 6x9, I am most interested in lens performance out to about 50mm off-axis.

If I were shooting 4x5, I wouldn't care about 40 lp/mm, or even 20 lp/mm but would focus on 10 lp/mm, and would be far more interested in performance farther off-axis. Most LF lenses are optimized for better performance farther off-axis and less emphasis on higher spatial frequencies since their primary audience is 4x5 shooters. This is why I look at the MTFs, to find lenses that work best for the MF situation which is not always the design goal.

Specifically, I try to determine at what aperture, the MTF values in the center of the field reach theoretical limits. Since I shoot MF, I like to use apertures like f/11 and f/16. Many LF lenses are not optimized to work well until f/22 or smaller.

Finally, I like to see how the performance falls off as a function of off-axis angle. Some lenses have flat curves that break sharply and fall off rapidly at some point. Other lenses, notably the Super Symmar XL lenses have MTFs that roll off more gently with off-axis angle. Since I shoot mostly landscape, I don't use alot of movements, so I am happy with flat curves and abrupt fall off so long as I have a reasonable usable image circle. If I were doing architecture, I might prefer a more gentle roll-off since I might need extreme movements in certain situations.

If I were shooting 4x5, I don't think I would worry much about MTF curves. Most LF lenses are optimized for 4x5, most do great for lower spatial frequencies, and I suspect most are at or near the limits of what is optically possible to produce. But since I am "swimming up river" so to speak, in using LF lenses on MF images, I use them to pick lenses that work well for that particular application.

As for numbers, I worry when modulation at 20 lp/mm drops below about 40%. That means 40 lp/mm is probably going below 30%. Given that most color transparency films are at about 50% modulation at 40 lp/mm, that means that the combined modulation is nearing 15% which is probably about the value corresponding to "resolution" of that frequency.

-- Glenn Kroeger (gkroeger@trinity.edu), January 07, 2001.


I am one of "the few" photographers to understand ISO9000. I have been working in the electronics industry for 20+ years and dealing with ISO9000 for the last 9. ISO9000 has nothing to do with quality. It was instituted by the European Union in order to discourage American competition in their home markets. It's only a paper trail, nothing more. If you make crap, and state that the purpose of your business is to make and market crap, you too can become ISO9000 certified. When I worked on the west coast the company that I worked for had "one-man" garage shop subcontractors that were ISO9000 certified. Let's not inject mumbo-jumbo into the debate.

-- Wayne DeWitt (wdewitt@snip.net), January 07, 2001.

Glen: Many thanks for your most informative reply. Your response answered my question faithfully, as posed. Of course, as a seismologist, MTF is your bread and butter the only difference being the inputs and the outputs. I prefer those (inputs / outputs) I get on film to nasty ones on your seismographs.

I have questions regarding three points.
a) Premise: Separated tangential and radial responses are the result of aberrations, , astigmatism being one of the culprits, but not the only ones. If you look at MTFs for the most highly corrected lenses as for example the Zeiss 250mm superachromat, (Zeiss's foremost accomplishment in photo optics according to a Zeiss source) its tangential and radial MTFs are virtually overlapping over much of the image, with slight separation occurring at the edge of the IC. Of course, those MTFs are at a relatively high level even at the 40 L/mm frequency.
Q. I expected you might have taken into consideration this tangential / radial divergence in your evaluations and I wondered if the omission was deliberate. This also brings out the issue of contrast and resolution.

b) Premise: High radial / low tangential responses lead to contrasty but not well defined images. Overlapping responses with medium contrast on the other hand lead to high resolution but low contrast images that can be improved by proper film selection or printing techniques to high contrast - high resolution images. With contrasty but low resolution optics the detail that went missing from the film is lost forever and can not be recovered by film or printing choices /techniques.
Q. Are these premises correct or not and what are your thoughts?

c) Premise: According to current practice, lens MTFs are not obtained from bar chart patterns but from Fourier analysis of imaged sinusoidal signals generated by knife-edges, pinholes, etc., and their resulting line spread and point spread functions.
Q. Can you elaborate on this as I do not quite understand the relationships between input/output and the Fourier calculations in between. I assume that bar chart patterns were abandoned for some reason in favour of knife-edges and pinholes but do not quite understand why. Could you elaborate?

Sorry for the lengthy dialog. Thanks also to all other respondents.



-- Julio Fernandez (gluemax@ora.auracom.com), January 07, 2001.


Hi all, isn't there a hypothesis and a simple emprical test hiding under this complex exchange? Or is this to pursue a deductive arguement? I'm kind of treading on thin ice here, but what do the MTF charts tell a person that a photograph of something with the lens can't provide? I mean, I want the best lens for my subject on my film and on to my paper that I can come by, and how will I know without trying the lens? Best, David

-- david clark (doclark@yorku.ca), January 08, 2001.

MTFs are very simple. They chart contrast (y axis) against spatial frequency (x axis). Spatial frequency can roughly be thought of as the old 'Line pairs per millimetre' resolution standard.
MTF curves are used because resolution figures alone don't tell you much about subjective 'sharpness'. Both contrast and resolution need to be combined to give any idea about the visual impression of sharpness from a lens.

However, having said all that, unless you know what you're looking for, an MTF chart means nothing. For instance, the sagittal and tangential MTF curves may track each other perfectly. This looks great on paper, but probably means that the lens suffers from curvature of field or image distortion.
Unless you're a buyer that won't actually ever use the lens, or you're specifying a lens for a particular technical purpose, forget about MTFs. Shoot some film with the darned thing and make your judgement on the result.

Lens designers just don't put bad design parameters into their computers these days, and the wide availability of glass types make the resulting computer output fairly easy to fabricate. So if a new lens performs unacceptably, then it's because it's been badly assembled or mishandled, not because the design MTF was poor in the first place.
Think about it. No manufacturer is going to issue MTF figures that show their lenses to be a pile of crap, are they?

-- Pete Andrews (p.l.andrews@bham.ac.uk), January 08, 2001.


Julio:

1. I disagree with this premise... radial and tangential curves THEORETICALLY separate off axis. So while their separation can be increased due to aberrations, they cannot coincide at all off-axis angles, unless the radial is underperforming. If you look at Hasselblad MTFs for 30mm CF, you will see this... the 40mm CFE implies they can converge, but I suspect this is because the radials are dying at the edge of the image circle.

2. Sort of agree, but I have come to believe that perceived image sharpness is much more a function of high contrast at moderate frequencies than "resolution" or moderate contrast at high spatial frequencies. It is the apparent "sharpness" of many digital images that has finally convinced me of this. With many digital images (either originally digital or scanned) the "resolution" is severely limited by the pixel count and yet the images look very sharp to most observers. Consider that the cameras such as the Nikon D1 or Canon D30 pack only about 2400 pixels across the wide dimension of the image. In 35mm terms (yes I know the image sensor is actually smaller) that is about 2400 pixels/ 36mm or 66 pixels per mm. That means that the absolute highest spatial frequency recorded is 33 lp/mm. Yet these cameras produce very impressive results even when enlarged to 11x14. This leads me to believe that the value in a 35mm lens that resolves 80 lp/mm is not resolution at 80 lp/mm but the correspondingly high contrast at 40 lp/mm.

3. In a MTF, the spatial frequency corresponds to a sinusoidal pattern at that frequency, not a sharp edged bar chart at that frequency. Those sharp edges are generated by higher spatial frequencies. So if a lens has a moderate modulation at a certain spatial frequency, the image of a bar pattern at that frequency will show light and dark bands, but the edges of the bars may be terribly soft, depending on how the lens does at higher spatial frequencies.

One way to generate the MTF curves would be to use lots of sinusoidal patterns at different frequencies. In practice, the image of a sharp edge or point source can be mathematically transformed into the MTF values for all frequencies. Actually, you could use any input for which the Fourier transform is known. The advantage of sharp edges and point sources is that they contain large amounts of all frequencies (the ideal point source is mathematically a Dirac delta function with an equal content of all spatial frequencies).

Again, however, most published MTF curves are not measured but theoretical. Hasselblad's may be measured. I know that photodo.com uses the same test setup as Hasselblad to produce their MTF charts.

As for other comments above, I always field test lenses. Side lit tin roofs (common here in central Texas) provide very good spatial frequency targets, and also a good target for spotting chromatic aberrations. I don't think I know exactly what MTF values will begin to show real visible image effects. I would love to have actually test MTF for several samples of a lens design to see what sample variation is like, and then to field shoot those lenses to see if those variations are visible in images. I suspect that we would be surprised at how much MTF can vary without observable differences in images... and I suspect that other factors such as film registration are in fact more important in actual image quality than sample to sample MTF variation. But I don't have the ability to test these hypotheses.

So, again, I use the MTFs mainly to make sure that the lens design doesn't overemphasize coverage at f/22+ at the expense of high spatial frequency performance and central field performance at f/11- f/16 BECAUSE I am using these lenses on roll-film.

So here is a current example. I have eagerly awaited the 80mm Super Symmar XL as a sharp but lightweight moderate wide-angle for roll film work. Schneider has published MTFs at f/8 and f/22. The MTF falloff with off-axis angle at f/8 is astounding and the falloff at f/22, although much better, seems correspondingly worse than equivalent off-axis angles for the 110XL. Based on MTFs alone, it appears that the 75mm f/4.5 Grandagon-N will be a much sharper lens for roll film work when used at f/11 or f/16, while for 4x5 work at f/22-f/32, the Schneider will be as good or better with a bit more coverage. On paper, it would seem that this is exactly the tradeoff I am trying to identify. But I don't have chromatic aberration graphs for the Schneider, and I was very impressed by the 110XL in action, so before spending $1100+ on either lens, I fully intend to do some side by side testing on film.



-- Glenn Kroeger (gkroeger@trinity.edu), January 08, 2001.


You might want to look at the Photodo web site. They have a discussion of MTF as well as measured curves (for smaller formats).

One thing to look out for when comparing MTF curves is that some people plot spatial frequency on the x-axis and other people plot distance from the center of the lens.

-- John H. Henderson (jhende03@harris.com), January 08, 2001.


There is no mumbo jumbo about ISO 9000. It is world wide. It is expensive to obtain and keep - so much so that it is unlikly a one man shop would be able to even afford to attain ISO 9001 certification.

It is also proudly posted on bill boards and banners in front of all types of American manufacturers and suppliers today.

You sound more like one who did not attain ISO 9001 or 9002. ISO 9000 is much simpler to attain.

-- Bob Salomon (bobsalomon@mindspring.com), January 08, 2001.



Bob stick to peddling overpriced cameras and do not attempt to remark on something which you know absolutely nothing about. "ISO9000" is a series of standards governing manufacturing/service practices which covers quality assurance and quality management. There are no "levels" as you imply. ISO9001 deals with design, development, production, installation and servicing of products. ISO9002 deals with production and installation. ISO9003 deals with final inspection and test. The "level" at which you gain certification has to do with the nature of your business, not the quality of it. The practices espoused in ISO9000 are nothing new - their roots can be traced back to post WWII and the work performed by J.M. Juran, W. Edwards Demming, Miles Feigenbaum and others. It is really just a unification of many pre-existing military and civilian specifications and practices. It has nothing to do with final "quality" of a product (it's not even addressed) - it only deals with the integration of management and testing systems. If I started an optical company tomorrow and I stated that the purpose of my company was to produce optics that would allow light to pass from one side of the lens to the other I could gain ISO certification. The fact that companies advertise that they are ISO-certified is because an ignorant public doesn't know any better. And it's only expensive if you do it wrong. The company that I was with which had the "one-man" certified subcontractors gained their certification through CSA at a cost of around $20,000 - the labor involved was done by the employees who produced all of the documentation inhouse (and written-off as overhead). I wrote most of the production/inspection procedures myself (I was the quality-assurance manager at the time).

-- Wayne DeWitt (wdewitt@snip.net), January 08, 2001.

Not sure about the point of your last post Bob - but I've got a quote for you.

"Our plants are QS9000/ISO9000 Certified and work every day to continuously improve their operations, which improves the products you receive. Several of our plants have also achieved ISO14001 Certification for their environmental efforts"...

It's copied off of the Firestone website. Caveat Emptor.

-- Wayne DeWitt (wdewitt@snip.net), January 08, 2001.


Wayne, that summed it all up! Nice job...

-- Bill Glickman (bglick@pclv.com), January 09, 2001.

Then it is fortunate that Bridgestone and its subsidiaries are not in the photographic manufacturing business.

-- Bob Salomon (bobsalomon@mindspring.com), January 09, 2001.

Anyone taken any nice pictures recently?

-- dave bulmer (dave.bulmer@bl.uk), January 09, 2001.

Thank you for your contributions to my questions. I too am familiar with ISO and Wayne's comments woke up my memory cells on those issues. I know Wayne is absolutely right but I know too that Bob deserves better. First, because ISO looks from the outside just as as Bob saw it and all those gilded words about quality sound very convincing. That is why ISO has been such a marketing success. Second, because Bob has been a most helpful contributor to this forum and he has helped many, including me. Lost in the debate is the fact that the German optical industry, whith its traditionally high standards did not need Juran Deming or Crosby to teach it quality nor does it need ISO to convince the world that they practice it. As for MTF, I learned much in the process yet I feel that there are many worthwhile technical issues that could benefit from further discussion. Perhaps those can be revisited at a later date, no ofense meant to all the empirical die-hards.

-- Julio Fernandez (gluemax@ora.auracom.com), January 10, 2001.

For what its worth, I have a friend who used to work in the photogrammetry field here in the United States and he worked extensively with Zeiss engineers and technicians on optical design. He decided to leave the field and become a full time fine art photographer. Before investing a large amount of money in a large format system, he asked the Zeiss engineers what manufacturer of large format lenses they would recommend. The Zeiss people stated that they "had the utmost respect for Rodenstock". They strongly believed (and observed - some had worked for Rodenstock) that Rodenstock's quality control and production standards were superior to those of other manufacturers assuring one of a production lens that met the design goals, tolerances and poublished information. As you can imagine, some manufactured lenses don't lie within the tolerances allowed and suffer from decentering, etc.

The Zeiss people stated that Rodenstock, like Zeiss, won't embark on manufacturing a certain lens design if they feel they cannot manufacture it to the riqorous standards they employ. Accordingly, they state that this is why you may see Schneider, etc., come out with a lens somewhat on the edge of optical design - 55mm, 45mm etc for 4x5. They want to be first to the market. This philosophy is not used at neither Zeiss nor Rodenstock. When Rodenstock if/and Zeiss come out with a cutting edge design, one can be assured that you won't be seeing a "new and improved version" for quite some time. If/when a new version comes out pushing the envelope further, it usually involves technology that is being worked on in the laboratory, not technology that exists, but neglected so one can get "first to market"

Ran across your forum, hope you don't mind if I jump in once in awhile.

P.S. Again, for what its worth, the Zeiss engineers he worked with also spoke very favorably on the Linhoff equipment, some of them investing in it for their large format systems.

-- Daniel Salvucci (danielsalvucci@netscape.net), June 25, 2001.


Moderation questions? read the FAQ