View Full Version : MTF determination: Rodenstock's answer

Bob Salomon
7-Feb-2001, 11:12
There have been unsubstantiated claims of how Rodenstock computes MTF curves with some claiming that the MTF curves are design points rather then actual performance curves. Obviously this is not a sensible approach to making curves available so we asked Rodenstock for their comments.

Their response is:

"calcuted. The reality is when you measure MTF curves on the MTF machine with a lens it could have a tolerance of -10 % at the most. In other words there is a difference when you measure the MTF together with the lens."

Hopefully this response will end the misguided, unsubstantiated and uneducated claims of theoretical curves.

Editor: See followup at http://hv.greenspun.com/bboard/q-and-a-fetch-msg.tcl?msg_id=004ZdS (http://hv.greenspun.com/bboard/q-and-a-fetch-msg.tcl?msg_id=004ZdS)

Pete Andrews
7-Feb-2001, 12:28
That response seems to indicate that the MTF curves exactly are theoretical (calculated), and that an actual lens can be 10% down from them.Doesn't it?

Bob Salomon
7-Feb-2001, 12:42
Let's be real. Obviously when a company publishes in a brochure or on the internet or anywhere else their MTF curves it would be impractical and impossible to publish the curves of every lens manufactured now, then or in the future.

Rather then what some were saying and drawing design curves. Rodenstock posts curves from a production lens. They state that any lens will be within -10% of that production lens.

That means that any lens will equal that curve or be - 1%, -2% or, at the worst, -10% from the posted curve.

That is pretty straight forward. Not engineering curves, but curves from an actual sample that tests within - 10% of every other lens of that type.

7-Feb-2001, 13:24

I disagree. Your quote specifically says "calculated." Now, if your quote had said "measured" or "a representation of a statisitcal sampling" they you could claim that the curves are from real measured data.

The way I read the quote, the Rodenstock comments appear to imply that the curves are precisely theoretical, and have been calculated based on the theoretical performance of the optical elements, etc.

Personally, I can't understand why you felt the need to use the tone and language you have used. It does nothing but create animosity and raise people's hackles. I don't believe anyone is trying to imply that theoretical curves are misrepresentations of the lens performance, and even if they were, since there are no measured curves out there, what's the point in debating this?

Both German lens manufacturers use theoretical curves, and the Japanese don't publish any curves that I'm aware of, so where does that get you? It's simple, and it's a point you have brought up in the past, Bob, you have to go out and test a lens under real-world conditions and determine if it is suitable for you own particular needs. That's the ultimate measure of a lens, not some curvey lines on a chart that nobody can compare to anyone else's since nobody uses the same units or metrics anyway.


Bob Salomon
7-Feb-2001, 17:12
Sorry Mike.

I said that a production lens is tested. And that curve will be within - 10% of the other lenses in the series.

Nothing theoretical. These are NOT the curves that the designers dream of.

They are the curves to be expected from an actual lens that is purchased across the counter.

What of this do you find difficult to comprehend?

QT Luong
7-Feb-2001, 17:44
I find it difficult to comprehend why they say "calculated" if your interpretation is to be right. Maybe you can ask them to elaborate a bit more, possibly show them the comments which were made here, and follow-up on this thread with the detailed answer ? This way, we will learn something instead of speculating on the obscure meaning of a paragraph.

Kerry L. Thalmann
7-Feb-2001, 18:46

Based on the quote you provide above, I still have no idea how Rodenstock arrives at their published MTF curves. The quote is incomplete, out of context and in broken English (I'm not blaming the source of the quote, since English is likely not his/her native language, but it makes it difficult for some of us to understand what is being said).

You claim, based on the above quote that the Rodenstock MTF curves are produced by measuring a single lens and that all others will then be within -10% of that measured lens. That may be true, but it is certainly not the conclusion I arrive at based on the above quote. All I get from the above quote is that something is (or is not) "calculated" and that lenses measured on an MTF machine can vary by as much as -10%. What is left to the imagination is what exactly is (or isn't) "calculated", and if the -10% tolerance is due to the lens to lens production variations, or the accuracy of the tester.

If as you say, one lens is measured and the curves based on that lens, how is that one lens selected? Is that lens the best (if so, how do they know if they don't measure the others), the worst, or "typical"? It has to fall somewhere on the curve, but where? Is the -10% figure then based on statistical samples (measuring more lenses) or theoretcial caclulations? Is the -10% (and why isn't it + or - 10%?) actually the variability in the performance of the lenses, or is it due to the test environment?

I'm not deliberately being a pain here, I really am interested in learning how this is done (or at least a clarification of the above quote). Without the complete quote, and the proper context, I am left with more questions than answers. For example, you started the quote with a truncated sentence ending in the word "calculated". It sure would be nice to know what the rest of that sentence said. Could be anything from "The MTF curves are measured, not calculated" to "This is how the MTF curves are calculated". You provide this quote as proof that the Rodenstock MTF curves are based on measured data and not theoretical models or computer simulations. Unfortunately, the evidence is incomplete and therefore proves nothing (BTW, I'm not saying what you have stated is false, just that it's impossible to tell based on the quote you provided).

Perhaps you could clear this all up by posting the entire exchange, including both the question and the answer, to put it all in context. I don't think the confusion is caused by any animosity or stupidity on anyone's part, it's caused by insufficient information to reach a definitive conclusion one way or the other. There was obviously more to the exchange than the brief quote you posted. Why not post the rest and hopefully enlighten the rest of us.


Bob Salomon
7-Feb-2001, 20:38

That is the complete applicable post as to the MTF question.

The remainder of the message does not address the question,

The statement is quite clear.

A production lens is tested.

All lenses of the series will be within - 10% and the published graph.

If you want specific answers to specific questions ask them.

But don't read between the lines.

The lens tested iis reular production and not some imaginary curve that some allude to incorrectly Please bear in mind when I post their direct answers that English to the factory is a second language. The answer is direct and to the point but thet don't speak English or write English the way a native American English speaker does.

Kerry L. Thalmann
7-Feb-2001, 21:04

It may be clear to you, but it is not clear to me. Hence my questions. No where in the quote does it state that a production lens is tested. That is your interpretation of the quote.

The exact quote is:

"The reality is when you measure MTF curves on the MTF machine with a lens it could have a tolerance of -10 % at the most."

To me, that just means there is a 10% tolerance in the measurements. It doesn't state the source of the 10% tolerance, and it certainly doesn't state:

"A production lens is tested.

All lenses of the series will be within - 10% and the published graph."

That may be your interpretation based on your conversaations with eth factory, and it may very well be true, but it's not what is stated in the quote. In fact, no mention of production lenses or MTF charts is even mentioned in the quote.

I did ask specific questions and you chose to ignore them. So, I won't bother to re-ask them.


7-Feb-2001, 21:08
>"calcuted. The reality is WHEN YOU MEASURE< - as opposed to calculate (not measuring).

>MTF curves on the MTF machine with a lens it COULD HAVE a tolerance of -10 % at the most.< - WE HOPE

>In other words there is a difference WHEN YOU MEASURE the MTF together WITH the lens."< - as opposed to not measuring. How would you MEASURE the MTF WITHOUT the lens?

Clearly they are stating that published MTF curves are calculated.

Bill Glickman
7-Feb-2001, 21:54
Rule number 1 - Bob can't be wrong. Rule number 2 - If Bob is wrong, see rule number 1.

Bob, all the posters raise very sensible points, instead of fighting everyone, why not just state the facts in clearly defined english - after you have confirmed it from the factory. Isn't it obvious you confused us all? Please, have a little respect for some of the extremely intelligent posters to this forum.

Glenn Kroeger
7-Feb-2001, 22:03
Well, I must be misguided and uneducated, but the quote does not speak to whether the published curves are measured or calculated, only that IF measured, real lenses have about a 10% variance. The latter is important information. Bob may be right that Rodenstock publishes a measured curve, or the average of many measured curves, but the evidence is yet to be presented.

7-Feb-2001, 22:59
Glen -

Sorry but I have to disagree with you also. And I have to thank you for pointing out the most obvious reason why in a statement that I overlooked. The statement "tolerance of -10 % at the most" indicates that the measurements are most definitely theoretical. If the data was from real measurements any tolerance would be stated as "+/-" not "-" only. Since you can only be worse than theoretical values, never better, the "+" cannot be used. If the measurements were of a sample they would state tolerances as "+/-" since the lens tested is subject to the same variation as the rest of the production, and could be worse than other samples. All indications point to the data being theoretical since I don't believe that the Germans would express themselves so poorly - even in a foreign language.

Glenn Kroeger
7-Feb-2001, 23:16
Wayne: excellent observation! I didn't catch the - sign.

7-Feb-2001, 23:32
Glen -

I just read the other answers and Kerry also noted the discrepancy in an earlier post. Since his response was so long (sorry Kerry) I skipped it the first time I read through the responses. Germans might struggle with english grammer, but math is math.

Julio Fernandez
7-Feb-2001, 23:47
My interpretation is this: Rodenstock publish "calculated MTFs" but they know from experience that when production lenses are actually tested their tested these tests show a 10% variation from the calculated MTF. Rodesntock does not MTF-test every lens produced, although by all rules of quality assurance, spot checks are carried out on a regular basis. Did Rodenstock's words say that? No, but it this is what I believe they meant and only Rodenstock can validate affirm my interpretation. Bob just tried to clarify unclear statements thought in German and written in English, and in the process put his own spin on them. As to taking pictures to see if the lens is OK, the infallibility and allure of this simple approach is too obvious and its allure too much for some to refuse. However, this approach succeeds only when the results are positive. It is succeptible to too many false-negatives and hocus pocus to be a scientifically valid and universally reproducible approach. There are sound, scienfic methods for evaluating lenses today and I personally believe those to be more reliable than joe's next door. If not, it would be quite OK to design lenses according to Joe's test methods. Personally I rather buy Rodenstock or Schneider lenses than Joe's. On another issue, the Japanese do not publish MTFs but the microscope and binocular division of one of those camera lens manufacturers do use MTFs for evaluating their lenses and making claims as to their quality. Of the lens manufacturers, only Zeiss's MTFs are for production units. When all is said, it would be helpful for Bob to address this issues with Rodenstock and ask Rodenstock to issue a position paper. Like many in this forum, I guess, we would rather hear it from the horse's mouth.

Kerry L. Thalmann
7-Feb-2001, 23:52
That's the problem with insufficient data. If you want to reach any sort of conclusion, you have to make assumptions. Based on the original quote, Bob had to make some assumptions to reach his conclusions. He chastised me to "not read between the lines", but that's exactly what he did (unless he has additional data he is not sharing). The problem is, there isn't enough information "in the lines" to support any conclusion one way or the other. Based on the incomplete, out of context and confusing quote that initiated this discussion, Bob can neither prove his conclusions, nor can we conclusively disprove them. We need more (or at least better) data. Since it is pointless to speculate without additional data, I suspend further commentary (but will not be holding my breath) until that data is provided.


8-Feb-2001, 22:52
I am a native German speaker and am afraid the statement quoted by Bob does not make any sense to me either - it certainly does not address the question as to whether the published curves are measured or calculated.

"The reality is when you measure MTF curves on the MTF machine with a lens it could have a tolerance of -10 % at the most." How do they know that a lens could be within 10% tolerance at the most unless they measure every lens? There must be occasional "lemons"! Perhaps what they are saying is that Rodenstock measures every lens and if it is below 10% of the published curves, it is discarded??? (This is pure speculation on my part).

Another question is what exactly do they mean by by 10%? Since MTF is expressed in %, this becomes ambiguous. Lets say the curves specify 40% MTF at a given frequency. Does -10% tolerance imply not worse than 36% MTF or not worse than 30% MTF?

"In other words there is a difference when you measure the MTF together with the lens." A difference compared to what? I thought that MTFs are always measured with a lens? If so, it must refer to a difference with calculated data, which would suggest (but I am guessing here again) that the Rodenstock curves as published are calculated, not measured.

Bob, I too would be very glad if you could clarify these issues with Rodenstock and post on photo-net:

1.) Are the published curves calculated or measured?

2.) If they are measured, do they reflect the very best lens they came ever across in their production line, or an "average" lens.

3.) What exactly does Rodenstock mean by "a lens could have a tolerance of -10%" at the most?

James Chow
9-Feb-2001, 03:42
The "10% at the most" deviation in the MTF for any random sample is also incomplete. As other posters mentioned, there is an occasional lemon. Of course there is! It's IMPOSSIBLE to claim that ALL lenses will deviate by at most some amount. Most manufacturers use a probability model to tell them the tolerances are within so much for some number of samples. It might be something like you get 1 lemon (MTF's are off by more than 10%) every 1000 units or 10,000 units.

Paul Coppin
25-Sep-2001, 22:01
Not to flog a dead cat with a radioactive lens, but there is another interpretation, (I think)- MTF data is calculated from theoretical design inputs for the lens - measured lenses fall within +90% of the calculated theoretical figure. ie, measured lens deviation is less than 10% typically. Did someone say that already?