Thanks a lot for your comments.
I'll be doing some editing in my reply -- for reference, my original post
can be found at http://lists.tako.de/html/Olympus-OM/2005-01/msg01802.html.
At 6:33 PM -0800 2005.01.26, Moose wrote:
> ... Careful definition of what one desires to measure is also
>important. Asking about hidden assumptions can help in resolving these
>issues.
Most definitely. In the interests of brevity, I decided to instead say
exactly what I used to get the numbers, hoping that this would bring out
any methodological flaws, and hopefully discussion of how to do it better.
As I said, I wanted to get some measurements involving just the lens,
trying to factor out the camera's TTL system, which is necessarily
involved when measuring exposure times in auto. I'm not saying this is a
bad way to do it, but I thought it might be interesting to have some
corroboration.
>
>Andrew L Wendelborn wrote:
>
>>Having been intrigued by this thread from last week, I tried to do some
>>measurements of different lenses to confirm how often there was a difference
>>of less than one stop between two consecutive f-numbers, and how great
>>the difference was.
>>
>>I wanted to do it without using the camera, to eliminate any effects of the
>>TTL system.
>>
>Generally a good idea, but see below.
>
>>After a bit of fiddling around I set up things thus:
>>
>>. bellows, with lens under test attached, but no camera;
>>. slide copier, attached to bellows in usual way via filter ring
>> i.e. only 49mm lenses can be tested;
>>. slide projector acting as light source;
>>. glass diffuser plate (from Durst Laborator 1000 enlarger) placed
>> upright in front of projector, so as to be fully illuminated and provide
>> a bright, even diffuse light source;
>>. copier / bellows unit mounted on tripod and positioned so that rear end
>> of lens abuts diffuser plate;
>>. the ground glass plate in the slide copier is then evenly illuminated;
>>
>So, as I understand it, the lens being tested is being tested backwards,
>with the light coming through the rear and out the front, where it is
>measured. Unexamined assumption 1 is that this will give the same
>results as using the lens in the direction in which it was designed to
>be used.
Yes, I realized that at the time. I regarded this as a sort of preliminary
study to see if I could get measurements that made some sort of sense, and
to work out something better if it seemed worthwhile.
I did in fact try two different ways to measure the lens "right way around".
One was to attach a varimagni, then try to get readings from the spotmeter
by placing it in the eyepiece of the varimagni. I couldn't get repeatable
readings this way; also, there is a mirror and screen in the light path.
Then it occurred to me that I could measure at the film plane by somehow
contriving to get a matte screen fixed there securely enough to take spot
measurements off the screen. (Matte because the fresnel affects the reading,
as implied in the letter from Olympus to Fernando.) This is then analogous
to the technique described by Walt in another thread for doing similar
measurements with a LF camera, where it much easier! But I figured the
risks of damage to the screen outweighed the uncertainty of significant
advances to scientific knowledge. Hence I came up with the more
convoluted apparatus whereby I could use the ground glass plate of the
slide copier.
> My assumption based on experience and intuition, but not direst
>knowledge, is that this will tend to be true of the rather symmetrical
>designs of 'normal' focal length lenses and may become less true s one
>goes farther toward WA and.or tele.
My thoughts also. Those results in particular need further corroboration.
The idea here was to point the way.
>
>Again as I understand it, no effort is made to use the lens in anything
>like a normal distance relationship to subject and film plane.
See comment above.
>Unexamined assumption 2 is that this will give the same results as using
>the lens 'in focus'. I realize that the subject is diffuse and
>measurement is through a diffuse medium, but still don't know that
>address the assumption.
I'm not sure either. I set out to ensure that the illumination measured
was as bright and even as possible, given the limited time and facilities
available.
> A quick test with the 200/5 shows that the light
>level shown by the meter of an OM-1 varies as a blank white subject is
>brought into and out of focus. So there is at least a potential effect
>to be considered.
I agree. I did do some spot checks for the effects of focus on the first
50/1.8 lens, and found nothing significant. But a thorough test would
address this.
>Another interesting assumption, that only light at the center need be
>measured. Here, the experiment runs up against the issue of what is to
>be measured and what the designers may have measured when setting the
>stop detents. Simple brightness at the center is a possible criterion,
>but not the only one.
Yes, agreed. In doing some web searching beforehand trying to track down
published results, I found very little. But I did find a passing reference
to t-stop measurement that implied it was usually taken at the centre.
Unfortunately I lost the reference :(
I did a few cross checks, and found edge measurements to give similar
results wide open. But again it was lack of time -- what I was doing at
each exposure setting was to take at least five readings to ensure
consistency. Interestingly, there was never more than 0.1 stop variation.
> Another measure could be total light delivered to
>the film plane, summed over the whole area.
Yes. At first I thought the only meaningful way to do the experiment would
be to design a light integrator. But then I figured I was getting numbers
in the right direction, and that I could use them at least to find out
where these lenses exhibit anomalous behaviour.
> Here we encounter the issue
>of vignetting, at least for tests from wide open to 1 stop down, but to
>some extent with lower stops. In a lens with significant vignetting,
>measures of the light delivered at the center at the 2 stops will differ
>by a different amount than will measures of unweighted average
>brightness over the whole film frame at the same 2 stops. And the
>direction of the difference will be in the direction of the 'error'
>under discussion and that you have measured.
Yes. I don't know whether or not this explains the differences observed
in timing tests, which use the camera's integrator.
I would be very interested to know about accepted methods of measurement
for this.
>
>>I tested a few lenses, taking several readings at each stop, and running
>>up and down the range several times.
>>
>>Here are summarized results, recording ERRORS i.e. deviations from expected
>>f-number differences (usually one, but not always, as in going from eg f5 to
> >f8, the expected difference is 1.3 stops). If the error is 0.1 or less,
>>it isn't recorded below.
>>
>>
>>50/1.8 f1.8 -> f2.8 error 0.5 stops
>>
>>28/3.5 f3.5 -> f5.6 no error
>> f11 -> f16 error 0.3
>>
>>50/1.4 f1.4 -> f2 error 0.6
>> f11 -> f16 error 0.3
>>
>>50/3.5 NO ERRORS
>>
>>200/5 f5 -> f8 error 0.7
>> f22 -> f32 error 0.3
>>
>>
>>Note errors are big for the 50/1.4 and 200/5.
>>
>>I've started to do some timing tests on a 2n.
>>
>Not quite sure what you mean by this. If using actual shutter speed of
>the TTL system, (using very slow film speed settings and a stop watch?),
Yes. Timed using "stop watch" in my mobile phone. Adjusted to a reference
point of 10 seconds to reduce effect of reaction times.
>this should give much broader averaging of light over the whole frame,
>depending on the evenness of coverage of the little lens on the mirror
>box floor. It may vignette too. I always thought TTL metering on the 2n
>was slightly center weighted, but have no info or proof either way.
I looked into that a while ago. Both my own measurement, and the scant
technical information that exists in books etc, suggest what you say --
mild centre weighting.
> If
>the lens being tested is at least roughly focused on the light source,
>the first 2 assumptions are also eliminated as possible sources of error.
>
>>For the 50/1.4, I get 1.5 increase from f1.4 to f2
>>(exactly as Wayne H did), which is an error of 0.4 stops,
>>a bit less than the 0.6 observed above. Not sure why.
>>
>>For the 200/5, I get exactly the same error as above. A bit disturbing.
>>
>That bothered me too. I did a much less elaborate test. Using an OM-1
>fairly recently CLAed, I measured exposure through a 200/5. I used an
>OM-1 for several reasons. First, it is easy to see small variations in
>the analog meter, giving what feels like finer graduation than on a 4,
>whether true or not. Second, its light metering is rather mildly center
>weighted. Third, the CdS cells respond slowliy enough that I could use
>my computer screen for the first test without the meter jumping around
>as on the later models. A fourth, non-logical reason is that it had the
>roll of film I was using at the time in it so I could also observe
>reaction with the 200/5 while out shooting.
Good reasons. My own tests with the 200/5 were a little rushed, but I
did get the reported error three times. Unfortunately I left the notebook
with the data records, at work and can't consult them now.
>
>Aiming at an almost blank, white subject, my computer screen, and with
>the lens not really in focus, but as close as I could get without
>getting out of my chair (This IS the meaning of armchair science, isn't
>it?).
Absolutely. I used the same technique!
> Moving from f5 to f8 and back, I observed a 1 stop difference in
>the meter. As the difference between those apertures is slightly less
>than 1 1/3 stops, that would indicate a difference of no more than 1/3
>stop, quite a bit less than you got. Out in the world taking pictures, I
>observed the same difference.
Interesting. On the 2n that I used last week, meter deflections on manual
and auto were both similar to what I measured.
Now on a 1n (with a different screen) I get deflections similar to your
observations.
Too late to figure that one! And besides I'm concurrently watching Federer and
Safin at the Aus Open ...
>
>>As I said, this is very much summarized, so happy to answer questions if
>>anyone is actually interested.
>>
>I'm actually interested, less than because I think it makes a lot of
>practical difference, but as an intellectual exercise in definition and
>measurement.
My interest was, firstly, to get a feel for this effect -- where it
occurred, and some idea of its magnitude. And to understand if it might be
significant in computing times for long exposures from a test exposure
taken wide open (as I mentioned in an earlier post). Probably not -- I
suspect it's a fudge factor that becomes indistinguishable from the many
others that exist in such exposures.
It's to some extent interesting to puzzle about the factors in lens design
that lead to anomalous behaviour at the last stop. Similarly to your
comments above, I expect engineering and cost aspects of say a wide and
fast lens lead to a faster f spec that t spec. This might be the
"curvature of the lens" comment in the Olympus letter to Fernando.
I also wondered about differences in practice between manufacturers. I did
a quick "deflection test" on a Leica CL with 40/2 Summicron-C. This gave
the same deflection fully open to 2.8 as for other one stop adjustments. But
the aperture ring actually rotates slighly past the f2 mark. So is it an
f1.9 or 1.8 lens speced as f2 to give the correct ratios?
> I'm not trying to put down your tests either, just
>exercising my own curiosity on the subject.
That is the spirit in which I have taken your comments, and offer mine.
It would be interesting to see some other measurements and thoughts ...
regards
Andrew
==============================================
List usage info: http://www.zuikoholic.com
List nannies: olympusadmin@xxxxxxxxxx
==============================================
|