Starting to get nervous ?

ed rader wrote in post #4242067
i don't recall the 20d or 30d having problems like this.

And yet they did have their share of problems - back focus on the 20D being a recurring theme in a Google search...

All's I know is I got my 40D last Saturday, went out shooting last Sunday, and I'm pretty damn happy with the results...

Keith R wrote in post #4242613
And yet they did have their share of problems - back focus on the 20D being a recurring theme in a Google search...

how about the 30....which was essentially an improved 20d? maybe the 40d is the 20d of the next generation.

ed rader

Stephen Scharf wrote in post #4242055
We can do a little Six Sigma experiment from the data we have from Shawn's poll:

If you take the data in the focus issues poll as a representative sampling, you can calculate a process sigma.

Adding all the votes reporting 1) no problem or 2) reporting a problem, there are 108 units out of 125 not reporting "defects" (front/back or AI Servo problems), and 17 reporting a "defect". A process sigma calculation using this data yields a process sigma of 2.60. With a process sigma of 2.60, you have a theoretically, an 86.4% chance of receiving a camera with no problems, and a 13.6% chance of receiving a camera with focus problems. The potential buyer can decide if that is a chance that is acceptable to them.

It should be understood that this is not a very big sample for testing binomial statistics-based attribute data (the attribute being that the camera either has a defect, or it does not), so the confidence levels are not particularly high. If you would like to have 90% confidence, the chances are that the defect rate, in fact, could be as low as 5.9%, due to the sample size we are using.

For example, if we run a Power and Sample Size 1-proportion test in Minitab, assuming for a moment that our observed defect rate of 13.6% is real and we have a sample size of 125, with 90% condifence level you get:
Testing proportion = 0.136 (versus < 0.136)
Alpha = 0.05
Sample Size:125 Power:0.90 Alternative Proportion: 0.0586381

The alternative proportion value of .0586 means that, with our poll sample size, you do not have sufficient sampling power to reject the (null) hypothesis that a 5.9% rate of defects is any different statistically than the observed 13.6% reported defects.

FWIW, if this data was representative, a process sigma of 2.60 is nothing to write home to mother about.

uh...yeah ! are you a black belt or sumthin' ?

ed rader

Hi Ed,
Yes, I was trained as a Black Belt in 2005. My "day job" is at a biotech company in the Bay Area. I've only known about Six Sigma for about three years; my background is as a human molecular geneticist. Now helping to teach Design for Six Sigma at my company.

Stephen Scharf wrote in post #4242055
We can do a little Six Sigma experiment from the data we have from Shawn's poll:

If you take the data in the focus issues poll as a representative sampling, you can calculate a process sigma.

Adding all the votes reporting 1) no problem or 2) reporting a problem, there are 108 units out of 125 not reporting "defects" (front/back or AI Servo problems), and 17 reporting a "defect". A process sigma calculation using this data yields a process sigma of 2.60. With a process sigma of 2.60, you have a theoretically, an 86.4% chance of receiving a camera with no problems, and a 13.6% chance of receiving a camera with focus problems. The potential buyer can decide if that is a chance that is acceptable to them.

It should be understood that this is not a very big sample for testing binomial statistics-based attribute data (the attribute being that the camera either has a defect, or it does not), so the confidence levels are not particularly high. If you would like to have 90% confidence, the chances are that the defect rate, in fact, could be as low as 5.9%, due to the sample size we are using.

For example, if we run a Power and Sample Size 1-proportion test in Minitab, assuming for a moment that our observed defect rate of 13.6% is real and we have a sample size of 125, with 90% condifence level you get:
Testing proportion = 0.136 (versus < 0.136)
Alpha = 0.05
Sample Size:125 Power:0.90 Alternative Proportion: 0.0586381

The alternative proportion value of .0586 means that, with our poll sample size, you do not have sufficient sampling power to reject the (null) hypothesis that a 5.9% rate of defects is any different statistically than the observed 13.6% reported defects.

FWIW, if this data was representative, a process sigma of 2.60 is nothing to write home to mother about.

So - long story short - even if the poll data were random (which they are not as the respondents are self-selected), the sample size is too small anyway to give a significant result at the 0.95 confidence level. In other words, the poll tells us nothing.

Depends on the question being asked. The process sigma calculation does not involve a confidence level. Also, over-reporting could be due to user error, and under-reporting could be due to users not stressing the AF system.

Stephen Scharf wrote in post #4242962
Hi Ed,
Yes, I was trained as a Black Belt in 2005. My "day job" is at a biotech company in the Bay Area. I've only known about Six Sigma for about three years; my background is as a human molecular geneticist. Now helping to teach Design for Six Sigma at my company.

hi Stephen -- i met a black belt who works for sun. she told me a little about being a black belt but in truth i didn't retain much .

ed rader

40d wrote in post #4198328
Wait till you get the camera in your hands and enjoy it. Many more people have been happy with the 40d I suspect than have had problems - of course those with problems are more likely to post about it. There are also a number of people knocking the 40d that have never so much as held one just adding to your fear.

Edit: Created a poll for 40d owners to state their happiness/disappointme​nt with the 40d here

but you could also say the same thing about the 1d mark III.

ed rader

Ummm - that would be your null hypothesis, surely. The "process sigma" you mention appears to be just the proportion of events in one class, so one would not expect it to have a confidence level associated with it, since it is a known fact related to the sample, rather than an estimate of the population proportion, and the experiment has insufficient power to make meaningful estimates about this (e.g. p> 0.95).

All good stuff.
We now return you to your normal programming....

Got my 40D and loving it.... no complaints, here!

I think we are losing what the poll was intended for: (reason was to see if Rob G's web comments were leading to a 40D...MarkIII...type problem)

1. Is there some "massive" focus problem?
2. If there is a problem, "what" have people noticed.

I do not believe there is a focus issue with the 40D. I believe that it does have a AI Servo learning curve when using multi-points. If the subject does not fill the frame and is on a complicated background the focus will lock on what it calculates it should and sometimes that is not the intended subject.

Other than this, I don't see anything but camera/lens/picky owner front or back focus issues.

OR

The there is a HUGE problem...but the self selected poll respondents don't use the camera's focus capabilities in a way that shows the HUGE problem.

I'd say there is NO ISSUE....

snappa wrote in post #4191973
I am beginning to worry a bit as I have ordered a 40D and I have read some very conflicting reports on here. I am a regular visitor here and realise that if something does go wrong the owners of the cameras do tend to shout the loudest !
Tales of having to return the camera, tales of trouble with the AF, tales of more or less everything are starting to bother me. I suspect I just need a bit of reassurance.
Helllllllllllllllp please.

You can't believe 1/2 of everything you read on the internet.

Hey, can you two Black Belts go karate chop each other somewhere else? None of what you have had to say has actually helped the OP, it has only given you guys a chance to come in here and flew your brains in front of everyone. Thanks.

P.S. I have 2 Black Belts (Six Sigma) in my family, my mother and my sister.

Tyreman wrote in post #4243738
You can't believe 1/2 of everything you read on the internet.

I'm still trying to determine whether to believe your post or not.