More on Canon Rebel XT noise at high ISO - 2 main new data points

Today Alan Browne commented courteously on the subject at
hand

1) /Both/ camera's, my "defective" one and his "working"
one showed the /same/ amount of noise at /all/ ISO from
100 to 1600 on /all/ 3 of my PCs. Naturally, no noise is
readily apparent until ISO 400 on both cameras.

Could you post these images at full size (JPG is fine).

Alan, I'll let you pick the place this time. I can post the
full-size JPGs to just about any binary NG.

But, before you and the others spend a lot of time on this,
let me say that I've progressed quite a ways with my new Rebel
XT in the last 3 days. I'm well on my way to understanding
what works and what doesn't, what causes noise in the peculiar
shooting situations shooting cars in muesuems and what to do
about it (I already knew that, but I needed to alter my modus
operendi a bit), etc.

I went yesterday afternoon to The Henry Ford Museum in
Dearborn, Michigan, and shot a couple hundred test pictures of
cars in various surrounding and various ambient lighting
conditions.

I tried Programmed Auto on the camera with my new Canon 430
external flash in eTTL mode (that I didn't have when I posted
this thread), I tried the same flash in full-power manual with
the camera on full manual, and I tried shooting available
light at various ISO settings. While on flash in any mode, I
also varied the flash exosure from 0 to +2 flash "EV".

Before I went, I "calibrated" the LCD on the back of the
camera to what I see on all 3 of my PCs at home, particularly
the 21" Samsung LCD I use for production, and found that the
Rebel's LCD needed to be made dimmer by one "notch" to help me
best judge exposures in the field.

I still see noise beginning at ISO 400, which is controllable,
getting worse at 800, and fairly bad at 1600. But, if I've
gotten the exposure right within a stop or two, the noise even
at 800 and 1600 is controllable in PSP 9 using either EPS or
DCNR.

Since I posted this particular thread, I've shot around 600-
700 images in a variety of situations, including the Christmas
display in my church this morning. I'm well on my way, as I
described above, but still have a lot to learn. I have until
Tuesday night to return the camera, but I think I'll keep it.

When I compare results I got previously with my now-broken
Nikon Coolpix 5700 and a Sunpak 433D external, the Nikon 8800
and SB-800 external I test drove last April, and the Canon, I
find that the flash exposure is fooled in very similar ways in
car museums.

Briefly, if the flash pulse hits /anything/ highly reflective
in my car scene, including WSW tires, paint just right, glass
just right, chrome just right, reflective license plates, or
signs in the foreground, the exposure goes into "black cat in
a coal bin mode.".

The Canon, unlike my Nikon 5700, though, is fully capable of
manual exposure, which works if I cannot get eTTL to work even
at +2 flash EV. Again, I am not there yet with the correct
procedure du jour, I'm getting better quickly.

With all that as a proviso, if you still want me to post my
images, here's some NG choices:

alt.binaries.photos
alt.binaries.amp
alt.binaries.paint-shop-pro (where I posted previously, not a
good place)

These are on my subscribed list. If you have other ideas, I'll
look to see if GigaNews gives me them. I'd rather not post
these test images to places people expect to see real
pictures.

Thanks for the offer of help.

--
ATM, aka Jerry

"I disagree with what you say, but I will defend to the death
your right to say it." - Evelyn Beatrice Hall under the
pseudonym Stephen G. Tallentyre

Today Prometheus commented courteously on the subject at hand

Good question, perhaps you failed to understand that the
subject was not 'average' and tell the camera that your
intelligence is better than its.

because the F__king camera made a mistake,

No, you failed to understand the scene you were
photographing was not "average' and the camera did what you
chose to let it do.

then compounded /its/ mistake by being noisy. Don't like
that, huh? Well, it is true.

The camera did what you told it to do, treat the scene as
average (or to be more accurate failed not to tell it to
do); when you tried to correct that in the image you
enhanced the noise.

I must have missed you standing there next to me while I was
shooting my test pics, Prom. Or, you're just blowing smoke.

You could always try offsetting the flash and using AFB in
such circumstances, one of a sequence at 0, +1, & +2 should
be correct. You will need to experiment a little and think
about the results and how to apply your new skill in
future.

Yes. See several of my replies to sane people today.

--
ATM, aka Jerry

"I disagree with what you say, but I will defend to the death
your right to say it." - Evelyn Beatrice Hall under the
pseudonym Stephen G. Tallentyre

So underexposure seems to be the problem. Look at changing the metering
to a narrower central area so that bright reflections don't throw off
the metering.

I also wonder whether it might be better to shoot full 8MP images and
batch reduce them, instead of letting the camera downsample.

Paul said:
So underexposure seems to be the problem. Look at changing the metering
to a narrower central area so that bright reflections don't throw off
the metering.

Yes, although some metering systems are easily thrown by even a small
reflection, and as posted elsewhere, I think the poor old XT can only
offer 'partial' centre-metering..?

An even better, and frequently suggested solution (from an old school
type like me!) is to simply shoot using manual flash/camera settings,
given the problematic subject matter. I can still do guide number and
distance calculations.. and it's not like the subject is moving fast..

But these solutions have met with similar disdain and derision. If
Auto isn't involved, it's not acceptable, apparently.

See also the following posts, all by ATM on this topic:

Nikon Coolpix 8800 when used with Nikon SB-800 external flash
rec.photo.digital

Want to buy a new digital camera to replace my Nikon 5700, big ...
rec.photo.digital

Nikon Coolpix 5700 severe flash underexposure problem
alt.comp.periphs.dcameras, rec.photo.digital

Looking for advice/opinion replacing Nikon 5700 with Nikon ...
rec.photo.digital

Postscript to my Nikon 8800 and SB-800 query
rec.photo.digital

...and so on. I don't think I've ever seen a topic so flogged to death,
and with such simple causes and solutions.

All Things Mopar writes:

1) /Both/ camera's, my "defective" one and his "working" one
showed the /same/ amount of noise at /all/ ISO from 100 to
1600 on /all/ 3 of my PCs. Naturally, no noise is readily
apparent until ISO 400 on both cameras.

The problem, then is that your expectations are not being met.

What *are* you expecting at ISO 1600? Usable photos? Or crisp,
clean, studio-looking photos? Are you expecting less noise than P&S
cameras and a lot less noise than film that speed has grain? Or are
you expecting a lot better than that? Because "a lot better than
that" isn't available yet.
--
David Dyer-Bennet, , http://www.dd-b.net/dd-b/
RKBA: http://noguns-nomoney.com/ http://www.dd-b.net/carry/
Pics: http://dd-b.lighthunters.net/ http://www.dd-b.net/dd-b/SnapshotAlbum/
Dragaera/Steven Brust: http://dragaera.info/

All Things Mopar writes:

Today G.T. commented courteously on the subject at hand

Known issue on Digital Rebel and Rebel XTs. Use + EC.

What means "+ EC"? Don't know it, didn't get anything
Googling, can't find it in the manual (unless I've blind). I
know what EV is...

He means "use exposure compensation to add (that's the "+" part)
exposure". I don't think his notation is widely used for that, I just
understand it because it's the obvious workaround for the problem (I'm
not very into relying on the auto exposure of my cameras anyway).
--
David Dyer-Bennet, , http://www.dd-b.net/dd-b/
RKBA: http://noguns-nomoney.com/ http://www.dd-b.net/carry/
Pics: http://dd-b.lighthunters.net/ http://www.dd-b.net/dd-b/SnapshotAlbum/
Dragaera/Steven Brust: http://dragaera.info/

Today Paul Furman commented courteously on the subject at
hand

So underexposure seems to be the problem. Look at changing
the metering to a narrower central area so that bright
reflections don't throw off the metering.

Paul, please don't take one small part of a much longer
epistle out-of-context to where it looms very large.

As is well known by everyone with even a modicum of digital
experience, including me, severe underexposre leads to high
noise. Some EVFs, like my old Nikon 5700, got noisy with just
a stop or two under, even at ISO 100.

What I'm saying is that the Rebel XT noise at /any/ given ISO
is proportional to underexposure, again as is well-known, and
that I tend to get mild-to-moderate and sometimes severe
underexposure due to excessive flash pulse reflection off some
part of the car and/or something like a sign in the
foreground.

This does /not/ mean that /all/ of my flash pics are under,
they aren't. About 1/3 are OK +/- 1 f/stop, another 1/3 are
over +1/under -2 stops, the rest are under 3,4,5 f/stops. It
depends. Without being able to easily post cropped examples
here, it is difficult to explain.

I also wonder whether it might be better to shoot full 8MP
images and batch reduce them, instead of letting the camera
downsample.

I tried that. Resizing is /not/ one of the culprits as best I
can determine. I understand, though, what you're saying, I
think. You're suggesting that I not allow the camera to do
image damage due to interpolation downward, which may
exacerbate the noise.

On the other hand, though, resizing down correctly in a
competant app such as PSP 9 can and does "compress" some of
the noise out, so I /always/ crop for best composition
horizontally and vertically at full 5.5 MP (specifically, 2496
x 1664) /before/ doing any noise reduction or sharpening, if
necessary.

In my case, my final image size is generally 1280 x 960 or
1400 x 1050, or a different aspect ratio final than 4:3 if the
composition is improved. Since the Rebel XT only saves in 3:2,
I allow about 20-25% more space at one end or both ends of the
viewfinder, and have a PSP "Preset" for 2218 x 1664 so I can
easily put a 4:3 crop rectangle on my image and crop before
further post-processing.

Last night, I spent several hours RTFM for both the camera and
the Canon 430 EX external flash and discovered that I was
incorrectly applying flash "power" or EC (which I discovered
means Exposure Compensation). I shot more tests with both
flash and available light, but haven't had time to evaluate
them.




--
ATM, aka Jerry

"I disagree with what you say, but I will defend to the death
your right to say it." - Evelyn Beatrice Hall under the
pseudonym Stephen G. Tallentyre

Today Ed Ruf (REPLY to E-MAIL IN SIG!) commented courteously
on the subject at hand

So underexposure seems to be the problem. Look at changing
the metering to a narrower central area so that bright
reflections don't throw off the metering.

Noo, not hardly the biggest culprit. Testing so far is
inconclusive but strongly suggests that changed the camera's
metering mode does /not/ help. Rather, changing the zoom focal
length on the Canon 430 EX helps some, but very, very little.
So, in flash mode, AE method isn't the cure. It /can/ be in
available light, of course.

I also wonder whether it might be better to shoot full 8MP
images and batch reduce them, instead of letting the camera
downsample.

I all ready suggested shooting full res, batch processing
for noise reduction with Neat Image or similar and then
down sampling way back when. Could all be done
automatically in batch mode. I've also suggesting using the
histogram to help correct for proper exposure. but no, it's
the camera's fault. just like his contention his CP-5700
couldn't manually expose properly with the SB-800, which I
long ago refuted after dusting my old 5700 off to give a
try with my SB-800.

Yes, it is the God Damn camera's fault because modern TTL is
just "too smart" to detect when a small part of the subject
excessively reflects the flash pulse resulting in severe
underexposure. If you don't understand what I mean in terms
specific to taking pictures of cars in museums, please try it
with your fav camera and flash before suggesting I am both
dumb and refusing to accept your advice.

As to the histogram, when it is good, I don't need to play
with it. And, when exposure is very bad, I don't understand it
well enough to help - yet.

As to noise reduction, I don't currently own Neat Image,
Noisee Ninja or those, because PSP 9 has 2 very good noise
reduction techniques - EPS for mild noise and DCNR for severe
noise. Don't know what those mean? Then /you/ need to do some
research before lecturing me.

And, until I more fully understand /all/ of the various root
causes for underexposure and/or excessive noise, batch
processing down from 8 MP using PSP 9's batch function with a
re-size script isn't going to help, it'll just confuse the
issue for my simple brain.

Back to the Nikon 8800 and SB-800, when I spent nearly a week
testing that back in April, which you're referring to, I had
not learned full manual nor the most effective EC techniques
before concluding it was not to my needs, and returned it.

Finally, to my 5700. I'm glad to hear that you not only own
one, but that you "dusted it off" and tried it again. I much
more highly value knowledge from someone who has facts,
instead of opinions based on what they've read.

Since my 5700 is now in 4 broken pieces - I fixed it with a
hammer after it died - I can't take any more test images to
prove what I say, but have thousands at the HF and WPC museum
that can easily be examined by looking at the degree of under
exposure and EXIF to get the shooting parameters, then looking
carefully at the car subject and what is in the foreground
that tricked the 5700's AE, which was determined solely by a
little sensor on the pop-up flash.

Now I ask instead of assert: Ed, have you personally shot
pictures of /cars/ in museums and dealer showrooms, or more
general subjects in museums? If you have, great, I am all
ears. But if you have not, I suggest that you have no
expertise because you have no direct experience with the
frailties of the car picture biz.

--
ATM, aka Jerry

"I disagree with what you say, but I will defend to the death
your right to say it." - Evelyn Beatrice Hall under the
pseudonym Stephen G. Tallentyre

Today Chrlz commented courteously on the subject at hand

So underexposure seems to be the problem. Look at changing
the metering to a narrower central area so that bright
reflections don't throw off the metering.

Yes, although some metering systems are easily thrown by
even a small reflection, and as posted elsewhere, I think
the poor old XT can only offer 'partial' centre-metering..?

the XT has 3 mode, evaluative, partial, and center weighted
average, but not spot exactly. However, careful testing has
shown me that changing AE sampling does /not/ help with flash,
but can help in available light.

An even better, and frequently suggested solution (from an
old school type like me!) is to simply shoot using manual
flash/camera settings, given the problematic subject
matter. I can still do guide number and distance
calculations.. and it's not like the subject is moving
fast..

Yes, I tried that Saturday at the Henry Ford Museum, however,
I was incorrectly using the flashes GN to calculate the
exposure, so my test results were unreliable.

But these solutions have met with similar disdain and
derision. If Auto isn't involved, it's not acceptable,
apparently.

No, these solutions have /not/ met with disdain and derision
/except/ when stated dogmatically, without specific reference
to the Rebel XT and car pictures.

See also the following posts, all by ATM on this topic:

Nikon Coolpix 8800 when used with Nikon SB-800 external
flash rec.photo.digital

Want to buy a new digital camera to replace my Nikon 5700,
big ... rec.photo.digital

Nikon Coolpix 5700 severe flash underexposure problem
alt.comp.periphs.dcameras, rec.photo.digital

Looking for advice/opinion replacing Nikon 5700 with Nikon
... rec.photo.digital

Postscript to my Nikon 8800 and SB-800 query
rec.photo.digital

..and so on. I don't think I've ever seen a topic so
flogged to death, and with such simple causes and
solutions.

Enjoy yourself, Chrlz. This is America, and you can be an ass
if you want to. It must really be nice to have been born with
high intelligence, outstanding education and training in
photography, experience with the cameras I have tried in the
car museum environments I have described, and have learned
enough through experience to have instantly developed
outstanding judgment. For the rest of the human race, though,
we must learn through the school of hard knocks.

I am smarter today than when I complained about my 5700
initially and the 8800 during a very abbreviated test drive,
and I am smarter this morning with the Rebel XT than I was
several days ago. I think "by the numbers" and do no have the
luxury of instataneous brights like you do.

Now, if you want to climb down from /your/ high horse, maybe
we can talk this through...

--
ATM, aka Jerry

"I disagree with what you say, but I will defend to the death
your right to say it." - Evelyn Beatrice Hall under the
pseudonym Stephen G. Tallentyre

Today David Dyer-Bennet commented courteously on the subject
at hand

1) /Both/ camera's, my "defective" one and his "working"
one showed the /same/ amount of noise at /all/ ISO from
100 to 1600 on /all/ 3 of my PCs. Naturally, no noise is
readily apparent until ISO 400 on both cameras.

The problem, then is that your expectations are not being
met.

What *are* you expecting at ISO 1600? Usable photos? Or
crisp, clean, studio-looking photos? Are you expecting
less noise than P&S cameras and a lot less noise than film
that speed has grain? Or are you expecting a lot better
than that? Because "a lot better than that" isn't
available yet.

I "expected" to see little or no noise at ISO 400, some noise
at 800, and noise, but controllable noise on a properly
exposed image at ISO 1600. Where'd I get that? On
dpreview.com, from my fav camera store manager who's much more
knowedable about why a DSLR is a lower noise solution than
even an expensive EVF, from rec.photo.digital, and in this NG.
And, talking to friends with DSLRs and looking at the relative
difference in high ISO results between people posting car
pictures to alt.binaries.pictures.autos with pro-sumer EVFs
and DSLRs like the former Canon Rebel and a Nikon D70, both
with kit lenses and a big external flash.

I learned all this primarily from lurking and listening to
others asking the questions. So, I was surprised and dismayed
to see as much noise as I have. And, as my earlier post
stated, just a few days ago, my camera store manager
pronounced my camera defective, yet his supposedly working
version produced /identical/ results as mine, right there in
his store. Wasn't hard to do that - just shot several
represeentative series with both cameras on Programmed Auto
and all other camera parameters factory reset, and looking at
the 5.5 and 8 MP images at home. Clearly (pun intended), one
cannot see noise on an LCD.

OK?

Now, what did I expect at ISO 800 and 1600? I expected /some/
noise that I could fix reasonably quickly with PSP 9's EPS or
DCNR, once I had the exposure within +/- one f/stop. What I
saw, though, was a scene painted on multi-color beach sand,
and that is /not/ what I was led to believe would happen.

--
ATM, aka Jerry

"I disagree with what you say, but I will defend to the death
your right to say it." - Evelyn Beatrice Hall under the
pseudonym Stephen G. Tallentyre