Anyone got a rational explanation why Canon underexposes?

Cosworth, I agree with YOUR WIFE's assessment!

If you want to give a try to relative brightness, do this....

Take a gray card and mount a telephoto lens on your tripod-mounted camera. Zoom in so that the gray card occupies 2/3 or more of the frame.
a. take a picture on Av mode with no flash at all
b. take a picture on Av mode with flash on, no FEC
c. take a picture on Av mode with FEC + 2/3 EV
d. take a picture with a digital P&S

See which one you like, as reproducing the gray card density the best. See which one your wife likes, as reproducing the gray card density the best. Compare each of those with what the dSLR does in Av mode with ambient light and no flash! Compare each of those with what the P&S does in Av mode with ambient light and no flash! Compare all of them with the gray card!

cosworth wrote in post #2389974
Here is a random sample.

Av mode, zero FEC, zero EC. Flash on ETTL.

ISO 250.

In my books this is slightly underexposed.

HTTP response: 404 | MIME changed to 'text/html' | Byte size: ZERO

Impressive subject. I think that the overall image is exposed well, but you would probably choose to increase exposure by about 1/3 stop to better expose the primary subject.

Just my opinion.

cosworth wrote in post #2390853
I think the first image is fine, but this is what has been moving me towards thinking "my shots are dark". I posted long ago about how digital P&S cameras had changed what is considered a normal shot. Someone even posted that Ansell Adams had to lighten his pictures to sell them in that thread. I had begun to exposure creep and I didn't even know it. Thanks everyone.

The pool table shot presents a situation where the overall exposure is balanced, but the shooter appears dark in the original image. In this case, I agree with your wife ( probably a wise move regardless ). The second shot exposes the shooter better, with the background being a bit too bright.

I will confess that I'm viewing these on my laptop, which is notorious for having brightness/contrast problems if I view the screen a bit off axis.

cosworth wrote in post #2390853
Exposure one right from the camera:

HTTP response: 404 | MIME changed to 'text/html' | Byte size: ZERO

What my wife thinks looks better:

HTTP response: 404 | MIME changed to 'text/html' | Byte size: ZERO

I think the first image is fine, but this is what has been moving me towards thinking "my shots are dark". I posted long ago about how digital P&S cameras had changed what is considered a normal shot. Someone even posted that Ansell Adams had to lighten his pictures to sell them in that thread. I had begun to exposure creep and I didn't even know it. Thanks everyone.

Again if you look at the original shot, the man's skin is about 53-57% grey and the red of the table 45-63%. This is what you might expect the meter to provide, taste is normally for white skin to be a bit lighter than that but the camera does not know what you are shooting so you need to compensate.

I mean no offence but to everybody thinking "about the good old days" (a trap I fall into sometimes) - feel free to go back to the good old days. What? You can't? Then adapt.. because you have no other option.

We can no longer meter at the film plane in TTL mode with the switch to digital, so camera manufacturers had to come up with a new system QUICK. Read the adventures of the early digital pioneers, TTL film plane metering didn't work - so they went back to thyristor autoflash. Yes, you can do that if you want to. Yes, you can drive a Canon speedlite in manual if you want to. Yes you can shoot film and go back to TTL metering. Oh wait... don't want to go back to film? Tough..

Otherwise most of the rest of us set +2/3 EC and live with it. Why are computers so quirky? Too many years in the business and I still can't figure it out.. I just live and let live (or occasionally throw 'em out and say good riddance ) When set at +2/3rd, I have only one occasion where results did not turn out satisfactory and that I feel was due to operator error - I set 1/200 M mode flash when I should have dragged the shutter to light the background better.

I would try Dellboy's trick of setting E-TTL2 to averaging and see how it works out.

Moppie wrote in post #2390763
Maybe thats the problem, we need "British" and "Non-British" metering modes?

TILK says: "Indeed!"

Lightstream wrote in post #2393314
We can no longer meter at the film plane in TTL mode with the switch to digital, so camera manufacturers had to come up with a new system QUICK.

Twenty-five years ago, Olympus was able to meter at the time of exposure and issue instruction to the flash to quench its output. They did it with a photosensor reading the shutter and/or actual film surface. Fifteen years ago, Bronica could read the film surface with a photosensor to quench the flash output at the time of exposure. OK, so now the sensor is too shiney to read its surface with a photosensor. Are the dweebs at Nikon and Canon so totally clueless that they cannot ask the SENSOR itself to let the camera know it should tell the flash to quite outputting light?!?!?! And regardless of the methodology of reading the exposure, do they have to program the camera to underexposure when flash is on and you don't have the insight to set FEC to some value greater than Zero?!?!?!

By the way: Fuji managed to meter flash TTL *off the sensor* in the S1/S2 IIRC
Worked with an SB24.

Every compact digital I know of today meters and focus's useing the sensor, although it is all done before and exposure is taken, it can be done. It's only another step required to be able to pull data out while the exposure is taking place.

I wonder if too many SLR designers/engineers are still thinking in film terms and not talking enough to the compact designers/engineers who have IMO made better use of some of the advantages of Digital?

Wilt wrote in post #2394262
Twenty-five years ago, Olympus was able to meter at the time of exposure and issue instruction to the flash to quench its output. They did it with a photosensor reading the shutter and/or actual film surface. Fifteen years ago, Bronica could read the film surface with a photosensor to quench the flash output at the time of exposure. OK, so now the sensor is too shiney to read its surface with a photosensor. Are the dweebs at Nikon and Canon so totally clueless that they cannot ask the SENSOR itself to let the camera know it should tell the flash to quite outputting light?!?!?! And regardless of the methodology of reading the exposure, do they have to program the camera to underexposure when flash is on and you don't have the insight to set FEC to some value greater than Zero?!?!?!

This can not be done with a sensor as the surface has specular reflections unlike that of film.

René Damkot wrote in post #2394598
By the way: Fuji managed to meter flash TTL *off the sensor* in the S1/S2 IIRC
Worked with an SB24.

Reflected off the sensor or live with the sensor?

Wilt wrote in post #2394262
Twenty-five years ago, Olympus was able to meter at the time of exposure and issue instruction to the flash to quench its output. They did it with a photosensor reading the shutter and/or actual film surface. Fifteen years ago, Bronica could read the film surface with a photosensor to quench the flash output at the time of exposure. OK, so now the sensor is too shiney to read its surface with a photosensor. Are the dweebs at Nikon and Canon so totally clueless that they cannot ask the SENSOR itself to let the camera know it should tell the flash to quite outputting light?!?!?! And regardless of the methodology of reading the exposure, do they have to program the camera to underexposure when flash is on and you don't have the insight to set FEC to some value greater than Zero?!?!?!

Moppie wrote in post #2394796
Every compact digital I know of today meters and focus's useing the sensor, although it is all done before and exposure is taken, it can be done. It's only another step required to be able to pull data out while the exposure is taking place.

I wonder if too many SLR designers/engineers are still thinking in film terms and not talking enough to the compact designers/engineers who have IMO made better use of some of the advantages of Digital?

Unfortunately, to do this on the CMOS sensors Canon uses won't work, since reading the sensor is essentially a destructive operation.

Lester Wareham wrote in post #2394830
This can not be done with a sensor as the surface has specular reflections unlike that of film.

Yes, my statement which you quoted acknowedges that fact!

Jon wrote in post #2394853
Unfortunately, to do this on the CMOS sensors Canon uses won't work, since reading the sensor is essentially a destructive operation.

Thanks for providing a rationale, Jon. Back to the drawing board...I can get rich if I come up with an idea and license it to Canon!

Lester Wareham wrote in post #2394837
Reflected off the sensor or live with the sensor?

Not 100% sure, but I'd say reflected off the sensor.

DPreview about the S2:

"Flash control: Controlled by five-segment TTL Multi Sensor
Automatic Balanced Fill-Flash with TTL Multi Sensor: 3D Multi-Sensor Balanced Fill-Flash compatible with built-in Speedlight, SB-28, 27, 26, 25 and D-type Nikkor lens; Multi-Sensor Balanced Fill-Flash with built-in Speedlight or Speedlight such as SB-29, 28, 27, 26, 25, 24, 23, 22s, 22, 20 and AF Nikkor other than D-type or Ai-P Nikkor lens (except for AF Nikkor for F3AF)
Standard TTL: With built-in Speedlight, SB-28, 27, 26, 25, 24, 23, 22s, 22, 20 and non-CPU Nikkor lens; or with built-in Speedlight, SB-29, 28, 27, 26, 25, 24, 23, 22s, 22, 20 and exposure mode set to manual or metering system to Spot" About the S1: "Note that Nikon's SB-28DX speedlight which was specially developed for digital SLR's will act as a normal SB-28 when connected to the S1 Pro".

Sounds like plain oldfashioned TTL to me.