Do that to the unit you intend to buy. The demo might just be a hand picked sample that doesn't exhibit any noise patterns. ;p
E-410/e510
- Thread starter howeikwok
- Start date
- Status
Do that to the unit you intend to buy. The demo might just be a hand picked sample that doesn't exhibit any noise patterns. ;p
is there other brands that can pass this test?
If no brand passes (unlikely)... well, I'll just have stay put for another year or two longer with my E-330.
Haha. I'm waiting for others too.
I've tested it.
And the result?
Non-uniform horizontal bands.
But you're probably going to get that with all sensors.
Maaaybe you got a lemon? Hahaha.
Define 'proper exposure'.
What if most of the histogram is towards the low end (ie. low key or 'darkish' image), and it is difficult/impossible to increase exposure to shift them brighter without overexposing other important areas?
Ah... this is good to hear.
Quite different behaviour from the E-330s.
Dark areas will get banding without exception at high ISOs.
But then again, what's the difference between underexposing a shot, and having a lot of dark areas? Either way, you'd have a histogram that has a lot of data clumped towards the low end.
When you purposely underexpose by 2EV and then amplifying by 2EV in RAW processing, you're actually forcing the software to add in data that wasn't captured by the sensor in the first place.
So in effect you're bullying the sensor into producing patterned noise, but even in this, the bands are non-uniform in thickness or in colour.
What Blu experienced with his 330 was totally different. Somehow his 330's image processing engine was producing bands of uniform thickness and shade.
So in effect you're bullying the sensor into producing patterned noise, but even in this, the bands are non-uniform in thickness or in colour.
What Blu experienced with his 330 was totally different. Somehow his 330's image processing engine was producing bands of uniform thickness and shade.
What if I want to make dark images brighter for personal taste? Nothing about doing that 'push' trick to gain pseudo ISO? Maybe just simple contrast increase to bring out apparent details in the subject?
I think software processing only makes more apparent what was already there, unless you're now saying Olympus Master/Studio is the real culprit behind the image artifacts?
Blu's E-330 was just plain faulty, as was my first.
BTW, in many cases on my E-330 I underexpose just simply because matrix metering just doesn't create the brightness that I feel fits the mood or actual situation, and I can't find a suitable spot for metering or can't be bothered to do spot metering. In those low-key images, bands can already be readily seen from camera LCD and I needn't do any more post-processing later to make them more any more apparent.
Well, if the bands on other E-410s, 510s, and the P-1 stay the same way or are better than Drakon's E-410s, I'd still say >ISO800 has been greatly improved already by Olympus. Though it came one product late...
Well, if the bands on other E-410s, 510s, and the P-1 stay the same way or are better than Drakon's E-410s, I'd still say >ISO800 has been greatly improved already by Olympus. Though it came one product late...
Well, pulling up the shadows will reveal details already there. If there is no readable data, the image engine will interpolate the data by comparing neighbouring pixel data.
What will fool interpolation is an highly amplified, underexposed, textureless, uniform-coloured surface, most of the time.
When I did my test earlier I really bullied the sensor, ISO 1600 at -2EV shooting at a plain wall, then +2EV at RAW processing, i.e. I forced the sensor to produce patterned noise.
I don't think any sensor can stand up to that kind of punishment.
Now the good news is that the bands are non-uniform, i.e. it is not a sensor problem.
Some of the ISO 1600 examples already posted on the various E-410 threads do show some very slight banding, but it is nothing that will show up on prints.
What will fool interpolation is an highly amplified, underexposed, textureless, uniform-coloured surface, most of the time.
When I did my test earlier I really bullied the sensor, ISO 1600 at -2EV shooting at a plain wall, then +2EV at RAW processing, i.e. I forced the sensor to produce patterned noise.
I don't think any sensor can stand up to that kind of punishment.
Now the good news is that the bands are non-uniform, i.e. it is not a sensor problem.
Some of the ISO 1600 examples already posted on the various E-410 threads do show some very slight banding, but it is nothing that will show up on prints.
Yours and Blu's cases were really bad ones; it may have been (and probably was!) a bad batch of NMOS sensors.
Well, I think the good news is that overall as a SYSTEM the bands are of really minor severity now. But I think if I drum up more noise (pun intended) about noise patterns in forums, I guess by the time E-420, 520, P-1b is out it'll be totally gone?
I'm not really sure what's behind the banding (though I'd bet my money on good-old stray EMF pickup or marginal EMF shielding) but if it's in the RAW data already it's most likely not an interpolation problem. I can think of only two major interpolation operations in the camera's image output flow: at the pixel-to-pixel level, for the Bayer photosite layout, and between groups of pixels, for the sharpening level. I think only the latter type of interpolation that's capable of creating image artifacts that are in the order of several pixels wide... are RAW images sharpened at any level?
But I don't know if Olympus was/is kiasu enough to throw in an interpolation algorithm so aggressive that it creates visible harm to the image alongside visible improvement over previous generation algorithms -and not notice it when outputting thousands (millions?) of test images during product development? :dunno:
- Status
Similar threads
