Panasonic blunder

[John Sheehy]

Look here - both images in same lighting, same Av and Tv values (manual exposure); the only variable is the ISO setting:



How does a company engineer a camera that uses an amplifier for higher ISOs, and get worse results than under-exposed ISO 100 does?

In any event, my comparison in the thread, "Those tiny, noisy pixels" was with the FZ50's ISO 1600, so had I under-exposed ISO 100 instead, those tiny pixels would be doing even better!

[Ray]

That is revealing! Can I suggest that this inability of the camera to produce cleaner images at higher ISOs using the same exposure is due to the fact that the FZ50 employs a CCD sensor. I wonder if all CCD cameras show a similar problem at high ISO. Perhaps not a worse image at high ISOs but one that's no better than at base ISO with the same exposure. I take it these shots were RAW?

I recall doing a similar test with my Canon D60 a few years ago. Having missed a few shots on occasions whilst stuffing around changing ISO (a procedure made more difficult because I'm long-sighted), I wondered just how much improvement there was between an underexposed shot at ISO 100 and a 'correctly' exposed shot at higher ISOs. I was surprised to find there was very little difference. There was some improvement in the shadows, at high ISO, but nothing to shout about.

[John Sheehy]

That is revealing! Can I suggest that this inability of the camera to produce cleaner images at higher ISOs using the same exposure is due to the fact that the FZ50 employs a CCD sensor.

If it has something to do with the sensor, they still could have tested against the other option of software pushing.  I can't see them choosing the option that gave the noisier result, if they actually compared them, especially since it requires less electronics!

I wonder if all CCD cameras show a similar problem at high ISO. Perhaps not a worse image at high ISOs but one that's no better than at base ISO with the same exposure. I take it these shots were RAW?

Yes.  I showed ACR in the JPG, but I got similar results with my hand-conversions (ACR's a lot faster).

I recall doing a similar test with my Canon D60 a few years ago. Having missed a few shots on occasions whilst stuffing around changing ISO (a procedure made more difficult because I'm long-sighted),

I'm heading there in the fast lane.  I now focus clearly only somewhat at about 20 inches, and only in bright light.  As a youngster, I could read the copyright lines on eye charts.

I wondered just how much improvement there was between an underexposed shot at ISO 100 and a 'correctly' exposed shot at higher ISOs. I was surprised to find there was very little difference. There was some improvement in the shadows, at high ISO, but nothing to shout about.
[a href=\"index.php?act=findpost&pid=100710\"][{POST_SNAPBACK}][/a]

Well, that's where most of the difference is going to be with a DSLR, as shot noise is relatively low, and the same no matter how you perform your exposure index.  Given a certain camera and lens, and the same subject in the same lighting, shot noise has nothing to do with ISO; it is determined by the Tv and Av values.

With the FZ50, shot noise is stronger at the pixel level, so noise is visible at all levels, and the high read noise of ISO 1600 adds further to it (it seems to be about 30x as high at ISO 1600 as it is at ISO 100 on the FZ50).

[Ray]

Well, that's where most of the difference is going to be with a DSLR, as shot noise is relatively low, and the same no matter how you perform your exposure index.  Given a certain camera and lens, and the same subject in the same lighting, shot noise has nothing to do with ISO; it is determined by the Tv and Av values.
[a href=\"index.php?act=findpost&pid=100728\"][{POST_SNAPBACK}][/a]

Perhaps I should have said, noise at higher ISO's with the D60 was only less apparent in the deepest shadows (same exposure). As I recall, when I did similar tests with my 20D, noise was better across the whole tonal range at ISO 1600 compared with ISO 100, although the improvement was clearly more significant in the deep shadows and lower mid-range.

[John Sheehy]

How does a company engineer a camera that uses an amplifier for higher ISOs, and get worse results than under-exposed ISO 100 does?[a href=\"index.php?act=findpost&pid=100693\"][{POST_SNAPBACK}][/a]

Well, from the responses I've received on DPReview, it seems that Olympus and Ricoh "prosumers" with RAW also have this same problem.  Some cameras are claimed to be better, under-exposed at ISO 100, even in JPEG mode.

[Graham Welland]

Well, from the responses I've received on DPReview, it seems that Olympus and Ricoh "prosumers" with RAW also have this same problem.  Some cameras are claimed to be better, under-exposed at ISO 100, even in JPEG mode.
[a href=\"index.php?act=findpost&pid=100918\"][{POST_SNAPBACK}][/a]

It was established practice with the Panasonic LC1/Leica Digilux 2 to do exactly this. Better to under-expose at ISO 100 vs shoot at ISO 400+. Not sure of the physics behind it all bt the noise performance was definitely better.

[BJL]

If the example is typical, and not an aberration or unusual situation, it is a bit of mystery why the engineers did not try various combinations of voltage amplification in the analog phase and digital "push", and choose the best. Many DSLR's seem to do this, achieving extreme ISO speed settings like 1600 or 3200 with a combination of analog amplification to ISO 400, 800 or 1600 plus one or two stops of digital push.

On the other hand, I can see a possible explanation for why too much voltage amplification (or is it current amplification, before current to voltage conversion?) gives worse results, which is that at very high amplification levels, amplifier performance gets worse and introduces additional noise. (This relates to a previous discussion, where my hunch is that there is nothing to gain and something to lose in analogue phase amplification so great that the floor noise level is amplified to  a voltage greater than the voltage difference between successive ADU output levels.)

[Ray]

If the example is typical, and not an aberration or unusual situation, it is a bit of mystery why the engineers did not try various combinations of voltage amplification in the analog phase and digital "push", and choose the best. Many DSLR's seem to do this, achieving extreme ISO speed settings like 1600 or 3200 with a combination of analog amplification to ISO 400, 800 or 1600 plus one or two stops of digital push.

On the other hand, I can see a possible explanation for why too much voltage amplification (or is it current amplification, before current to voltage conversion?) gives worse results, which is that at very high amplification levels, amplifier performance gets worse and introduces additional noise. (This relates to a previous discussion, where my hunch is that there is nothing to gain and something to lose in analogue phase amplification so great that the floor noise level is amplified to  a voltage greater than the voltage difference between successive ADU output levels.)
[a href=\"index.php?act=findpost&pid=100923\"][{POST_SNAPBACK}][/a]

Once again, this seems like a marketing decision that has predominated over an engineering recommendation. To design a camera that produces a result at a higher ISO that is no better than the same exposure used at a lower ISO, is one thing. We see this in Canon DSLRs like the 20D and 5D where ISO 3200 is no better than ISO 1600 with the same exposure (or perhaps hardly better. I haven't tested this thoroughly). But it's quite another thing to give the customer a worse result with same exposure at higher ISO.

I can imagine the conversations between marketing and the engineering department. Engineer: "There's no point in amplifying the signal to this extent to give the camera ISO 1600. We can't get decent image quality without redesigning the camera and making it heavier and more expensive". Marketing: "Never mind. Do as you're told. Our competitors are offering ISO 1600. Consumers expect ISO 1600 so we'll give it to them. Most of them probably won't notice and won't care if it serves no real purpose. They expect over-all image quality to be worse at high ISOs".