10 mp better than 6? Don't be too sure!

[jlw]

I had an interesting shooting experience last week that I thought I'd share with the group.

I was able to shoot two successive nights of dress rehearsals for the same production (Creighton Dance Company's Coppélia) and took the opportunity to try two different cameras: my R-D 1 the first night, and my newly-acquired Nikon D80 DSLR the second night.

The Nikon is a 10-megapixel camera, whereas the R-D 1 is only a 6-megapixel camera... and of course more megapixels is always better, right? After all, that's what the salesman tells you! Well...

Here's a picture from the first night, with the R-D 1, cropped the way I wanted it and scaled down for convenient RFF viewing:

Now here's roughly the same moment from the second night, with the D80:

I think even at these reduced sizes you'll be able to notice that the D80 shot looks substantially more "granular." Just to make sure, here are some 100% crops; first, the R-D 1 shot:

And now the D80 shot; since it has a higher pixel count, 100% crop covers a smaller area of the scene:

Again, notice the greater "granularity" of the D80 shot. Both were made at EI 1600, by the way. There were some differences in exposure, lighting, etc., but I think the overall impression is clear: The R-D 1 image just looks smoother, despite its lower pixel count.

To level the playing field, I tried reducing the D80 shot to a 6-megapixel image, to see if downsampling it smoothed it out. Here's the result:

You can draw your own conclusions, but I still don't think it looks quite as smooth as the R-D 1 shot!

Now, I'm not saying the D80 isn't a good DSLR -- I'm certainly not going to send it back just on the result of this test! It's a good value in terms of features, and I expect I'll see more benefit from its higher pixel count when I'm able to try it in better light at lower EI settings.

For now, though, this little experience just reminded me: All other things being equal, more pixels should be better... but all other things seldom ARE equal! Or to put it another way: In the real world, pixels ain't everything!

 

[Nikon Bob]

Would that be an example of stuffing more pixels into the same sized sensor creating more noise?

Bob

 

[Keith]

It would be interesting to do the same exercise with a D70 as I think I read somewhere that they have the same sensor as the Epson. Then again maybe not ... I have a D70 and I avoid shooting at anything over 400 due to noise issues!

Apparently the Epson is regarded highly for it's lack of noise at higher ISO settings!

 

[Nachkebia]

How about downsizing 10 megapixel file to 6 megafixel file?

 

[zuikomatt]

I had an interesting shooting experience last week that I thought I'd share with the group.

I was able to shoot two successive nights of dress rehearsals for the same production (Creighton Dance Company's Coppélia) and took the opportunity to try two different cameras: my R-D 1 the first night, and my newly-acquired Nikon D80 DSLR the second night.

Which lenses did you used?

 

[Gray Fox]

10 MP better than 6?

10 MP better than 6?

Thom Hogan in his just released review of the D80 described basically the same thing, commenting that it was the result of cramming 10 MP onto the same sized chip. The D50 is known for its better than D70 high ISO jpeg performance, and so I bought one for night football shooting with a long 2.8 lens. It worked well for me, anyway. I have a D70, but damaged it in a fall and needed a replacement, but it did not have the good high ISO performance, as noted above. I've only recently started shooting RAW with the D50, and have found that in the CS2 RAW converter that I can up res the image to 11.2 mp with later sharpening on output with good results. I've also found that the noise filter in CS2, while not up to Noise Ninja levels, does well enough for me, even at high ISO. Bottom line is I'm now questioning my D80 lust I guess that if I had any Leica glass (only Contax G2) I'd be one of the many looking for an RD-1s. Sure wish there was one I could use with the G2 lenses!

 

[jlw]

How about downsizing 10 megapixel file to 6 megafixel file?

Did that -- that's the last example photo in the group. It does look smoother, but still not as smooth.

 

[Nikon Bob]

Thom Hogan in his just released review of the D80 described basically the same thing, commenting that it was the result of cramming 10 MP onto the same sized chip. The D50 is known for its better than D70 high ISO jpeg performance, and so I bought one for night football shooting with a long 2.8 lens. It worked well for me, anyway. I have a D70, but damaged it in a fall and needed a replacement, but it did not have the good high ISO performance, as noted above. I've only recently started shooting RAW with the D50, and have found that in the CS2 RAW converter that I can up res the image to 11.2 mp with later sharpening on output with good results. I've also found that the noise filter in CS2, while not up to Noise Ninja levels, does well enough for me, even at high ISO. Bottom line is I'm now questioning my D80 lust I guess that if I had any Leica glass (only Contax G2) I'd be one of the many looking for an RD-1s. Sure wish there was one I could use with the G2 lenses!

I have had a similar experience with an Olympus C5050 vs a Panasonic FZ50. I will take the Oly 5mp output any day over the Panas. I wish they had downsized the the Panas sensor not upped it. Having said that, the Pana is great to use ie control placement wise and at low iso. Sometimes more is not better.

Bob

 

[jlw]

Which lenses did you used?

The R-D 1 picture was made with a 50/1.5 Nokton; the D80 picture with an 85/1.8 AF Nikkor. I sat a bit farther back in the theater on "Nikon night," so the image areas were about the same, although the R-D 1 picture did need a bit more cropping to get the framing I wanted.

I should have noted in the original post that I picked these two images only for comparing the "granularity" of the image; you should NOT try to judge other image-quality attributes from them, such as sharpness.

For example, I did notice in the crops that the Nikon picture doesn't look as sharp and detailed as the R-D 1 picture -- but I have no way of knowing how much of that is from the greater accuracy of rangefinder focusing vs. AF, how much from the difference in quality of the lenses (although my impression of the 85/1.8 is that it's very sharp overall), how much from mirror vibration in the DSLR (can be a bigger factor than you'd think) and how much of it is just poor photographic technique and/or bad luck on my part!

The only conclusion I feel safe in drawing from this little pseudo-experiment is that with the two specific cameras I used and under these specific conditions (dim tungsten lights, EI 1600) the 6-megapixel camera delivered results which appear subjectively better than the 10-megapixel camera.

That's a result that's both counterintuitive and contrary to "conventional wisdom," and I think such cases are always worth reporting -- even if they're not broadly applicable, which I'm not sure this one is!

 

[jlw]

Would that be an example of stuffing more pixels into the same sized sensor creating more noise?

It certainly could be, although there could be a lot of other variables involved, such as:

• 
  Manufacturer's choice of default settings for in-camera noise reduction

• Manufacturer's choice of default settings for in-camera sharpening

• Photograpic technique (I noticed in the EXIF data that I shot the R-D 1 picture at 1/125 and the Nikon picture at 1/250; the reduced exposure could have made the Nikon's shadow noise worse.)

All I'm really prepared to conclude from this so far is that of the two digital cameras I own, the R-D 1 continues to be a better choice for low-light shooting, even though the Nikon is newer and has more pixels. Not quite what I expected, but good to know.

 

[Nachkebia]

depends on algorythm you used for downsizing.... anyhow I need to see raw file to agree, maybe you have underexposed, god knows

 

[akptc]

Thank you for posting this, very interesting, practical test.

 

[Nikon Bob]

It certainly could be, although there could be a lot of other variables involved, such as:

• 
  Manufacturer's choice of default settings for in-camera noise reduction

• Manufacturer's choice of default settings for in-camera sharpening

• Photograpic technique (I noticed in the EXIF data that I shot the R-D 1 picture at 1/125 and the Nikon picture at 1/250; the reduced exposure could have made the Nikon's shadow noise worse.)

All I'm really prepared to conclude from this so far is that of the two digital cameras I own, the R-D 1 continues to be a better choice for low-light shooting, even though the Nikon is newer and has more pixels. Not quite what I expected, but good to know.

That is so true. I think manufacturers use aggressive nr and sharpening settings to combat/mask the noise issues associated with adding pixels while keeping the sensor size the same. That does nothing for image quality. When you add possible underexpose to it then you really compound the situation. I hope manufactures rethink adding mp to small sensors at this point in time. Seems that 5/6 mp is a sweet spot and the technology is not there yet to go higher with the same file quality. All the present bandaid solutions are stop gap at best causing image/file degradation.

Bob

 

[jlw]

depends on algorythm you used for downsizing.... anyhow I need to see raw file to agree, maybe you have underexposed, god knows

I used the regular bicubic interpolation (not "bicubic sharper" or "bicubic smoother") in Photoshop CS2.

The point of my post was not to offer a proposition with which people would either agree or disagree --- only to point out that when working the way I normally work, this particular situation didn't turn out the way I had expected.

 

[marbrink]

It certainly could be, although there could be a lot of other variables involved, such as:

• 
  Manufacturer's choice of default settings for in-camera noise reduction

• Manufacturer's choice of default settings for in-camera sharpening

• Photograpic technique (I noticed in the EXIF data that I shot the R-D 1 picture at 1/125 and the Nikon picture at 1/250; the reduced exposure could have made the Nikon's shadow noise worse.)

All I'm really prepared to conclude from this so far is that of the two digital cameras I own, the R-D 1 continues to be a better choice for low-light shooting, even though the Nikon is newer and has more pixels. Not quite what I expected, but good to know.

Would be strange if the D80 didn't show more noise then. There's a full stop difference between them. From what I've heard the D80 has low noise and is better in that department compared to the D70 and D100 which means that it should also have less noise than the R-D1. A noise comparison between two cameras in low light with a full stop difference in exposure is totally worthless.

 

[kaiyen]

So wait - you actually did expose the D80 1 stop less than the Rd1? Or just that you use a faster shutter speed, and compensated with the aperture (equivalent exposure)?

If you shot at the same aperture, but with the higher shutter speed on the Nikon then indeed the test is worthless. "Granularity?" You mean noise, right?

Not meant to be rude. Awaiting clarification.

allan

 

[jlw]

So wait - you actually did expose the D80 1 stop less than the Rd1? Or just that you use a faster shutter speed, and compensated with the aperture (equivalent exposure)?

Feel free to download the sample photos and look at the EXIF data for yourself. All I know is that I had the R-D 1 set to 1/125 and the Nikon set to 1/250, per the EXIF info.

Before you go all Doctor Science about this, keep in mind that I shot the photos on two successive nights. During a dress rehearsal week it's not at all uncommon for the lighting designer to make changes from one rehearsal to the next, so we can't be sure that the light levels on the two evenings were identical.

All I really know is that I shot each picture at what appeared to be a "correct" exposure as indicated by each camera's metering system. Suppose I had set the Nikon to 1/125 and had gotten blown-out highlights as a result: what would that have proved? I used each camera the way I would normally use it in an effort to make a good picture.

If you shot at the same aperture, but with the higher shutter speed on the Nikon then indeed the test is worthless.

Which is why I very explicitly avoided using the word "test," instead referring to my observations as an "experience." All I can say is that I used both cameras the way I normally use them in this type of situation, then compared the results according to the criteria I normally use for deciding whether a picture is good or not.

"Granularity?" You mean noise, right?

I can't be sure, because I don't know what YOU mean by "noise." I chose the term "granular" to describe the appearance of the image, without speculating on possible technical causes for that appearance.

The purple speckles in the dark background drape on the Nikon photo certainly look like "noise pixels" to me, but even the brighter areas have an appearance that's different from the appearance of the same areas in the R-D 1 photo. They appear more textured, more, uh, granular.

I have no idea why this is the case. It could be exactly the appearance that Nikon intended in designing the camera's imaging system. All I can say is that I, subjectively, prefer the more smoothly-toned rendition I got from using the R-D 1 in a similar picture-taking situation.

Not meant to be rude. Awaiting clarification.

Just a lucky accident, then? Oh, well, to clarify: This set of pictures was NOT a scientifically controlled experiment, or even an organized test.

What I did was to try two different cameras under similar (although not identical) conditions, using them the way I normally would use them, and then give my subjective observations of the results. The observations ran counter to my expectations, and I thought that was interesting. That's it. Need I say, your mileage may vary?

 

[kaiyen]

So I wrote this giant reply, then thought better of it and deleted it. I think I took a lot of offense to that "lucky accident" comment. But I'll try to stay out of it.

allan

 

[jlw]

So I wrote this giant reply, then thought better of it and deleted it. I think I took a lot of offense to that "lucky accident" comment. But I'll try to stay out of it.

Probably what I should have done as well. I realize now that you really weren't intending to be rude -- it just struck me that way at the time, hence the "lucky accident" comment (which was also rude.)

It's been said a zillion times, but since online forums lack the cues of old-fashioned communication (facial expression, body language, tone of voice, etc.) it's easy to take offense when none was meant. I really ought to be more careful about that.

Anyway, sorry I yanked your chain. Your comments and questions were very appropriate, and I should have just accepted them as an opportunity to clarify that I was not presenting my observations as a formal test.

 

[marbrink]

jlw,
What you did is like comparing a ISO 1600 shot to a ISO 3200 shot..