CCD worse than CMOS in ISO performance?

[passerby]

I don't have a camera with CMOS engine so I can't judge the CMOS
ISO performance. The nikon D40 I have here is CCD. I took these
5 shots barely minutes ago to ask you if they are worse than CMOS?
Sure it is cheapest DSLR therefore I expect no miracle.

You can see the ISO number right in each shot, and I cropped
the ISO 3200 to get as close as possible.
Is ISO 3200 are really bad here?

ISO 200

ISO 400

ISO 800

ISO 1600

ISO 3200

ISO 3200 cropped

 

[Garbz]

Firstly you photos are scientifically invalid. The last photo shows HighISO NR is turned on and will mask the real noise result from the sensor.

Secondly no there's no such comparison. In the right engineers hands both CMOS and CCD can be made wonderfully resilient to noise. I believe CMOS are cheaper to produce these days because of economies of scale.

 

[passerby]

Firstly you photos are scientifically invalid. The last photo shows HighISO NR is turned on and will mask the real noise result from the sensor.

Thanks man, you right here. The built in camera info states it as Hi-1, but when I look at Picasa it's blank. So is there a method to push it to it's limit? the 3200 is visibly grainier than 1600.

 

[RyanLilly]

Hey Garbz, not to thread jack, but what exif reader do you use? I most of the basic info but nothing about noise reduction.

As far as the photos, My 3200 on my canon 20D at first glance I would say is better, but anything up to 1600 in real 4x6 prints the noise is probably very close, you may see a difference in larger prints but its hard to say. Hard to say without a side by side comparison.

 

[Bifurcator]

I am under the very strong impression that CMOS sensors are inherently less noisy than CCDs. It may be due to the physical properties of the device or it may be due to the maturity of the technologies and engineers now more focused and active in developing CMOS process designs than the older, more expensive CCD types. It could also be the quality or layout of the surrounding system components as Garbz taught me a little over a month ago now.

The images in the original post are really just showing the efficiency of the in-camera noise reduction algorithms - which I believe can be beaten with any of several Photoshop plug-ins. If you want to compare noise between a specific CCD and a CMOS chip or just show the noise level of the camera you have then I guess you need to show 100% crops of the unprocessed RAW files.

 

[Bifurcator]

BTW, yes, those are pretty terrible. By 1600 the in camera NR had to be so strong that all the brick detail was erased in the process. At 3200 not only was the detail destroyed but the NR algorithm could no longer even do it's job.

I guess for great high-detail photos the top ISO that is still usable in the d40 is 400??

 

[Garbz]

Oh exif reader? Didn't use it this time I could see that the image had NR applied to it because of a lack of colour noise (something most prevalent and yet most easily removed from high-ISO shots)
I do use the firefox plugin FxIF though which shows the exif data in the standard properties dialogue when I rightclick on an image in the browser.

 

[notelliot]

I think CCDs generate more heat, which would result in more noise. but at lower ISOs (the CCD sweetspot) generates less noise. I could be wrong, but I know I'm at least close..

 

[table1349]

Hey Garbz, not to thread jack, but what exif reader do you use? I most of the basic info but nothing about noise reduction.

As far as the photos, My 3200 on my canon 20D at first glance I would say is better, but anything up to 1600 in real 4x6 prints the noise is probably very close, you may see a difference in larger prints but its hard to say. Hard to say without a side by side comparison.

FYI - Opanda IExiF viewer does show noise reduction was on. Free plugin from Opanda.com.

 

[RyanLilly]

Oh exif reader? Didn't use it this time I could see that the image had NR applied to it because of a lack of colour noise (something most prevalent and yet most easily removed from high-ISO shots)
I do use the firefox plugin FxIF though which shows the exif data in the standard properties dialogue when I rightclick on an image in the browser.

Oh, you you just used your brain rather than a computer program.
Yeah I already use that firefox plugin as well. Just have to get my eyes-to-brain plugin fixed I guess.

FYI - Opanda IExiF viewer does show noise reduction was on. Free plugin from Opanda.com.

Cool Good to know. Thanks.

 

[Bifurcator]

Here's what I see in Safari:

• File name: 1600-iso.jpg
• File size: 406773 bytes (800x532, 7.6bpp, 3x)
• EXIF Summary: 1/1000s f/16.0 ISO1600 55mm (35mm eq:82mm)

Camera-Specific Properties:

• Equipment Make: NIKON CORPORATION
• Camera Model: NIKON D40
• Camera Software: Ver.1.10
• Maximum Lens Aperture: f/5.7
• Sensing Method: One-Chip Color Area
• Color Filter Array Pattern: 836
• Focal Length (35mm Equiv): 82 mm

Image-Specific Properties:

• Image Orientation: Top, Left-Hand
• Horizontal Resolution: 72 dpi
• Vertical Resolution: 72 dpi
• Image Created: 2008:07:25 11:40:31
• Exposure Time: 1/1000 sec
• F-Number: f/16.0
• Exposure Program: Manual
• ISO Speed Rating: 1600
• Exposure Bias: -1.3 EV
• Metering Mode: Spot
• Light Source: Unknown
• Flash: No Flash
• Focal Length: 55.00 mm
• Color Space Information: sRGB
• Image Width: 800
• Image Height: 532
• Rendering: Normal
• Exposure Mode: Manual
• Scene Capture Type: Standard
• Gain Control: High Gain Up
• Contrast: Normal
• Sharpness: Hard
• Subject Distance Range: Unknown
• ISO Speed Used: 1600
• Color Mode: COLOR
• Image Quality: FINE
• White Balance: AUTO
• Image Sharpening: MED.H
• Focus Mode: AF-A
• Flash Setting: NORMAL
• Flash Compensation: 39.7 EV
• ISO Speed Requested: 1600
• Tone Compensation: NORMAL
• Lens Type: Nikon D Series
• Lens Range: 18.0 - 55.0 mm; f/3.5 - f/5.6
• Auto Focus: Closest Subject, Center Selected, Top Focused
• Shooting/Bracketing Mode: Single Frame/Off
• Color Mode: Landscape sRGB
• Lighting Type: NATURAL
• Noise Reduction: OFF
• Camera Actuations: 4978
• Image Optimization: VIVID
• Saturation 2: ENHANCED

For the ISO 3200 image however it says:

• Noise Reduction: FPNR
• Camera Actuations: 4979
• Image Optimization: VIVID
• Saturation 2: ENHANCED

Other Properties:

• Resolution Unit: i
• Chrominance Comp Positioning: Co-Sited
• Exif IFD Pointer: 216
• Compression Scheme: JPEG Compression (Thumbnail)
• Horizontal Resolution: 72 dpi
• Vertical Resolution: 72 dpi
• Resolution Unit: i
• Offset to JPEG SOI: 30620
• Bytes of JPEG Data: 8939
• Chrominance Comp Positioning: Co-Sited
• Exif Version: 2.21
• Image Generated: 2008:07:25 11:40:31
• Image Digitized: 2008:07:25 11:40:31
• Meaning of Each Comp: Unknown
• Image Compression Mode: 4
• Comment:
• DateTime Second Fraction: 40
• DateTimeOriginal Second Fraction: 40
• DateTimeDigitized Second Fraction: 40
• File Source: Other
• Scene Type: Unknown
• White Balance: Auto
• Digital Zoom Ratio: 1
• Saturation: High
• Nikon Note Version: 2.10
• Auto Flash Mode:
• Flash Used: No
• Digital Vari-Program:

 

[Garbz]

I think CCDs generate more heat, which would result in more noise. but at lower ISOs (the CCD sweetspot) generates less noise. I could be wrong, but I know I'm at least close..

Most definitely. However a sensor rarely has time to heat up during a 1/100th of a second. These things take time. But to take your analogy further yes regardless of how the sensor is implemented in a design if the shutter speed was slow enough to allow heat to play a role (think 10 minute plus exposures) a CMOS will win hands down.

Off topic: Funny this should come up because today my advisor and I was talking about a professor from QUT who's going to Antarctica for some experiment and was too stubborn to spend money on a decent camera and is taking a Fuji P&S. One of the photonics guys in the meeting turned and said well if his camera works at least he'll get noise free shots from it for a change

 

[TamiyaGuy]

I was wondering the same question, so thanks for the answer guys. I guess it's really difficult to actually see whether CCD is better or worse than CMOS in real-life circumstances, as there is no "identical" pair of a CCD and a CMOS sensor.

But IMO, the D40 actually has very little noise from ISOs 200-1,600, especially considering their price. I've printed a few ISO 1,600 photographs, and they actually looked very good.

 

[Village Idiot]

I was wondering the same question, so thanks for the answer guys. I guess it's really difficult to actually see whether CCD is better or worse than CMOS in real-life circumstances, as there is no "identical" pair of a CCD and a CMOS sensor.

But IMO, the D40 actually has very little noise from ISOs 200-1,600, especially considering their price. I've printed a few ISO 1,600 photographs, and they actually looked very good.

Nikon was generally considered pretty lousy at high ISO performance as they used CCD sensors in most of their cameras. They've been using CMOS now and are on par with Canon for the most part with high ISO performance.

Most CCD's (all?) are used with electronic shutters and have the added plus of being able to not have cutoff while using a flash over the manufacture's stated x sync. In fact, the OP could fire his D40 with a remotely triggered flash at 1/4000 a second and pretty much over power the sun at 12 noon on a sunny day as long as he was using a powerful enough flash.

 

[passerby]

If you want to compare noise between a specific CCD and a CMOS chip or just show the noise level of the camera you have then I guess you need to show 100% crops of the unprocessed RAW files.

It can't be done. The photobucket does not accept raw format btw, so it ends here.

I view the raw 3200 and they are at the equivalent slightly less than A4 paper on my monitor. The noise is there but only for pixel peepers. And it maintained in that form upto 30% magnification. By the time it reach 50% the grain start noticable, but still very good. At 75% magnification the grain start condensing to 100%.

With 1600 it is different, there is no visible noise increase at 50%. It become noticable at 100%. Yet it is quite acceptable when it passed the 300% magnification. The way I see it here the 300% plus percent magnification of 1600 ISO seems better than 100% magnification of 3200.

Anyway, this CCD and CMOS jargons to me is just way to distract what our eyes see things. The pictures are there to see regardless where they came from and how the camera processed them. We are getting too technical here for no good reason really.

Shutter Speed and Image Quality are two things that concern photographers all the time. Over the years the camera manufacturers succeed in constantly increasing the SS. But that achievement came with the price tag. So when the D40 managed to double the SS to achieve ISO 3200 with the minimal loss of quality and low price tag - than that is quite achievement. The JPEG fine in D40 is so good that firing at ISO 800 seem no different with ISO 200. ISO 1600 is only look noisy for the people who look at photos to find noise.