Because of all the interest that 950 owners have in upgrading to the 990, I thought that I would run some side by side comparisons. I continue to update this, check back for updates! Special thanks to Dan Lauring for his input and discussions.   First posted May 2, 2000. This is revision  number 13, last updated September 16th 2000 to reflect changes in firmware 1.1 for the CP990.

Overview

I will admit that I was skeptical of the 990 at first. After using it for 4 months now, I have been  won over. I have not picked up my 950 now for almost 2 weeks, although I am keeping it as a backup/car camera. Even things that bothered me at first like the new "look" are starting to look good to me. I have revised some of the material below because I feel that I was too critical of the 990 in some spots just because it was "different" than what I was used to using.

Look and feel.

A very subjective topic, of course. My first impression was that the 990 felt cheaper than the 950.  The 950 feelt hard and metallic, solid, and retained the traditional Nikon Black and Red colors. The 990 feelt warm like a piece of plastic, and was less dense. The reason for this was twofold, the camera isn't actually cheaper at all. The reasons that it felt cheaper initially were that :

1) The camera is physically larger, but weighs a little less, so it feels less dense and solid.

2) When you grab the 950, your hand rests on cool metal, and you know it is solid metal. With the 990, there is a larger rubber handgrip, so only your thumb touches the metal. The case is actually no less solid though.

The doors and knobs on the 990 are all plastic or rubber which does not help the impression. The 990 focus and telephoto mechanisms are much noisier and more grindy; this isn't unique to my camera I understand. The screen on the 990 is smaller, but is much easier to see in sunlight. the 950 is smaller, which can be an advantage when you are carrying the camera, espeically if you want to try stuffing it in your pocket.

The only thing that I actually don't like about the 990 is that the new command wheel has a sharp lip on the underside that is protruding. My thumb catches on it constantly, and I would hate to break it off. Overall though the camera is actually quite solid.

Winner: None. I realy don't prefer the design of either camera over the other.

Limited buffer size on the 990 reduces functionality.

Early reports were that the 990 was to have a 32MB buffer. This is absolutely not the case, the buffer size appears to be closer to 32 Mega BITS, however I'm waiting for an official size. Whatever size the physical buffer is, it appears to be exactly the same size as the buffer in the CP950 by my own tests. As an example test, I set the 990 to continuous mode, and took as many shots as possible with BASIC  compression. It took 8 shots. The total size of all of these shots was 3240KB. Doing the same test on the CP950, I got 26 shots. The shots are smaller, but the total size was 4032KB. The 990 buffer is approximately the same size as the 950 buffer.

This small buffer has several consequences.

1) . Update! Limited BSS shooting is fixed in firmware 1.1, you can now take 10 shots in Fine mode using BSS instead of about 3. The camera does not buffer all 10 shots, it must be picking the "best" to save as you go.

2) It limits "shoot ahead while images are storing". For example on the 950, I can take a shot, and then after 1/2 to 1 second I can take another shot... in Fine mode... up to 6 times. With the 990, you can do this only 2 or 3 times (Up to 3 times with the right flash card).

3) This also limits the number of shots that you can take in continuous mode, and the size of the movies.

In all the limited 990 buffer is a step back, with the same size buffer, the 950 can get off more shots because of the lower resolution.

Winner, Tie.

Shooting speed.

The 990 is not faster than the 950 in most tests. It reaches focus in about the same time as the 950, the full cycle time is about the same at about 0.9 seconds. For example holding one camera in each hand, both set to Fine mode, I press the releases. The 990 finished just maybe 0.1 second before the 950. BUT!  If I now put a faster flash card in the 950 (Lexar 8x), it finishes first by maybe 0.1 second. I think that the full cycle time (from press to finished storing) is very close. The only thing is that you will be able to buffer fewer full size shots on the 990. Firmware 1.1 seems to make the camera just slightly faster, but they are still pretty close.

The 990 DOES have a new "press the button right away" mode that lets you take a second shot right away, that the 950 does not have. To access this mode, you have to set the camera to "Preview Only"; it took me three days of owning the camera to figure this out. This will eliminate the part of the shooting cycle where you can pause and delete the picture you just took, but you will be able to take another shot right away. The buffer on the 990 still limits you to 2 or 3 shots in Fine mode though, so this isn't super-useful. Firmware 1.1 adds a new "shoot another shot as soon as the buffer empties enough" feature to all modes, which is very nice.

Winner: 990 with firmware 1.1 by a hair

Image quality comparison.

To me, this is the part that really counts. Because you will need to look at the original images to really compare, and because they are so big that I can't fit them on my ISP, I have put them up in a PhotoPoint album. The album is here! Take a look at it as you follow along. The first set of pictures could stand to be rotated, sorry, but I wanted to keep them strictly unmodified.

Outdoors in good light:

This is the first set of images in the album involving the mailbox. Great subject, eh? I picked this one because it included the Blue flowers and truck. Exposures and f/stops were similar, both cameras were set to ISO 100 for all shots (note that the 950 defaults to ISO 80, I changed it to 100 for the test shots)

Outdoors in good light, the 990 images are superior. The 990 has a little bit better resolution (but not dramatically more). Also the 990 has more true to life colors, especially blue and purple. For example the 990 captures the color of the purple flowers (purple, not blue) and the truck (blue-green, not blue) best. However, shots with test charts while helping to develop the 990 color plugin for Qimage show that the 990 actually has less accurate Blues, although it seemed to do better in this test shot. The 990 also seems to have the ability to reach into the shadows better and pull out detail; look at the near side of the mailbox, for example. Whether this is because the 990 exposed the scene differently, or whether it has greater dynamic range, I don't know. As I've used the camera more, I feel that it is using a different transfer curve than the 950 did that preserves shadow detail better. Because of this, at first the 990 shots can appear too "bright" or "washed out" if you are used to the 950, but after a while I've begun to appreciate them, they have more latitude. even if they do compress the available range more to fit more information in the shot.

The 990 did not seem excessively noisy in the outdoor shots either, and was in fact less noisy than the 950. The extra 990 resolution was visible even on 4x6 prints made in my HP970 printer. It is not a lot, but it is there.

Another outdoor shot:

This is the second set of images in the album. You can clearly see the enhanced 990 resolution; you can for example tell on the 990 shot that that is a robin out there on the lawn. Also, note the house wall to the left in both shots, on the 950 shot the shadow is excessively blue, while the 990 reaches into the shadow to bring out some detail with natural color. Possibly the 950 shot could be color balanced better, I used auto white balance on both cameras.

Indoors low light:

These are the curio shots. I took these to address the "noise in low light shots" issue, sorry if the subject matter is lame. I felt that 1/4 second was as slow as I would really care about in normal shooting.  I tried to match the exposures as closely as possible (1/4 sec), but in order to fix the shutter speed on the 950 I had to use ISO 80, which should be considered when looking at the shots, the lower ISO should give the 950 lower noise. I think that these shots show two things, first, that the 990 is not overly noisy at all at this light level, it is difficult to say which shot is noisier. Secondly, you can see the slightly increased resolution of the 990.

   Also, I've now taken a longer exposure set (the lava lamp set). This was 1 second at ISO 100 with incandescent white balance selected manually (the Red lamp was messing up the auto white balance). Personally, I think that the 990 shot is LOWER noise than the 950 shot. Take a look for yourself. Sorry the 950 got tilted on the tripod a few degrees due to the funny tripod socket on the 950. At least at 1 second, the 990 is looking pretty good, maybe better than the 950, and certainly no worse.

About image noise with the 990:

There have been some pretty noisy shots posted from the 990 and a lot of talk about noise. In all cases, I suspect that the noisy samples have been taken at either AUTO ISO, or higher ISO. The camera is noisy even in good light at ISO 400, and ISO 400 starts to look very noisy in low light. The problem appears to be that in the AUTO position, the camera will boost the ISO as high as 400. This is hard to prove because the camera only writes "AUTO" in the EXIF file. You can prove to yourself that the camera uses ISO 400 in AUTO by checking the exposure, for example in a situation where I had to manually use ISO 400 to get a 1/4 second exposure, the camera was able to also get 1/4 second in AUTO. The camera's logic for using AUTO ISO appears to be similar to the 950's logic, which is that it will allow the exposure time to fall to 1/4 second, then will boost the ISO as high as possible to try and hold it at 1/4 second, this can result in a very noisy shot when you aren't expecting it. The work around is to lock the ISO at 100 or 200 to avoid the noise. I plan to leave my 990 set at ISO 100 unless I really have to have it higher.

   I think that Nikon has made a mistake having the AUTO ISO extend up to 400, for example the Sonly S70 limits auto ISO to about 280. The camera is so noisy at 400 that it should really not be used except in brighter light to get a shorter exposure where the only alternative is to have a blurry shot, and the ISO 400 shots are making everyone think that the camera is always noisy in low light. Using ISO 400 at 1/4 second and longer is a mistake. Probably the AUTO mode is a reaction to all of the "my shots are blurry" postings with the 950, they decided to try to try to hold the shutter speed up as long as possible. Most people are probably using the AUTO mode, so when the take a long exposure test shot... ugh.

   The way to get very noisy shots with the 990 are the following settings. ISO 400 (or Auto ISO in low light), + contrast, High sharpening. The camera really picks up the extra noise, even on daylight shots if you have extra sharpening and extra contrast turned on, the camera is just sharpening the noise!  

Conclusion:

There still more to do with this comparison, but I think that the 990 has shown considerable strength. I was very skeptical of the 990 at first, especially about the low light issue, but after making my own side by side shots I feel a lot better about it. I'm not sending mine back.

Winner: CP990

Ease of use/functionality/features.

The 990 has a lot going for it that make it easier to use. A lot of thought went into the redesign. The controls have been rearranged to make them easier to get to. The rocker joypad switch thing is nice. The tripod socket is metal, and the surface is flat. The camera, while no faster than the 950, still operates at a pretty good speed, right up there with the fastest cameras available today, especially considering that the camera has to work with 50% more data. The new LCD displays enough information now that you don't have to turn on the monitor unless you want to. The LCD monitor is much more visible in sunlight. Except for the brightness factor, the LCD monitor is really a step down. It has about 25% less surface area, and where the 950 has 130,000 pixels, the 990 monitor only has 110,000 pixels; both combine to make things harder to see.

 One very nice new feature is the ability to dial up other exposure settings that are properly exposed in P mode.   This lets you bias your shots the way you want them; towards faster shutter speeds for example.

Another nice feature is the Bulb setting, which lets you take very long exposures.

Another thing that I like is the way that you can operate the menus, like for example for white balance, while still shooting pictures; the menus are overlaid on the image. It all adds up to faster easier operation.

Another nice thing, the flash intensity adjustment, and it works with external flashes too.

Winner: CP990

Add on lenses.

One thing that always bothered me about the TC-E2 on the 950 was that the manual focus settings were off with the lens on. Mine focused at infinity at about the 10 foot manual setting. The TC-E2 appears to suffer from a similar problem when on the 990. For example an object 15 feet away focuses best at the 3.8 foot setting. I guess that this is just too hard to get right.

Using the lens settings for the wide angle lenses still turns off the flash, so I will probably not use the settings as on the 950; the lenses work fine without them.

Winner: Tie!

CCD very low light noise comparison and long exposures.

There has been a lot made of this on the various discussion groups.  My initial comparisons show that there really isn't a problem for typical shots, but it is something to be aware of. In any case, it is best to keep the camera at ISO 100 whenever you can, and avoid the AUTO setting which might jack your ISO up to 400 without asking. The 990 is VERY noisy at high ISO and long exposures. I like astrophotography, and I'm not sure if the 990 will be useful or not for star pictures, it will work great for planets and the Moon. My first star pictures attempts were OK, but the 990 was not able to reach as far down to as faint a magnitue for stars at the 950 could. An few initial lens cap on shot look promising. 1 second at ISO 100 is perfectly clear. 8 seconds at ISO 100 has one hot pixel, and several warm pixels, but not as bad as the 950 would be. 8 seconds at ISO 400 looks like the explosion of a bag of sugar, much worse than anything obtainable with the 950. Only ISO 100 is going to be usable with longer exposures.

In direct head to head comparisons with the 950 at ISO 100 or a dark subject ( low light shots of the lava lamp in the PhotoPoint album), the 950 shot is the clear winner on longer exposures in spite of a few very hot pixels that are easily removed. The 990 shot is just plain noisier. In addition, the 950 still produces usable shots at ISO 320, the 990 would have to reach out to 24 seconds to equal that light gathering power at ISO 100. I don't think that an uncooled 990 will be able to beat a 950 at low light exposure that requires more than a few seconds, but consider if this is really important to you. At shorter exposures, the 990 does just as well as the 950, and in brighter light it is superior.. The 990 also does quite well at longer exposures when there is enough light, in other words, when there is enough light to expose the shot brightly. In that case, the excess noise is completely masked. It is only on a very dark exposure that you can see the 990 noise.

Where the 990 has the advantage is in being able to go beyond 8 seconds. In fact, it is possible to go out to 60 seconds and still get an image if you limit yourself to ISO 100. The camera shows quite a bit of noise at these longer exposures, and some have even reported cooling their cameras for lower CCD noise. I have made an exposure of the inside of my laundry room (very dark) for 55 seconds, then subtracted a 60 second dark frame from it with OK results, sure it is grainy, but it was taken in almost total darkness and reveals more that I could see by eye.

There is some kind of reciprocity failure/gain issue also. In other words, a 4 second shot at ISO 400 is not as bright as a 16 second shot at ISO 100, and a 4 second shot at ISO 100 is not half as bright as an 8 second shot at ISO 100. It seems like there might be some fall off or charge dissipation on longer exposures (similar to film where some electrons return to the non-excited state), so that you get less and less gain, and more noise on longer exposures. It also seems like the higher gain settings fall off faster and are less effective. Once I figure out how to quantify this, I will.

At first I was tempted to declare the 950 the low light winner, but now I'm having second thoughts. The ability to take longer exposures is an advantage for the 990, and the excess noise isn't visible on shots that end up being exposed properly, such as a lighted building, only dark shots like of the clear night sky.

The thing about the noise on the 990 is that it is constant for a given exposure duration and ISO no matter how much light ends up entering the lens. Whether or not it is objectionable depends on the how much light the exposure takes in to swamp out the noise. For example, I measured the "hot" pixels on an 8 second dark frame exposure, they are RGB 5,5,5 on the average, with the worst being maybe RGB 10,10,10 (if White). In LAB color the luminance part is about 2 to maybe 10 on the very worst hot pixels, most are 2-3. What determines if you will see the noise is how bright your shot will be in the final exposure.  If it is completely dark, then yeah, you will see the noise. For example, on the lava lamp sample shot the luminance of the wall near the
chair  on the left is 25 to 35, and the 10% variation of the luminance 2-3 typical hot pixel just can't be seen even in that dim area ( and the 2-3 value was derived from an 8 second exposure, this is 1 second). If I turned off the light, then you would be able to see the noise, so it all depends on what you are going to take a picture of.

Winner: Tie?

990 resolution limited at smaller apertures by diffraction and small CCD pixel size.

This possibility was raised on the newsgroups. I've run a test, and yes, it is true. The camera is a little less sharp at f/11. This is probably due to diffraction through the small opening, combined with the smaller pixels in the 990 CCD. Tests with the 950 do not show this effect.

Crop from an f/4 shot

Crop from f/11.1 shot

These have suffered a bit from being re-compressed, on the originals the difference was more apparent.  These were telephoto shots of a newspaper on the wall.. I used a tripod and the self timer to minimize shake, ISO 400 to keep the shutter speed up, and manual focus. I found no difference at wide angle, where the lens is softer anyway, so the problem is only really seen at the smallest apertures and telephoto. You might want to avoid them, probably f/4 is the sharpest. In tests, I could start to detect softening at maybe F9. Consider that even the softest 990 shot shows more detail than the 950 can due to the extra pixel resolution though.

Winner: tie

Focal length measurement.

The 990 is billed as having slightly longer focal lengths than the 950. However, something appeared to be wrong when I tried setting both cameras to 20.4 mm (max telephoto on the 950) and taking pictures of a resolution chart, ( I set the 990 by trial and error), the chart was obviously smaller in the 990 shot. Next I photographed a yardstick with both cameras at the same distance at full telephoto.

990

950

You can see the result above HEY! the fields of view are actually identical, measuring 12 3/8 inches at that particular randomly chosen distance. Obviously, the effective focal lengths are wrong with one of the cameras. I repeated this at wide angle also, not shown, with the same results. There is no significant effective focal length difference. I can only imagine that this was done on purpose, but then the published specs for one of the cameras is wrong, the 35mm effective focal lengths are the same. What must have been done is that the actual focal length of the 990 lens system was increased a bit to make up for the slightly larger CCD. This gave identical effective 35mm focal lengths. I measured the actual angular size of the 990 field of view at wide angle (measured diagonally in a 3:2 frame), and it did correspond to what would be expected for a 38mm lens.

Resolution test

I photographed an old Air Force resolution chart with both cameras, then calculated the number of lines of resolution manually. The 990 showed 1530 lines horizontally, and the 950 showed 1145 lines horizontally. Enlarged crops of the central portion of the chart are shown below. This is in line with the linear resolution increase expected, about 30 percent.

Winner: 990

950

990 (normal sharpening)

Focusing/Low Light focus:

The 990 and 950 have similar focusing speeds in side by side tests. The 990 is not faster, even with firmware 1.1, dispite the initial expectations.

I used the cameras side by side in dark conditions, pointing them at various objects and adjusting the lighting. The 990 was every bit as good as the 950 for low light focus, but not markedly better. I did find one situation where the 990 would focus and the 950 would not, so it may be just a touch better.

Macro focus speed was identical on both cameras, but sometimes one camera or the other would decide to do a long crank out of the lens.

Winner: Tie

Infra-Red:

The 990 is not as sensitive to Infra-Red light as the 950 is. See the comparable 1 second exposures below (950 was ISO 80, 990 was ISO 100). While this is probably a Very Good Thing for 99% of all users (reduced spurious IR response in the visible which is where you take most of your shots), it does mean that you will have a harder time taking an Infra Red shot if you want to. The key to getting it to work appears to be to have a filter that has a cut off very close to the visible. Cory Shubert has been able to take some nice shots with 1+ second exposures at ISO 100 using a type 89B IR filter. He reports that the common type 87 filter that a lot of people use on the CP950 does not work. Look at his page of shots.

Winner: If you want good blocking, plus some ability at IR, then the 990. If you want to take short IR exposures, the 950

CP990

     CP950

Summary:

Well, I'm keeping the 990. For me, the enhanced ease of use, plus the addition of new features that I wanted, plus the increased image resolution, plus the more accurate color and better white balance, put it over the top for me. I don't mean to say that the 950 isn't still a great camera, just that for me, an obvious enthusiast who has to have the latest, the 990 was worth the money. Sure, the 990 has a few areas in which it could be improved, but if you want a 3 Mpixel sensor, the 990 is a great choice.

Unfortunately the 990 is not obviously better in all areas, so if you already have a 950, keeping the camera becomes a judgment call.  I can't fault those that choose to remain with the 950 at all.

I'm looking at the 990 as a series of trade-offs that are (for me) overall positive.

The 990 is/has...

* Worse on long low light (over 1-4 seconds), sorry, but balanced by the Bulb feature.
* OK on low light that I probably care about like the 1 second lava lamp sample shot and certainly the 1/4 second curio sample shot. I can take shots up to 8 sec at ISO 100 that look OK.
* No faster but no slower focusing and taking shots, but less buffering of shots. Lower shot-to shot time is now possible because the 990 can take advantage of faster flash cards (see this page), and you can set the "Preview only" mode to allow you to take another picture right away, but you loose the ability to delete a picture right after you took it, and you can only shoot one shot ahead because of the limited buffer size. Firmware 1.1 improves the overall shot flow even more by letting you take another shot as soon as there is room in the buffer, even if the buffer is not empty yet.
* No better but no worse at low light focus, more focus options and assists, the manual spot focusing is nice..
* Much improved interface, faster operation, much faster in play mode.
* A real tripod socket that is flat ( I really like this)
* Bursting with extra nice features and user settings.
* Higher resolution is visible in some shots. May be slightly limited by diffraction in bright light, but not less resolution than the 950 in any case.

* I revised my opinion of the color issue August 2000. I had previously said that the 990 had better color than the 950, but that is not entirely correct. After working with Mike Chaney on the Qimage plugins for both of these cameras, the really surprising thing is that the color reproduction on the 990 is actually not as accurate as the 950, which is contrary to what I had thought. Apparently the 990 has been adjusted to be more "pleasing" to the eye, but is NOT more accurate than the 950. In fact the 950 produces very accurate color, and it needs little correction. The 990 on the other hand, produces a warmer image, and really has a problem with Blue being rendered as Purple. So it is a trade off between good looking but wrong color and correct but drab color. The 990 is helped by better white balance on most shots.

Complicating the issue is the white balance issue. The 950 sometimes produces an off white balance shot under a lot of conditions. In fact, I had a hard time getting the white right on the test chart for Qimage, and had to use manual white balance with the 950. Because of this, some 950 shots look too Blue, and that just complicates the color issue. The 990, on the other hand, seems to err on the side of too warm when it makes a white balance mistake, and that seems more acceptable to the eye than too cool.

* Better IR blocking in the 990. I consider this a plus because I don't take many IR photos.
* More visible LCD in bright light, but the 950 LCD was  bigger and had higher resolution.