HP945 - Hewlett Packard Photo Smart 945 Digital Camera - a review

by Bob Atkins; created 2003

A short while ago I wrote a brief review of the HP945 which you can find here: HP945 Review. That review was written based on the use of a a pre-production sample and HP requested that full size images not be posted since changes might be made before the camera was finally released. This extension of the earlier review is based on a production sample and represents what you would receive if you purchased the camera today. I won't repeat all the things in the previous article, so if you haven't read it yet, you might want to do so before reading this follow-up article.

Pre-production vs. production differences

The main operational difference I noticed between the pre-production model and the production model was that the electronic viewfinder now works much better! Earlier I had commented that the electronic viewfinder didn't turn on reliably when the camera was held up to my eye and didn't turn on at all in vertical mode. With the production model there was no such problem. The electronic viewfinder turned on as soon as the camera was brought up to eye level, whether in horizontal or vertical orientation. I didn't notice any other operational differences. There may have been some small software tweaks. For example, I believe that the production model takes advantage of the ISO 200 setting over the ISO 100 setting in auto ISO mode when the shutter speed becomes low enough that there is a chance of camera shake affecting image quality.

What's in the box?

• HP945 digital camera -5MP, 37-300mm zoom (35mm equivalent)
• 32MB SD card
• 4xAA Photo Lithium Batteries
• USB Cable, camera to PC
• USB Cable, camera to printer
• User Manual
• Neck Strap
• Lens cap cord
• Dock insert for use with (optional) 8881 camera dock
• CD with photo & imaging software

The inclusion of 4 AA Photo Lithium Batteries is a nice plus over the lower capacity AA alkalines some manufacturers supply. The Photo Lithiums seem to last forever, though I'm sure they really don't! They certainly significantly outlast AA alkalines, NiCads or NiMH cells.

The software is best described as basic. Not many frills here, especially the image editing software. You can crop and rotate images, adjust contrast, brightness, sharpness and color plus remove "red-eye", but that's about it. The color adjustment is a little weird. Here's a quote from the help file:

"Type a number between -180 and 180 [or use the slider] in which to increase or decrease the color. The color values correspond to the angles on the color wheel. Increasing the color by 1 changes each color by one color unit. For example, if you increase the color by +60 degrees, green becomes cyan, blue becomes magenta, and red becomes yellow"

This is fine if you want to experiment with strange color effects, but pretty useless for things like correcting overall color casts. As far as I can tell you can't correct a color cast using the supplied HP software. I'd say this is a problem if you have no other image editing software. Thankfully there are quite a few freeware and shareware programs out there that you can use. For example IrfanView is a very useful program

Digital Flash

In my earlier review I had commented that I didn't quite see what advantage the "digital flash" feature had over post exposure processing using an image editor. "Digital Flash" (also called "Adaptive Lighting Technology") isn't really flash at all. It's an image processing mode where the darker parts of an image are lightened - somewhat in the manner you might get from using conventional fill-flash. The difference is that (1) you don't actually use a flash and (2) it's effective at any distance.

The advantage of using the in-camera processing is that the full 12-bit image depth is used. Since the camera saves images as 8-bit (per color) JPEG files there's a clear advantage in doing any manipulations in the camera at the 12-bit stage.

Does it actually work? Yes, indeed it does! Here are two shots taken of Mt. Katahdin in Baxter State Park, Maine. They are both taken with the same exposure, one has "digital flash" enabled, the other doesn't.

1/433s at f4.2, ISO 100, Full auto mode, digital flash off

Above is the shot without "digital flash". The mountain is well exposed, but the trees in the foreground are rather dark. Below is the same shot, but this time with digital flash enabled. As you can see, the exposure on the mountain is the same , but the trees are now lighter and show much more detail. These images are exactly as they came out of the camera except for resizing. No color or density corrections have been made.

1/480s at f4.2, ISO 100, Full auto mode, digital flash on

Conclusion - digital flash works and can be an effective tool.

Battery Life

I started out with a fully charged set of Rayovac 1600mAh NiMH cells. These powered the camera for 10 days during which time I filled a 128MB SD card with 44 images. I shot more than that (maybe 55) but I deleted some of the images. I also reviewed the images on the camera LCD display. I only used flash for one shot. The in camera battery indicator showed that there was still some battery capacity left (~10%).

This isn't bad performance. Higher capacity NiMH cells are available (up to 2000mAh) so you can do better with better batteries. As mentioned earlier, AA photo lithium batteries (which HP supply with the camera) last significantly longer, but at around $12 for a set of 4 they can be expensive if you use the camera a lot. NiMH batteries are much more economical, but you'll probably want to have a spare set available if you're shooting in the field a lot.

HP literature gives the following information on energy consumed by various operations in terms of shots:
(1 shot = 1 no live view/zoom/strobe picture)

• Boot (zoom camera) = 0.5 shots
• Zoom (full) = 0.5 shots
• Strobe (full) = 0.5 shots
• Live Preview (5 seconds) = 0.5-1.0 shots
• Video (5 seconds) = 3 shots [think of a 60 second clip as 36 shots!]
• Instant Review (2 seconds) = 0.1 shots

Stuff I liked

I liked the digital flash which does a good job in bringing up the detail in darker areas of the image. The wide range zoom (37mm-300mm 35mm equivalent) was nice too. Generally exposure seemed accurate. Since there is no histogram display you can't tell exactly when you have the best exposure, but you can get a reasonable idea from the image playback. Exposure compensation is available in full auto, aperture priority and shutter priority modes (there is no fully manual mode). I used the camera in full "idiot mode" most of the time (full auto with default settings) and in general the camera did a pretty good job.

Stuff I didn't like quite as much

The view in the viewfinder freezes while AF is operating, so it makes tracking a moving object a little tricky. One you have AF lock the view becomes "live" again, but you can have a second or so of dead time during AF. Normally this isn't a problem, but I'm spoiled by my EOS 10D I guess!

I'd like to have seen some sort of "RAW" mode which saved the full 12-bit data produced by the sensor, but I guess that the typical purchaser of this camera probably wouldn't miss it. Again I'm somewhat spoiled by using a EOS 10D.

The supplied image editing software leaves something to be desired. It's OK for very basic adjustments and printing but you won't be happy with it if you want to any real image processing. Every camera will turn out some images that need some processing. I expect to process just about all my shots, but I know some consumers are used to film, where "what you see is what you get" (with a little help from the printer!). For example take a look at the following two images:

On the left is the image from the camera. It would have benefited from a little + exposure compensation and perhaps a manual white balance, but it was a grab shot on full auto and this is what I got. On the right is the processed image - lightened and shifted a little to green - and it looks pretty good. This processing was done in Picture Publisher 8.0 but it could have been done with almost any basic image editing program on the market - commercial, shareware or freeware. However it was impossible to get this degree of correction using the HP supplied software. Not all images need correction of course, but for those that do, you'll probably need to use something other than the software that HP supplies.

Stuff I didn't test

Since I don't own a compatible HP printer, I didn't test any of the "print from camera" functions which are available. I didn't test the HP Instant Share functions either which allows you to automatically send your images to preset destinations (email addresses or websites) whenever you connect the camera to your computer.

Image quality

Image quality is pretty good. I know that's very subjective but here's an example of the same scene shot at 300mm (35mm full frame equivalent) taken using the HP945 and an EOS 10D using a 75-300IS lens. Both were shot using a tripod at an aperture of f5.6 and a shutter speed around 1/60 at ISO settings of 100. The EOS 10D image was processed to be as close to the HP945 image as possible in terms of density and color. The 10D image was also sharpened since in camera sharpening is very mild in the 10D. The HP945 image is straight out of the camera with no corrections. It looks like the default settings use fairly aggressive sharpening.

The full frame image from the HP945 (no corrections). The red box sections are shown below at full size

As you can see, in the center of the image the quality of the HP945 shot stands up well when compared to the shot from the EOS 10D. This is a very good result.

Towards the edge of the image, and on high contrast features, the HP945 doesn't do quite as well. Color fringing is observed and the effects of more aggressive sharpening are evident. There may also be some DOF issues at play here since the smaller sensor of the HP945 results in significantly greater DOF. However remember we're comparing a 6.3MP SLR system (body + lens) costing $2000 with a much smaller and lighter 5MP digicam costing around $500. No, it's not as good as a DSLR, but it's still quite decent.

Some geometric distortion can be seen as shown in the images below:

As is clear from the images, at the shortest focal length (37mm equivalent for full frame 35mm) significant barrel distortion is present and is quite visible in images which contain straight lines near the edges of the image (architecture for example). However it's not always noticeable as can be seen in the two examples shown at the bottom of this article which were shot at this focal length. At 300mm equivalent focal length, distortion is lower and is pincushion in nature. Again, though visible in images with straight lines near the edges, this degree of distortion may not be noticeable in most "real" images as can be seen in the "garden" shot which appears above.

As with all digicams using small sensors and hence small pixels, noise starts to show up as the "ISO" speed is increased. The following 3 images show the noise levels at ISO 100, 200 and 400

Though it's not really a fair comparison, the 10D noise levels are shown below. These samples are 100% crops from the full image.

This pretty clearly shows what happens when pixels become very small - noise goes up. This isn't just a feature of the HP945, all cameras using very small pixels - which pretty much means all small non-SLR digicams - show similar noise levels. It's one of the tradeoffs you have to make. In actual use (when not imaging a grey card!), noise is less noticeable. Here are a couple of 100% crops from shots made at ISO 400 with the HP 945. As you can see, the noise is much less obvious than the grey card test.

Here are a few more sample images. All were shot in full auto mode except for the waterfall which was shot using aperture priority. All images were shot with "normal" settings for saturation, contrast and sharpness with centerweighted average metering and auto white balance.  These images link to medium size versions. One link is given to a full size image. Please limit downloading of the large image since it places a significant load on the server, but if you need to see full size image quality, feel free to download it.

1/352 @ f8.2, 61mm focal length Full size image (2.4MB)	1/362s @ f8, 37mm focal length	1/929s @ f4.0, 99mm focal length
1/1052s @ f4.2, 297mm focal length	1/193s @f12.3, 186mm focal length	0.769s @ f8, 37mm focal length

Conclusion

I liked the camera. It does a creditable job as a "do it all" pocket digital camera with a 37-300mm zoom and sufficient manual control to allow the creative photographer to override the fully auto functions. Exposure compensation, shutter and aperture priority, digital flash, manual white balance selection and average/center/spot metering give enough flexibility to enable you to get the image that you want if and when you don't like the results you get in fully auto mode. The HP945 would be a good choice for those looking for a high end consumer digital camera and especially for someone interested in shooting sports or nature where the 300mm effective focal length would be very useful. It also makes a good "backup" digital for more advanced shooters using a DSLR as their primary camera.

Readers' Comments

Add a comment

Gurpreet Singh Bhasin , October 23, 2003; 05:26 P.M.

Thanks Bob. This is the first time I have seen a noise comparison between a digicam and a DSLR. At least from your images, it is clear that the noise from the 10D at ISO 1600 is LESS than that from the 945 at ISO 100!!!!.

And digital flash on the 945 does look very promising.

Arthur Yeo , October 23, 2003; 06:47 P.M.

That Digital Flash thingy seems to make ND filters obsolete, doesn't it? Cool feature. Thanks, Uncle Bob, again for the tips.
BTW, would you be doing a review on the Kodak DX6490 any time soon? I'd be interested to see how that Schneider Variogon perform in that box.

Bob Atkins , October 23, 2003; 07:44 P.M.

I probably need to revisit the noise tests again. I'll shoot a Kodak Grey card under controlled conditions and look at the histograms. I don't think the 10D at 1600 is better than the HP945 at 100, even if that's the way it looks from the samples I posted.

As for future tests, none are currently scheduled since we haven't had offers of products to test. I hope to do something about that soon though!

Bob Atkins , October 23, 2003; 08:49 P.M.

ISO noise comparison - test 2

Here's a set of revised ISO test images. They were shot defocused using a Kodak grey card and flash. The were then adjusted to the same density in each case to make comparisions easier.

Looking closely at the originals I'd say the HP 945 at ISO 100 is about the same as the 10D at ISO 400. The HP at ISO 200 is somewhere between 800 and 1600 on the 10D and ISO 400 on the HP945 is probably similar to ISO 3200 on the 10D.

This isn't unreasonable. The HP945 pixels are aproximately 2.76 microns square, while the 10D pixles are 7.35 microns square. The pixel areas are 7.6 sq microns for the HP 945 and 54 sq microns for the 10D, which makes the 10D pixels 7x larger. If you assume noise is proportional to area (reasonable to a first approximation), 7x is almost 3 stops (8x), so you might naively assume that the 10D would have a 3 stop advantage. In practice it looks like about 2 stops at ISO 100 and almost 3 stops at ISO 200 and 400.

It should be noted that this isn't just a characteristic of the HP945. ALL cameras with small pixles (which means just about everything that isn't a DSLR) suffer from the same noise penalty.

In practice, shooting regular subjects, ISO 100 performance with the HP 945 is excellent, ISO 200 performance is pretty good and even ISO 400 performance is OK. When judging noise it's much better to look at actual images shot at various speeds (see the examples in the article) than defocused flash shots of grey cards which represent a "worst case" situation.

Gleb Baida , October 24, 2003; 01:16 P.M.

What I SEE from the two tests here is that noise of HP's point-and-shoot at ISO 100 is about equal to that of the Canon dSLR at ISO 800 or even between 800 and 1600. A reviewer should not argue too much with the obvious. Otherwise this site will turn unto just another photomag where everething is great until next generation comes. Where cameras with small sensors could compensate somewhat for their low effective ISOs is relatively high apertures especially at long end of zoom lens.

Bob Atkins , October 24, 2003; 03:10 P.M.

Cropped from full size image shot at ISO 400 with HP 945

Well, it's always possible that what you can see from the small sample images posted, and I what I see from looking at the full size, full resolution images and histogram functions of the variation in luminance etc. are somewhat different.

My judgement is that the noise levels of a 10D at ISO 400 and the HP 945 at ISO 100 are pretty similar and both are capable of yielding excellent quality images - see the 2.4MB full size file for an ISO 100 HP945 sample.

This isn't meant to be a scientific investigation of noise, just a rough guide for users. The bottom line is that noise isn't an issue at ISO 100, isn't much of an issue at ISO 200, but can become visible at ISO 400. There's a reason they stop at ISO 400 - as do just about all small sensor digicams - and that reason is noise may become unacceptable at higher ISO settings.

You can do noise filtering which will improve image quality (e.g. NeatImage), but it's very computationally intensive and on a 5MP image it would take too long to do "in-camera". Even on a fast PC it can take a minute or two.

I've included an example I just shot at ISO 400. This is a small, unprocessed, section cropped from the full size image. I don't think noise is an issue, even though if you look very carefully at an enlarged section of the uniform grey walls of the birdhouse you can see some noise.

So again I'd say the bottom line is that whatever ISO setting 400 on the HP 945 corresponds to on a 10D, 1Ds, D10o or D2H, most users wont' find too much of a problem with it, and for the purposes of this review, that's what counts I think.

Bob Atkins , October 24, 2003; 03:17 P.M.

Full frame version of ISO 400 test image

Just for perspective, here's the full frame shot.

Jerry Squires , October 24, 2003; 03:49 P.M.

Bob, The auto print auto email functions (whatever they are called) both work very well. The auto print gives print size choices and can print 1 image or all images on the memory card. Plug the camera into the printer and whabing! a beautiful picture. The auto email works in conjunction with the HP PhotoSmart download software. As you download images, the SW also emails through your email application.

Hope that helps.

Jerry Squires

Florian Vielhaber , October 27, 2003; 03:33 P.M.

Bob, I tested the HP 945 for two weeks and I liked the camera. For me it is wrong to compare HP 945 with a four times more expensive Canon D10. It is the same than you compare a Volkswagen New Bettle Cabrio with a Mercedes SL. I think the competitors are more an Olympus 750, a Canon G5 (just "normal" zoom) or a Nikon 5700. But these cameras are here in germany more expensive (the nikon even 300€) and the results are not better or the cameras are faster. The digital flash is a very nice option. May be I buy the HP after christmas for 400€ or so. That would be a really good deal.

Bob, do you have any ideas how the image quality would be if I use a wide angel converter (Olympus WCON 07) on the HP 945? Starting at 28mm would be very interessting for me. But I am not quite sure if the barrel distortion would increase too much.

Regards Florian

Derek Bumstead , October 31, 2003; 04:55 P.M.

The digital flash is a usefull tool. However much of the same functionality can also be obtained using a curves tool in post imaging. Given though that it is nice to be able to do such things in camera.

My image is not quite perfect, but I suspect that is due more to the lake of the original 12bit colour data and JPEG compression.

Image Attachment: Curves (digital flash).jpg

Bob Atkins , October 31, 2003; 11:30 P.M.

That's the whole point of "digital flash". It does the work on the original 12-bit data, which is otherwise unavailable to the user. You can get a similar result using curves on the 8-bit Jpegs, but you'll end up with more noise in the dark regions.