In Part 3 and Part 3s of the shootout, I compared ability of the DP1 to capture detail to that of the Canon 5D with 28mm f/2.8 lens and Olympus E-420 with 14-42mm lens. Based on those comparisons, it seemed that the DP1 is capturing a good bit less detail than the 12MP 5D combination. However, would a 12MP small sensor compact camera also capture more detail than the DP1? That is the question I set out to answer here in Part 7. As in Part 6, the DP1 will be compared to the 12MP Canon Powershot G9 and the 10MP Leica D-LUX 3 (Leica's version of the Panasonic LX2).

I tested each camera in turn to see what f-stop would be optimal for sharpness across the frame. For both the Canon and Leica, the best results were obtained at f/4. Stopping down to f/4.5 of f/5 (f/4.9 with the Leica) resulted in very slightly better performance in the extreme corners with a slight drop in performance in the center of the frame. Stopping down beyond f/4 did not appreciably help further eliminate C.A. in the Canon or Leica images. The Sigma was a consistent performer across a variety of apertures, and I finally settled on f/5.6. The Canon was at ISO 80, the Leica and Sigma at ISO 100. Each camera was allowed to determine shutter speed and white balance automatically.

The tripod location was fixed for the Leica and Sigma images and then moved for the Canon image in order for that capture to have a similar field of view to the others, given a change from 28mm equivalent (Leica and Sigma) to 35mm equivalent (Canon) angle of view.

The Canon and Leica files were processed in Adobe Lightroom, latest version, with default settings. The Sigma file was processed in Sigma Photo Pro with +0.5 Exposure to give a similar apparent exposure to the other files. The only other change from default settings was a -0.4 Sharpness adjustment in order to decrease the apparent sharpening artifacts. I experimented with sharpening settings from -1.0 to 0 and determined that -0.4 did not adversely affect overall detail captured in any way. The DP1 file was processed at native size (4.6MP).

Leica and Sigma files were upsized to 109.35% and 160.71%, respectively, using Genuine Fractals 5. This resulted in an output size of 12MP, matching the native output of the Canon G9. Genuine Fractals maintains detail well during upscaling, sacrificing perhaps a touch of detail when compared against Photoshop Bicubic. I tried both methods for this comparison, and the only significant difference was that the Bicubic method resulted in more jaggies along diagonals in the DP1 image. The All three files were given the same final sharpening step in Photoshop CS3 (Smart Sharpen, amount 50, radius 0.3). This resulted in a similar degree of sharpening (all being a bit oversharpened) amongst the three images, as determined by comparing halos throughout each image.

The DP1 and D-LUX3 samples shown below represent 100% crops of images that have been upsized to match the native G9 output size (12MP).
The specific area of examination is depicted by the yellow box in the resized photo preceding each set of crops.





Some C.A. is present in the G9 crop, and a trace in the D-LUX crop as well. Shadow detail is smeared in the D-LUX3 crop. The G9 crop is holding the most detail.





Again the G9 is showing a bit of C.A. around the sign here, but the more impressive difference is just how much detail the G9 is capturing even out towards the edge of the frame.





The DP1 is the most consistent here in the extreme corner, but the G9 is still capturing more detail. The Leica has fallen well behind here, though it clearly faces the greatest challenge of the three given the 28mm equivalent diagonal angle of view with a 16:9 aspect ratio. To put this in perspective, Canon representives have stated in the past that Canon's advanced compacts currently do not go wider than 35mm equivalent because of technical difficulty maintaining performance beyond that point.





A bit more softness is apparent in the upper left extreme corner of the Canon image. Once again, the DP1 is consistent.





Here the Canon again has captured the most detail, but again it is affected by C.A. in high contrast areas, as is the Leica. Both the G9 and D-LUX3 are also showing color aliasing effects here, whereas the DP1 is showing a bit of luminance aliasing.





Moving towards the left edge, we can see again that the G9 and D-LUX3 suffer C.A., and the G9 captures the most detail. Additionally, this crop nicely demonstrates that the G9 shadows are the noisiest, the DP1 shadows the cleanest, and the D-LUX3 shadows the most smeared of the three.





More of the same here. G9 has the most detail, C.A., and shadow noise.




Now just left of center, the D-LUX3 and DP1 are capturing a similar amount of detail. The G9, meanwhile, goes a step further.

A DP1 review at DPReview recently concluded "Class leading detail in ISO 100 images" and "Resolution is as good as it gets in compact cameras." However, based on my tests, the DP1 simply cannot match the G9 (at its widest setting) for absolute detail captured at base ISO in good light. Which compact delivers the best overall image quality? Easy, in my opinion, DP1. Most detail resolved? G9.

As always, the RAW files are available for your interest/investigation:

Read More......

In this part of the shootout, I'll be presenting a rough but practical comparison of dynamic range based on highlight and shadow detail retention in the RAW files. The methodology will be similar to that in Part 2 of the shootout with a few exceptions mentioned below.

All cameras were set to f/4.5 and ISO 100, except for in the last comparison, which was shot at ISO 400. Since the G9 was at 35mm equivalent and the other two cameras at 28mm equivalent, the subject distance was increased for the G9 images in order to keep the field of view similar. This caused the composition to change slightly, though this did not affect the results in this comparison.

In order to truly push the limits in terms of both the highlights and shadows, highlights were tested at 1/125s and 1/250s, shadows at 1/500s and 1/1000s. Thus highlights and shadows were not tested from a single exposure as was done in Part 2. All images were taken within the same short time frame at noon on a cloudless day, so there was no significant variation in the available light.

G9 and D-LUX3 images were processed from RAW using Adobe Lightroom, latest version (same engine as ACR). All settings were at default except for specified EV adjustments. DP1 files were processed from RAW using SPP (latest version, Mac), all settings at default except for specified EV adjustments. G9 and D-LUX3 16-bit TIFFs were downsized to match the native size of the DP1 output (4.6MP) before cropping. This downsizing had a positive impact on the apparent shadow detail/noise ratio for those files.

The test scene is below. The approximate areas examined for shadow and highlight performance are depicted by the yellow boxes.



We'll start by looking at the highlight performance. The following crops were taken at 1/250s. Each crop is accompanied by a second crop to the right, showing what highlight detail could be recovered with a -2EV adjustment during RAW conversion.



As you can see, the G9 clipped badly in the grass but was able to recover a good amount of highlight detail at -2EV. The D-LUX3 and DP1 had a somewhat better showing here. Now lets see what happens when we get a stop brighter at 1/125s:



At this point, all three cameras blew out the grass. However, the DP1 and D-LUX recovered partially at -2EV, whereas the G9 fared less well. This is a good time to point out that the DP1 files are somewhat oversharpened at the default SPP settings. This oversharpening isn't helping with the DP1 highlight performance in the grass. Also, don't be fooled into thinking that the DP1 is capturing the most overall detail! The differences in sharpening make a big difference in apparent detail, an area of performance that will be addressed in a subsequent part of the shootout.

Next, we'll turn our attention to the shadows. The following crops were taken at 1/500s. Each crop is accompanied by a second crop to the right, showing what shadow detail could be recovered with a +2EV adjustment during RAW conversion.



The DP1 is clearly outclassing the competition here. The D-LUX 3 shadows are darkest and seem to contain the least information, but they are also significantly less noisy than the G9 shadows. This is consistent with earlier reports that Panasonic has implemented on-chip noise reduction in the shadows of LX2/D-LUX3 images. Next we'll see what happens when we get a stop darker at 1/1000s:



The G9 digs deep but comes up empty on shadow detail, and the D-LUX3 seems to have given up altogether. Meanwhile, the DP1 continues to find some reserve at this setting. As I mentioned in Part 3, the degree to which this "pushed" DP1 shadow detail lacks luminance noise suggests that SPP is doing some noise reduction here, though there is no setting for this. Some patchy chroma noise is evident in the pushed DP1 crop.

Finally, let's see how these cameras handled the same scene at ISO 400. First the highlights, before and after a -2EV adjustment in RAW:



At this ISO, the DP1 is the only one of the three able to largely recover highlight detail at -2EV. Interestingly, the DP1 highlight recovery in in RAW at -2EV was better at ISO 400 and 1/500s than at ISO 100 and 1/125s (same apparent exposure). This suggests, as I've mentioned before, that the high ISO DP1 files represent a somewhat "pushed" state. Again, the D-LUX3 has outperformed the G9 in highlight detail capture. Below is the shadow detail from the ISO 400 images:



Some desaturation is present in the ISO 400 DP1 crop, but this is easily addressed by increasing saturation in SPP.

As always, the RAW files are available for your interest/investigation:







Stay tuned for Part 7, coming soon.

Read More......

In the next few sections of the shootout, I'll be comparing the DP1 to two of the more popular advanced compacts, the Canon Powershot G9 and the Leica D-LUX3 (Leica's version of the Panasonic LX2). Here is a size comparison of those cameras, along with the smallest currently available DSLR, the Olympus E-420:





As you can see, from tallest to shortest E-420 > G9 > DP1 > D-LUX3.
From thickest body to thinnest body, again E-420 > G9 > DP1 > D-LUX3.
Including the lens, thickest to thinnest, E-420 > DP1 > D-LUX3 > G9.

In practice, based on which camera I can fit in which cases and pockets, the size order is E-420 largest, DP1 and G9 about the same, and D-LUX3 smallest. Note that that Ricoh GRD2 would have been easily the most compact amongst this company.

For the comparisons to come, I'll be comparing at equal output size based on total pixel count. That's a bit different than comparing at equal diagonal dimensions, which was the approach in Parts 1-4 of the shootout. The reason for the change is that the 16:9 native aspect ratio of the D-LUX3 is such that it has a significantly lower megapixel output than the G9 when both are output at the same diagonal angle of view.

The following images represent 24% views of images which have been downsized to match the native output pixel count of the DP1 (4.6MP). They give an idea of the field of view (FOV) coverage of each of the three cameras with the Canon and Leica at their widest zoom settings (35mm and 28mm equivalent, respectively). In addition, they give a preliminary sense of the colors one can expect at default settings when using these cameras with automatic metering and white balance. The DP1 image was processed using Sigma Photo Pro (SPP) and the G9 and DLUX3 images were processed in Adobe Lightroom (latest version), which should give the same results as Adobe Camera RAW (ACR).

First the G9 image:



Next the DP1:



Finally the D-LUX3:



In case anyone would like to play with them, here are the RAW files with the caveat that although a tripod was used, multiple takes were not done to assure the sharpest possible result:



Part 6 is coming soon.

Read More......

In Parts 1-4 of the Shootout, I compared to Sigma DP1 performance to that of the Canon 5D with EF 28mm f/2.8 lens and Olympus E-420 with ZD 14-42mm lens at 14mm. The Fuji F31 was the sole compact competitor in those tests and lived up to its reputation for excellent low light performance.

In the upcoming parts, I will test the Sigma DP1 against the Canon Powershot G9 and Leica D-LUX3.

Read More......

Regions of DP1 images which have been significantly overexposed or underexposed are often recoverable using Sigma Photo Pro. I'll give a couple real world examples of this remarkable exposure latitude.

I was at a friend's wedding yesterday. We were in the dark hall of a castle, waiting to toast the bride and groom. Thinking I had set my DP1 for ISO 800, I dialed in f/4 and a shutter speed of 1/10th of a second, which is about as long as I can hold the DP1 steady. Even if I had set the ISO as planned, the resulting photo would have been at least a stop underexposed. Unfortunately, I had left the camera at ISO 100. Here is the resulting image, straight out of SPP with no manipulation other than resizing:

(Click image for intended viewing size)
With any other compact digital camera I have owned in the past, that file would have been a complete loss. After pushing the X3F exposure maximally, processing using the SPP monotone conversion, adjusting levels/curves in Photoshop, and resizing, here's what I got:

(Click image for intended viewing size)

With the DP1, I was able to pull enough out of the shadows to preserve the memory of the occasion.

So that's the shadow latitude, what about the highlights? In Part 2 of the DP1 Shootout, I gave a sample of the DP1 highlight retention at ISO 100. However, the highlight recovery at higher ISOs can at times be shocking. Here's one example from Satuday afternoon at the National Museum of Natural History in Washington, DC. Shot at ISO 800, f/4, 1/10s. Original image with no manipulation other than resizing:

(Click image for intended viewing size)
After processing in SPP:

(Click image for intended viewing size)
I suspect that the DP1 is using an "underexpose and push" method, similar to Canon's Highlight Tone Priority, to achieve high ISOs. No other reason comes to mind for why the high ISO files show far greater highlight recovery potential than do the ISO 100 images. Whatever the explanation, the DP1, as compared with either the Canon 5D or Olympus E-420, seems both less prone to blow highlights at high ISO and more able to recover them. The downside is that the DP1 shadow recovery is somewhat compromised at high ISO. It's a tradeoff I'm happy to accept, with results that are generally more "film-like" than I am used to seeing from digital cameras.

Read More......

In Part 4a of the shootout, I presented photos taken in low, incandescent lighting, a sort of "worst case" scenario for the DP1. In this next example, the samples were shot in a mixture of natural and incandescent light. The technical details for this set are otherwise the same as those in Part 4a with one exception. I forgot to move the tripod for the F31 shots, so the F31 crops in this comparison represent a different field of view corresponding to the difference in angle of view (35mm equivalent vs 28mm equivalent) at a given perspective.

Here is the overall test scene with the specific regions examined identified in the yellow boxes:



ISO 400 Crop 1:



ISO 400 Crop 2:



ISO 400 Crop 3:



ISO 400 Crop 4:



ISO 800 Crop 1:



ISO 800 Crop 2:



ISO 800 Crop 3:



ISO 800 Crop 4:



ISO 1600 Crop 1:



ISO 1600 Crop 2:



ISO 1600 Crop 3:



ISO 1600 Crop 4:



As was the case in Part 4a, color fidelity is a problem for the DP1 in this low light comparison, and the problem grows worse with each increase in ISO. However, looking at Part 4a, you can see that the problem was more severe when incandescent light was the sole source of illumination. For example, the shift and desaturation in the brick region shown in Part 4a was already severe in the DP1 ISO 400 image, whereas the similarly colored red curtain in this comparison is better preserved. With a good source of natural light and willingness to postprocess, better results than those shown here are possible. However, a simple boost in saturation will not solve the color problems that affect high ISO, low light DP1 images.

Addendum: After reviewing the RAW images a few more times, I decided on a different custom white balance for the DP1 images to best match the colors of the other cameras. It's a tricky call because matching colors from some regions of the image will throw off the match in others. Having arrived at a better overall white balance choice, I have now replaced the DP1 crops in the comparisons above. As always, feel free to play with the RAW files.



Thanks again to Serhan for lending me the Fuji F31 used in this test. Part 5 of the shootout will be coming soon.

Read More......

The technology behind the Foveon sensor at the heart of the DP1 has been well covered elsewhere. One interesting aspect of the three-layered Foveon design is that the quality of color representation in low light is partially dependent on the spectrum of light available. For example, low light provided by natural light (i.e., sunlight through a window) will typically produce better colors than will a similar level of light coming from an incandescent source. I decided to do this first high ISO comparison as a sort of "worst case" demonstration. Here I will show crops from the Sigma DP1, Fujifilm F31, Canon 5D with 28mm f/2.8, and Olympus E-420 with 14-42mm f/3.5-5.6 lens. The photo compared was taken in low light, indoors, with an incandescent light source.

The competitors:



The Sigma was shot at f/5.6, Fuji at f/4, Canon at f/11, and Olympus at f/5.6. Shutter speeds were manually determined to match exposures. For detailed EXIF information, please consult the files available for download at the end of this post. A sturdy tripod was used in all cases. The subject distance was altered for the F31 shot in order to keep the field of view consistent.

An important thing to note is that the crops are grouped by ISO, which may not be relevant to actual practice. For example, the Fuji F31 has a one stop advantage in maximal lens speed compared with the DP1 (f/2.8 vs f/4). Therefore, one might compare the ISO 800 F31 crops with the ISO 1600 equivalent (ISO 800 pushed one stop) DP1 crops. Likewise, the Canon lens has a one stop advantage over the DP1; thus one might want to compared the ISO 400 DP1 crop to an ISO 200 5D crop. However, using the Canon lens at f/2.8 will result in a more shallow DOF than will using the DP1 at f/4. Therefore one might choose to shoot the Canon at f/6.3-7.1 for acceptable DOF in a situation where f/4 would do with the DP1. In such a case, the more relevant comparison would be DP1 ISO 400 to Canon ISO 1000 or 1250. I will simply present the crops according to ISO and leave it to you to decide which ones should be compared.

All images were resized using Photoshop Bicubic Sharper to match the diagonal pixel dimension of the native DP1 files. Therefore, the DP1 crops are at 100% of the original, and the 5D/E-420/F31 crops have been downsized as needed. All were shot RAW and processed using the settings specified in the introduction. The exception was the Fuji, which does not offer RAW. Although I custom white balanced all cameras at the time of use, all required further adjustment during RAW conversion to get a closer white balance match. I also applied a 25% cooling filter in Photoshop.

Here is the overall test scene with the specific regions examined identified in the yellow boxes:



ISO 400 Crop 1:



ISO 400 Crop 2:



ISO 400 Crop 3:



ISO 400 Crop 4:



ISO 800 Crop 1:



ISO 800 Crop 2:



ISO 800 Crop 3:



ISO 800 Crop 4:



ISO 1600 Crop 1:



ISO 1600 Crop 2:



ISO 1600 Crop 3:



ISO 1600 Crop 4:



I think these crops speak for themselves. With this quantity and spectrum of lighting, the DP1 high ISO color fidelity is poor. Boosting saturation in SPP (not shown) can help restore some punch, but it does not restore color accuracy. Nor does the use of custom WB use at the time of the shot help in any way that I could discern. On the positive side, high ISO DP1 files are clean, so black and white conversions can produce high quality results. The degree to which they are noise free raises the question as to whether SPP is applying some mandatory noise reduction. I look forward to trying 3rd party RAW processing applications as they become available and hope to read some analysis of the RAW files by those with expertise in that area.

What about the little Fuji? I think it can stand proud after this comparison. Sure there's some mandatory noise reduction going on with the F31, and yes it shows the least usable detail in this group. In my opinion, the good color fidelity of the F31 translates to better overall image quality than the Sigma DP1 for color, high ISO, small prints in this type of lighting. Especially so considering the one stop advantage of the Fuji lens (f/2.8 vs f/4), i.e. comparing F31 ISO 400 vs DP1 ISO 800 or F31 ISO 800 vs DP1 ISO 1600 (equivalent).

I realize that this comparison gives a somewhat different impression than the preliminary one I shared in the discussion area. Comments and suggestions are, as always, appreciated.



Special thanks to Serhan for lending me the Fuji F31 used in this test. In Part 4b of the shootout, I'll show the results from a similar comparison conducted in mixed low light.

Read More......