Sony A350: Full-Time Live View at 14.2MP

113 Comments

View All Comments

• danddon - Wednesday, April 2, 2008 - link

  Lordy, Lordy, Lordy.
  What a mess.
  
  1. The re-do shots of the K20D and 5D look like they were made with a flash compared to the original a350. Did someone turn on a few more lights?
  2. The re-do shots were not made using the same camera location/position as the original a350. They appear to be lower. Check the yellow "5" on the front of the Corsair box. It can be clearly seen on the a350, but not on the K20D re-do. The 5D re-do has the same problem.
  3. The P5K Deluxe box appears to have been moved in the re-do shots. It is out of focus on the original a350, but not the re-do's. In the a350 shots, it appears to clearly overhang the box below, the Nvidia 780i box. In the re-do shots, it appears to be flush.
  
  Please, please, please - give some thought to a proper test setup, and one that can be re-produced accurately. Not a bunch of boxes....
  
  The same is true of the camera position. It must be exactly the same for all shots.
  
  Nice try, and thanks for the effort - but, this should never have been published by AnandTech.
  
  IMHO and FWIW.
  
  

• Wesley Fink - Wednesday, April 2, 2008 - link

  One more consideration in this and future reviews. Most DSLRs advertise 1.5X magnification factor, but the Canon is 1.6X, Sigma is 1.7x and Olympus/four-thirds is 2.0X. We can understand why full-frame and 1.5X would vary, but even among the group of 1.5X cameras there is still some variation among the actual multipliers since the sensors vary a bit in effective size. This will show in review crops with slight variations in image area even with camera brands claiming the same multiplier.
  
  Rather than correct the camera crops so all reveal the same image, we plan to show a constant pixel size in crops of 230x300. This has the advantage of showing the true impact of the multiplication factor on the image you capture. Constant 1:1 pixel crops are also always examining the same number of captured pixels.

• Wesley Fink - Wednesday, April 2, 2008 - link

  You make good points but you are not completely correct.
  
  1. The light is exactly the same and in the same location in all shots. No flash was used and the EXIF info in the full image will show you that. It is a single 100 watt tungsten bulb high and to the right of the image. We do want to provide a reasonably consistent image in AT reviews but it is not our intent to duplicate a studio setup for our reviews. We do have the studio lighting but there are already enough sites that do that.
  
  2. Good catch. You are correct about the lower point of view. We received the sturdier tripod and remote release that were on order for our future reviews and used it them to reshoot with the k20d and 5d. The main point of the reshoot was to show more meaningful crops and archive for future reviews, so we set the new tripod up to be reproducible in future shots. I have been able to arrange another day with the A350 so I will reshoot those from the same point of view and same constant aperture as the current 5D and K20D images.
  
  3. That box was jostled and bounced, but it is not part of the crop area so we didn't worry too much about it. In the reshoot we will make sure everything is in the same place for the A350.
  
  LAST we have found the issue with the noise variations we see in the crops not showing in your web images. They are all too visible now. PLEASE keep in mind these are ATUAL PIXEL CROPS and they are equivalent to a 1"x2" area on a very large 16x20 enlargement. They are actual pixels to emphasize noise variations.
  
  We are listening to your suggestions and making the necessary changes to improve this review and future camera reviews. We have also been working with a company that produces photo test procedures and analysis tools to develop more test procedures for the future that communicate quality differences among digital cameras without going to the extreme levels of some other dedicated photo sites. Readers have told us many of these sites are intimidating and they want a simpler but dependable source for digital camera info. I apologize for the learning curve in accurately displaying images in the review that convey what we found in a lot of shooting with each camera and looking a many images to reach our conclusion. We will improve those. However, my conclusion has not changed with anything we have reshot as the issue is how to best convey to you what we found in examining the camera and the images.
  
  As you probably guessed the 5D and K20D are two of my personal cameras and I can reshoot anything needed with them. Since I will have the A350 again for just a day do you have other suggestions for the reshoot?
  

• simla - Saturday, April 5, 2008 - link

  This article is a shambles...if more of these Anandtech-trying-to-be-an-all-in-one supersite continue, there's a very good chance you're going to turn away a lot of the readers who have stuck by this famously detailed and well-informed site (just look at some of the INQ-style content thats being posted on DT nowadays).
  
  In my honest opinion, AT should stick to what the reviewers know best: PC hardware analysis. As someone posted earlier, this is very much an opinion piece worthy of a blog post - and no more. For camera reviews - this doesn't even come close to that of DPReview standards, and just makes AT look amateurish.
  
  Wesley, I enjoy your PC articles and you're a good writer, but this one misses the mark by a long way.
  
  

• danddon - Wednesday, April 2, 2008 - link

  First, let me thank you for your considered reply. I suspect you have been busy since this article came out, and I appreciate the time you spent.
  
  There must be something wrong with the a350 pictures. They are visibly darker than the K20D or 5D. For example, the EXIF tag on the 800 ISO JPEG for “BrightnessValue” shows –0.75. Perhaps this is part of the exposure problem, perhaps not. I do not have access to the camera – you do.
  
  I do not think you have done anyone a favor with respect to the Sony a350 by publishing these results. I can’t believe these noise levels are truly representative of the camera and what it can do.
  
  If you can’t obtain a dedicated room to set up and properly maintain your test facilities, then you may want to reconsider this undertaking. You have already pointed out the problem with shooting a bunch of boxes.
  
  A good first step would have been to talk to that “photo test company” before embarking on any tests like this. But, better late than never.
  
  Some suggestions:
  1. A dedicated facility, that is managed by you. This means a separate room with a locked door.
  2. Proper studio lighting, along with as many “temperatures” as you care to deal with, such as daylight and tungsten.
  3. A real photographic resolution chart, properly mounted on a wall. Noise can be seen on this kind of chart, as well as resolution.
  4. Marking of tripod leg locations on the floor, so that photos can be reproduced, and differences between cameras can be minimized.
  5. Marking of tripod height, for the same reason as number 4.
  6. Making sure all cameras are set at factory defaults, to minimize possible problems due to an unfamiliar setting being left activated in a borrowed camera.
  7. A separate light metering facility, to ensure that the lighting stays constant for each shooting session, and is maintained over the weeks and months you plan on doing these tests.
  8. Try to get as many internal reviews as possible of your material before putting the article on the web site.
  
  I would not bother making these comments if this were not AnandTech. You guys have a huge and well-deserved reputation that carries a lot of weight. Please don’t dilute that reputation by a “low-ball” approach here, however well meaning the intent.
  

• halcyon - Wednesday, April 2, 2008 - link

  "good news about the sensor is that output is very clean and noise remains low up to ISO 1600."
  
  Are you smoking crack?
  
  Look at the shots already at ISO100. They are horrible! They are like from a tiny sensor P&S camera.
  
  Even my Fuji F30 fairs better.
  
  The noise and dynamic range is indeed apalling.
  
  Already the Pentax leaves it fair behind.
  
  Canon 5D cannot be even mentioned in the same week. It is an order of magnitude if not several better.
  
  Please, if you continue reviewing cameras, try to make a better effort of it in the future.
  
  

• Wesley Fink - Wednesday, April 2, 2008 - link

  The crops are equivalent to pixel-peeping as they are 1:1. They are an actual 230x300 pixels cropped from a 4592x3056 image. That is equivalent to a 0.5”x0.6” spot on an 8x10 enlargement. The 1:1 crops are to emphasize the observed noise. There is also the issue that the shallower depth of field at f2 causes the non-perpendicular image of crop area to be slightly out of focus across the crop – which is not an issue for f4.
  
  Before making pronouncements you need to compare a similar 230x300 crop of your F30 image apples to apples. You really need to look at the full images by right clicking to get fair comparisons of noise at various ISOs.
  
  I have been able to arrange to get the A350 back for a reshoot, and I will update those images as soon as they are complete late tomorrow.
  

• cputeq - Wednesday, April 2, 2008 - link

  Are we all looking at the same shots?
  
  The Sony crops in the article look completely out of focus ><, which would negate any sort of noise testing.
  
  Granted, I haven't taken a look at the complete files, but from the review crops it looks like the focal point of the Sony shots wasn't on the same spot as the Pentax / Canon shots.
  
  -------------------------------------
  
  That being said, overall good effort on the review, but I hope you guys learn to standardize and review your testing methodology. You don't have to get all nit-picky like a lot of the photo sites do, since Anandtech isn't really a photo-oriented site, but some standard guidelines should do.
  
  I would make a few suggestions:
  
  1) Make a *standard* setup for testing noise levels, something you can duplicate across many cameras. The setup in this review doesn't look like it would be easy to recreate 2 years down the road!
  
  Of course, if AT doesn't plan on reviewing cameras that often, this may be a moot point.
  
  2) On a camera of this class (entry/enthusiast DSLR) you might want to give two sets of shots : Shots showing the *same* settings in a full manual mode with the lens not all the way open (maybe 1-2 stops down at least) and another set with the camera in a P or Auto mode.
  
  Seeing as how these cameras are targeted to the first-time SLR buyer, a brief look at how the various cameras handle situations in their "auto" or "program" modes might be helpful. Food for thought
  
  3) Post both JPEG results and exported RAW results (crops, of course). Of course this doubles the workload when comparing image quality, but sometimes there is a substantial difference between the JPEG engines in a camera that doesn't carry over as bad to the RAW side of the equation.
  
  

• Wesley Fink - Wednesday, April 2, 2008 - link

  I reshot the K20D and 5D at a constant f4 as baselines for future reviews. The crops and full k20D and 5D images have been updated. As we stated on the first page and on p.6 we the images are now shot at f4. We were not able to reshoot the A350 since we no longer have the camera. As a result of the shallower f2 Depth of Field the focus for the A350 images is very slightly off. Because of that, we also warned in the update comments that the A350 shots were f2.0 and NOT f4 and to keep that in mind when viewing.
  
  The real issue is not the f2 versus f4, however. The crops we see show much greater differences in sharpness and noise than you can see on the web. We are trying to determine why the noise differences mostly disappear in a web post. Perhaps it is some image processing that is being done behind the scenes in our web posting program. That is beyond my control, but we are trying to find answers to more accurately present the images as we see them.
  
  In the mean time, you can right click to view or download the full image. In looking at the full images you can clearly see the differences described in the article between the A350, k20d, and 5D at the various ISO sensitivites.
  

• JarredWalton - Wednesday, April 2, 2008 - link

  Crop images were incorrectly linked in my HTML coding. I have addressed this. Check page five and you will now see a MUCH greater difference in noise levels.