Messages

1

EOS Bodies - For Stills / Re: EOS 5D Mark III Firmware 1.2.1 Available for Download

« on: April 30, 2013, 03:15:54 PM »

Magic Lantern is working on a true RAW and 14 bit video recording along with 4:2:2 to the card, something we'll never see from Canon.

And they've already said it will never be something that will work for video people...best they've gotten so far is 720 at 24fps for 2s...the reality is you'll never get long recording times because the camera has a buffer to deal with. 1080p wont happen.

Really, it's an upgrade for time lapsers and may have some cool other uses.

If this is the case, the soft output of the 5DmkIII compared to other cameras, as documented on many of his tests, may in fact be impossible to fix with a firmware update, at least one from canon.

That said, what they did find interesting was that the resolution of the DNGs they got from their RAW mode were way higher resolution than just the traditional h.264 video stream from Canon. Not sure if that means that Canon intentionally is crippling it, or if the h.264 encoding loses a lot of resolution, but, that could be useful for some purposes.

Of course, if you want Raw video at 2k, you'd just spend the same amount on a BMCC instead and get a much nicer workflow.

I wonder if they can feed that 'raw' DNG stuff into the h.264 compressor and get better quality out? Or if they can push it out frame by frame over the HDMI and delete each frame from the buffer afterwards so nothing would ever overflow and you could at least get an 8bit or maybe 10bit crisper image out over HDMI? It seems like it should be possible but it's hard to know the internal Canon subsystem, it might not be, or even if it is, it quite likely might require all sorts of knowledge far beyond what they have been able to hack out so far and might only be reasonably doable by Canon people with full docs and access?

I checked out the DNG file posted on eoshd.com. Each file is about 5MB. Therefore, to get 30fps, you need a sustained write bandwidth of 150MByte/s (that is 1.2Gbits/s). That will be quite challenging even for the fastest notebook SSD. A 64GB CF card, even if it were fast enough, would only store about 7 minutes of video.

Therefore your suggestion to use the HDMI out is potentially the solution. HDMI 1.0/1.2 supports a bandwidth of 4.95Gbit/s and HDMI 1.3 supports about 10Gbit/s. BTW, does anyone know which version of HDMI is supported by 5D3?

Ofcourse, I am no camera engineer. These are just rough calculations that I came up with.

Lastly, pardon my ignorance in video, doesn't the 'uncompressed HDMI' output as provided by firmware v1.2.1 mean raw? Is the ML DNG just a higher resolution raw compared to the one provided by firmware v1.2.1?

2

PowerShot Cameras / Re: Patent: Integrated Lens for an APS-C Compact Camera?

« on: December 27, 2012, 03:30:21 PM »

Amazon is selling the G1X at $629 and the RX100 at $648

Anyway, I took quite a few thousand photos with the G1X and loved all the images I got from it. My main camera is the 5D2 and I have a few Canon glass. It does not make sense for me to invest in another mirorrless lens system for my leisure camera.

The G1X is a great one lens solution for me. I do hope Canon would address the macro distance and short battery life issues in the next version though.


 

3

EOS Bodies / Re: Big Megapixels Coming Soon? [CR1]

« on: August 05, 2012, 12:21:43 PM »

There is all the saying about how much ahead Sony sensor is vs Canon and that the 5D3's sensor has not improved from the 5D2.

Firstly, the 5D3 sensor uses a gapless microlens design which was not used before. Numerous test have shown that it has improved over the 5D2 by at least 1-2 stops. So it is quite unreasonable to say that the sensor has not improved.

Secondly, I noticed that most of the test showing the D800 having a better noise performance than the 5D3 often involved downscaling or upscaling the image to the same size... When showed at exactly the same size, it is often obvious that the D800 is noisier than the 5D3. When downscaling is involved, even the Nokia Pureview 808 showed remarkable improved noise performance. Now, is this purely due to sensor technology?

Given Canon's 18mpx APS-C sensor technology, they can produce a 46mpx FF sensor without having to do too much. At the noise level of the 60D and if one were to downscale the 46mpx image to 36mpx (or 22mpx), I am very sure the noise level will be very comparable to the D800 or better. Shadow noise is a real concern for Canon however. Hopefully they can start putting in an on-chip ADC soon.

Thirdly, DR. Perhaps someone can show me proper photos showing what those 3-4 EV advantage that Sony sensor supposedly boasts as compared to a photo taken with a Canon camera. I would be convinced then. Right now, all I hear is 'scientific' tests with a bunch of numbers. A proper set of photos would convince me what advantage a Sony's sensor has in terms of DR. So far, all I have seen is a bunch of numbers.

Lastly, the whole point of all this megapixel. In DPReview's tests, they admitted that getting that observable resolution advantage from the D800/D800E required a lot of care and effort. Only in lab condition and at very specific lens type and setting can they achieve that resolution advantage. All this while, one has to carry around that extra bits of file size and run out of drive space consistently.

If we look at the AF speed comparison on imaging-resource.com, it is very obvious that the 5D3 focuses almost 2x as fast as the D800 and do not suffer as badly from delays in buffer clearing, etc. That in itself makes the 5D3 a much more usable all-rounded camera.

So Canon should make a high mpx camera, by all means. Landscape and studio photographers would love that. It would definitely not affect the target market of 5D3 - People who need an all-rounded camera.

4

EOS Bodies / Re: After DPREVIEW of D800 would you still get the Canon 5d mark iii?

« on: May 11, 2012, 12:34:35 PM »

There are all these claims that the D800 is better than the 5D3, most attribute this to DR and resolution. So far, the DR claims have mostly been a bunch of numbers or graphs. I have yet seen any photos showing what difference that XX ev has resulted in. Not even lab photos.

As to resolution - yes, it is clearly visible in the photos. However, this statement from dpreview illustrates just how difficult it is to get that kind of resolution advantage (even in a highly controlled lab environment).

"We should note, however, that we had to work quite hard to get this amount of resolution. We used flash to eliminate any risk of blurring due to vibration, we focus-bracketed in extremely fine increments, and we used an excellent lens (the Nikkor AF-S 50mm f/1.4 G) at an aperture optimal for central sharpness of F4.5."

Moreover, this resolution advantage - achieved through considerable efforts - carries a price. Larger files mean less space for photos on your memory card, lower frames per second, longer upload time, longer processing time, more hard drive space required, etc...

On the other hand, the 5D3 has a faster frame per second, less moire and aliasing in video, more advanced in camera movie functions (although some professional had to remove the AA filter to achieve an acceptable resolving power), an advanced AF system from the 1D series and a free bundled RAW conversion software.

Looking at the two cameras from this perspective makes one wonder which is really the better one...

5

EOS Bodies / Re: More Full Frame Cameras on the 2012 Horizon? [CR2/CR1]

« on: April 27, 2012, 12:35:58 PM »

From Canon's recent launches, it is obvious that they are trying to build a portfolio around DLSR based movie cams. I would not be surprised to see a 5DC... Perhaps with a much weaker AA filter to improve resolution, better codec... After all, the only real complain about the 5DIII's video capability is the resolving power.

6

EOS Bodies - For Stills / Re: DxOMark scores for 5DMkIII out - total score 81, 5DMkII had 79

« on: April 19, 2012, 03:04:00 PM »

apple designs the iphone, foxconn makes it. Nikon designed the sensor, sony makes it.

To be correct, Nikon designed some supporting circuitry around the sensor, not the sensor itself...

I believe the relationship between Nikon and Sony is more like Apple and LG (or whoever their screen supplier is now), not Apple and Foxconn

LG develops the technology in display, manufactures it according to some specification and Apple uses it in their products. No shame and not issue in that in my opinion.

What really annoys me is this DXOMark scoring thing...

They come up with a bunch of numbers, without any correlation to any pictures to illustrate any perceptible differences in image quality as a result to whatever score difference they claim between cameras.

They claim the D800 is so much better at high ISO than the 5D3 when all the sample pictures point at most to a draw. They claim the D800 has so much better DRange than the 5D3 when changing the modes of the camera brings about such a huge difference in DRange  measurements. One only has to play around with dpreview's DRange comparison charts to realize that.

They claim that their measurement is purely on the sensors and does not take into consideration image processing, lens and other peripherals.

How would one take a picture without these mentioned peripherals, may I ask? And if one uses all these peripherals to take a picture in real life, what is the purpose of throwing out a bunch of numbers which may not even be relevant in real life?

I think all these arguments about DXOMark scores is embarrassing to say the least. It is like a bunch of kids arguing over whose father is taller. If there needs for an argument, photographers should go out and take photos and then argue about whose pay check is bigger 

7

EOS Bodies / Re: Canon 5D Mark III - Price Drop Details

« on: March 06, 2012, 11:09:32 PM »

1. When will Canon drop the price of the new 5D Mark III?
When the 5D Mark IV is released.

2. What price will it drop to eg. $2999

Depends on the exchange rate of the USD to JPY at that time. If it were 10JPY to $1USD, then the price would be $10,000.

3. Your thoughts

I think the 5D3 will be a great upgrade to the 5D2. The 5D3 samples at imaging-resource are great compared to the 5D2's. Check out the ISO3200 (and higher) samples... much better than the 5D2... and may even be better than the D4.

I think it is great that Canon let the images show the true upgrade from 5D2 to 5D3 rather than a bunch of specs and numbers, which is usually the easy path for manufacturers. It would definitely have been easier to just write a headline such as "The 5D3 has 3 times as many pixels as the 5D2".

8

PowerShot Cameras / Re: More Super G Series Talk

« on: December 26, 2011, 03:24:06 PM »

As was mentioned before in previous posts, the so called 'mirrorless market' is so loosely defined that it includes cameras with/without viewfinder, popup flash, hotshoe and includes cameras with various sensor sizes, etc.  These permutations continue to be tweaked frequently by manufacturers.

Yet, I keep reading comments that Canon do not have a mirrorless strategy. In the midst of the various permutation that all the camera manufacturer throw out, why would Canon's move to throw a fixed lens into the mix be less of a strategy than the others?

As for zoom range, the G series already support tele and wide angle adapters... If the new G series camera would indeed have a 6.5x zoom lens (24-156mm equiv. according to some rumors), a set 0.7x and a 2x adapter would increase the range to ~ 16-312mm (sufficient for most people). Provided Canon makes the high quality adapters themselves.

On top of this, if Canon is able to improve the image quality and get it closer what the V1/J1 and m4/3 can do (judging from the 1/1.7" CMOS sensor of S100, I believe this is possible), maintain an f-stop of about f3.5 or f4 at the tele range so that it is not too slow when using tele-adapters and reduce the overall size of the camera, I think the G series is a viable competitor in the 'mirrorless market'. Oh yes, and add faster AF and operation speed to the list too.

As for sensor size, I am more practical. There is no way Canon can use a 4/3 like sensor without adding an elephant trunk of a lens to the camera. Thus, a 2/3" or 1" sensor would be enough to make me queue up overnight to buy the camera.

9

EOS Bodies / Re: Thoughts From Canon on a Mirrorless Interchangeable Lens Camera

« on: December 22, 2011, 01:43:46 AM »

The whole marketing hype of mirrorless cameras is the ability to have a camera with DSLR quality images at the size of the point and shoot. The closest camera to that promise the the GF3X, but that would mean to maintain that size profile, you will need to stick to pancakes or that one lens. Still the GF3X is bigger than the G12.

I have yet seen any photos comparing the NEX-5N + SEL 18-200mm vs DSLR. Would it really be that small?

If we could just take a step back from all that marketing hype that all the mirrorless camera manufacturers poured onto us and take a clear look at the true value proposition of mirrorless cameras, we would realize that this is still a highly un-established market segment. If not for their ability to use classic MF lenses, note that mirrorless lenses are often poor in quality and high in price. Now, do I want a collection of those lenses, definitely not.

10

EOS Bodies / Re: EOS-1D X Canon USA Press Release

« on: October 18, 2011, 08:28:02 AM »

Dual DIGIC 5+ image processors capable of delivering approximately 17 times more processing speed than DIGIC 4

Having the same MP as the 7D, shouldn't this camera have say... 17 times faster burst rate?
  Now... before anybody shoot me, let me make it clear that I'm not asking for a 136 fps. I personally think 14fps is just right. I just like to know how the 17 times faster processor being utilized in this camera.

I think it is unlikely that the Image Processor is the limiting factor for frame rate. Mirror, shutter, Card/Buffer IO, etc could well be. My guess is that the IO buffer is a critical aspect for fps and shot to shot time.

Note that RAW shooting with larger files is almost always the same or slower than JPEG, although less processing is done. For JPEG, although you need to do all the conversion, compression, processing, your files are smaller and the speed is faster. So my guess is that IO Buffer plays a significant role in fps. To see the pure image processing capabilities, you need to look at JPEG, where smaller files results in less IO but more image processing is happening.

The fact that this camera can do 14fps JPEG while performing more calculations in terms of CA correction,  iTR AF, AF Tracking, etc, shows its processing prowess. Also note that processing 61 AF points instead of the 19 from 7D adds a multiplier to the processing requirements of the AF features I have mentioned as well.

11

United States / Re: Why a shutter curtain?

« on: September 10, 2011, 04:02:30 PM »

Electronics shutters are used in almost all Point and Shoot cameras, including the Powershot G12 and S95. As we are all aware, the G12 can hit a shutter speed of 1/4000, so there should be no limitations in terms of the speed of sensor read out.

So why not dSLR?

This could answer your question:
http://www.steves-digicams.com/knowledge-center/why-digital-cameras-have-mechanical-shutters.html

"Cameras, typically smaller point-and-shoot cameras, that use no mechanical shutters typically use an interline transfer sensor. An interline transfer sensor dedicates a portion of each pixel to store the charge for that pixel...
....Interline transfer sensor's typically have higher noise levels and lower sensitivity than the full frame sensor's used in high end digital SLR's."


I am truly impressed with the amount of technology that goes into our point and shoot really... Despite all that pixel cramming into a 1/2.5" sensor, you've got to deal with interline transfer... blah blah"

12

Canon General / Re: Canon 3 Layer Sensor (Foveon Type?) Patent

« on: July 09, 2011, 03:26:20 PM »

I believe a back-side illuminated, 3-layered, CMOS sensor is nothing new. Panasonic published a patent last year on this:

http://egami.blog.so-net.ne.jp/2010-07-19-english

My guess is that the innovation here seems to have something to do with microlens and the difference in sizes of the BGR layers with layer B being the largest...

It is still not clear how that solves color bleeding though... and is Canon the only company using microlenses on their sensors?

13

EOS Bodies / Re: Will Canon Withdraw from the Megapixel War?

« on: June 21, 2011, 02:59:27 PM »

Dr Croubie, you brought up a very interesting point. In the past Mhz = Performance, until the multi-core strategy came in and changed everything.

Maybe the future, higher resolution may not be dependent on Mpx alone. Hasselblad just introduced a sensor shift technology that takes a 200mpx picture from a 50mpx sensor.

I would think the future of high resolution sensor technology may employ techniques such as sensor shift, multi layer sensor (ala Foveon) or multiple sensor (ala video cam)... or maybe not.

Note that while the Mhz war has sort of taken a different course for Intel/AMD, the number of transistors packed into a microprocessor has been doubled every two years as predicted by Moore's law without change...

Since we are in this topic, here are some articles discussing the application of Moore's law to camera sensors

http://www.luminous-landscape.com/essays/moore-rebuttle.shtml

14

EOS Bodies / Re: Will Canon Withdraw from the Megapixel War?

« on: June 21, 2011, 02:03:52 PM »

Once upon a time, Bill Gates allegedly said 640kByte RAM is enough for everyone. Not too long ago, the Olympus president said 12 Mpx is more than enough for everyone... see the similarity?

Once upon a time, people complained that there was no need for more powerful processors, Moore's Law will not persist in the next 5 years.. well, back to the future, Moore's Law held and processors kept packing more transistors with no signs of slowing down... And guess what? We kept finding new uses for all that processing power.

This is technology... with it, we are meant to overcome the shortcomings while improving the capabilities of our equipments. Without it, we are no different from our banana eating, tree swinging cousins.

Haven't anyone noticed the wonderful innovations that have gone into our sensor technologies. The Nokia N8 with its glorious 12mpx sensor will blow the old 2mpx sensors in my old Sony Ericsson phone out of the water in terms of noise performance and Drange. The BSI sensors will easily out do old 4mpx camera sensors in terms of DRange and noise performance. And the great 7D sensor with its gapless microlens and reduced photosite distance is a great break through in sensor design worth admiring.

So diffraction is a problem now? Is it an indication that lens design might be falling behind sensor design? Well, lens engineer go solve it... Is this a limitation with laws of Physics? Did they not say that microprocessor will face an inevitable limit with laws of Physics at xx nm and there is no way to go beyond? ... well, we are well into 20nm territory now and going strong....

So I say, go for 100 mpx... solve the issues, improve Noise Issues, improve DRange issues, innovate and make a better camera each year.

The issue of prioritizing Drange vs Mpx is a slightly more tricky one... However, if we look at the Fujifilm Super CCD technology which groups pixels into two groups, each taking the same scene at different brightness and then combining them to form a HDR image, it would seem like having more pixels might help your Drange after all. Back to my imaginary 100 mpx  sensor. I may be able to group them into four groups which capture a scene at 4 different exposure. I can then use these 4 sets of images to cancel out random noise and create a HDR photo. This would give me a glorious 25mpx HDR image which has very low noise...

So why is more megapixel bad?