Sunday, 30 December 2012

GH2 vs GH3, AF speed comparison

All new premium Micro Four Thirds cameras come with a claim to have the fastest AF ever, and this applies to the Panasonic GH3 as well. The GH2 improved upon the GH1 by adding 120fps AF readout from the sensor (in bright light), and the GH3 ups this to 240fps. However, this feature is only available with the Lumix X 12-35mm f/2.8 and 35-100mm f/2.8 zoom lenses, and then only with the most recent firmwares. Probably, this only works in bright light, otherwise, one would guess that the camera chooses a slower sample rate.

I've tried to measure the focus speed of the GH3, compared with the GH2. I did so in two lightning conditions, moderate indoor light (6 EV), and very low indoor light (3 EV). I also used two lenses, the Sigma 30mm f/2.8 EX DN and the GH3 kit lens Lumix X 12-35mm f/2.8.

I tested them by setting the cameras on a tripod, with centre spot focus mode enabled. And I put a small figure on the table in front of the cameras, with a distance of about 40cm. When turning on the camera, the lenses are usually reset to around infinity focus.

I video recorded the process using a GH1 camera. To measure the time the camera needs to focus on the figure and take the picture, I've analysed the audio, to find the time from my finger taps the shutter button, until the camera snaps the picture.


Here are some example tests in moderate light (6 EV):



And the results are:

Lens / CameraGH3 (6 EV)GH3 (3 EV)GH2 (6 EV)GH2 (3 EV)
Sigma 30mm0.54s0.63s0.61s0.52s
Lumix X 12-35mm @ 35mm0.23s0.74s0.43s0.45s
Lumix X 12-35mm @ 12mm0.18s0.25s0.27s0.28s

In the table, we can note that in the GH3 excels in good light with the Lumix X 12-35mm zoom lens, especially in the wide setting. This is as expected: The GH3 can use the higher framerate for the AF sensor output in bright light.

In dim light, though, the results are a bit unexpected: The GH2 comes out the best. This may be down to the fact that the GH2 is more mature, the firmware has been updated some times for better AF performance. I guess the GH3 will too, and probably the AF performance will be tweaked for the better in the future.

There is also the odd result that the GH2 and Sigma 30mm combination performs better in very low light than moderate light. I would attribute this to some random variations. After all, these are just single observations.

Conclusion


The GH3 performs very well in reasonable light, especially with the Lumix X 12-35mm f/2.8 kit zoom lens. However, in very dim light, it is not as good as the GH2. However, keep in mind that the GH2 has a more mature firmware.

Thursday, 27 December 2012

GH3 electronic shutter

One of the interesting new features of the Panasonic GH3 is the possibility to use an electronic shutter. It's not correct to say that this is a new feature, though, as the Panasonic G5, launched some months earlier, also has an electronic shutter. Probably, these cameras share some sensor components.

The electronic shutter has the advantage of being silent, of course, since no mechanical shutter is moving at all. It probably also has a quicker response time, however, I have not tested if this is actually a fact.

On the downside, the electronic shutter suffers from heavy rolling shutter artefacts, making it virtually unusable with moving subjects, or with a long lens. It also does not work with a flash, and cannot be used with high ISO or a very slow shutter speed. I'll get back to all of this in the article.

Silent operation


As said initially, the electronic shutter is interesting because it is totally silent. Normally, when making an exposure with the mechanical shutter, what happens when you press the shutter is this:

  1. The camera focuses (unless you have selected manual focus, MF)
  2. The camera stops down the aperture, unless you have set the largest aperture
  3. The mechanical shutter closes, and the sensor is made ready for an exposure
  4. The mechanical shutter opens for the exposure
  5. After the exposure is finished, the mechanical shutter closes
  6. The mechanical shutter opens again for continued live view
  7. The aperture opens up again

The focusing creates some noise. However, with most lenses, this is pretty much negligible. I have measured the focus noise for various lenses here, and it appears that only the Lumix G 20mm f/1.7 pancake lens has excessive focus noise.

Also, changing the aperture makes some noise. This is, by the way, why some users have reported the clicking noise when using the premium Lumix-Leica 25mm f/1.4 lens. I have measured the noise when changing aperture on various lenses, and they are mostly quite comparable.

Finally, what makes the most noise, is the mechanical shutter. Probably because it travels very fast. With the electronic shutter mode, you avoid this noise. Since it is so dominating, most people probably don't notice the noise of the aperture change, anyway.

When using the electronic shutter mode, the camera skips steps 3-6 in the above list.

For the ultimate in silent operation, I took this image, out of a series of exposures, using the Lumix X 12-35mm lens at 12mm f/2.8. When using the maximum aperture, the lens does not change the aperture before the exposure, and there is no noise whatsoever, beyond the initial autofocus sounds. In retrospect, I could have stopped down the lens some, as I notice that the foreground is a bit out of focus. Probably, the crowd would not have noticed anything anyway:


Also, when not using the mechanical shutter, there is less vibrations, which is good for the stability.

Limitations


The electronic shutter cannot be used with a flash. Not the built in flash, nor an external flash. Not even with the external flash in non-TTL auto mode, or in manual mode.

You cannot use the mode with exposures longer than 1s, or with an ISO higher than 1600.

Due to the latter limitations, one could speculate that the image quality is worse for the electronic shutter. To test this, I took the same picture using both shutter types, both images at ISO 1600 and 1s exposure:



Mechanical shutterElectronic shutter

And here are 100% crops from both images:


As far as I can tell, the images are identical. So there appears to be no reason to worry about the image quality when using the electronic shutter.

Rolling shutter artefacts


So, the silent and vibrationless operation is the advantage of the electronic shutter. What about the disadvantages?

I have examined the rolling shutter effects of the GH3 video, compared with the GH2, and found that the GH3 has less artifacts during video recording. On the other hand, I also saw that the electronic shutter mode of the GH3 produces very pronounced rolling shutter effects.

The rolling shutter artefacts of GH3 video is largely "academic". You only see it when you deliberately try to provoke it, e.g., by excessive horizontal panning during video recording. For normal use, this is not much of a problem.

The electronic shutter rolling effects are a problem with real life use, though. Here is a pair of example images taken using both shutter types, at f=32mm, f/2.8, 1/320s, ISO 800:



Mechanical shutterElectronic shutter

Setting a faster shutter speed would not help. You would still get the skewed car using the electronic shutter. This is because the speed of the sequential sensor readout is independent of the shutter speed.

Moving subjects is not the only problem with the electronic shutter. Camera shake can also be a big problem. Even if you use a very fast shutter speed, the sequential readout from the sensor will be rather slow. Any camera shake during this time means that the image gets a "wobbly" look.

Below are four example images I took using the Lumix X 45-175mm f/4-5.6 lens at 175mm, f/5.6, 1/200s, ISO 1600:



I was holding the camera at arm's length, to get pronounced effects. Even with a shutter speed of 1/200s, which one would normally think is safe with this lens at 175mm, there are heavy rolling shutter artefacts.

So, in this example, I exaggerated the effect by holding the camera at arm's length. But with normal use, I often note similar effects, although less pronounced.

Measuring the rolling shutter


It's easy to forget that the mechanical shutter is also a rolling curtain type shutter. It just moves so quickly that it behaves like a global shutter for most practical purposes.

We know the speed of the mechanical shutter. The camera's flash sync speed is 1/160s. That is the slowest shutter speed in which the whole sensor is exposed at one time. This means that the shutter curtain travel over the sensor at 1/160s, or probably slightly faster for some margin of error. So let's say around 1/180s as a guesstimate of the mechanical rolling shutter speed.

What about the electronic rolling shutter speed? Based on the example images, we know that it is slow, but just how slow? To measure it, Technic Lego again comes to the rescue.

I made this rotating propeller setup. The yellow propeller rotates exactly nine times faster than the red part to the right. By video recording the rotation, I find that the red part rotates twice per second, meaning that the propeller rotates 18 times per second:


Since the yellow propeller is three-pronged, this means that one single blade passes 54 times per second.

To illustrate what an "ideal" picture would look like, I take an exposure using the mechanical shutter first. It turns out like this, 1/1000s, ISO 6400:


So even using the mechanical shutter, we get some rolling shutter artefacts. We just need a very fast moving item to see it. This reminds us that the mechanical shutter is not a perfect global shutter, in fact, it is also a rolling curtain type shutter.

And based on this image, I can try to estimate the speed of the mechanical shutter. The blade has moved around 30° during the exposure. This is 1/12 of a full circle, and the blade rotates 18 times per second. This makes 1/216s passing during the rolling shutter movement. Since the flash sync speed is 1/160s, my estimate makes sense. The shutter is probably slightly faster than the flash sync speed. So my method appears to have merit.

What remains then, is to photograph the upper part of the propeller using the electronic shutter, and count the number of times it passes during the exposure. Here is one example image, taken at ISO 1600, 1/250s, f/2:


A propeller blade passes just above five times, let's say 5.2 times, just to name an estimate. Knowing that one single blade passes 54 times per second, this means that the electronic exposure takes 1/10s. Hence, the flash sync time using the electronic shutter, if it was possible to use the flash, would have been 1/10s. Which is very, very poor. This is the reason why we get the horrible rolling shutter artefacts.

By the way, this is consistent with an interview I read with some Panasonic engineers. They said that the electronic rolling shutter speed was about 0.1s, which is what I measured as well.

Compared with the GH2 electronic shutter


The Panasonic GH2 also has an electronic shutter option. However, it only gives 4MP images at the maximum. My study here reveals that the rolling shutter properties of the electronic shutter mode is exactly the same as the video mode. Hence, it is my belief that the electronic shutter in the GH2 simply takes the video output. That explains why the resolution of the 4MP images are poor, even for a 4MP image. It also explains why the GH2 could not produce images larger than 4MP when using the electronic shutter: This would have required an additional sensor readout process which was not yet available, until the G5 came along.

Comparison with other systems


The Nikon 1 mirrorless cameras were designed with electronic shutter in mind from the start. The Nikon 1 S1 10MP camera has a 1/60s readout, six times faster than that of the Panasonic GH3. The Nikon 1 J3 14MP camera is even better, with an electronic shutter capable of reading the whole image during 1/80S.

Only the Nikon 1 V2 camera features a mechanical shutter at all, usable when you want to be sure the image comes out without any rolling shutter artefacts. Further, the camera can take full 14MP images at a staggering 60FPS rate using the electronic shutter, opening up for very interesting uses.

From the Sony NEX line, some of the recent cameras, e.g., NEX-5R and NEX-6 have an electronic front curtain shutter. This means that the mechanical shutter is not closed prior to the start of the exposure. The only mechanical shutter is the one which closes when the exposure is stopped, the rear curtain. The advantage is obvious: It reduces the vibrations before the exposure commences, and also reduces the number of mechanical movements overall, while still avoiding the rolling shutter artefacts discussed above.

Conclusion


The electronic shutter feature is very interesting. But it's usefulness is limited, since it requires that both the target and the camera are very still. You must keep the camera still for 1/10s, which is pretty much impossible. However, when photographing organic objects like nature or people, small wobbly effects will probably not be noticed. Photographing geometric objects using the electronic shutter can be a problem, though.

Panasonic is marketing this function to be used for action photography. Here is a part of a screenshot from panasonic.net regarding the Panasonic GH3:


From panasonic.net, image copyright Panasonic

However, I doubt that those images could have been taken using the electronic shutter mode, as the movement would have caused significant wobbling effects. They could be video captures, though, as the video rolling shutter artefacts are much smaller.

Despite this, I think the electronic shutter mode of the GH3 is quite interesting, and I use it a lot myself. Unless you are photographing very square objects, like urban architecture, and especially when using a long lens, it is a very good feature to have.

You could even turn the electronic shutter to your advantage, and use it for creative purposes. Here is a picture of a moving car. Since I used the electronic shutter, it appears to lean backwards, while the rest of the scene is the right way up:


This way, you can recreate the famous photo of a racing car taken in 1913 by Jacques Henri Lartigue using a 4x5 Speed Graphic camera:



The shutter moves relatively slowly on this camera almost 100 years old camera, when compared with modern SLRs, which gives the distortion of the racing car. The distortion is especially visible in the wheels, which appear to be leaning forward. This effect was later copied by cartoonists when they wanted to give the impression of speed.

To get the same effect using the GH3, you need to hold the camera upside down. Otherwise, the car will lean backwards.

Real life example


This picture was taken at dusk, at f=28mm, 1/13s, ISO 1600, f/3.2, with the electronic shutter mode. I was panning with the movement of the car to keep it sharp, while blurring the background:


Since I used the electronic shutter, the background buildings are a bit skewed in this picture. They appear to be leaning to the left. This is due to the rolling shutter effect. Had I used the mechanical shutter, the image would have become exactly the same, except that the buildings in the background would not have been skewed. So, is this effect a problem? Not really. I doubt that many would notice the skewed buildings in the background. Perhaps you could say this even makes the image more lively. So my advice is: Unless you are using a very long lens without a tripod, or taking pictures of architecture, go on and use the electronic shutter.

There are more GH3 electronic shutter examples here.


Wednesday, 26 December 2012

Lumix X 12-35mm f/2.8 sharpness evaluation

The Lumix X 12-35mm f/2.8 is the new premium standard zoom lens from Panasonic. It was launched before the Panasonic GH3 camera, but obviously intended as the kit lens for their highest quality cameras. With a steep price tag to boot, there is much anticipation to the image quality from this lens.


A viewpoint commonly found in online discussions is that a prime lens is always better than a zoom lens. With this in mind, it is sensible to compare this zoom lens with some prime lenses. I chose to compare it with the Lumix G 14mm f/2.5 pancake lens and the Sigma 30mm f/2.8 EX DN, which are both covered by the zoom range of the Lumix X 12-35mm lens.

At 14mm


Using a tripod, and with the GH3 camera at the base ISO 200, I took the same picture with both lenses. The focus distance was very short, about 30cm/one foot. The pictures are out of camera (OOC) JPEGs. I made sure to focus on the same item in both images, the lone branch in the upper left corner.

Lumix G 12-35mm @ 14mm f/2.8Lumix G 14mm @ f/2.5

Now, let me be the first to admit that there are many problems with this test setup. For example, the lenses have different lengths, so that the field of view turns out to be different in the two cases, even though the camera is at exactly the same place in both examples.

Also, since I am focusing on an item close to the border of the frame, what is in focus could be quite different in the centre of the frame, due to different focal plane curvature. However, despite these shortcomings, I think we will be able to see examples of items that are in focus in both pictures, and, hence, be able to do some comparison of their relative sharpness.

Here are some enlargements (100% crops) from the top left corner, where I set the focus point.


And from the top middle:


Finally, from the centre:


Based on these images, keeping in mind that the focal plane may be slightly different in the two last crops, I think we can conclude that they are pretty much equally sharp.

The Lumix X 12-35mm zoom lens appears to give a bit nicer bokeh, but on the other hand, the Lumix G 14mm f/2.5 handles the flare better. These observations are perhaps not surprising: We already know that the pancake lenses are not optimal in terms of bokeh. And when it comes to flare, we would normally expect that the simpler lens, with fewer lens elements, handles flare the best. The Lumix G 14mm pancake lens has only six lens elements, while the zoom lens has a whopping 14 lens elements. Flare is usually caused by unwanted stray reflection and refraction between the lens elements.

I think we can conclude that both lenses perform very well here.

At 30mm


This time, I focused in the single leaf in the centre of the frame. The focus distance was about one meter (three feet), suitable for a portrait photo.



Lumix G 12-35mm @ 30mm f/2.8Sigma 30mm @ f/2.8

While I thought that I focused on a single leaf in the centre, it turns out that they were two different leaves, at some distance, and the Sigma lens focused on the left leaf, while the Lumix lens focused on the right leaf.This just shows that extreme caution must be taken when doing these studies. Still, I think the 100% crops from the centre are interesting:


We see that both lenses are very sharp in the centre. We also see that the Sigma 30mm lens is more resistant to flare, again probably because of the simpler construction using only seven lens elements.

Here are some crops from the lower left side, to look at the bokeh:


Based on these examples, the bokeh looks just fine for both lenses.

Conclusion


My experience so far indicates that the Lumix X 12-35mm f/2.8 zoom lens has very good optical properties. I plan to make more comparisons later.

Despite using "nano surface coatings", the lens is prone to flare, though, but it is not unexpected for a lens of this class.

Tuesday, 25 December 2012

What's in a mirrorless camera?

In the Deutsche Technikmuseum in Berlin, there is this exhibition showing the individual parts of a Canon EOS D30 from 2000, one of the first Digital Single Lens Reflex (DSLR) cameras:


Seeing this exhibition had me thinking: How would this look for a mirrorless camera? What's in a mirrorless camera, anyway? Obviously, it contains a lot of the same items: The battery, the imaging sensor, an LCD/OLED display for viewing images and menus, shutter buttons and wheels, a tripod mount, and so on.

It's probably easier to answer the question: What's not in a mirrorless camera. For a start, it doesn't have a mirror, that's given by the name.

Here are a couple of very simple illustrations of DSLR and mirrorless cameras, illustrating that the DSLR has a mirror between the lens and the sensor, and a pentaprism. While, on the other hand, the mirrorless camera has a shorter register distance, the distance between the lens mount and the sensor plane.

DLR camera with lens:


Mirrorless camera with lens


But the lack of the mirror is not the only difference. Let's try to make a summary of the main differences between a DSLR camera, and a mirrorless camera:

FeatureDSLR cameraMirrorless camera
MirrorYesNo
Pentaprism/pentamirrorYesNo
Mechanical shutterYesYes
Phase difference autofocus (PDAF) sensorsYesNo, the imaging sensor is used for autofocus
Light metering sensorYesNo, the imaging sensor is used for metering
TTL flash sensorYes, can be the same as the light meterNo, the imaging sensor is used for TTL flash metering
Contrast detection autofocus (CDAF) processing, liveviewMostlyYes
Video recording and microphoneMostlyYes
Electronic eye-level viewfinder (EVF)Only Sony SLT camerasSome

Mirrorless cameras use CDAF for focusing, and liveview for viewfinding, rather than the optical path through the mirror and pentaprism, in the DSLR cameras. However, DSLR cameras are expected to do the same nowadays, so it's not like this is exclusive to mirrorless cameras. The same goes for video recording: This is also expected from DSLR cameras.

Some mirrorless cameras add an EVF, an eye-level electronic viewfinder. Examples include the Panasonic G and GH series, the Olympus OM-D, the Nikon V1 and V2, the Sony NEX 6 and 7. This technical feature is generally not found in DSLR cameras, with the exception of the Sony SLT family of cameras, which use a fixed, semi-transparent mirror, and replaces the pentaprism viewfinder with an EVF.

Hence, we see that mirrorless cameras lack some technical features found in DSLR cameras, but they don't add any new. For that reason, they are simpler, and contain less components. And they should be cheaper to produce and require less maintenance.

The latter is not to be ignored. Since the digital revolution, you can find old second hand SLR and rangefinder cameras available at low prices. However, while they may appear to operate flawlessly, they may require very expensive cleaning and lubrication to operate safely over some period.

With less mechanical components inside the mirrorless cameras, they can be expected to require less maintenance. Some time in the future, they will come with "global electronic shutters", and can drop the mechanical shutter as well, further simplifying the design.

So why are mirrorless cameras not cheaper to buy, since they are simpler? Interviews with industry insiders say that the DSLR mirror box is "mature technology". In business-speak, this means that they are very cheap to produce. They don't add much to the production cost.

On the other hand, DSLR cameras have been around a long time, people know what to expect from them. Mirrorless cameras, on the other hand, are more of an unknown, and for that reason, there has been a smaller market for them. Add to this the research and development cost for the new mirrorless technology, and you have the reason why mirrorless cameras are still not cheaper than DSLR cameras.

In the future, though, as mirrorless cameras grow in volume, they should be cheaper and require less maintenance, which is good for us consumers.

One big problem with mirrorless cameras so far, is that they generally have an appalling continuous autofocus (AF-C) performance, as compared with premium DSLR cameras. The Nikon 1 series of cameras try to fix this by adding PDAF sensors on the imaging sensor. However, I don't think this makes much of a difference yet.

In the mean time, people who need a very good AF-C performance, e.g., for photography of sports or birds, must still stick with DSLR camers.

Thursday, 20 December 2012

GH3 guide to video modes

The Panasonic GH3 comes with a large choice of video modes. In this article, I try to make some sense of the choices. This writeup is not intended for experienced video users, since most likely you already know what modes you prefer.

PAL and NTSC versions


The camera comes in two versions. One for the PAL market, and one for NTSC. The PAL market includes Europe, China, Africa, Australia and New Zealand, while the NTSC market includes the Americas and Japan.

The difference is the set of framerates the cameras can capture in their video modes. PAL cameras can do 25fps and 50fps, and NTSC cameras can do 30fps and 60fps. In addition, both camera models can do 24fps.

The difference in framerates is to match the different TV broadcasting standards. Today, this seems like a strange difference. Not that many people use TV sets anymore, and besides, many modern TV sets can handle both PAL and NTSC input.

If you have a PAL version of the camera, you are stuck with the 24fps, 25fps and 50fps framerate options. And if you have an NTSC camera, you can choose between 24fps, 30fps and 60fps.

There is also a 30 minute video recording limit in some markets, due to tax reasons.

24fps


The 24 frames per second option is useful if you are shooting video for use in motion picture movies. These have used 24fps as a standard for years, while 25/50 (PAL) and 30/60 (NTSC) are more useful for video and television use.

Some newer movie productions even double the framerate to 48fps, for example the movie "The Hobbit".

I would use 24fps only if you know that you are going to be shooting for motion picture/cinema use. Otherwise, use 25/50 (PAL version) or 30/60 (NTSC version).

Video formats


The GH3 gives you the choice of three video formats: MP4, AVCHD and MOV.

MP4 This mode gives you a single video file with the .mp4 extension. In this mode, you get the most compressed files, i.e., the smallest video files. This is good for uploading videos to YouTube, for example, when you are not going to be editing the videos.

AVCHD In this mode, the camera creates a directory catalogue structure with your video files inside. The actual video files have a .mts extension. In terms of compression, this mode sits in the middle, having a compression level comparable to the predecessor GH2.

MOV For the very best video quality, use this mode. Each video capture gives you a single .mov file, and with a high frame rate, the files can become rather huge.

Using the MOV format requires that you have a fast SD memory card. A card with the description "Sandisk Extreme 45MB/s U1 Class 10" does work, also with the 72Mbps All-Intra mode. A card called "Sandisk Extreme 30MB/s Class 10" did not work with the MOV format at all, not even with the 50Mbps mode. That is my experience, anyway.

Interlace and progressive


The interlace video mode originates from older, analogue TV transmission technology, and gives the appearance of a quicker frame refresh rate by updating odd and even numbered rows in every second frame.

This is a rather obsolete technology by now, and I would advice against using interlace video with the GH3. Use it only if you positively know that you need interlaced video, for some reason. The interlace video modes are designated with a lower case i, like here:


Rather, use the progressive mode, designated by the lower case p:


All Intra


The GH3 adds a totally new option, to use the All Intra compression scheme in the MOV video mode. This is the highest bitrate option available, with a staggering 72Mbps output, about three times the maximum of the GH2.

What's special about this mode, is that it does not employ any between frame compression, only inside each frame, hence the name All Intra. It is designated with a capital I:


For the ultimate video quality, should you use this mode? It may be useful if you need a strong consistency between each frame, for example for green screen video shooting. Otherwise, I think it is largely overkill to use this mode. The massive files will clog up your hard drive, with a fairly marginal quality improvement.

Also, there is evidence to support that the normal 50Mbps mode has better video quality, despite having a lower bitrate. The reason is that the All Intra mode is less efficient, it requires a lot higher bitrate to be similar in quality.

So the bottom line on All Intra is: Only use if you speficially need consistence between each frame. Otherwise, the 50Mbps mode is the best for most uses.

720p and 1080p


There are two choices of resolution: High Definition (HD), 1280x720 and Full HD (FHD), 1920x1080. Given this choice, most people would want to use the largest resolution. After all, you can always scale down from FHD to HD, but you cannot scale up from HD to FHD.

Or can you? Some would argue that with the 50/60 fps All Intra (capital I) mode in 720p mode, you could scale the footage up to 1080p, and it would still be very sharp.

Most of the time, though, I would go for FHD 1080p.

Summary


The table below sums up the major video modes available, and the maximum bitrate

HD 1280x720FHD 1920x1080
MP4 25p/30p10Mbps20Mbps
AVCHD 50p/60p28Mbps
AVCHD 50i/60i17/13Mbps24/17/13Mbps
AVCHD 24p24/17Mbps
MOV 24p/25p/30p/50p/60p50Mbps50Mbps
MOV 24p/25p/30p All Intra72Mbps72Mbps
MOV 50p/60p All Intra72Mbps

As you see, the MOV video format gives the highest bitrates, with the All Intra mode at the top with 72Mbps. The All Intra mode is not available for FHD 1080 @ 50fps or 60fps.

Conclusion


If you want a good video quality, for uploading to YouTube and other video sharing services, you can safely use the MP4 format, for reasonably sized video output files. For the very best video quality, use the MOV format.

Personally, I would use FHD (1080 rows), and select 25/30 frames per second when the light is dull, and 50/60 frames per second in bright light. The lower frame rate allows for using a slower shutter speed, hence it is more usable indoors and in other low light situations.

In my experience, the continuous autofocus during video recording works best in the high frame rate, i.e., 50fps or 60fps, depending on if you have a PAL or NTSC version.

On the other hand, note that the high fps modes (50/60) will not work with the Extended Tele Conversion (ETC) mode. ETC is useful for extending the reach of your lens, while retaining the resolution. For this mode to work, you must set a lower frame rate, 24, 25 or 30 fps.

Starting video recording


On a related note, there are two ways to start the video recording. The first is to use the dedicated red video button on the rear. The second is to set the dial mode to "Creative Movie Mode", and then use the shutter button.





These two are quite different, so lets look more closely at them:

1. The red video button. When using the dedicated video button, the camera ignores your mode dial settings, and does it's best to give you a sensible exposure. This usually involves using the lens wide open, unless you have a fast or wide lens, and it is very bright.

This is usually all well, but the camera will often set a very fast shutter speed. When recording videos, the ideal is often to use a 180° shutter, meaning that the shutter speed is twice as fast as the frames per second rating. So if you have a European camera used at 25 fps, you will want to set the shutter speed to 1/50s. The reason why 180° shutter is often used, is that moving objects will have some motion blurring, making the video look more fluid.

This is impossible to control when using the red video button, and, in fact, the camera does not say what shutter speed it uses. To control this, you need to use the Creative Movie Mode.

One advantage when starting video recordings using the red button, is that you can press the shutter during video recordings, to take still images. If you select 4MP picture size, this is done seamlessly, and there are no gaps in the video when taking still images. Using larger still images, though, there will be a small gap in the video stream when you take a still image.

2. Using the Creative Movie Mode. In this mode, you need to use the shutter button to start and stop video recording. Hence, it is not possible to take still images during video capture.

The advantage, though, is that you can set the exposure mode to P, A, S and M, just like when taking still images. In the M mode, you can set the aperture, shutter speed, and ISO settings manually for full control. To change the exposure mode setting, you can touch the LCD display in the top left corner, in the movie camera icon.

Thursday, 13 December 2012

Lens cap tip

Front lens caps typically come in two categories: The pinch side type, and the pinch centre type:


The left cap is the slimmer, and came with one of the pancake lenses. The right cap is the pinch centre type, which I prefer.

The reason is that it is easier to remove and insert when you have a hood mounted, as illustrated with the Lumix X 45-175mm lens:


In this case, inside the hood, the slimmer pinch side type cap would be impossible to remove when you want to take pictures.

However, I often find that the pinch centre type hood grips are a bit slippery still. Especially in cold weather, it can be hard to get a grip on the levers that loosen the hood. To improve the usability, I tend to glue some small plastic parts on them, to increase the friction. Here is an example:


I took the yellow plastic parts from the locking ring of a soda bottle cap, however, any plastic part that has some texture can be used. Adding these strips of plastic makes it easier to feel where to grip the cap, and easier to hold it. The yellow colour also makes it quicker to spot the grips in dim light.

Why the manufacturers don't add some grip surfaces on the lens caps is beyond me: That would surely improve the ergonomics out of the box.

Saturday, 8 December 2012

GH3 video example @ ISO 6400

The Panasonic GH3 camera improves upon the GH2 in many ways. One improvement is that you can record videos at ISO 6400 sensitivity, while the GH2 only went up to ISO 3200.

This is quite useful. Some months ago, I was recording fireworks, and found that even with ISO 3200, I had to set the shutter speed to 1/12s, and record at only 12 frames per second. With the GH3, I could have set ISO 6400, and used 25 frames per second, a very common FPS setting.

I have tested the ISO 6400 setting at a dimly lit concert. I used the Lumix X 12-35mm f/2.8 lens at 35mm f/2.8. The video mode was the 1080p, 50 fps, and I set the shutter speed to the slowest possible, 1/50s. Here is the result:



In this case, I could probably have gotten better results by using 25 fps, allowing for 1/25s shutter speed, and reducing the ISO to 3200. But here I wanted to test the ISO 6400 setting.

Note that the lightning at the concert was very difficult, with a lot of red coloured lights. With better lightning, the video outcome would have been vastly better.

Here is an extracted frame from the video. I used Kdenlive in Linux to edit the video:


The video mode I used is denoted "MOV FHD 50p" in the camera, and gives 1920x1080 image frames at 50fps. On an NTSC camera, the fps would be 60. This mode gives you 50Mbps bitrate, about twice as much as the highest bitrate the Panasonic GH2 could produce.

For the ultimate in image quality per frame, you could choose the "MOV I FHD 25p" mode, which encodes the video as ALL-INTRA, with a bitrate of 72Mbps. Again, on an NTSC camera, this would be 30p, not 25p.

GH1, GH2 and GH3 @ ISO 200

Whenever a new camera is launched, there is usually a lot of discussions about what the true ISO values of the camera is. Is the ISO rating conservative, meaning that the true sensitivity of the camera is higher than the rated sensitivity? Or are the manufacturer cheating by stating that the sensitivity is ISO 200, for example, when in fact it is ISO 100?

To try to shed some light on this question regarding the Panasonic GH series, I have taken a picture of the same scene at the same time of day with the three cameras. I used the same lens, the Lumix G 20mm f/1.7, and I set the aperture to f/1.7. The other settings were: ISO 200, 1/250s, sunny white balance. Here are the out of camera JPEGs:

GH1:



GH2:



GH3:



When isolating only the wall part of the image, and looking at the histograms, it reveals that the ISO ratings are probably a tad bit different:


This indicates that at ISO 200, the Panasonic GH1 appears to be a bit more sensitive, while the GH2 and GH3 appear to be fairly comparable. This is consistent with the DXO ratings, which show the GH1 to be more sensitive than the GH2.

Note that the above results are based on the out of camera JPEG images. I have also developed the RAW images using Lightroom 4.3 RC with no exposure compensation, and gotten these histograms:


These results indicate a similar conclusion as the JPEG tests: The GH1 has the higher sensitivity at ISO 200, while the GH2 and GH3 are fairly similar.

Using the RawDigger program, I extracted the average exposure from the wall area of the images, again from the RAW images. The results are here:

CameraGH1GH2GH3GH3(*)
Red215153267123
Green618427470326
Blue430287346202
Green2612422471327

The current version of RawDigger does not set any default black level offset for the GH3, since it is a new camera. For the GH1 and GH2, it sets a black level offset of 15. Some say that the Panasonic GH3 should have a black level offset of 144, in which case you get the rightmost values in the table, denoted by "GH3(*)". The other GH3 column is taken without any black level offset.

For comparison, here are the images converted from the RAW files using Lightroom:

GH1:



GH2:



GH3:


Friday, 7 December 2012

Bug in GH3 JPEG conversion

While testing the Panasonic GH3 camera, I took a series of nine identical pictures using the Lumix X 12-35mm f/2.8 lens in the high speed drive mode, with electronic shutter. They all came out fine, except for one single image, which came out without the pincushion in-camera geometric distortion correction. It also has some corrupted parts in the upper part of the image.

Here is one picture which came out ok, and the one out of nine which came out wrong:



I opened the JPEG file in several image processing programs to verify that the mistake was in the file, not in the editing client.

The RAW image file still came out fine, so this is not a big issue. I wouldn't recommend to avoid the camera because of this finding. It is probably normal that new cameras need some firmware upgrade to behave better. My camera has the original 1.0 firmware.

I later tried to take another long series of images, and they all came out just fine.

To verify that the corrupted image indeed has pincushion distortion, I have superimposed the edges from the JPEG image (in red) on the edges from the RAW image (in blue):


We see that the JPEG image has pincushion distortion, while the image based on the RAW file has the correct rectilinear appearance. If the camera worked well, both images should have the same geometric properties.