GH1 vs GH2: Rolling shutter evaluation

Rolling shutter is the name of a type of mechanical or electronical shutter mechanism. In this type of shutter, the whole film or sensor is not exposed exactly at once, but rather, a shutter rolls across the frame and exposes it bit for bit. Rolling shutter is also the name of the distortion associated with this shutter implementation.

You can see this distortion for example when panning heavily during video recording. It looks like the scenery is "leaning" towards one side when panning. I tried the Samsung NX10, and immediately noticed that the viewfinder had much more rolling shutter artifacts than the Panasonic G-series.

This shutter implementation also leads to distortion to rotating elements. To test which camera has the most rolling shutter artifacts out of the Panasonic GH1 and GH2, I made a simple LEGO contraption which rotates a propeller at a constant speed. Then I videofilmed this with both cameras, using the Olympus 50mm f/2 lens.

Here are the two video streams:

GH1, ISO 1600, f/2, 1/500 second, 25p, 1080

GH2, ISO 3200, f/2, 1/1000 second, 24p, 1080

To more easily compare the rolling shutter artifacts, I have made similar framegrabs from both:

What we see here, is that the distortion is slightly smaller in the GH2 video stream. The shorter the distance between the two prongs to the left, the more the distortion.

So my conclusion is that the GH2 handles rolling shutter at least as good as the GH1.

For real life usage, rolling shutter is not a problem with GH1 or GH2 video. You can generate these effects by filming a rotating propeller, like I did here. Or by panning heavily. But most types of video footage will not display any noticeable rolling shutter artifacts.

As a side note, the amount of rolling shutter artifacts depend on the speed of the rolling shutter (in a mechanical implementation), or the speed of the sequential image data readout for a digital shutter. It does not depend on the shutter speed itself. Here is an illustration, where you can see three different shutter speeds generating the same amount of rolling shutter artifacts. But the amount of motion blurring is of course different.

Another term widely used is global shutter. This refers to a system in which the exposure values from the sensor are read all at once. Since the values are not read out sequentially, there are no rolling shutter artifacts with a perfect global shutter.

Before launching the GH2, a Panasonic representative was quoted saying that implementing a global shutter in Micro Four Thirds cameras is not coming soon: At the very earliest with the GH3. In retrospect, the the Panasonic GH3 did not introduce global shutter. And now, it looks like the GH4 (aka GH4K) is going to be more about 4K video than global shutter.

Comparison @ 14mm

There's a number of Micro Four Thirds lenses which include the 14mm focal length. So it's natural to compare their properties at this focal length. I've compared these:

Lumix G 14mm f/2.5 pancake

Lumix G 14-42mm f/3.5-5.6 Mega O.I.S.

Lumix G HD 14-140mm f/4-5.8 Mega O.I.S.

I did the comparison by taking the same photo with all three lenses, at various apertures. The pictures were taken with the Panasonic Lumix GH2 camera on a tripod, at ISO 200. I used the self timer to avoid camera shake during the exposure, and disabled image stabilization. The pictures below were rescaled and sharpened. Here they are:

Video quality, hacked GH1 compared with GH2

I've tried to test the GH1 (hacked) and the GH2 under the same conditions. To make sure the conditions were the same in both tests, I made a rotating platform using Technic LEGO, and left it under artificial lights. To better be able to judge the quality, I put a sock on the platform. The sock has some texture, which will come in handy in the comparisons.

The lens used was the Panasonic Leica 45mm f/2.8 lens. The focus was put on the centre of the image, and I used f/6.3, ISO 400 for some depth of field.

Both videos are filmed in 1920x1080 pixel resolution. The GH1 is hacked to a maximum of 44Mbit max bandwidth, while the GH2 has the native firmware. The GH1 video stream is 25p, the GH2 is 24p. I used the new Cinema 24p mode on the GH2.

The GH1 stream is 11 seconds, 38.430.720 bytes, 3.5MB/s.

The GH2 stream is 10 seconds, 27.998.208 bytes, 2.8MB/s.

So there is more information in the hacked GH1 stream. However, this does not necessarily mean that it's better.

Here's the GH1 (hacked):



And the GH2:



It's not easy to make an objective comparison based on these YouTube videos. So to be more fair, I have made frame grabs from the video streams.

GH1 (hacked):

GH2:

To make the comparison even better, here are 100% views of two similar framegrabs (click for a larger version):

Conclusion

First of all, it is easy to conclude that the colour balance is different. I have the same saturation settings in both cameras (+1), and the lightning was the same. So the difference must be due to different implementation of auto white balance. Overall, I feel that the GH2 colours look more natural.  It's not strange that the cameras struggle with the colour balance here.  After all, there is artificial light, and the subject colours are rather strange.

When it comes to the sharpness, I think the conclusion is that the GH2 has better image quality. Even though the bandwidth is a bit smaller than in the hacked GH1.

I have also made a comparison of the GH1 video quality before and after the hack, in which case I found that the video quality was slightly better after the hack.

GH2 touch screen AF (superzoom example)

I've previously tested the touch screen AF of the GH2, using the 45mm macro lens.

Here's another example using the touch screen AF with the Lumix G HD 14-140mm superzoom lens. The lens was set to 100mm focal length.



As you can see, the response is still somewhat slow, but the actual focus action of the lens is faster than when using the Lumix Leica 45mm f/2.8 macro lens.

I'm using the A mode on the mode dial. Using the iA mode typically gives me the focus tracking mode, in which the response time from touching the screen until the focus changes is a bit shorter.

When using the shutter button to trigger focus, the focus is very responsive, just like in my study of the GH2 autofocus speed.

When recording videos, you can change the focus areas using the touch screen as well. The response and speed is similar to this example.

I have also made an analysis of the focus speed of the GH2 for a number of lenses, and compared with GH1 for some of them.

GH2 touch screen AF (macro example)

One of the new features of the G2/GH2, is touch screen LCD. Some fans have been negative about this feature, since they are afraid of losing the hard buttons on the camera body. It turns out that the touch LCD is an additional feature, and that it does not replace any hard button functions. So those who dislike a touch screen can ignore it, without losing any vital functionality.

One application where the touch screen makes some sense, is macro photo. It is rather quick to move the focus point around in the frame. Here is an example usage, with the Panasonic Leica 45mm f/2.8 lens:



You can doubleclick on the video to be able to view it in other sizes.

As you see, the response is not very quick. It takes a moment from changing the focus point until the camera actually starts changing the focus.

I'm using the A mode on the mode dial. Using the iA mode typically gives me the focus tracking mode, in which the response from touching until the focus changes is a bit faster.  So if you want the focus response to be quicker, set the dial to iA.

In the video, I am sliding my finger over the screen now and then.  This is not necessary. You can just point where you want the focus point.  My sliding finger action is probably something I've learned from using an Iphone.

This can also be done during video recording. The response speed is the same, as far as I can tell, during video recording.

Of course, the PL45 lens is not the quickest around in terms of autofocus. Using a quicker lens will make the response quicker as well.

I have also made an analysis of the focus speed of the GH2 for a number of lenses, and compared with GH1 for some of them.

Panasonic GH2 autofocus speed

Update: Since this test, I have compared the GH1 and GH2 head to head under exactly the same conditions.

I have tested the autofocus speed for the Panasonic Lumix GH2 with various lenses. The test was done by turning on the camera (which leaves the lens at near infinity focus), and then pressing the shutter to see how long time it takes to reach focus. This was filmed using a GH1, and I played back the video in a video editing software, to read out the timings.

There is fairly dull indoor lightning. The lightning corresponds to around EV 6, the same as in my previous test of the GH1. A LEGO figures is placed in the middle of the camera frame, at around 45cm distance. The focus mode used is centre spot.

Here is an example test. In this test, I have the Olympus 9-18mm f/4-5.6 Four Thirds lens mounted to the camera, using an adapter. (Not to be mistaken with the Micro Four Thirds version of the lens.)



This combination gives a rather slow autofocus.

Another example using the Lumix G 14mm f/2.5 pancake:



As you see, the Lumix 14mm lens focuses very fast.

Another example, using the Leica Lumix DG 45mm macro:



This lens does not focus as fast as some of the other lenses, but it's still fair to say that the focus is not slow.

Summary

Here are the timings. I have included a column with the similar autofocus times for the GH1. You should note that the GH1 timings were done with early camera and lens firmware. Later firmware have improved the AF speed.

Lens	Focus time GH1	Focus time GH2
Lumix G 8mm fisheye	Not tested	0.24 seconds
Lumix G 14mm	Not tested	0.20 seconds
Lumix G 20mm	1.23 seconds	0.44 seconds
Leica Lumix DG 45mm macro	Not tested	0.60 seconds
Lumix G 14-42 @ 14mm	Not tested	0.28 seconds
Lumix G 14-42 @ 25mm	Not tested	0.20 seconds
Lumix G 14-42 @ 42mm	Not tested	0.44 seconds
Lumix G HD 14-140 @ 18mm	0.53 seconds	0.16 seconds
Lumix G HD 14-140 @ 50mm	0.40 seconds	0.40 seconds
Lumix G HD 14-140 @ 140mm	0.68 seconds	Not tested
Lumix G 45-200 @ 45mm	0.33 seconds	Not tested
Lumix G 45-200 @ 100mm	0.36 seconds	Not tested
Lumix G 45-200 @ 200mm	0.87 seconds	Not tested
Olympus ZD 4/3 9-18 @ 9mm	2.90 seconds	1.44 seconds
Olympus ZD 4/3 9-18 @ 18mm	1.50 seconds	1.36 seconds
Olympus ZD 4/3 50mm f/2	No AF	4.76 seconds

Conclusions

What we can see, is that the GH2 autofocus speed is very good. The Four Thirds lenses used on adapter are still slow, but they behave better. The Olympus 50mm f/2 macro lens can actually autofocus on the GH2, but the speed is very, very slow. It could be used for stationary objects, but for photographing people, the AF is more or less useless.

The Panasonic Lumix GH1, and the whole first generation of Panasonic G cameras, could not autofocus with the Olympus Four Thirds 50mm f/2 macro lens. This also applies to a host of other Four Thirds lenses that are not optimized for CDAF. The newer GH2 can autofocus with most Four Thirds lenses. However, the focus can be rather slow, as we have seen in this example.

Even the Four Thirds Olympus 9-18mm f/4-5.6 wide angle zoom lens, a fairly recent CDAF optimized lens, has an annoyingly slow autofocus. The AF is usable, but not as fast as we have become used to.

I am also rather happy that the autofocus speed for the Lumix 20mm f/1.7 pancake lens has improved a lot with the GH2. After all, this is a very good lens, and I find the AF speed to be important when photographing people.

As expected, the Lumix G HD 14-140mm superzoom lens remains very fast in terms of AF. It seems that the close focusing range differs for various focal lengths. The close focus range is specified at 50cm, however, it can focus a bit closer in the wide end of the zoom range. I was not able to test the lens in the longer range this time, since it would not reach focus there at 45cm distance.

The new kit zoom Lumix G 14-42mm is also a very competent lens when it comes to focus speed. I think this was to be expected, after all, it is a brand new kit lens, and focus speed is one of the important factors when people buy camera kits.

Lens hoods

Frequent readers of my blog will know that I like to tinker with lens hoods. Here is a summary of some of the lens hood solutions I like.

Panasonic Lumix G 20mm f/1.7 pancake

This lens does not come with a hood at all. While a hood is probably not much needed from a stray light perspective, I still like to put a hood on my lenses for protection against objects accidentally touching the front lens element.

My favourite solution is to put a 46mm to 37mm step down ring as hood. This gives some basic protection of the front lens element, while keeping the overall package compact.

I've checked that there is no extra vignetting caused by the step-down ring.

In addition, you'll need a 37mm front lens cap. Both can be gotten from various auction sites for about US$10 in total.


This lens features a traditional focus mechanism, in which the whole lens assembly moves back and forth. Adding extra weight to this assembly is generally not a good idea. However, the step-down ring doesn't add that much weight. Less than a glass filter would, anyway.

Another solution is to get a 46mm metal screw in hood designed for Leica Summicron:

While this hood looks stylish, I think it adds too much bulk. Besides, you might not be able to fit the front lens cap inside it.



Panasonic Leica Lumix DG 45mm f/2.8 1:1 macro

This lens does come with a hood, however, I am sad to say that I find the hood completely hideous:

This hood does look stylish, and reminds me of the older Leica hoods. However, it is much too wide, and could have been used for a wide angle lens. But this lens is a short tele. So the hood does not do a good job of keeping stray light out.

I prefer a hood that is a narrow as possible, while still not inducing any extra vignetting. I found the solution I like by adding three extra elements to the front lens thread: First a 46mm stand off ring (glassless filter), then a 46mm-37mm step-down ring, and finally, a 37mm-28mm step-down ring:

I was almost a bit surprised when I verified that this combination does not add any extra vignetting. But the front lens element of the 45mm macro has a rather narrow diameter, so I suppose it makes sense.

This "hood" does not add any extra diameter to the lens, which is good. In addition, you'll need a 28mm front lens cap. Such a cap is pretty inexpensive.


Olympus M.Zuiko Digital 45mm f/1.8

In a move which has annoyed quite some fans, this lens is not sold with a hood included. There is a hood you can buy from Olympus, which fits in to the bayonet mount. However, it is quite expensive.

I bought a collapsible rubber hood for screwing into the 37mm front threads. It works great:

This rubber hood cost me US$13, including shipment from China.



Panasonic Lumix G 14mm f/2.5 pancake

Just like the 20mm pancake, this lens does not come with a hood. Again, I've used the same solution as with the 20mm pancake lens: A 46mm to 37mm step down ring as hood:

In this case, I had bought a grey ring by a mistake, so I had to paint it black with enamel paint. No big deal.


This lens features internal focusing, so I am not afraid of putting some extra stuff onto the front lens thread.

Panasonic Lumix G 45-200mm f/4-5.6 Mega O.I.S.

This lens does come with a hood, and I think the hood is very well designed:

The only problem I can see, is that it is awkward to insert and remove the front lens cap.

I still changed this hood, since it was a tad bit too long for my camera bag. I got a 52mm screw in metal hood:

This alternative hood is probably not as good for protection, but is is more practical to use, and it's easier to add the front lens cap after use.



Other lenses

The kit zoom lenses generally have well designed hoods, and I see no reason to replace them.

Testing the effect of the GH1 hack

For a while now, it has been a well known fact that the Panasonic Lumix GH1 firmware can be adjusted to yield higher video bitrate, among other things. This is referred to as using "the hack", or "hacking" the GH1. Similar adjustments can also be done to the GF1 and G2 cameras.

I wanted to test the effect of the hacked GH1. Is it possible to measure the improvement of the video after installing the hack? I know many people have praised the quality of the hacked GH1 video before, but I wanted to make sure this test has exactly the same moving subject and lightning before and after installing the new firmware.

Test setup

Here is my test setup:

The camera is on a tripod, with the Leica Lumix 45mm f/2.8 macro lens. The camera is set to f/5, 1/60s, ISO400. The LEGO figure is on a platform that rotates at a constant speed.

I recorded one video file before installing the hack, and one after. When adjusting the firmware, I upped the max bandwidth from 16,000,000 to 50,000,000. Both videos were recorded at Full HD, 1920x1080 pixels. This can only be done using the AVCHD format. My camera is the PAL version, meaning that I get a framerate of 25fps.

The first video is 11s, and 20,754,432 bytes, 1.89MB/s. The second video is also 11s, and 31,395,840 bytes, 2.85MB/s, 50% higher than before the hack. So already here there is an indication that the hacked video contains a higher bandwidth.

Video examples

Here are the videos. You can double-click on the video to get to the YouTube page, where you can see them in higher resolution, up to 1080p.

Before the hack:



After the hack:



Of course, just watching the videos on YouTube isn't enough to judge the quality. For an extra test, I have made some framegrabs from the videos, to compare the quality.

First, two full size framegrabs from the video.

Before:

And after:

It's still difficult to compare the quality of the images. Here are 100% crops from the 1920x1080 framegrabs. I chose four consecutive frames from each video stream. Click on the image for a larger version.

Conclusion

Based on these screenshots, I believe it looks like the there is some more clarity and sharpness in the hacked video stream. Also, the colours are more washed out without the hack. The hacked video stream is better. But from my experiment, I don't see a dramatic difference.

Other effects

After recording the second video, I was not able to playback the videos any more. Just showing the thumbnail in the viewer was enough to freeze the camera. To reset it, I needed to remove the battery. Just flipping the on/off switch was not enough. This is, of course, quite annoying.

Lumix 14mm, insufficient distortion correction

Just like a host of other Micro Four Thirds lenses, the Lumix G 14mm f/2.5 pancake lens utilizes in camera distortion correction.

However, when using the lens on the Panasonic GH1 camera, I've noticed that the distortion correction is not sufficient. There is still some residual barrel distortion after the in camera correction.

Here are a couple of examples. First, let's look at an example at a close focus distance, close to the minimum focus distance of the lens:

Corrected JPEG	Uncorrected RAW

In this example, even when looking at the small image (above, left), it is easy to see that there is still some barrel distortion in the upper, horizontal line. The line is not straight. You can click on the image to view a larger version of it.  The corrected image to the left is the JPEG output from the camera.

The right image shows the image as captured by the sensor, without any distortion correction at all.  I used the UFRaw RAW processing software.  But any software that allows for disabling the distortion correction could be used for this purpose.

And one example with a longer focus distance:

Corrected JPEG	Uncorrected RAW

In this latest example with a longer focus distance, it appears that the remaining distortion is not so significant. I added a red, straight guideline to the left in the JPEG image, above left. You can see that there is some barrel distortion still, but not much.

Conclusion

It appears that when using the Lumix G 14mm f/2.5 pancake lens on a Panasonic GH1 camera, there is some residual barrel distortion. This is most apparent at shorter focus distances.

One could speculate why this is so. I don't think it is a firmware issue, since I have the newest firmware available for my camera.

Another speculation is that Panasonic chose to not correct all the barrel distortion, since that would have decreased the diagonal angle of view. The diagonal angle of view for this lens is specified to 75°. Further correction of the barrel distortion would reduce this figure, albeit with a small margin.

The Lumix G 14mm f/2.5 pancake lens uses internal focusing, and it is not uncommon that these designs lead to changes to the distortion properties, and field of view, at closer focus distances.