GH2, does ETC affect the video quality?

ETC, or Extended Tele Converter, is an interesting feature with the Panasonic GH2. I've previously looked at how it can be used to get closer video recording of the moon using the Lumix G 45-200mm lens.

Essentially, ETC is a digital zoom for video. During normal video recording, the entire sensor area is scaled down to 1920x1080 pixels. When enabling ETC, the camera crops the 1920x1080 pixels from the centre of the sensor. This achieves full HD resolution, as well as a digital zoom. See the illustration:

Since 2800/1080 = 2.6, the ETC mode gives a tele conversion effect of 2.6x. If you use the 720p video mode, the conversion effect becomes larger still, 2800/720 = 3.9.

A question that remains is to see if the video quality is still good using the crop mode. In theory, it could be somewhat worse, since the camera doesn't have the option of scaling down from a larger number of pixels.

On the other hand, we know that the camera doesn't actually sample all the 16MP of sensor pixels for each frame in the video.  That would require too much bandwidth and processing power.  Just how many are sampled is unknown, but given that the electronic shutter 40fps continous mode has 4MP, one can guess that about that many are sampled during video recording.  This is, of course, just speculation.

To study this, I have used the kit zoom lens, Lumix G 14-42mm f/3.5-5.6. Since it has a zoom ratio of 3x, using ETC in the short end should give the same field of view as the longer end without ETC. This is a good basis for comparing the video with and without ETC.

I used 40mm without ETC, and 15mm with ETC. Since 15mm times 2.6 is approximately 40mm, these two modes give about the same field of view. I recorded video using ISO 160, 640 and 2500. Here are the recordings in sequence:



To better evaluate the quality differences, I have compared still images from the video streams (click for a larger version):

It appears that the video stream recorded using ETC has worse image quality than the normal stream, at all ISO values. There is some lack of sharpness and contrast, and more noise in the ETC video streams.

The reasons for these differences could be the sampling from a higher number of pixels when not using the ETC mode, as discussed above. Or it could also be related to the lens sharpness, see the discussion in the Appendix, below.

This does not mean the the ETC videos are useless, far from it. But it is fair to say that there is some image degradation when using ETC. As a rule of thumb, I think you should avoid using ISO higher than 640 when using ETC.

ETC is still a very good tool to use when you need extra reach during video recording. If you don't have any longer lenses, ETC is the only way to get quality videos at longer reach.

Appendix, some technical aspects

I used the new 1080p, 24fps cinema mode. This is the video mode on the Panasonic GH2 which gives the highest bitrate. It is generally recommended to use this 24ftp progressive mode, except when you have significant movement in the picture frame. When there is movement, people generally recommend to use the 60fps (NTSC) or 50fps (PAL) modes.

When the lens was set to 15mm, I stopped down the aperture to f/4. In ETC mode, the camera only uses the centre of the image, where lenses are generally very sharp. I've previously looked at the sharpness of the Lumix G 14-42mm kit lens, and found it to be good wide open at 14mm. So I don't think the centre sharpness at 15mm f/4 is a significant limiting factor when using the ETC mode.

But there is probably some impact on the sharpness due to the lens performance when using ETC. Remember that the ETC mode is more demanding on the lens sharpness, since there is no downscaling. When video recording without ETC, the image is downscaled from a larger sensor area, and the lens doesn't need to be tack sharp. It is like putting the lens designed for a 16MP sensor on a 2MP sensor camera.  It's not very likely that you can identify a lack of sharpness with a 2MP sensor.

In real life use, it is of course silly to use ETC in the short end of the Lumix G 14-42mm f/3.5-5.6 kit zoom. It is better to just zoom in. But I did this to get comparable images with and without ETC.

It is much more relevant to use ETC with the longer end of the Lumix G 45-200mm tele zoom lens. However, many have concluded that it is not very sharp in the longer end, so it may not be sharp enough to get a good video quality with ETC. You may want to stop down the aperture to f/6.3 or f/7.1 for the best result.

Geometric distortion correction

Many Micro Four Thirds lenses feature in-camera geometric distortion correction. Two examples are the Panasonic pancakes, Lumix G 14mm f/2.5 and Lumix G 20mm f/1.7. Both have pretty similar distortion correction needs. When converting the sensor output to the out of camera JPEG image, around 10% of the pixels in the border area are lost.

This is somewhat controversial. Some feel that a quality lens should not require further software correction. In fact, the lack of geometric distortion is a traditional sign of a high quality lens.

I think that this is mostly a non-issue. By allowing some aspects of the image to be adjusted in software, the lens designers can focus on issues which cannot be corrected in post processing. This has the potential of making the lenses better, at a smaller size, and potentially a smaller cost. Panasonic Micro Four Thirds lenses are adjusted for geometric distortion and some chromatic aberrations. The geometric distortion is also corrected in Olympus Micro Four Thirds cameras. At this time, though, Olympus does not correct chromatic aberrations.

To illustrate the geometric distortion done with various lenses, I have photographed a tiled wall with them, and shown the sensor output compared with the corrected JPEG output.

Here is an example pair from the Lumix G 20mm f/1.7 pancake lens:

uncorrected RAW output	JPEG image

Note that this is in no way a criticism of using RAW images. There are many RAW image converters which will do the distortion correction automatically and seamlessly, and you will never notice that there was any geometric adjustment done at all. I am using the RAW images to visualize the initial image captured by the sensor, as it is the only way to access it.

Here is a comparison of the uncorrected and corrected images for some lenses.  Since I am only interested in the geometric distortion, I have increased the contrast so that the images become monochrome.  I also superimposed the corrected out of camera images (black) onto the original uncorrected images (red).

I have also included the appropriate adjustment needed. The adjustment numbers in percent refers to the "Lens Distortion" filter in The Gimp.

Lumix G 20mm f/1.7: -11%

Lumix G 14mm f/2.5: -16%

Lumix G 14-42mm f/3.5-5.6 @ 14mm: -18%

Lumix G 14-42mm f/3.5-5.6 @ 30mm: 0%

Lumix G 14-140mm f/4-5.8 @ 14mm: -17%

Lumix G 14-140mm f/4-5.8 @ 30mm: -4%

Lumix G 45-200mm f/4-5.6 @ 45mm: +1%

Conclusion

Normal zoom lenses pretty consistently feature barrel distortion in the wide end. The tele zoom Lumix G 45-200mm appears to have some very small pincushion distortion, but very minor.

Some lenses that do not feature any geometric distortion correction are the Lumix 8mm f/3.5 fisheye and Panasonic Leica Lumix DG Macro-Elmarit 45mm f/2.8 1:1 Macro lens.

Bokeh comparison @ 14mm, take two

I have previously looked at the bokeh of the Lumix G 14mm f/2.5 pancake and Lumix G 14-42mm f/3.5-5.6 lenses.

That comparison featured an image without any highlights in the out of focus background. This time around, I figured I would try a high contrast night exposure, with bright highlights out of focus. Again, I set ISO 160 on the Panasonic GH2. The camera was on a tripod, and the exposures lasted around 1-8 seconds.

Here are the full images from both lenses:

Lumix G 14mm @ f/2.5	Lumix G 14-42mm @ 14mm f/3.5

The focus was set on the post to the left, which is around 30cm from the camera, close to the minimum focus distance for these lenses. I used the A (aperture) exposure mode.

To evaluate the bokeh, let's enlarge some of the areas from the image.

From the centre of the image:

And from the top of the image:

(Click for larger images.)

I think the bokeh from both lenses is a bit "dirty" and "swirly". The 14mm pancake lens has the most ringing in the out of focus rendering.

Conclusion

My conclusion, which is a bit unexpected, is that the zoom lens has the best bokeh. Or perhaps it is better to say it has the least displeasing bokeh, since neither are very good. But the difference is small.

The 14mm pancake lens has non-circular bokeh even at the largest aperture, which is not so usual. The Lumix 20mm f/1.7 pancake lens also exhibits non-round bokeh. So perhaps it is related to the pancake design?

Keep in mind that these enlarged images are 100% views from the 16 megapixel GH2 sensor. most people will probably downscale the image some, in which case the dirtiness and ringing of the bokeh becomes a smaller issue.

Also, bokeh and wide angle is not that much of an issue anyway. To get some out of focus rendering at 14mm focal length, I had to focus close to the minimum distance, and look at objects in the far background. It is not so likely that you'll find the same situation in real life.

For example, if you're photographing people, you'll want to keep then at a distance of around 1 meter or more, to avoid perspective distortion. At this focus distance, you're unlikely to see much bokeh with a 14mm lens, even at f/2.5.

So even if the bokeh for these lenses at 14mm could have been better, you should not see it as a fatal problem.  It is, at worst, a minor annoyance for some types of images.

Using protective filters on lenses

A topic of much debate is whether or not to use clear glass filters on the front of lenses, for protection. Some would argue that putting on a clear filter protects the lens against objects touching it, and against water, dust, and so on. If something hits the front of the lens, they would say, it breaks the filter and not the lens itself.

Others argue that the front lens element typically is strong, and does not need extra protection. Some would even say that adding a filter may be negative, that it breaks more easily, and that the glass shards could add more to the damage of the front lens element.

Another argument is that an extra glass filter deteriorates the optical properties of the lens.

Now, I wouldn't want to drop my lens on the pavement to examine if using a filter helps protect it or not. Besides, as all lens accidents are different, one drop would not prove much anyway.

But to look at the resulting images when using a filter, as opposed to not using a filter, is possible. To examine the effect of using a clear glass protective filter, I took the same picture with and without a filter.

Example @ 200mm

Here is an example picture taken using the Panasonic GH2 with the Lumix G 45-200mm f/4-5.6 lens at 200mm f/5.6.

without filter	with filter

This is an image with fairly low contrast, and on first glance, the two images are pretty similar. Let's look at an enlargement:



So is there any difference? Not really, I would say. Perhaps one could say that there is a tad bit more contrast in the left image, without the filter.

Example @ 14mm

Another example taken using the Panasonic GH2 with the Lumix G 14mm f/2.5 lens at f/5.6.

without filter	with filter

In this case, we barely need to look at an enlargement to see the differences. Clearly, there is more flare when using a filter. The right image also has some odd "phantom lights" in the frame, from light reflections between the filter and the lens. Here is one example in the centre of the frame:



Beyond the flare, it's hard to argue that the right image is worse in terms of sharpness, for example.

Example @ 45mm

Finally an example picture taken using the Panasonic GH2 with the Lumix G 45-200mm f/4-5.6 lens at 45mm f/4.

without filter	with filter

In this case, it is very clear that the presence of the filter deteriorates the image. Not only is there much more flare, there is also a severe reduction in the contrast and sharpness across the whole frame.

Here is an enlargement which demonstrates this:



Conclusion

In low contrast situations, it is hard to find much evidence of negative impact from using a filter. However, when there is more contrast, e.g., a strong light source in front of the camera, the presence of a filter generates significantly more flare, and potentially also less contrast.

People who are proponents of using clear protective filters may still not agree with me. They could argue that if I had used higher quality multi coated filters, I would not have gotten these negative impacts. And perhaps they are right.

Since I don't believe in using filters for protection myself, I haven't invested in the most expensive filters. So it could be that higher quality filters would have reduced the deterioration.

Still, I believe it is fair to say that adding a filter will generate some deterioration of the image quality. But if using a protective filter gives you more peace of mind when using your expensive lens, perhaps that is an acceptable cost to you.

Personally, I prefer using hoods for basic protection of the front lens element.

Appendix

The filters used were: "Kenko Digital Filter UV" (52mm) for the Lumix G 45-200mm f/4-5.6 lens, and "Green Digital Filter UV" (46mm) for the Lumix G 14mm f/2.5 lens. Both filters are littered with terms like "High Quality" and "Premium Quality", but I would take that with a grain of salt.

Bokeh comparison @ 14mm

The Lumix G 14mm f/2.5 pancake and Lumix G 14-42mm f/3.5-5.6 are both rather new lenses. What they also have in common, is that they have received a mixed reception online.

The pancake lens has some disappointed over lack of sharpness, and not being as fast as the Lumix G 20mm f/1.7. While the comment about not being as fast is correct, I have found the 14mm to be very good optically.

The new kit zoom lens is said to be worse than the lens it replaces, the Lumix G 14-45mm f/3.5-5.6. While it is probably true that the old kit lens was better, I have found the new one to be good, too.

In this article, I have taken a look at how the lenses render out of focus parts of the image, the bokeh. I did this by taking the same image with both lenses. The lenses were focused on the checkered pattern in the lower right part of the image, which is about 30cm from the camera. (click for larger images):

Lumix G 14mm @ f/2.5	Lumix G 14-42mm @ 14mm f/3.5

The pictures were taken with ISO160, base ISO for the Panasonic GH2.

To better evaluate the images, here are 100% crops from the left, centre and top parts of the images, respectively. The images were taken with apertures ranging from f/2.5 to f/8:

(Click for larger images.)

Due to diffraction, one should avoid using apertures smaller than f/8, i.e., avoid larger aperture numbers. This is especially true if you are going to be studying enlargements from the images. If you plan on publishing the images in web size only, go ahead and use apertures as small as f/22. The diffraction will not be a problem if you are going to downscale the images that much.

Conclusion

I think this test does not reveal any bokeh problems with any of the lenses. Perhaps the test could have been made more challenging by having some strong highlights in the background.

The patterned textile in the first compilation image is in focus. We can see that it very sharp already from the smallest aperture. This appears to verify my initial claim that both lenses are in fact quite good, despite their rather bland reception.

The Lumix G 14mm f/2.5 pancake lens appears to be sharpest. This was to be expected, I think, since it is not a zoom, and hence has fewer optical compromises. Further, it also lacks an OIS lens group, and has a simpler construction with only six lens elements. The 14-42mm zoom lens has twice as many lens elements. Generally, more lens elements can lead to worse optical performance, for example reduced contrast. However, modern quality zoom lenses can retain superior optical performance with 15-20 lens elements.

TTL flash metering and flash delay

Flash metering has come a long way the recent decades. TTL flash metering for SLR cameras was first introduced by Olympus in the mid 1970's. TTL refers to Through The Lens. The camera measures the amount of light coming onto the film through the lens during the exposure, and cuts off the flash as the exposure is sufficient.

Film based SLR cameras

For film based SLR cameras, this is usually implemented by having a flash light meter in front of the film plane. The amount of light reflected off the film from the flash is metered, and the flash is turned off when there has been a sufficient amount of light for the desired exposure. See the illustration below.

Film based SLR camera with lens

In this illustration, the mirror is raised for exposing the film.

This generally worked well, at least as long as the subject was not too dark or too light, in which case you needed to manually adjust the flash exposure.

Digital SLR cameras (DSLR)

With digital cameras, this does not work well, since the imaging sensor, replacing the film, is not reflective enough. To overcome this problem, most DSLR cameras fire a pre-flash before raising the mirror, and then fire the main flash after exposing the sensor.

The pre-flash is used to determine the amount of flash needed for the exposure. With this method, the TTL flash meter is no longer needed, the camera's ordinary light meter is used. See the illustration.

DLR camera with lens

There are some DSLRs that still measure the amount of light reflected off the sensor chip, and avoid the pre-flash. The Fujifilm S1 and S3 does this.

Mirrorless cameras

As you know, Micro Four Thirds is a mirrorless camera system. The camera has no light sensor anymore. The imaging sensor is the light sensor. So to find the correct flash exposure, a pre-flash is triggered while the sensor is exposed. Then the camera must make the sensor ready for a second exposure, and fire off the flash with the correct amount of light. This typically takes a bit more time than with a DSLR. The DSLR used the separate light meter for the pre-flash, and could expose the main imaging sensor only once.

Here's a basic illustration of a mirrorless camera with lens. It is much simpler, since there is no mirror, pentaprism, or light meter.

Mirrorless camera with lens

Flash and pre-flash timings

To examine the pre-flash and main flash timings, I have video recorded operating the Panasonic Lumix GH1 and GH2 cameras. I also measured the Pentax K10D, which is an older DSLR from 2006. I turned off autofocus, to measure the flash delay only, and not also the autofocus delay.

I used 50fps when recording, which gives an accuracy of approximately 0.02s.

	GH1	GH2	K10D
Time to pre-flash	0.16s	0.12s	0.08s
Time to main-flash	0.12s	0.14s	0.12s
Total flash delay	0.28s	0.26s	0.20s

The first timing is the delay from pressing the shutter until the pre-flash is fired. The second is the delay from the pre-flash until the main-flash. The third figure is the sum of the two first: The total delay from pressing the shutter until the main flash is fired.

You can see the recordings here. I uploaded them as 25p videos, so they are not as good for verifying the actual timings.



Conclusion

One could say that the GH2 improves slightly on the GH1 in terms of flash delay. However, the difference, 0.02s is not significant with my way of measuring. So we can only say that they are comparable.

When comparing with the older Pentax K10D, we see that the GH1 and GH2 perform almost as good. The difference between 0.20s and 0.26s is not very large. Probably, the autofocus speed is more important to the average user, and we have seen that both the GH1 and GH2 perform very well in terms of autofocus.

One way to avoid the TTL pre-flash, is to use a flash in Auto mode.

Now, we should not conclude that the extra flash delay with Micro Four Thirds cameras is exclusively a bad thing.  As opposed to DSLRs, which have a limited number of light metering sensors, the Micro Four Thirds cameras essentially take one full picture to determine the correct exposure.  This means that the camera has at least 12 megapixels of information available.  It probably doesn't use all of this information.  But what it can potentially do, is to use the information about where in the frame faces are, to enhance the exposure.  Also, the camera knows which areas are in focus, and can make sure that these areas are properly exposed.  The extra information the camera has can be put to good use to give you a better exposure.

Self portrait with a fisheye lens

Most of what I write in this blog is non-personal. While the technical analysis is generally based on fairly objective observations, it is supplemented with subjective comments. But it's still far from personal.

With this post, I take a turn towards the more personal side with a self portrait:

It was taken with the Lumix G 8mm f/3.5 fisheye lens. It has an amazingly short minimum focus distance of 0.1 meter. This is measured from the sensor, which means that the distance from the front lens element is around 2cm, or approximately one inch. The field of view is still very wide, so it cannot be called a macro lens. Here is one example where I used the lens to photograph a LEGO figure up close.

At this distance from my face, the perspective gets very distorted. You can see this in the self portrait, since my nose and eye looks enormous compared to the rest of the face. Of course, the fisheye lens is far from what you would call a portrait lens.

To get as much as possible in focus, I set the aperture to f/9. There was still enough light to get a shutter speed of 1/20 second at ISO320.

Composing the image was fairly easy. I could just flip out the LCD screen on the Panasonic Lumix GH2 and see how the resulting image would be, while holding the lens towards my eye. I used autofocus, which is very fast with this lens, even at this close distance.

Example video capture, GH2+Lumix-Leica 45mm f/2.8 macro

Here is an example video capture using the Panasonic Lumix GH2 camera with the Panasonic Leica Lumix DG Macro-Elmarit 45mm f/2.8 1:1 macro lens.

The camera was handheld, and I left the aperture at f/2.8, shutter speed around 1/100 second. I used the 720p mode, which, since I have the European camera model, gives me 50fps. It was downmixed to 25fps using Openshot video editing software.



I left the autofocus on, using the face detection focus mode. Initially, the camera chose to focus on the toy men on the train station platform, which suited me fine.

As the train enters the frame, the focus starts wandering off, which doesn't really help in this case. However, the focus returns to the toy figures afterwards, which is desirable. In this case, it would probably have been better to turn off autofocus during the video capture, since it doesn't help much at all.

Other lenses, like the Lumix G 14-42mm f/3.5-5.6 would probably have focused faster. However, with a maximum aperture of f/5.6 at 42mm, it would not have given the nice blurred effect of items out of focus. On the other hand, a large part of the frame is in fact out of focus in this video. So perhaps f/5.6, giving more of the subject in focus, would have been a better choice here anyway.

Chromatic Aberration and lens correction

Chromatic Aberration (CA) is a type of lens distortion. It is caused by light of different colours being refracted differently by the glass lens elements.

This is one reason why lenses often consist of pairs of lenses grouped together: The two lens elements in the pair are made of different glass types, and have different optical properties. The aim of the construction is to neutralize the effect of chromatic aberrations, so that all visible colours focus in the same place.

Here is an example image illustrating CA. The image was taken with a Panasonic GH1, using the Olympus Zuiko 9-18mm f/4-5.6 wide angle zoom lens at 9mm f/4:

An enlargement of the lower left corner reveals the CA artifacts:

You'll see that there is red and green fringing off the high contrast areas, where the white and black paint meet. These artifacts are typically seen in the corner of the frame, while the centre of the image is generally free from them.

In the extreme corner, the artifacts take up about 2-3 pixels, which is a moderate effect.

CA artifacts are generally found in the corner, especially when using wide angle lenses.  The lack of CA artifacts indicate a high quality lens design.

Automatic lens correction

Panasonic Lumix G lenses are automatically corrected for CA artifacts by the camera. In the JPEG output files, the camera has adjusted the images with the intention to remove these distortions.

When using Panasonic Micro Four Thirds lenses on Olympus cameras, the CA corrections are not done with current camera models. Future cameras from Olympus may employ the same technology as Panasonic uses, and adjust for these effects.

We can still see the original exposures by opening the RAW image files in a converter which allows for not implementing the CA adjustments. One such RAW converter is UFRAW. By using this converter, we can compare the original exposure with the out of camera JPEG output, to see what kind of adjustments were done.

I am fully aware that there are RAW converters which will do the CA corrections as well.  So this is not a test of the RAW converter, but rather a look at the image before the CA correction, to see what kind of CA artifacts the lenses generate.

I have done these comparisons for three lenses. The images were taken with the Panasonic GH1 at ISO 100:

Lumix G 8mm f/3.5 fisheye

Here is the full image, take with a wide open aperture at f/3.5:

And the upper right corner at f/3.5 and f/5.6:

In this example, we can see that the fisheye lens is very sharp in the corner, even wide open at f/3.5.

In the extreme corner, there is some red and green fringing where white meets black.  Perhaps around 2-3 pixels of colour artifacts.  They are corrected well in the JPEG output, though.

Lumix G 14mm f/2.5 pancake

Here is the full image, take with a wide open aperture at f/2.5:

And the lower right corner at f/2.5 and f/5.6:

We see here that when using the lens wide open, there is some vignetting in the corner, and also, the sharpness is not optimal. This is quite common for any lens, really, and I wouldn't say it is a problem.

Looking at the CAs, it looks like there is about 1-2 pixels of red and green fringing in high contrast areas in the original RAW image. In the adjusted JPEG image, there is still some purple fringing.

Lumix G 20mm f/1.7 pancake

Here is the full image, wide open at f/1.7:

And the lower right corner at f/1.7 and f/5.6:

My comments here are mostly the same as for the Lumix G 14mm lens: There is vignetting, and there is dullness in the corner at f/1.7. Again, this is not uncommon, and especially so for a low light lens with a large aperture.

Before conversion to JPEG, there is some small amount of green and red fringing. In the converted JPEG image, the green fringing is gone, but there is still some purple fringing.

Olympus Zuiko Digital 50mm f/2 1:2 macro (Four Thirds lens)

This lens is rather well known for it's CA artifacts. Here is a video showing the lens being used on a Panasonic GH1 camera (with an adapter). When focusing manually close to the minimum focus distance, you can see clearly black text on white background is black only when in perfect focus. Focusing a bit longer gives a green outline. And focusing closer gives a red outline. Doubleclick on the video to get a larger view, up to 720p is possible.



Here is a 100% view of a photo taken using the Olympus 50mm f/2 macro at 1:2 magnification, and f/2 aperture. As you see, text which is beyond the focus point has green fringing, while text nearer has red fringing.

The Panasonic-Leica 45mm f/2.8 macro lens for Micro Four Thirds does not exhibit these kinds of artifacts.

Conclusions

When using Panasonic lenses on Panasonic cameras, some CA artifacts are corrected automatically. However, there are still some purple fringing left in the corners.  It is generally restricted to around one pixel width, which is not much.

Even before the CA correction, the CA artifacts are very moderate on the Lumix G 20mm f/1.7 and the Lumix G 14mm f/2.5 pancake lenses. The Lumix G 8mm f/3.5 fisheye lens has somewhat more CA effects in the extreme corner, but it is well corrected by software.

All in all, I think that CA artifacts are nothing to worry about with these lenses. Even when using the lenses on Olympus cameras, without in camera CA correction, this should not bother you much.