Most of my home videos contain clips shot with a handheld Panasonic camcorder and a GoPro helmet camera. Since most of my videos are of action sports, I like to record all of my footage at 60 fps progressive scan to avoid interlace artifacts or multi-imaging.
My Panasonic camcorder is capable of recording a full HD image at 1920 x 1080 pixels at 60 fps progressive. In the past, my helmet cameras have been limited to 1280 x 720 at 60p. Furthermore, I frequently stabilize my helmet camera footage, which gives up an additional 10% of the resolution.
Thus, I was extremely excited when GoPro introduced the new Hero3 Black Eddition because it is capable of recording 1920 x 1080 at 60p. I will still loose some resolution when I stabilize the footage, but it should be significantly better than the Hero2.
My enthusiasm for the new product was dampened somewhat when I read of all of the stability problems that others were having. But in April of 2013 GoPro provided a firmware update that resolved many of the issues. Other issues are likely the result of poor quality micro SD cards. So, I took the plunge and purchased a Hero3 Black Edition.
This report contains some of my early findings with the new camera.
One of the downsides of the newer Hero3 cameras is the smaller battery that has a shorter operating time. I generally record many short clips over the course of a few days while dirt biking or river running. I can recharge the camera each night at camp, but I need the camera to stay operational all day long. The Hero2 worked well for me – I never had the battery go dead while out for the day.
I had read other posts that suggest the Hero3 Black would provide about 75 minutes of battery life when operating in 1080 60p mode. My initial test gave me a battery life of 60 minutes while continuously recording. Under my more normal conditions of turning the camera on and off, I suspect I will get something less than 60 minutes of battery life. The GoPro website indicates that battery life should improve after the battery has gone through a few power cycles.
Currently I do not use WiFi or any of the new fancy features, and I have my camera set in “one button” mode. I hope this will be sufficient, but it does cause me some concern.
Update: On a recent outing I only got 24 minutes of video before the camera went dead. After recharging the battery at camp, I got 44 minutes and still had battery life left. I noticed that at the beginning of the first day the battery icon indicated only about 1/2 charge even though I charged the battery before leaving home. After returning home I have been investigating this and found that sometimes the battery does not fully charge. The red LED goes out, leading you to believe it is fully charged. But I noticed that the LED goes into a slow blink mode. I do not know why it does this, but it stops charging the battery. So far it appears to be an issue with my wall charger, even though it works fine on my older GoPro cameras.
Update 2: I learned that you need about 1 Amp to correctly charge the camera battery. Many phone USB charges do not have enough current to do the job, so the camera can go into weird modes. Be sure your USB source has sufficient power to do the job.
I have concluded that around one hour of battery life per charge is insufficient for my needs. And swapping batteries mid-day is impractical for me. So I have ordered the battery bacpac. This will make the system bulkier and heavier, but it should give me sufficient battery life for a day’s activities. It is disappointing that I have to buy ‘extra’ stuff to make this camera work as well as prior units.
Unfortunately the Hero3 uses micro-SD cards, so I could not use any of the SD cards I already own. Furthermore, I knew that the higher resolution of the Hero3 Black Edition would require more memory than previous versions. I therefore purchased two 64 GB memory cards. These turned out to be much larger than I need.
During my 60 minutes of recording, the camera broke the video file into four chunks no larger than 3.8 GB, or 17.25 minutes. Based on my calculations a single 64 GB card can store approximately 4.8 hours of 1080 60p video. That is more video than I will want to edit for any single adventure. For my needs, a 32 GB card would be sufficient.
If you intend to use the professional grade “Protune” option, a larger memory card may well be in order.
Update: While out on our adventures we noticed that sometimes the camera would not start recording (in one button mode). Sometimes it works, sometimes it does not. After getting home I did some testing and found that when it does not record, I momentarily get an error message indicating “No SD”. So for some reason it sometimes will not recognize my 64 GB SanDisk cards. According to the GoPro user forums, this appears to be a common problem with 64 GB cards. I get the same results with two different 64 GB cards on two different cameras. I will return the 64 GB cards and buy some 32 GB cards, which will hopefully avoid this problem.
Update 2: With the 64 GB cards I would receive the “No SD” error about 50% of the time. With the 32 GB cards it dropped to about 20%, but the error still occurs. I have read good reviews on Samsung cards, so I plan on trying those next.
GoPro support told me how to reset the camera, which may have resolved some of my strange problems I encountered on my last trip. Here is the proceedure:
- Remove the battery.
- Remove the micro-SD card.
- Hold down the shutter button and keep it down.
- Insert the battery (keep the shutter button down).
- Turn on the camera with the power button (while still holding the shutter button down).
- Resort your preferred camera settings.
I haven’t yet done a detailed comparison of the image quality with that of my Panasonic camcorder, but the video is certainly crisper than the Hero2. I am fairly impressed with the crispness of a frame grab taken from the video while I was traveling at about 32 mph on my son’s motorcycle.
From this early look, the color quality seems to be pretty accurate and the sound quality seems to be better than the Hero2.
Another annoyance of most helmet cameras is the rolling shutter, or “jello”, effect. This adds distortion to a captured frame of video. The frame is not captured instantaneously – it is scanned vertically over a portion of the frame time.
Consider the following illustration where the rectangle represents a video frame. The solid line represents a vertical line within the scene. If the camera had no rolling shutter effect, the line would remain vertical even if the camera were panning quickly to the side. The dashed line might represent where that vertical edge would lie on the next frame of the video.
Now consider what happens on a camera with rolling shutter artifacts. Because the camera is panning while the image is being captured, the position of the vertical edge moves as the shutter scans down the image. The movement of the edge from top to bottom relative to the movement from one frame to the next gives a percentage that indicates the amount of rolling shutter introduced by the camera. The smaller the number, the better.
In practice, it is a little more difficult to actually measure the rolling shutter percentage. The fish-eye barrel distortion of most helmet camera lenses is particularly challenging to work around. Also, it is difficult to set up a test scenario that has a perfectly vertical edge. It is therefore useful to capture a reference still image without the camera moving and to measure the data when the vertical edge is very near the center of the screen. Here are the frame grabs that I used for my measurements.
In this still frame reference, notice that the column is not perfectly vertical – which is why I need a still reference frame.
In this next shot you can see that the column appears to be leaning to the side. This is the rolling shutter distortion we are trying to measure.
This next image was captured from the frame immediately following the prior image. The movement of the column from one frame to the next defines 100%.
Once you have captured these three images, you can use a photo editing tool such a Paint to find the x pixel location of the top and bottom of the vertical edge in each image. You can then compute the rolling shutter as follows:
Rolling Shutter = 100 * (Fb – Ft + St – Sb) / (Nt – Ft)
Where F = first frame, S = still frame, N = next frame, t = top, and b = bottom.
I took measurements for both a right high-speed pan and left high-speed pan. My rolling shutter calculations came out at 92% and 96%, for an average of around 94%. These numbers are probably not precise due to the barrel distortion of the lens. The next image shows how severe the distortion can be near the edges of the frame.
And now for my main objective – high-quality video that has been stabilized. The following video has also been down-converted to 1280 x 720 @ 30p, but it looks much better when viewed on my big screen TV on my BluRay player at 1920 x 1080. The quality isn’t as good as my Panasonic, but it is far better than my Hero2. Furthermore, the quality is pretty good for sharing on the Internet at 720. I am happy with the results.
In this test video, the left half of the screen is the original footage while the right half has been stabilized with the Deshaker plugin to VirtuaDub (see my blog posting for details). Notice that the stabilized video is much smoother (even while riding on pavement – it makes a huge difference for dirt bike videos). Also notice that the image is scaled slightly larger – this is the 10% resolution I give up in order to stabilize the image. For me, a stabilized helmet camera is much easier to watch.