I love to see a better video auto focus system. DFD has already removed the first wrong direction start of the focus pull, but I like to address the second wrong direction autofocus pull, the "focus confirmation" rack at the end of the auto focus pull.

I'm not an engineer but I like to understand what is happening what, how, where, when and why.

At this moment the Auto Focus system for video looks the same as for stills. And for stills it works fast and well, accept in continuous high when it has a hard time keeping up with fast action. For video (akin to a very high number of stills) the AF is way to slow, and the focus confirmation pull is very distracting and jarring.

If DFD is able to measure the blur amount, and therefore can measure how far the subject is, the I think a reprogramming of the focus system can help to get rid of that jarring focus confirmation refocus.

I can think of this focusing system:

Focus pull situation:

1) DFD measures the amount of blurr, 2) compares it to the current focus distance, 3) estimates subject distance, 4) smoothly starts racking focus, 5) measures amount of blur half way racking focus, 6) checks the blur shift amount of the first and second measurement, outputs second focus target distance, 7) at the last part of the focus rack the focus speed eases off to arrive at the second estimated focus distance.

Continuous AF situations:

1) DFD measures the amount of blurr, 2) compares it to the current focus distance, 3) estimates subject distance, 4) smoothly starts racking focus, 5) measures amount of blur half way racking focus, 6) checks the blur shift amount and compares it to the first measurement, calculates subject movement, 7) rack focus to second focus target distance, 8) checks the blur shift amount and compares it to the previous measurement, calculates subject movement, 9) repeat 7 and 8 until no blur.

10) If subject leaves the frame maintain the current focus distance for x time before starting with 1 on the new subject in the assigned focus area.

Maybe DFD blur measurements are hard & slow, so having measurements while the camera is auto racking focus is harder to do, but I prefer a slower focus much over racking past the focus and racking back in video.

As a compromise DFD could be used to make a 1 time focus distance estimate, the focus racks smoothly (s curve) to the estimated distance, and stops (without racking past the estimate focus distance AND racking back). This would be good enough for a G85, and the more processing heavy higher DFD sample rate for the high end GH5. I must admit this all hinges on DFD accuracy in measuring the blur and estimating the subject distance.

What do you people think? Love to learn how the current contrast and DFD actually works. Love to hear your suggestions of how to improve AF.

Setup - In order to achieve the higher bitrates I had to use M from the dial atop the GH2. The camera icon with the M would not allow this. The tradeoff was Auto ISO with M, and adjustable ISO with the camera icon with the M. I use mostly Canon FD lenses.

1a) Bitrates - I have tried a few hacks with varying success. The best I have got is about 100MB with Flow Motion v2.02, however, most of the shots I got with this hack seem to be from 70MB-90MB. I tried Flow Motion, Driftwood's Moon T7, and two Driftwood hacks (176MB & 220MB) from a pack made by the user Sohus on this forum.

Is it common to be well below the stated bitrate amounts?

1b) AVCHD vs MJPEG - I never seemed to get very high bitrates when using 1080i, despite the hacks stating to be much higher. They would range from 15MB-40MB. To achieve higher bitrates with Flow Motion, I used MJPEG which boosted them much higher.

Shouldn't I be achieving high bitrates with AVCHD?

2) Auto ISO - Whenever I use M (from PASM) the ISO would always be set to Auto ISO. I could adjust the ISO when using the camera icon mode (with the M), but then the bitrates would suffer. I tried changing the ISO before shooting, but Auto ISO would kick in no matter what.

How do I change the ISO while in M mode?

3) EV - I know that EV depends upon shutter, aperture, and ISO, but I can see the EV when I push the Display button so I was hoping there was still a way to change it.

Am I correct to assume that because I am using Canon FD lenses I cannot use EV? Even if using a Panasonic lens, how do I change the EV (I tried, but failed).

4) Modes - I have only been able to use the hacked bitrates - and only to a certain degree - by using M with MJPEG.

What modes should I be using to use the hacked patches? The camera icon with M? High Bitrate? 24P? 1080i?

Sorry for all the questions. I am not unhappy with the camera, but I really want to know why the bitrates are not working? And moreso, why the Auto ISO can't be changed (as I don't like the amount of noise it can introduce)? Thank you!

Also I'd love some input from others on colours and noise ... This was shot using a cinelike D mode with everything set to 0 and edited in Prem Pro 2014 , If memory serves I was on 800 ASA which is high but I expected better results for the noise ... The "black" bars are not the same black as the black surrounding them on my monitor... I output the clip using the factory preset for Vimeo which presumably is RGB as opposed to 709, so surely the black bars should match the black on my screen?? (the bars are a shade lighter)

Aside from the focus issue I'd be very grateful for comments / advice / guidance on noise especially in the blacks which are always a bit crappy ..

Thanks

Alan

Attention, noob questions ahead !

I would like to know a few things : - Which is the smoothest hack on the GH2 ? (All the detail, no macro-blocking) - What does "Intra" or "AVC-Intraa" mean ? - What is Auto Quantitizer ? I'm not sure if a higher number means a better image (lower compression) or it's the inverse. What is better, AQ0 (all to motion), or AQ4 (all to details) ? - What does GOP stand for, and there is "Low-GOP", but then does that mean there's "high GOP" ? What are 12/15 GOP, 3 GOP and 6 GOP ? - What are P-frames, B-frames... ? - What is "Matrix" and "444 Matrix" ? - On some settings there's written "14-16 QP", or "QP16". What does QP mean ?

I've not seen much smooth looking GH2 videos on Youtube (unlike the canon videos), lots of them have a lot of macroblocking or whatever and causes the video not to be smooth, and sometimes there are some sort artifacts (I think it's because of the compression or something...). I'm a bit of a pixel-peeper :P Could someone explain me the causes of the un-smoothness, and why are some videos really clean compared to others ?

Example of smooth videos I've found :

PS : Can someone give me the link to download the 244mb/s hack ?

And here on EOSHD there's an article about Intra :

http://www.eoshd.com/content/4452/gh2-gets-avchd-intra-at-176mbit

But I still don't understand the difference between the 176mbit so-called "Intra" hack and any other one ! I'm confused...

Please answer !

Thank You.

I meant to ask this for a long time but kept forgting about it.

My 14-140mm Lumix lens is not behaving itself in AFS mode on my GH2. It just can not focus. It is working fine with GH1, it is working fine in AFC mode. But won't work with GH2. No other lens have this issue on GH2 so I don't think it is a camera body issue.

I don't use Auto Focus for filming and rarely use it for stills anyway but it is an annoying issue when you need it...

Anybody had a similar problem? Any thoughts?

Thanks a lot

I thought the videos were good in the past. However, the lack of image stabilization, accurate Continuous Auto Focus, and a good wide aperture zoom lens were always mentioned as limitations with my videos.

Finally with the GH3 I think we are over that hurdle. The video below was shot entirely hand held(No tripod or monopod at all) by 1 person in continuous Auto Focus mode with the 35-100mm F2.8 lens. I also only used one battery and still had some juice left after more than 4 hours of filming.

I would say that the focus was spot on for pretty much every clip. Even the focus pull at the beginning from the guy’s Ipad to his son climbing worked perfectly. That was a one take shot. If I didn’t get it right I couldn’t ask the guy and the climber to go back and do it again.

I simply slid my finger from the Ipad to the climber on the camera’s LCD screen. I think it gave the video a very nice effect.

I had other lenses with me like the 25mm F1.4 and the 14mm F2.5. However, I never felt like I needed them. At F2.8 I was almost never above ISO 3200 even with the darker lighting they used for this round.

I think the IS was terrific in the 35-100mm lens as well. Yes I would rather have had the OMD’s 5 axis stabilization. However, I think this is the next best thing. Try holding a camera up for 4 hours straight with a 200mm 35mm equivalent focal length and see how shaky it is without good IS. I also like to take images while I record video and instantly upload them to the web so that spectators can follow along with the action with real-time pictures. With the GH3 I put it in the real time wireless upload mode and shot at 1080p @ 60 FPS in the stills mode. That way I could take stills during video and instantly upload them.

The only issue I had with that is that every time you press the shutter button to take a still it will initiate a focus adjustment. That actually made the focus go off more times than not. After noticing this I just decided to save images directly from the videos in playback mode(Press pause during video playback and then save an image). The camera will upload those images in real time as well.

I have tested the quality of images recorded during video recording and images saved during playback and they are indistinguishable in my opinion.

I also mount my GoPro 3 Black on top of the GH3 and shoot simultaneous wide angle 720p @ 120 FPS video. However, I scrapped all of the GoPro footage because it simply was not up to the caliber of the GH3 video. It is great for super slow motion but the 120 FPS mode indoors produces so much noise that it would not mesh well with the excellent video from the GH3.

I also was impressed with how well the youtube encoding turned out. I normally post to Vimeo as well because it is better for compression. However, in this case I thought the 1080p youtube encode looked great.

I didn’t alter the video at all. I did not do any color grading or post processing on it. I shot in scenery mode with -2 Contrast, +2 sharpness, Zero Color, and -5 Noise Reduction and simply pieced the videos together with a cheap($50) program called AVS Video Editor.

It was really nice to be able to produce the video with the footage as it was shot in camera and still get what I thought was a decent result.

There were plenty of other professional film crews there with much more sophisticated equipment. The little 11 year old girl in the last clip is probably the 2nd best female climber in the country(At 12 years old) and she has a film crew that follows her around everywhere she goes. I wonder how much better their footage looks?

So what do you think? Is the GH3 finally the camera that would make you put down the traditional video camera if you shoot both stills and video?

But whenever the environment is not under control, a well-working continous auto focus can be really, really helpful.

Yet all the recent implementations of continous auto focus in cameras using contrast based focusing suck - they just don't work good enough to be useful. They are either to slow or do too much visible "focus hunting".

I understand that any focusing based on contrast only will require some sort of "hunting", because the direction in which to adjust the focus plane cannot be predicted based on contrast only.

But continous focusing is relevant mostly for video (not for still images), and the resolution of the sensors is usually much much higher than the resolution of the recorded video. So it seems like a very obvious idea to adjust the focus plane only by the smallest possible amount (according to the step motor of the lens), one step back, two steps forward, one step back, and if properly sized, these steps are small enough that the change of the focus plane is not perceivable in the (relatively low) resolution of the recorded video, while it is still big enough to yield a significantly different average contrast in the (relatively high) resolution of the sensor.

I am pretty convinced that my old Sanyo hybrid camera did precisely that: With just one sensor chip and no phase detection, it could only base focusing on contrast. While shooting video, you could hear a very faint noise that was indicating the lens was focusing back and forth just a little - not as much as to be visible in its 720p output file.

Now I wonder: Why don't they just implement this simple method in current cameras? Why do they follow this primitive approach of "accepting some variation in contrast without moving the focus plane at all, then suddenly hunt focus from scratch"?

I see my GH2 with a pancake lens, shooting away while sitting discreetly on a sock-full of rice, or attached to my belt. I've starting to read the manual and find I'm missing out on a lot of what the GH2 can do, automatically, to free me up to work faster and better - not to make a silk purse out of a sow's ear, but to become a more modern videographer.

So, within reason, I'd like to hear from fellow users like me.

[Edit] The stinger, "You don't buy a dog and then go around barking, chasing cars and pissing on trees" is an Australian variation on, You do not get a dog and bark yourself.