Raw HDR-FX1 mpeg2 files are posted.

[Donal Briard]

"Kaku, do you remember trying out any special enhancement features such as when filming the "mini.m2t" or "fersshwwndw.m2t" files?"

Not sure if this has any relevance, but both are shot through store windows... Polarizing filter maybe?

Kaku. any chance of getting a daylight clip with 0db hain on a tripod? Thanks

[Mark Grant]

"They come through fine on I and P frames, but they're randomly less intense on the B frames, sometimes much lower in intensity! And worse, when the dark gray interior becomes visible around frame 590, its intensity is pulsing like mad, perfect on I and P, and darker than it should be on B frames."

To me it looks like perfectly normal lighting flicker when the camera is recording at a different frequency to the lights: I've seen it often with PAL cameras in a 60Hz country, or film cameras with 50Hz lights. It might also be a result of using a non-standard shutter speed (e.g. 60fps but not 1/60 of a second).

If it was a general problem with the compression, it should be visible in all the clips, but I can only see it in clips with electric lighting... so I doubt that it is.

[John Jay]

Yes Ill buy the lighting flicker, a lot of shops use mini HMI, so its the gas discharge rate which may be causing flicker. I believe the FX1 has a switch to remove this.

The fine noise I am seeing is almost certainly due to mpeg2, hard to do without - I have even seen very clean digibeta get noisy when coded to full band DVD quality.

Algolith has an excellent filter to minimise this

Most of all I would like to thank KaKu and the folks at dvinfo for giving us a great insight to the cam

[Steve Crisdale]

Firstly, let me thank Kaku for taking the time and effort (and no doubt expense) to post these m2t files for us all to enjoy.

Secondly.....I get the feeling that not everyone who's checking these clips out has got the relevant hardware and software to be making the sort of critical judgements that they are.

I'm currently processing a couple of the clips to different formats to see how they respond. Captured still from the 'crosswalk' clip looks excellent. BTW, the audio of the guy saying "that's a bad-ass camera" indicates the FX1's street appeal!!!.

The low-light capabilities of the camera are evident and impressive. No way could you expect such quality results from a HD10 under such twilight and night light conditions.

[Lorin Thwaits]

Mark and Donal -- You guys were on to something here! After pondering polarization and electric lighting, I focused my research on these and found another trace of this phenomenon on the daytime GAP sample ("eevgap.m2t"). Starting at frame 167 and going to the end, the red shirt worn by the girl on the fluorescent-lit sign there flickers, but only slightly. An important factor here is that I and P frames come up reliably every 1/10 of a second. With that section of video I conducted a little test:

If you take every frame that is an I or P (so every third frame) from frame 167 to the end of the GAP video and put just those in an uncompressed AVI, there is no variation whatsoever in intensity. I then made the same kind of "every 3rd frame" sequence, this time consisting of only the first of the two B frames in every group of 3. Those were all the same intensity. Finally I tried the second of the two B frames in each group of 3, and that also was consistent. Considering that each frame is 1/30 of a second, whenever we take every third frame it will always be on 1/10 second intervals.

Checking out the frequency of electricity in Japan from this page: http://kropla.com/electric2.htm, it says that Eastern Japan is 50Hz, and Western Japan is 60Hz. Kaku is from Shibuya, which is in the Tokyo area, where electricity is 50Hz. The least common multiple of 30 (our framerate) and 50 (the power company frequency) is 10, and that's the frequency at which everything is always consistent.

So the issue originally seemed tied to the codec only because the I and P frames represent a predictable 1/10 second timing.

I've never taken my NTSC video camera to Europe or Japan, so I hadn't ever seen this anomily. Now I know!

-Lorin

[Mark Grant]

Ah, I wasn't sure if that was the cause because I thought that all of Japan was 60Hz, now you've settled it!

[Donal Briard]

<<<-- Originally posted by Lorin Thwaits : Mark and Donal -- You guys were on to something here! After pondering polarization and electric lighting, I focused my research on these and found another trace of this phenomenon on the daytime GAP sample ("eevgap.m2t"). Starting at frame 167 and going to the end, the red shirt worn by the girl on the fluorescent-lit sign there flickers, but only slightly. An important factor here is that I and P frames come up reliably every 1/10 of a second. With that section of video I conducted a little test:

If you take every frame that is an I or P (so every third frame) from frame 167 to the end of the GAP video and put just those in an uncompressed AVI, there is no variation whatsoever in intensity. I then made the same kind of "every 3rd frame" sequence, this time consisting of only the first of the two B frames in every group of 3. Those were all the same intensity. Finally I tried the second of the two B frames in each group of 3, and that also was consistent. Considering that each frame is 1/30 of a second, whenever we take every third frame it will always be on 1/10 second intervals.

Checking out the frequency of electricity in Japan from this page: http://kropla.com/electric2.htm, it says that Eastern Japan is 50Hz, and Western Japan is 60Hz. Kaku is from Shibuya, which is in the Tokyo area, where electricity is 50Hz. The least common multiple of 30 (our framerate) and 50 (the power company frequency) is 10, and that's the frequency at which everything is always consistent.

So the issue originally seemed tied to the codec only because the I and P frames represent a predictable 1/10 second timing.

I've never taken my NTSC video camera to Europe or Japan, so I hadn't ever seen this anomily. Now I know!

-Lorin -->>>

Jesus, we're nerds...

[Kaku Ito]

I'm going to try to report one by one....

*Cineframe 24 is not supporting shutter speed of 48.
Fisrt of all, I though I was stupid to leave the shutter speed to 60 when I was shooting 24p, but when you select "Cineframe 24", FX1 does not give you any slower shutter speed than 60. Maybe this will be different in Z1? When I shot the 24p footage, I selected the preset that came with for shooting 24p then I though it was giving me the necessary change in the shutter speed, but there's no way I could adjust since you can't select lower than 60. What a bummer.

*About the show window issue
What Mark Grant mentioned is most likely the problem. I was just using the cam alone, no filter nor wide conversion lens or what so ever. They were behind the window as all of you guessed. The gain could have been up little higher. John Jay mentioned about the flicker elimination and FX1 seems to be able to tunr the elimination on or off, there's no switch between 50/60 like Sony is mentioning on Z1.

Lorin,
I might have had ND filter on to check how I can get the details on clouds and shadows in the tree at the same time. But by lookng at shots from today, this noise issue could be MPEG2 problem. With the gain set to 0db, no ND filter on, cloudy but enough brightness, when I go all the way to wide side and have a lot of small details from the woods, then it looks not as sharp. Under the same circumstances, when I zoomed into subject it looks a lot more sharper and detailed. More things in the screen would be more likely to be noisier.

I posted a small file with 0db/F6.2. Sorry, but no tripod since I had other bike things to carry.

[John Jay]

that last footage certainly 'floats my boat' :)



Kaku,

If you find time can you please try this test for me

1) hook up FX1 to a HDTV in 1080i mode via the D analog connector

2) remove any cassette in the FX1 (now using the camera head only)

3) wave your hand in front of the lens and tell me if you see any delay on the HDTV

I am trying to find out whether the analog out is mpeg2 compressed

this test will nail this question down


Many thanks

[Kaku Ito]

<<<-- Originally posted by John Jay : that last footage certainly 'floats my boat' :)



Kaku,

If you find time can you please try this test for me

1) hook up FX1 to a HDTV in 1080i mode via the D analog connector

2) remove any cassette in the FX1 (now using the camera head only)

3) wave your hand in front of the lens and tell me if you see any delay on the HDTV

I am trying to find out whether the analog out is mpeg2 compressed

this test will nail this question down


Many thanks -->>>

John,

What I could try at this moment is to connect it to PVM-D20 by component.
I see not much delay.

[John Jay]

Kaku

thanks, can you repeat the test but this time with the cassette loaded in the camera and tell me if the delay is the same or more

thanks,

[Kaku Ito]

It's the same.

[Donal Briard]

"I posted a small file with 0db/F6.2. Sorry, but no tripod since I had other bike things to carry."

All the noise I was complaining about is gone. Sure, there are MPEG 2 compression artifacts (little squares) here and there in the trees, but it's minimal and completely acceptable. If you look at all th eother daylight shots (except GAP), they had so much noise on the trees it almost looked like it had been shot in super-8!

P.S. I hope the sun comes out soon in Japan so that you can give us a few shot on a bright, sunny day. I bet THAT footage will be awesome!

[John Cline]

For what it's worth, after renaming Kaku's HDV clips to have the ".tp" extension, they play perfectly through the MyHD MDP-120 HDTV card hooked up to my Sony XBR910 34" CRT HD monitor. The MyHD card is capable of playing MPEG2 transport streams that have been encoded at up to 40 Mbits/sec. The MyHD card can also upconvert and downconvert to various resolutions in hardware and I viewed these clips at 1920x1080i where the card was scaling the 1440 horizontal pixels to 1920. The clips look excellent and better in some cases than the HDTV broadcasts I receive here. (HDV is 25 megabits/sec vs. 19.4 for broadcast TV.)

Kaku's clips do exhibit a little grain in the nighttime stuff, but otherwise no serious motion artifacts of any kind. If there were going to be any problems, the XBR910 would probably have revealed them.

In case anyone is interested, here is one place to get the MyHD card:

http://www.digitalconnection.com/Pro.../mdp120_pb.asp

John

[Michael Pappas]

Hello Kaku, could you film close ups of people talking etc. Medium shots as well. Also please at 0db gain and no high shutter speed above 1/60th and no lower then 1/60th.

Kaku, I am curious if the 1/30th or lower shutter modes do not discard a field like the ( vx2000-2100-pd150 ) do. Kaku can you try this. Please if you could, let us no what the simple settings are like gain, shutter speed and aperture on the clips you are showing. Thanks

Michael Pappas