NanoFlash features wish list

[Steve Kalle]

With an EX1 or EX3, can you record out via firewire(25Mb HDV only) AND SDI to Nano?

[Mike Schell]

Originally Posted by Gints Klimanis

Dan, so does the Nanoflash architecture not allow storing full resolution frame grabs? I was asking about the possiblity of a 15-60 fps "still" camera that writes to uncompressed TIFF or PNG files.

Hi Gints-
This is outside the capability of the current design. But, if you capture at 160 Mbps Long-GOP, you can grab individual frames from the timeline (and save an tiff, png or jpg). These captures really look quite stunning to my eye. I seriously doubt you would gain very much quality going to an uncompressed format.

Best-

[Gints Klimanis]

Bummer. It would be cool if future video recording devices would allow for writing an uncompressed frame to the flash card.

Architecture issues aside, wouldn't an uncompressed frame from a 3 CCD Sony EX1 rival that from a 6-8 Megapixel Bayer sensor from a decent DSLR camera in spatial detail? I'm just curious about a low cost, high speed still camera with very high frame rates for the purposes of capturing bursts but selecting the single, perfect frame. Even high frame rate DSLRs such as the Nikon D3 at 9-11 fps still aren't able to go fast enough.

[Dan Keaton]

Dear Friends,

If one needs to record both SD and HD at the same time, it currently takes two nanoFlashes, and a good downconverter to convert the HD-SDI output to SD-SDI for the second nanoFlash.

This may make sense for some applications, but for general use I would not recommend it.

Downconverters have to be very sophisticated to be really good. Our experience has shown that some are just not up to the task.

In other areas:

I doubt that we could record a still image and record HD-Video at the same time.

When one records at any of our high bit-rates, say 100 Mbps or higher Long-GOP, and records progressive images, the individual frame captures can look very good.

One interesting side note:

For most recorders, when it says 100 Mbps, you may get 100 Mbps for one format/frame rate, but get a far lower bit-rate for other conditions.

With our recorders, you get 100 Mbps, if you are doing 1080i60 or 1080p24. Thus, if you are using 1080p24 (23.976), you actually get more bits per frame.

So, as Mike said, if you record progressive and use our 160 Mbps Long-GOP, you may find the images to be very good.

[Aaron Newsome]

Originally Posted by Gints Klimanis

Bummer. It would be cool if future video recording devices would allow for writing an uncompressed frame to the flash card.

Architecture issues aside, wouldn't an uncompressed frame from a 3 CCD Sony EX1 rival that from a 6-8 Megapixel Bayer sensor from a decent DSLR camera in spatial detail? I'm just curious about a low cost, high speed still camera with very high frame rates for the purposes of capturing bursts but selecting the single, perfect frame. Even high frame rate DSLRs such as the Nikon D3 at 9-11 fps still aren't able to go fast enough.

Gints, a video stream really is just a collection of still pictures. 220Mb/s or 100Mb/s, either way you look at it, you'd be hard pressed to tell the difference between a frame grab from the video stream and an XDR that could write TIFF images.

If you want to get the still images, export the video file to stills or do frame grabs in your editor.

I've done a ton of stills with my XDR and they look as good as any high end DSLR you can get.

[Gints Klimanis]

Aaron, while I understand the difference as I work in audio (primarily) and video (lightly)signal processing, I am contesting the difference, be it slight, between stills from 4:2:2 and 4:4:4 and am inquiring about getting closer to "DSLR quality" than 4:2:2 allows with its half rate chroma subsampling, which is a cause of color bleeding and other false color artifacts. A rough analogy to this difference would be choosing an 8 megapixel camera over a 4-6 megapixel camera.

[Aaron Newsome]

Sounds like a challenge. I'll supply the uncompressed and XDR frames and you all can pick which is uncompressed and which isn't.

[Gints Klimanis]

I'll take your word that they're close, but I have no opportunity to see this for myself. So, thank you for your efforts.

How are you able to provide exactly the same frame?

[Mike Schell]

Originally Posted by Gints Klimanis

Aaron, while I understand the difference as I work in audio (primarily) and video (lightly)signal processing, I am contesting the difference, be it slight, between stills from 4:2:2 and 4:4:4 and am inquiring about getting closer to "DSLR quality" than 4:2:2 allows with its half rate chroma subsampling, which is a cause of color bleeding and other false color artifacts. A rough analogy to this difference would be choosing an 8 megapixel camera over a 4-6 megapixel camera.

Hi Gints-
The nanoFlash can only record a 1.5GHz HD-SDI, which is 1920x1080 4:2:2 8/10-bit at frame rates up to 30p or 60i.. So, anything outside this spec is simply not possible. Given this restriction, I think the compressed images are visually indistinguishable from the original uncompressed source, assuming compression of 100 Mbps Long-GOP (or higher).

That's why we are reluctant to add uncompressed single frame captures, since the incoming source resolution or color depth can not be increased (in the nanoFlash).

Best-

[Aaron Newsome]

Originally Posted by Gints Klimanis

I'll take your word that they're close, but I have no opportunity to see this for myself. So, thank you for your efforts.

How are you able to provide exactly the same frame?

hi Gints. For the challenge I would run one of my camera's SDI into an XDR and one into an uncompressed recorder. I would post Identical frames from each.

If my uncompressed recorder could do 4:4:4 uncompressed I would do that but my third SDI out would not be identical frames since it would be a processed signal.

Doing a single camera test for 4:4:4 uncompressed compared to XDR would really only work for my setup if it were locked down.

[Daniel Symmes]

I would suggest it be shot against a properly lit green screen with an appropriate foreground subject (possible with motion blur).

Any significant issues will turn up in the key, under magnification.

This would HOPEFULLY settle some of this.

Possibly CD will have a page on their new Web site to show the results.

The 8/10 and data rate issues otherwise will continue to clog the pipes here.

[Gints Klimanis]

Originally Posted by Mike Schell

That's why we are reluctant to add uncompressed single frame captures, since the incoming source resolution or color depth can not be increased (in the nanoFlash).

Thanks, Mike. In my enthusiasm for the Nanoflash, I overlooked the fact that HD-SDI was 4:2:2. Ok. Finally, I get it.

[Gints Klimanis]

Originally Posted by Aaron Newsome

hi Gints. For the challenge I would run one of my camera's SDI into an XDR and one into an uncompressed recorder. I would post Identical frames from each.

Aaron, don't worry about it. Mike Schell reminded me that the Nanoflash will only record 4:2:2, so there is no way to record better. I was asking about the Nanoflash doing uncompressed 4:4:4 frame grabs for superior spatial resolution, but that's not possible with my Sony EX1. So, this issue is settled for me. I'll be happy with 4:2:2 still pulled from the MPEG.

[Billy Steinberg]

Originally Posted by Gints Klimanis

Aaron, while I understand the difference as I work in audio (primarily) and video (lightly)signal processing, I am contesting the difference, be it slight, between stills from 4:2:2 and 4:4:4 and am inquiring about getting closer to "DSLR quality" than 4:2:2 allows with its half rate chroma subsampling, which is a cause of color bleeding and other false color artifacts. A rough analogy to this difference would be choosing an 8 megapixel camera over a 4-6 megapixel camera.

You do know that with the exception of one DSLR with a foveon sensor, all the digital sensors used in DSLRs are single chip bayer filter affairs (not three chips with a prism like a high end video camera). And that none of them have as many blue and red assigned pixels as green? As in color sub-sampling...

Billy

[Gints Klimanis]

Agreed. A modern DSLR (Nikon D90, D300, D3) has about 12 Million photosites, of which 1/4 are blue, 1/4 are red and 1/2 are green. So, that means full color resolution is something between 3 and 6 Million.

A 1920x1080 full raster sensor has about 2 Million photosites, and the Sony EX1 has three sensors. Meaning, the EX1 starts with 2 Million tricolor pixels or the near equivalent of a 6-8 Million Photosite Bayer sensor. Color subsampling by a factor of 2 reduces that to a rough equivalent of a 3-4 MP Bayer sensor. Sure. It's not totally accurate to compare the color spaces this way. Anyone would rather have the 4:4:4 still than the 4:2:2 still.