HV20-Intensity Pro-AspectHD

[Ed Hecht]

Hello All,
Feel free to redirect me here. I'm sure I'm not the first to ask. I am using the HV20 and capturing HD footage via HDMI with BM Intensity Pro. Using BM's Media Express gives me Motion JPEG avi's that Premiere Pro CS3 seems to choke on (though I'm not sure I have the right project settings). Then I discovered I have HDLink for Cineform capture! Well, all the HDLink capture settings are pretty intimidating for me. Can someone give me a rundown of optimal settings for 1920x1080 capture in HDLink? (And what differences, if any, are there for footage shot using the 24F mode?) Thanks in advance!

[David Newman]

Note: Aspect HD is only licenced to 1440x1080 16x9, which is still plenty for the HV20 (we have proved in the past it is near impossible to tell the difference.) If you want 1920x1080 you will need either Prospect HD, or NEO HD (if you not using Premiere.)

In HDLink under Prefs:

For 24p footage, check removal 3-2 pulldown. Nothing else needs to be set. For 60i, leave pulldown removal off.

So the only controls you need to think about is the quality level (start with High) and capture location.

That is it.

[Ed Hecht]

Great. Thanks for this simple, but valuable info, David. Your response time and overall great support for your products is exemplary!

[Mike Thomann]

Originally Posted by David Newman

Note: Aspect HD is only licenced to 1440x1080 16x9, which is still plenty for the HV20 (we have proved in the past it is near impossible to tell the difference.) If you want 1920x1080 you will need either Prospect HD, or NEO HD (if you not using Premiere.)

Actually the tests I performed in this thread show that full 1920x1080 is worth it when you want to get the best live footage out of the HV20:
http://www.dvinfo.net/conf/showthread.php?t=94079

Test results:
http://thomann.net/hv20/interpolation/

EDIT: That test was almost a year ago and back then I surmised that even if one were to down-rez to 1440x1080 it would not be noticable, but as display technology improves, a 25% direct loss of detail will increasingly be considered significant. Strangely enough, even Sony's touted CineAlta HDCAM is only 1440x1080.

[David Newman]

This test doesn't represent loss of detail. Tests have shown that all single senor bayer cameras can resolve 75% of the sensor resolution, with the best camera equipment, and this a consumer camera. So if this had a great lens, the best optical low pass filter and high end demosaicing, then the camera will resolve 1440x810. Your test just seem to demo a small artifact in the upscaler, who's up-scaler did you use? Here is a thread on someones else finally concluding 1440x1080 was sufficient http://hv20.com/showpost.php?p=48989&postcount=82

[Mike Thomann]

Originally Posted by David Newman

This test doesn't represent loss of detail. Tests have shown that all single senor bayer cameras can resolve 75% of the sensor resolution, with the best camera equipment, and this a consumer camera. So if this had a great lens, the best optical low pass filter and high end demosaicing, then the camera will resolve 1440x810.

My test was based on Derek Green's first real comparison of the HV20's live uncompressed HDMI to HDV, NeoHD, and NeoHDV. It was actually back then when it was concluded that NeoHD was very much a noticable improvement with the HV20, and you agreed that it would be the best performer.

My test represents an exact 25% loss of detail of the final live HDMI output. So even if there is a 75% loss of detail due to lens and internals before live HDMI output, by going 1440 we are directly adding an additional 25% on top of that. When the HV20 first came out there was a well-known shootout at the time with the top cameras where resolution tests were performed and it was found that the HV20 resolved more detail than about all of them when well lit. If they warranted more detailed codecs, then the HV20 is worthy of them as well.

Originally Posted by David Newman

Your test just seem to demo a small artifact in the upscaler, who's up-scaler did you use?

This was a comparison of the samples provided by Derek. In that thread I compared upsale possibilites and even the best clearly showed the difference.

Originally Posted by David Newman

Here is a thread on someones else finally concluding 1440x1080 was sufficient http://hv20.com/showpost.php?p=48989&postcount=82

The video in that comparison is misleading because it too is 1440x1080, and when the video is viewed at 1:1 pixel map the aspect ratio is only 12:9 and is horizontally squeezed, so the very thing we are trying to distinguish in that video has been introduced in both samples. Also that video is recompressed to only <7mbps.

The example jpg is not lit well enough to find many comparable areas and the jpg itself is too small and introduces additional compression artifacts. Just look at the jpg compression artifacts around the white text where jpg is typically more accurate...

http://thomann.net/hv20/rhkfilm-test...-artifacts.jpg

Even so, it's easy to see the difference in the few areas where there is enough detail, like the buttons on the shirt and the stripes. I overlaid the two images for direct comparison and the difference is easily apparent at 100% on a 20" screen and at 200% on a 20" screen (comparable to 100% on a 40") the difference is considerable. Now imagine the difference on a home theater projected onto a 100+" 1920x1080 screen. I'm reluctant to post an illustration of this comparison because I don't like the fact that the low light and jpg compression introduced are misleading as to the true extent of the difference...

http://thomann.net/hv20/rhkfilm-test/

A much better test to illustrate the difference would be with something with small lettering, like various bottles under decent lighting, or how about a resolution chart.

[Mike Thomann]

After revisiting Derek's test, I made a comparison of NeoHD and NeoHDV with a bottle that has small text. Although the difference is not overly extraordinary, NeoHD does appear at least 25% sharper than NeoHDV. Note the difference in sharpness on the text and on the edge of the bottle over the green screen.

Will the average person see this? moreso as screens get bigger
Will a production company? a good chance
Is the difference useful in post? yes
Useful for keying? yes
When you want the most? yes
Reasonable for the effort? absolutely

What's really incredible is that I didn't even have to include the uncompressed sample in the comparison because NeoHD is nearly identical to it!

http://thomann.net/hv20/dereks-test/

[David Newman]

We been through all the before with others, your numbers are not correct, and remember we do stand to benefit if you are. I do not believe there is huge gain capturing 1920 vs 1440 with the HV20. Blow an image up to 400% pixels is not a good test, is all you doing is showing apparent sharpness, not detail. Sharpness and detail are separate factors, you are confusing them. The scaling tends to reduce apparent sharpness, but if the detail is not really there, it is not lost (don't for confuse noise for detail either, that is easy to add back.) None of the 25% further loss nonesense. If you correct for apparent sharpness you will see very little detail difference. Of course if you must have 1920, we have a product for you. If you don't intend to correct for sharpness, and want the camera output as is, then you should go for 1920, but sharpness control should be is part of the finishing workflow (if you are using HV20 for highend work.)

[Peter Moretti]

David, one thing I don't understand about Neo HDV versus HD is the whole 8-bit versus 10-bit color argument.

I realize that 10 bits leaves more room for rounding during color correction. But if my NLE supports 10-bit projects, can't I just specify a 10-project project even with 8-bit source media?

[David Newman]

Peter, that is correct, and we aren't saying anything different as the issue is upon render or 10-bit+ color correction; 99% of all exports formats are 8-bit and can introduction banding. The trick is to keep all intermediate renders in a 10-bit format and then dither from that to 8-bit if needed for distribution.

[Mike Thomann]

Coming from a pro photo background, blowing an image up is standard practice to more easily compare detail. As home theater screens become bigger and more refined, what we see when we blow things up now will become more and more apparent in the future. All numbers and calculations aside the images tell the story. I'd agree that much of the difference is attributed to sharpness, but there are many places in the footage where I can clearly see more "detail". Moreover, on a large screen NeoHD looks vibrant and with good depth where NeoHDV looks flatter in comparison. Here is an example which shows that while retaining sharpness may play a large role, there is a detail difference as well. The example pretty much speaks for itself, but I will point out that in this comparison the uncompressed clearly retains more detail, enough to be able to distinguish the separate white lines that form the circle of the logo. More detail can also be identified within the highlights. And retaining original sharpness has its own merit, as it shows more vibrant and with more depth...

http://thomann.net/hv20/dereks-test-2/

[David Newman]

I suggest you do you our tests, using today's images, not something from last year posted by someone else (Derek made a few errors in his testing, had some field handling issues and compressed the CineForm images more heavy than we recommend) and try some of the things I'm saying. With sharpness matching this two images have pretty much the same detail. If not, please purchase NEO HD or Prospect HD. ;)

[Mike Thomann]

Yes, nothing is definitive from just one set of tests. Derek's samples have offered more than anything else available. One thing is for sure, everything considering here is way better than HDV. Actually I'm eagerly awaiting CineForm SOLID for the EX1! Thanks.

[Ed Hecht]

Originally Posted by David Newman

Note: Aspect HD is only licenced to 1440x1080 16x9, which is still plenty for the HV20 (we have proved in the past it is near impossible to tell the difference.) If you want 1920x1080 you will need either Prospect HD, or NEO HD (if you not using Premiere.)

In HDLink under Prefs:

For 24p footage, check removal 3-2 pulldown. Nothing else needs to be set. For 60i, leave pulldown removal off.

So the only controls you need to think about is the quality level (start with High) and capture location.

That is it.

David,
I finally got around to doing this and the captured footage looks primo (in Windows Media Player). So for Premiere Pro CS3, what should my project settings be for 24F and 60i respectively? Thanks.
Ed

[David Newman]

A CineForm Preset that much your footage.

For Aspect HD - 1440x1080 at 24p or 1440x1080 at 60i.
For Propsect HD - 1920x1080 at 24p or 1920x1080 at 60i.