SD vs HD Comparison

[Fred Hess]

I've been wanting to create a side by side comparison between my old Tobin machine (which I've used for many years) and my DIY project using a modified Sankyo Dualux 1000 projector and a Panasonic HDC-HS9 camera. I thought I would share the results with all of you.

My biggest concern is being able to focus the DIY as well as the Tobin. As you can see, the focus on the Tobin (on the left) is much better. It's quite obvious around the 28 second mark. I just can't seem to match that with the DIY (and I've fiddled with it for many hours).

I was also very curious about how SD compared with HD. They really are fairly close in my opinion. Better detailing with HD of course but I think either way works as a finished project for customers. With some work in DaVinci Resolve either one could probably be improved even further.

FYI, the Tobin was captured into VirtualDub as a lossless AVI with Neat Video added. The DIY was captured into OBS (non lossless). In Resolve, I used the AutoColor option on both videos. That's it for the editing.

Anyway, I'd love to hear your comments and any advice you can offer regarding improvements to either method.

https://www.tonalproductions.com/A-v...sidebyside.php

[Bill Hauss]

I would assume it's the AutoColor option in resolve that blew the highlights out however if not please correct me on this... I advise against neat video, denoising is destructive to fine detail and overall makes the image worse imo

[Alwyn Adkins]

Like all denoisers, they can be destructive if overused. For an almost one-click option though, Neat Video is fantastic but does need to be wound back a bit at times. The secret is to find a clear area, with no colour variations or objects, to build the profile on. The clear sky, a plain wall, such as at 47.45 (the sky) yields the best profile results.

My biggest concern is being able to focus the DIY as well as the Tobin.

I must admit this has been a disappointment for me too. Overall, I don't like the HS9; the manual focus function is very difficult to use. Stan Jelavic has some good focus tips here:
https://8mmforum.film-tech.com/vbb/f...9772#post79772
I've found that using the RPI lens to slide in and out allows very tight focus (with the HS9 focus all the way in and zoomed just enough to fill the screen). I have put a texta stripe along a piece of white leader film which I use to focus on before I do an actual capture run with film.

Re the redness of the HS9, I'd try the different white-balance settings. Also, a substantial amount of red can be knocked off with Colormill in VDub. A basic colour wheel adjustment away from Red will also improve it (I don't use Resolve but perhaps a manual adjustment would be better than "auto").

To be picky, there's a fair amount of interlacing in the Tobin capture that could be cleared up.

You're probably not achieving anything by exporting at 60fps; they both have duplicated frames so it won't be smoother at 60 vs 30.

Overall though, at full speed, they look good, although the Tobin in your sample is definitely nicer to look at, but with the red reduced, the DIY comes close.

Out of interest, what is your DIY workflow? Cabling? Digitiser/capture card?

Edited to add: probably a more accurate comparison would be using VDub for capture for both workflows. A lossless capture will always produce a better outcome than a direct-to-MP4 workflow using OBS.

[Fred Hess]

I would assume it's the AutoColor option in resolve that blew the highlights out however if not please correct me on this... I advise against neat video, denoising is destructive to fine detail and overall makes the image worse imo

The AutoColor could very well be doing that. I'm not well-versed in color correction yet, but I am beginning to experiment with it. I hope to get better using color correction so that I can improve the issues that come up.
I believe Neat Video has it's place. While it is destructive, there are many videos I've done where the end result is much better using it. There are tradeoffs and I always compare the results to see if the end result merits it's use. Sometimes yes, sometimes no.

Alwin:
I will continue to play with the focusing issue. Thanks for the tip.

I would prefer to capture everything in VDub, however, I purchased an AVerMedia capture card which doesn't work in VDub. Works fine in OBS so that's what I'm using now. I'm looking for an HDMI interface that VDub recognizes. Know of any?

Yes, I'm sure the AutoColor isn't the answer. Just tried it to see the results. I'm just beginning to play with color correction. Hopefully I'll get better in time.

Regarding interlace/de-interlace, I'm still confused about how/when to de-interlace. The more I read up on it, the more confused I get. Most of my customers want the video on a flash drive that they can plug into their TV. As I understand it, TVs will de-interlace automatically so I've never worried about it. And then there's the issue of the best method to de-interlace and when you should or shouldn't de-interlace. Perhaps someone could explain that to me better when it comes to film.

For the DIY, I'm using the HDMI out from the camera into the AVerMedia card and capturing in OBS. The Tobin has S-Video out which is connected to a Pinnacle USB-710 and captured in VDub. In most cases, after editing, I use Hybrid to produce the final file. I'm looking forward to capturing both in VDub in the near future.

I'd be very interested in hearing about other's methods for capturing and rendering. How do you go from film to the final product (in my case, MP4) with a real-time setup? I know there are lots of ways to do it. Just wondering what others consider a good/correct transfer method.

[Bill Hauss]

Regarding interlace/de-interlace, I'm still confused about how/when to de-interlace. The more I read up on it, the more confused I get. Most of my customers want the video on a flash drive that they can plug into their TV. As I understand it, TVs will de-interlace automatically so I've never worried about it. And then there's the issue of the best method to de-interlace and when you should or shouldn't de-interlace. Perhaps someone could explain that to me better when it comes to film.
.

If you capture in progressive than this is not an issue, deinterlacing is a video artifact and if you're capturing on virtualdub it should not be an issue, I see no interlacing at first glance.

[Fred Hess]

Another comparison. Instead of the Panasonic, I used a Canon HF-20. I can focus a little better with this it seems. This is what the Canon and Panasonic look like side by side. Canon is on the left. No color correction used this time. Any thoughts?

https://www.tonalproductions.com/A-v...idebyside2.php

[Bill Hauss]

Highlights still seem a bit blown out, you can always increase highlights but recovering blownout details is impossible.

[Fred Hess]

How does one go about stopping blown out highlights when capturing film? Is it in the camera settings or the capture software? Many times I get 5 or 7 inch reels and have no idea what's coming next from 1 splice to the next. Can you make adjustments to prevent the brights and darks from overwhelming the sensors?

[Alwyn Adkins]

deinterlacing is a video artifact and if you're capturing on virtualdub it should not be an issue, I see no interlacing at first glance.

Interlacing is a fundamental method of displaying video, not an artifact. If you "burn in" the jaggies by incorrectly processing it, then it becomes an artifact.

VDub will capture interlaced video if that is what it is fed.

Interlacing in Fred's cine capture:

[Alwyn Adkins]

I'm still confused about how/when to de-interlace. The more I read up on it, the more confused I get. Most of my customers want the video on a flash drive that they can plug into their TV. As I understand it, TVs will de-interlace automatically so I've never worried about it. And then there's the issue of the best method to de-interlace and when you should or shouldn't de-interlace. Perhaps someone could explain that to me better when it comes to film.

Yes, a bit of a head spin. TVs do a very good job of deinterlacing but I prefer just to deinterlace with QTGMC regardless. You're using Hybrid already; it has QTGMC incorporated and it will also double the frame rate (29.97 to 59.94) when it splits the fields, resulting in noticeably smoother video.

All this depends, of course on whether you're actually capturing Interlaced. If you're using OBS, you can set it to deinterlace during capture (which it doesn't do without the occasional minor error). Also, the HDMI output from your camera (HS9, at least) can be set to "i" or Progressive. I haven't worked out what is going on with my PAL HS9; it is set to 1080i output, but it looks to me like it is outputting Progressive.

How does one go about stopping blown out highlights when capturing film? Is it in the camera settings or the capture software?

Either/Or, but it is critical that the camera doesn't have too-high levels. This can be done by having the manual aperture control displayed so you can blip it when the scene changes. Alternatively, you can use the Proc Amp controls on your digitiser, but you must make sure the levels coming in from the camera are low enough not to cause blowouts initially. You may find this on my website interesting.

I see a definite red tinge in the HS9 video again. That's interesting because mine produces a definite green tinge. Perhaps these cams are starting to fail.

I'm looking for an HDMI interface that VDub recognizes.

Try AmarecTV. It is my go-to capture program for this type of work. For HDMI, I use a Startech USB3HDCAP. It has a good range of input formats covering all the usual combinations. The cheaper HDMI > USB boxes have less of a range of capture combinations and more noisy quality. But I would stick with the Avermedia initially until you get the capture settings right, then decide.

It all depends on how much time and effort you need/want to put into a project. A well-balanced Progressive capture with a decent bitrate in OBS will most-probably satisfy most people. Capturing Interlaced, then properly post-processing should yield better results but take more time. And then again, capturing to the SD card is probably the quickest but with little control over things like bitrate, and then you may have to re-process an already-lossy format.

[Fred Hess]

Thank you Alwyn. I'll post more later, but for now I downloaded AmerecTV and got it configured, but when try to capture I get an error message. "Can not support colorspace by video codec". I have HuffYUV installed (and works fine in VDub). I'll try lagarith and see if that works, but wondering how that error message can be corrected.

[Alwyn Adkins]

"Can not support colorspace by video codec"

Hmmm... Never seen it myself, but it has popped up a few times on the 'net. The codec doesn't appear to be relevant. Reading between the lines, I would suspect the digitiser drivers as the problem. I just tried HDMI from my HS9 with my 4 different HDMI>USB thingos and HuffYUV and all captured with no issues.

For what it's worth, here are my Huff settings with the Startech (note the version is "HUFFYUV 2.1.1"):

[Bill Hauss]

Interlacing is a fundamental method of displaying video, not an artifact. If you "burn in" the jaggies by incorrectly processing it, then it becomes an artifact.

VDub will capture interlaced video if that is what it is fed.

Interlacing in Fred's cine capture:

I was having a brain moment when I wrote this, forget it.

Now that I actually watch it the interlacing is visible.

How does one go about stopping blown out highlights when capturing film? Is it in the camera settings or the capture software? Many times I get 5 or 7 inch reels and have no idea what's coming next from 1 splice to the next. Can you make adjustments to prevent the brights and darks from overwhelming the sensors?

Exposure would be the setting I go to, this can also be prevented

[Fred Hess]

I did some more comparisons with different cameras, an HD Canon (HF20) and a 2K Canon (HFG70) and found that the Panasonic produced the best results. The Canons both had a very noticeable yellow tint. Details were pretty good on all 3 and focusing was about the same. But the Panasonic just seemed to have a much more realistic look to it. I'm wondering if that's due to the Canons being CMOS. I think the Pany will be my go-to camera in this set up.

I tried both HuffYUV and Lagarith in AmaRecTV with the Avermedia card and couldn't get either to work. So I picked up a Startech card and it worked fine. Couldn't get it to work with VDub but that's okay. The AmaRecTV is easy to use and produces lossless avis which was my goal to begin with.

Alwyn, maybe you help with this. Resolve won't import the avi file. I loaded it into VDub and hit Save as AVI (very time consuming) to produce a file that Resolve will open. How do you work with the files that AmaRecTV produces? I'd sure like to cut a bunch of time off the process.
Also, Resolve tells me that the file is Progressive, but it seems I see quite a bit of interlacing. Or is it just the blurred frames that I'm seeing? I never noticed it with the Tobin/VDub process. Maybe the better camera is showing me some new things?

Edit: The StarTech seems to be giving some bad capturing. I'm noticing lots of jagged lines (that I mistook for interlacing) whenever there is movement in the video. It does the same thing in OBS, whereas the Avermedia doesn't. Is there a way to make adjustments? Plus, Resolve crashes when trying to render the video. Arrggg!

[Alwyn Adkins]

Resolve won't import the avi file..... How do you work with the files that AmaRecTV produces? I'd sure like to cut a bunch of time off the process.

I had a brief look at some YT videos of Resolve and decided nah, it wasn't for me. Too complicated for my needs. I'm using Magix Movie Studio and the MagicYUV codec (Movie Studio is not keen on HUFF or Lagarith; sometimes they're Ok and other times not, with jerky play and weird frame inserts).

I capture and then run the file through AVISynth (QTGMC (deinterlacing), SRestore (duplicate and blends removal), speed change to 18fps and RIFE (frame rate increase up to 50fps). I then export that from VDub as a MagicYUV AVI and import it into Movie Studio, do my colour adjustments, cropping, reshaping (if my capture is a bit oblong) and resizing to 1440x1080, then export to MPEG4. The killer is the processing time creating that AVISynth AVI; processing the 1440x1080 video from the HS9 is painfully slow (4fps). I'm now experimenting with an SD capture from a Panasonic GS300 (which interestingly has exactly the same optics as the HS9). For 720x576, I get around 35fps and to be frank, the difference between the SD 720x576 from the GS300 and the 1440x1080 from the HS9 is almost indistinguishable.

For Resolve, you could try capturing in say HUFF, LAGS or UTVideo, import that into Hybrid to do your stuff there, then export it out of Hybrid in FFV1, which looks like Resolve will import .

If you wanted to go directly into Resolve, VDub 2 has FFMPEG ProRes available as a capture codec (not sure if it works though), otherwise you might be stuck with Uncompressed.

Resolve tells me that the file is Progressive, but it seems I see quite a bit of interlacing.

Is this out of VDub after deinterlacing? If it's the raw capture file, they have no real metadata so programs guess what the file is. For example, Hybrid does the same and you have to manually set the file type to Interlaced, Top Field First.

The StarTech seems to be giving some bad capturing. I'm noticing lots of jagged lines (that I mistook for interlacing) whenever there is movement in the video. It does the same thing in OBS, whereas the Avermedia doesn't.

The easy way to tell what the file actually is is to open it in VDub and apply the Deinterlacing filter (use Yadif and Double Frame Rate, TFF). For an interlaced file, you'll see the jaggies in the left pane but clean, crisp frame in the right pane. If you still see jaggies, they've been burnt in somewhere along the line and you've got big problems.

It sounds to me that the Startech is doing the right thing; you should be getting interlaced HD video from your HS9 (Setup, HDMI output 1080i).

A lot to it but when you've go it set up, you will wizz through the captures in realtime and depending on what you want to do to the AVI, you can have an MP4 published reasonably quickly.