The old "war of opinion" - Capture 8mm tapes in DV or lossless?

[Sharc]

OK, depends how we read it. Not to argue, but I understood that the OP wanted to use the DV converter of his videocam via Firewire to PC versus using the camera's S-video output and following the analog lossless capture route from there. A different setup from yours as I see it. In your comparison the tape player is the VCR and the DV converter is the SW in your PC, so you always have the USB live2 in the loop, no?

[Marvolo]

OK, depends how we read it. Not to argue, but I understood that the OP wanted to use the DV converter of his videocam via Firewire to PC versus using the camera's S-video output and following the analog lossless capture route from there.

Yes, that's exactly what I plan to do once the equipment is here.

Currently I have trouble finding a splitter that reliably removes HDCP protection. A lot of those on amazon or ebay turn out to NOT do this. From what I read, it seems to be a game of luck. Anyone got recommendations?

[DB83]

There are always 'two sides to a coin'

Now if the OP had D8 tapes there would be no discussion. But he has s8 tapes which are analogue. So the first substitution is a s8 tape to a VHS - both are analogue.


Now the D8 camera plays that analogue tape and outputs a DV signal by means of an internal conversion. I guess that not all D8 cameras were created equally and some could well give a better output than others. Now we can substitute the D8 camera with a VCR and a ADVC which equally converts the analogue signal internally and outputs DV. That is what an ADVC is - AnalogueDigitalVideoConverter. Again, not all ADVCs are created equally. I just happen to have a 300 which will have some internal filtering. Both devices now send the DV over firewire in to a PC. No digital conversion is done by the PC which simply records the incoming stream with the appropiate software. For my example I used WinDV.


Now using a D8 camera and a s8 tape would be a more 'authentic' test. But even then, as I state above, not all cameras are equal. So the only 'real' test is with the OPs equipment.


Now the OP does appear to want to test the analogue route. But if he does go down the rabbit-hole then he can not argue that he was not forwarned about the perils. And I am not sure if his choice of capture device is appropiate since BM cards have, as I have read on here, their own 'quirks'

[Alwyn]

Anyone got recommendations?

Yes, I put the one I have (that works) in post 18:

https://www.amazon.com.au/dp/B091GG183D?ref_=pe_2361882_297267902_302_E_DDE_dt_1

This looks like the same one:

https://www.amazon.de/Incutex-Splitter-Switch-Port-1080P/dp/B00LCUJX9G/ref=dp_prsubs_s...0LCUJX9G&psc=1

[Sharc]

Sorry, maybe I missed something, but we did not see a sample of the OP's new DV captures from 2023 with the blown out whites, right?
From the first post:

The only slight difference I noticed is that my cam seems to have a slightly higher brightness / gamma compared to 2008. But really only slightly. But resulting in a loss of fine structures such as details of clouds that now are just plain white areas and in 2008 you could still see a little more detail inside the cloud.

If the super whites are not clipped but just extending into the 235...255 luma range it would be easiest to readjust the levels in post processing for recovering the brights, like the details of clouds. Would be much simpler than 'going down the rabbit hole' of an analog re-capture with new gear.

@Marvolo: Worth a test, I think, but we would need a short sample of such a scene with the lost details of clouds.
The downside being that it would eventually require a (lossy) re-encode, unless the filesize doesn't matter in case of lossless or 'visually lossless' re-encode.

If the super whites are however clipped the damage is irreversible and nothing can be recovered.

[Marvolo]

Sorry, maybe I missed something, but we did not see a sample of the OP's new DV captures from 2023 with the blown out whites, right?
From the first post:.

Not yet, but I'm already preparing the files so I can upload both here later. Do you prefer them to be both in one video side by side, or each version as a separate file? So, the 2008 version by the unknown TV engineer and my own attempt from 2023.

[DB83]

Maybe I also missed somat inasmuch that I do not see a single sample from the OP in any post.

He mentions percieved detail change but no actual visual evidence of it.

[Sharc]

Sorry, maybe I missed something, but we did not see a sample of the OP's new DV captures from 2023 with the blown out whites, right?
From the first post:.

Not yet, but I'm already preparing the files so I can upload both here later. Do you prefer them to be both in one video side by side, or each version as a separate file? So, the 2008 version by the unknown TV engineer and my own attempt from 2023.

Separate files would be fine. A few seconds duration including the critical scene should normally be enough for a quick analysis.

[Marvolo]

@Sharc & @DB83, and everybody else interested:

like I wrote in my previous post, I haven't provided sample files yet. But here they are. They are matched by frame. I put both captures in my NLE on top of each other and matched them frame by frame. That way I'm able to compare them frame by frame by muting one track and see how the image changes after muting.

That way, I can see that my own 2023 capture seems to be a little bit brigther / the gamma seems to be a little bit higher than the 2008 one. Interestingly, such differences are only visable in certain scenes (usually outdoors). Other scenes are literally identically with no change between 2008 and 2023 whatsoever.

I rendered a couple seconds out in DV-AVI (making sure my NLE doesn't re-encode as the target format is the same as the input-format, my NLE calls this "Smart Rendering") for each captures:

2008 capture: Here
2023 capture: Here

The good thing about re-capturing, even though most of the time I don't see a clear difference in picture, is that I can eliminate errors such as at 20 seconds in the 2008 capture (small horizontal line, visible for the length of a frame) which usually aren't there in the 2023 capture. So I gained new information from the 2023 capture that I can use to replace faulty frames with in the 2008 one and vice versa.

I did the 2023 captures with a Sony Digital8 TRV-730E (which also has analogue-in and digital-in - so could be used as an A/D converter for analogue signals coming in and outputting them either via DV or analogue).

[Sharc]

Your 2023 capture is as good as the 2008 variant IMO.
Here just a quick and dirty levels alignment of your 2023 capture to make it look similar as the 2008 variant; nothing optimized yet.
I left it interlaced as the original capture, re-encoded to AVC x264 and removed the audio to keep the file size small.

[Marvolo]

What I find weird is those fine parallel white lines throughout the entire image. It doesn't seem to be a player thing as both captures (with presumably different players) have it. It seems as if those white parallel lines must have been recorded onto tape by the original cam for some reason.

Thanks for the modification file.

Now let's see what analogue re-capturing via S-Video and the Panasonic player and the Blackmagic Intensity card may yield..

Does anyone know what that slightly purple banding on the right hand side of the video is? In some captures it's green, here it's purple.

[Marvolo]

Your 2023 capture is as good as the 2008 variant IMO.

Despite there being 15 years between the last capture. Seems like the tapes haven't deteriorated visibly. Also seems as if that "professional engineer" in 2008 didn't use magic equipment after all when I can more or less yield identical results with a simple camcorder-Firewire-Setup... We paid a lot of money back then to have the tapes transfered professionally.

[Sharc]

What I find weird is those fine parallel white lines throughout the entire image.

Yes, can be improved as well. I just made it look "most similar" to the 2008 without applying any more fixes or improvements, nor did I deinterlace.
Do you still spot differences in details retention?

[Sharc]

Does anyone know what that slightly purple banding on the right hand side of the video is? In some captures it's green, here it's purple.

Typical border effects of analog video. Normally one would crop or mask all these border effects (left,top, right, bottom).

[Marvolo]

Do you still spot differences in details retention?

If I compare the original 2008 and your modified version frame by frame I can see very slight differences in colour. Especially in the clouds: yours seem to be a tad more blue and the 2008 original one seems to have soft purple specs:

2008:
[Attachment 71856 - Click to enlarge]

2023 modified:
[Attachment 71857 - Click to enlarge]

[Sharc]

You may always tweak the colors as you like. Your NLE probably comes with its own color grading tools.
As I wrote I just shifted the excessive luma (Y) a little back (not as much as one should actually), which already brought the details (e.g. the sand on the beach) back, and encoded to AVC (x264) with CRF 15. That's all.

Btw I have 30 years old VHS tapes which are still in good condition. Depends on the storage conditions, and non-excessive wear of course.

[Marvolo]

How come that my camcorder makes the image a tad brighter than the one used in 2008? Is this just a player kind of thing? I looked, but I can't seem to be able to set brightness or hue manually for playback... I can only set TBC on and off as well as Digital Noise Reduction.

I played with these 2 options, but none of these had any visible effect. No matter if it was turned on or off.

[Sharc]

How come that my camcorder makes the image a tad brighter than the one used in 2008?

Different factory alignments (analog tolerances), different firmware, different chipsets, different taste, different model/brand ...... you name it. My Panasonic Handicam has a significantly different tint than my SONY etc. Also, you don't know which tweaks the 2008 shop applied, maybe you advised him to tweak it a little......and so on. And you don't know which one is "right" or "correct", do you? Have you seen 2 color TVs in a shop producing exactly the same picture

[Marvolo]

What do you mean by knowing which one is "right" or "correct"?

As far as I remember, no orders about tweaking whatsoever were placed in 2008. Only the mere capture process. Not even the images were cropped to cut out the unwanted parts of analoge video. I assume it was just recorded right off the tape and that's it.

[Sharc]

What do you mean by knowing which one is "right" or "correct"?

What is your reference? How do you calibrate your monitor, TV, settings etc. How blue should the sky be? How greenish or blueish the water....? How dark or bright the shadows? Contrast? It's a lot about personal preference. Trust YOUR eyes.

[Marvolo]

What do you mean by knowing which one is "right" or "correct"?

What is your reference? How do you calibrate your monitor, TV, settings etc. How blue should the sky be? How greenish or blueish the water....? It's a lot about personal preference. Trust YOUR eyes.

Yes, I have no idea how it should look. I have no reference monitor or callibration tools. I don't even know if it is within the color space range of analogue video.

[Bwaak]

Also seems as if that "professional engineer" in 2008 didn't use magic equipment after all when I can more or less yield identical results with a simple camcorder-Firewire-Setup... We paid a lot of money back then to have the tapes transfered professionally.

Capturing into DV is about simplicity and reliability. The "engineer" should have been professional enough to use a working machine and a correct cable OTOH, interlaced DV is not what most people want now.

[johnmeyer]

[QUOTE=Sharc;2694291]

The lossless Huffyuv looks flawed to me: Blended fields.
<snip>
Edit: Just noted that the DV is also blended. Maybe the movie should have been IVTCed at some stage instead of poorly framerate converted?

In my one post above, I mentioned that I sensed that something was wrong with the motion in some of the captures and wondered if they had been converted during capture to progressive. I warned that this is a really bad idea to do during capture (do it later, if you must), and if it was done with field blending, then that confirms my fears.

[Sharc]

What do you mean by knowing which one is "right" or "correct"?

What is your reference? How do you calibrate your monitor, TV, settings etc. How blue should the sky be? How greenish or blueish the water....? It's a lot about personal preference. Trust YOUR eyes.

Yes, I have no idea how it should look. I have no reference monitor or callibration tools. I don't even know if it is within the color space range of analogue video.

Of course there are technical restrictions and conditions which must be fulfilled. For example consumer video should be "limited range" or "TV range" in YUV (Y'CbCr) color space, means for example that the luma (Y) should be in the 16 (black) ...... 235 (white) range. Means the waves in the picture below shold not extend much into the brown bars (representing luma 0...15 and luma 236....255 respectively). This condition was not fulfilled with your 2023 capture, see the picture below. The luma (Y) went up to 255 which was the reason why you lost details in the brights when the YUV signal got converted to RGB for viewing on your monitor.
Your NLE has most probably color tools to visualize the luma and/or RGB parade. Otherwise it's readily checked with Avisynth's histogram() function (see attachment).
It can be easily fixed with Avisynth for example; that's the modification I did to bring the details back.

As I said, your 2023 DV capture is basically ok, just set the brights level a few (say 12) notches lower. It can be done with proc-amp adjustments during capture or in post processing (as I did).

[Marvolo]

Anyone got recommendations?

Yes, I put the one I have (that works) in post 18:

https://www.amazon.com.au/dp/B091GG183D?ref_=pe_2361882_297267902_302_E_DDE_dt_1

This looks like the same one:

https://www.amazon.de/Incutex-Splitter-Switch-Port-1080P/dp/B00LCUJX9G/ref=dp_prsubs_s...0LCUJX9G&psc=1

Hi Alwyn,

thanks for the recommendation. I'm a bit worried if this type of splitter may work together with my Panasonic DMR-EH575 which I bought cheaply off ebay. In the German video forum, they reported that splitters which supported a higher version of HDCP such as the most recent 2.2 version are not downward-compatible to older versions of HDCP / HDMI such as 1.3 or 1.4. Users didn't get any image from these splitters when connected to a player that had a lower version of HDCP.

That's why I checked the details in the (German) amazon link you posted: the splitter there supports HDCP 1.0 up to 1.2.
My panasonic above was produced some time early 2007 or so I read. In 2006, HDMI standard 1.3 was released. So I assume the player most likely uses 1.3 standard as well.

The splitter above only covers up to 1.2.

Assuming that HDCP versions and HDMI versions are the same thing??

[Alwyn]

That's why I checked the details in the (German) amazon link you posted: the splitter there supports HDCP 1.0 up to 1.2.
My panasonic above was produced some time early 2007 or so I read. In 2006, HDMI standard 1.3 was released. So I assume the player most likely uses 1.3 standard as well.

I have an EH-57, sticker on the back says September 2007, and an EZ-48, sticker says September 2008, and both work fine with my splitter. Taking an educated guess, I'd say you'll be fine with a earlier model splitter.

Here just a quick and dirty levels alignment of your 2023 capture to make it look similar as the 2008 variant; nothing optimized yet.
I left it interlaced as the original capture, re-encoded to AVC x264 and removed the audio to keep the file size small.

@Sharc, I don't want to appear ungrateful but in my view there is little point in posting example files without telling us how you did it. We're here and willing to learn, but unless you show us how you did it ie posting the code, we're just scrabbling around in the dark trying to achieve the desired result. I'm not saying "here, fix this for me" (unless I've tried myself and can't do it) I'm saying "how did you do that", because that then allows me to do it myself. Jagabo's explanation of ColorYUV on another topic is an example. That has set me off on a whole new voyage of discovery.

Posting your method/code also allows others to decide if that method is suitable for them. If it entails a hundred lines of AVISynth code, some may be inclined to use a simpler method with say VDub. But if we don't know how you did it, we'll never know how difficult it is.

For example:

just set the brights level a few (say 12) notches lower. It can be done with proc-amp adjustments during capture or in post processing (as I did).

How are you setting the brights down? Using Tweak?

How does it go, "give a man a fish, it will feed him for a day. Teach him how to fish, and he can feed himself for life".

[Sharc]

Posting your method/code also allows others to decide if that method is suitable for them. If it entails a hundred lines of AVISynth code, some may be inclined to use a simpler method with say VDub. But if we don't know how you did it, we'll never know how difficult it is.......
How are you setting the brights down? Using Tweak?

Oh sorry, I thought this has been explained so many times I would only pour a pint of water into the ocean.
It can be done in Avisynth with Tweak() or ColorYUV() or (Smooth)Levels(). I used SmoothLevels here:

AVISource("2023_DV.avi")
Smoothlevels(input_high=255,output_high=243)  #bringing the high level 12 notches down
Histogram() #to show the waveform for adjusting the parameters

As you see from the picture below there is still little luma reaching out into the brown bar, hence the output_high should actually be set a bit lower than 243 (like 235), recovering few more details in the brights. I used 243 in this case just to make it look much the same as the OP's 2008 'professional' capture.

Added:
And, btw., to eliminate the horizontal stripes one could use:

AVISource("2023_DV.avi")
Smoothlevels(input_high=255,output_high=235) #setting the max level correctly to 235 here
separatefields().vinverse().weave()  #remove the horizontal stripes
Histogram() #to show the waveform for fine tuning

See Attachement 2 with the horizontal stripes removed (still interlaced though).

[DB83]

With respect, we do not know the level of experience the OP has with post-capture processing. Maybe he just wants the 'best' from the initial capture since avisynth has a steep learning curve.

I am more curious as to what the new setup will achieve on its own and how it compares with what he can already achieve.


But if I read that detailed guide correctly he will NOT get out of it what his topic is all about >> a comparison between DV and lossless. We know that DV is 4:2:0 (same as dvd btw) 13 gb per hour. Lossless should give you 4:2:2 at 35-40 gb per hour. I just await the reaction when he asks why his files, and still 4:2:2 are so much bigger.


And why use a sledge-hammer to crack a nut ?. The author is so wrong when he states that only his set up will give correct analogue transfer. Totally dismissive of a usb method whereas the quality ones will give desired result and at a fraction of what I dread to think he is expending at present.

[lordsmurf]

capture seems to be a little bit brigther / the gamma seems to be a little bit higher than the 2008 one

This is very often due to lack of TBC.

For most people who aren't perfectionists, given that they are starting with VHS/SVHS/8mm/Hi8mm which are all pretty lousy to begin with,

This is an excuse, not a reality.

To state "perfectionist" is honestly a BS putdown again quality methods.
As in:
- "only a perfectionist cooks on the stove, all we need is a microwave"
- "only a perfectionist needs to brush his/her teeth, all we need is gum"

It's just damned lazy, and uncaring of consequences. Call it like it is.
Sometimes cheap -- but ironically usually more expensive for lesser gear, lesser quality.

DV is old tech, 1990s old, add noise, alters color, loses color. it is what it is. At the time, it was all we had. Now it's extremely inferior.

With respect, we do not know the level of experience the OP has with post-capture processing. M.

I agree, info overload. Worry about getting the quality capture first. Walk before run.

[Sharc]

Totally dismissive of a usb method whereas the quality ones will give desired result and at a fraction of what I dread to think he is expending at present.

Sure. But to each his own, and it is not forbidden to experiment and to strive for the 'best', whichever criteria the 'best' and 'quality' are based upon
Most important is that one should understand what one is doing and draw one's own conclusions.