DV conversion vs lossless capture - comparison (Hi8) [WARNING: auto-load!]

[Brad]

NTSC Hi8 played via Sony DCR-TRV340 with DNR Off & TBC On.

Screenshots vertically padded to 720x486 to match the second lossless capture, then LanczosResize'd to 1440x1080. The pre-resized shots are also attached at the bottom of the post.

In the DV transfer, the video was shifted up by one line after the first "scene cut" on the tape. I tried twice and this happened both times. I've shifted it down here, to match the others.






As you can see, the camcorder's DV output and the first lossless capture have raised blacks and blown-out brights.

I've included two video samples. It's the same portion of the tape as the first image set. I can add more videos if desired, but I think these tell the story fairly well.

Script to align the attached samples for pixel-peeping:

AVISource("DV vs lossless - Wedding P 13-30 - WinDV.avi",pixel_type="RGB32")

# Entire DV frame is shifted up shortly after first scene cut
AddBorders(0,1,0,0,color_white).Crop(0,0,-0,-1)

DV = last

###

AVISource("DV vs lossless - Wedding P 13-30 - DMR-ES25.avi")
ConvertToRGB().AddBorders(1,0,0,0,color_white).Crop(0,0,-1,-0) # Match DV horizontal framing

Y2 = last

###

Interleave(DV,Y2)

[jagabo]

Thanks for the samples but I don't understand the captures. I assume the WinDV video is the Hi8 tape played on the camcorder and captured as DV. What exactly is the DMR-ES25 video? Tape played on the camcorder, the DVD recorder used as a passthrough then captured with a raw YUV device, compressed with UT?

I'd prefer you NOT resize the PNG images before uploading. Resizing can be done locally, in the browser, using screen magnifier, or after downloading the images.

[Brad]

Does Firefox offer in-browser upsizing? I thought it only did zoom-out.

WinDV.avi = what you said
DMR-ES25.avi = camcorder <YC> Panasonic DVD recorder <HDMI> Blackmagic Intensity Pro 4K

So, Panasonic as ADC only without the final DAC passthrough.

[jagabo]

Does Firefox offer in-browser upsizing? I thought it only did zoom-out.

Yes, Ctrl+= to zoom in, Ctrl+- to zoom out, Ctrl+0 to go back to 1x. Same as IE.

WinDV.avi = what you said
DMR-ES25.avi = camcorder <YC> Panasonic DVD recorder <HDMI> Blackmagic Intensity Pro 4K

So, Panasonic as ADC only without the final DAC passthrough.

Thanks for the clarification.

The DV cap is losing a bit of fine detail. Even after running noise reduction filters like TemporalDegrain() there's more detail in the UT cap.

[Skiller]

As you can see, the camcorder's DV output and the first lossless capture have raised blacks and blown-out brights.

After a levels correction recovering the superbrights (235...255), I find that the DMR Capture is not really missing any details for the first scene (same for the DV cap), actually the third cap seems a little too dark (there is no white, the brightest value is at about 235, whereas it should be 255 for a PNG screenshot).

So I applied a Levels correction of 16...255 to 16...235 to the DMR-ES25 capture, comparing it to the third capture (NHX-E2) using your PNG screenshot.

AVISource("DV vs lossless - Wedding P 13-30 - DMR-ES25.avi")
ConvertToYV24().AddBorders(1,0,0,0,color_white).Crop(0,0,-1,-0) # Match DV horizontal framing

SmoothLevels(16,1,255, 16,235, Chroma=0, HQ=True)

ConvertToRGB()

Subtitle("DMR-ES25, Levels corrected 16...255 –› 16...235")
ShowSMPTE()

Addborders(0,5,0,1, color_white) #486

Now the ES-25 doesn't seem to be missing any details in the brights anymore (at least in this scene).


However, the second scene might be better suited to judge this ultimately (raw sample needed^^).

[jagabo]

After a levels correction recovering the superbrights (235...255), I find that the DMR Capture is not really missing any details for the first scene

I thought that was too obvious to mention. Anyone who's going to the trouble of capturing lossless is going to be filtering so adjusting levels to get brights out of the 235-255 range is a non-issue. There is a tiny bit of clipping in the very brightest areas, maybe 1 or 2 Y units.

[Brad]

I find that the DMR Capture is not really missing any details for the first scene (same for the DV cap)

Well, the DV and ES25 both hard-clip at Y=255/254.

I see the NHX-E2 images are also hard-clipped. The waveform monitor built into Edius is quite misleading, and I wasn't focused on levels for this comparison. I'll have to redo the capture using its Manual Gain control instead of its Contrast control, or use another capture device to show what's missing.

Ctrl+= to zoom in, Ctrl+- to zoom out, Ctrl+0 to go back to 1x

I like to open each image in a new tab and scroll between them. So for me at least, doing a zoom-in six times on each individual tab isn't a time savings vs downloading the huge PNGs.

[jagabo]

I like to open each image in a new tab and scroll between them. So for me at least, doing a zoom-in six times on each individual tab isn't a time savings vs downloading the huge PNGs.

It's not a matter of time saving. It's a matter of viewing the actual captured images, not some distorted, resized, version. This time you included the originals so it's not a problem. But in some of your other comparisons there have only been resized images.

[Brad]

However, the second scene might be better suited to judge this ultimately (raw sample needed^^).

I've attached samples of this scene ("24:32") to the OP now. The levels in the NHX-E2 recapture were in range during the wedding portion, but here they're a little over. It doesn't really matter; ultimately, none of the three methods grabs discernibly more bright detail than its competitors.

The playback camcorder seems to be to blame for this. Here's a field from a Video8 tape that I had captured earlier. With TBC Off, there is a little less clipping. But it's clear that the analog output always goes through a digitization stage regardless (chroma is always DNR'd, and hard borders are masked onto both sides). Perhaps a non-TBC Hi8 player would reveal a little more at the upper end.

[dellsam34]

Why comparing apples to oranges, You should be comparing DV compression vs MPEG2 compression, For the average user that difference don't mean nothing, for the pros that do it for a living they don't use DV devices so I would like to know what is the purpose of this comparison.

[KarMa]

@vaporeon800 post #9
What are you using to get that scope information? If it's avisynth can you post the code to get all that?

Why comparing apples to oranges, You should be comparing DV compression vs MPEG2 compression, For the average user that difference don't mean nothing, for the pros that do it for a living they don't use DV devices so I would like to know what is the purpose of this comparison.

There are people on this board that advocate DV for transfer/archival purposes. No harm in comparing, seeing how much you may lose. And DV uses 4:1:1 while MPEG2 encoder chips usually use 4:2:0, so that's also apples and oranges.

[dellsam34]

@vaporeon800 post #9
What are you using to get that scope information? If it's avisynth can you post the code to get all that?

Why comparing apples to oranges, You should be comparing DV compression vs MPEG2 compression, For the average user that difference don't mean nothing, for the pros that do it for a living they don't use DV devices so I would like to know what is the purpose of this comparison.

There are people on this board that advocate DV for transfer/archival purposes. No harm in comparing, seeing how much you may lose. And DV uses 4:1:1 while MPEG2 encoder chips usually use 4:2:0, so that's also apples and oranges.

Only for NTSC, PAL DV is 4:2:0

[Skiller]

It doesn't really matter; ultimately, none of the three methods grabs discernibly more bright detail than its competitors.

I agree, so far there isn't more detail captured in the brights in any of the three.

One thing that would bother me is the (albeit slight) posterization in the NHX-E2 cap. It's not visible to the eye in the samples because these are bright scenes; and even in dark scenes there is probably enough noise to mask it, but it's there.



One could say the NHX-E2 capture is not using full 8 Bits per channel.
I found that this happens with most capture devices if you put their "Proc-Amp" settings on anything other than the default.
It seems that these settings do not affect the levels during the analog-digital conversion but rather after (which renders them pretty useless because you cannot recover any clipped brights this way; only squeeze, expand and shift what you already have). Actually the only capture hardware I know of where the Proc-Amp settings actually do affect the levels during ADC is the Canopus NX.

[FLP437]

What hdmi setting was used for transfer to the black magic, 480i ? have you tried also the usb3hdcap any difference ?

What exactly has been your last workflow, camcorder <YC>NHX-E2 .I assume you have captured svideo in the NHX-E2 with or without the dvd recorder in pass- through or a external tbc.

Have you arrived to a final conclusion namely related to the 2 last captures using the dvd recorder as ADC and the NHX-E2 as ADC. Any of them could be an optimal setup ?

[Brad]

What are you using to get that scope information? If it's avisynth can you post the code to get all that?

It's a combination of a few different scopes in Avisynth, using a script I hacked together. I've posted it before, but not the current version, which I've attached. If the input video is any width other than 720, one of the text labels will be misaligned.

Actually the only capture hardware I know of where the Proc-Amp settings actually do affect the levels during ADC is the Canopus NX.

I'm not sure if it's clear, but the NHX-E2 is the PCIe version of the Canopus NX. The VC500's Contrast control recovers clipped brights without adding notable banding, IMO. I assume what it's actually doing is 10-bit A/D -> proc amp -> 8-bit D/D.

What hdmi setting was used for transfer to the black magic, 480i ?

Ah, well, this is a tougher question to answer than you would expect. Short answer: I would use 480i if I could.

Long answer: The lowest the DMR-ES25 will go is 480p. This can be reinterlaced to 480i. The Intensity Pro 4K accepts 480p, reinterlacing and letterboxing it to 486i. Now you have to be very careful about reinterlacing because you want to grab all-original fields rather than all-interpolated fields. Which fields the IP4K grabs depends on which frame happens to hit the capture driver at the moment I start the capture. To ensure I'm grabbing originals, I start off the captures with the DVD recorder tuned to an analog TV channel displaying pure noise. This makes it painfully obvious whether it's grabbing original fields, or interpolated garbage. If I started the capture at the wrong time, I just re-try until I get what I want.

It's annoying, but actually less-so than filling up my HDD with 480p and reinterlacing after the fact.

have you tried also the usb3hdcap any difference ?

There is an odd difference; see here.

What exactly has been your last workflow, camcorder <YC>NHX-E2 .I assume you have captured svideo in the NHX-E2 with or without the dvd recorder in pass- through or a external tbc.

Direct from camcorder to the card.

Have you arrived to a final conclusion namely related to the 2 last captures using the dvd recorder as ADC and the NHX-E2 as ADC. Any of them could be an optimal setup ?

I would not use the NHX-E2 unless I really cared about grabbing those extra NTSC scanlines for some tape, if for no other reason than the Edius program's quirks. At the start and end of playback (when the camcorder switches between blue screen & actual video), and most times there is a "scene cut" on the tape, Edius creates separate "divide" files. There is no way to force it to capture everything in a single segment. Additionally, it sometimes skips frames at these points, and sometimes repeats random frames from earlier in the tape before stabilizing.

In general, the Panasonic isn't optimal either, because it clips brights (the info on that is in the link above; my response to Skiller). This appears to be a general problem with overexposed video sources, including PAL. Here, that didn't turn out to matter, as it seems to me the player is clipping them before anything else has a chance to.

[FLP437]

Interesting, as soon as I have time I will upload some samples using my LG HDD DVD Recorder RH199HS ( accepts firewire /svideo/composite and output almost everything from any of the inputs , Hdmi (720(1440)x576i, 720x576p, 1280x720p, or 1920x1080i ) component I or P , scart RGB, svideo or composite )


The LG HDD/ DVD recorder is a 2006 device probably one of the best LG dvd recorders I think only sold in Europe and probably Asia with one of the news LSI chips (IC 1101 DMN8653 LSI) and a Philips 7137d a/v decoder . I will try the captures/ transfers with this device from svideo or firewire outputs from a digital8 camcorder playing a video8 tape and will do the transfer via hdmi 576i (PAL) through an startech usb3hdcap to compare with a direct capture from a Ati hd 600 usb or an Osprey 210e

[dellsam34]

Does it have a TBC function?.
If you are playing an analogue tape with a digital camcorder using the firewire port the output to your DVD recorder is going to be DV format, try to capture using analogue ports.

[FLP437]

Yes it as . I sometimes use it instead of the avt 8710 as a svideo pass through. It stabilizes the video and there are no drop frames any more .In terms of characteristics is very similar to the Panasonic DMR-EH65 , however the internal circuitry of both and final results I don’t know.

Well, use of the camcorder ADC to do the A/D conversion , transfer the DV output to the dvd recorder, transcode internally and after record the hdmi stream is a sort of esoteric workflow more a “proof of concept”. Pal DV is 4.2.0 and svideo is 4.2.2 at the very least I will be loosing in terms of colour sub-sampling . Even so I am interested to see if the dvd recorder is able to improve or not the process.

However the only workflow with interest will be the camcorder digital8 / svhs deck running analog tape- svideo – Dvd recorder – Hdmi576i – startech usb3hdcap/black magic shuttle/4k – virtualdub – Avi /Huufyuv and analyze results with the ones from established workflow and capture cards .

[FLP437]

I made two samples captures

workflow 1
Digital8 camcorder sony DCR-TRV238E tbc on, nr off - video8 Pal tape
LG HDD DVD recorder RH199HS set to hdmi 576i (PAL)
Startech usb3hdcap

camcorder->svideo->LG DvD rec -> Hdmi-> usb3hdcap->Vdub -> Huufyuv

workflow 2
Digital8 camcorder sony DCR-TRV238E tbc on, nr off- video8 Pal tape
AVT 8710
ATi Hd 600 usb

Camcorder -> svideo-> AVBT-8710-> ATi 600 usb -> Vdub -> Huufyuv

the video8 tape is a 23/24 years old tape not in a particular good shape due to have been extensively played
In the ati capture the three frames inserted as been due to the cancorder activation hiccup, this situation didn´t occurred in the LG workflow

[jagabo]

There's a fair amount of horizontal jitter in both k23 caps.

[FLP437]

Ok, can the horizontal jitter be tracking cancorder issues?

There is any evidence of benefit related to chroma , noise ,.. due to dvd recorder filter function in the specific workflow?

[Skiller]

Yeah, there's still jitter in both samples but since there is no perfect static part in it (like a date code for example), the right black border is my only orientation in terms of jitter.

But what's overwhelming here is: both differ greatly in contrast and color (ATI600 cap has a green tint – I suspect the AVT 8710 TBC is the cause of that).
Both caps have banding issues which the histogram reveals (AVT/ATI600 cap is worse here).

[jagabo]

Ok, can the horizontal jitter be tracking cancorder issues?

The horizontal jitter is a result of the heads of the VCR spinning faster and slower with time. This is common to all consumer analog video tape formats. You need a line TBC to reduce it.

There is any evidence of benefit related to chroma , noise ,.. due to dvd recorder filter function in the specific workflow?

The ATI cap has less contrast, less saturation, and more noise. But you can probably adjust that with the proc amp controls. The chroma channels are 1 or 2 units higher too. That could be adjusted in post without any loss (ColorYUV(off_u=1, off_v=1) but it's not clear which cap is correct. It also has some mild vertical stripes in the luma channel (most visible in the sky).

[FLP437]

I will try to put samples using a stable svideo signal with test patterns from a dvd player and make an alternative vhs/svhs capture using a jvc svhs set to see if it´s easier to identify if this alternative workflow could have any benefit , namely in reduction Chroma noise, improve color and better levels record adjustment and could in this case be a viable alternative to the conventional workflow.

The idea is to see if the DMN8653 LSI and Philips 7137d a/v decoder integrated in the dvd recorder could give any advantage in terms of filtering and as an ADC. Eventually JVC and Panasonic units could have better implementations in terms of circuitry and algorithms but if this one prove to work probably we only get best results with better dvd recorders if it´s the case.

[Brad]

Please post additional samples in a new thread directed toward such a comparison, as this one was meant to be DV vs lossless.

To see whether that LG RH199HS has any sort of line TBC, you would need to use a source that hasn't been TBC'd already. And in my comparisons, the DCR-TRV340 seems to defeat the DMR-ES25's TBC even when its own TBC is supposedly "Off". I guess that's related to its output always being digitized.

[FLP437]

Sorry , it was not my intention to hijack your thread I thought the matter was somewhat related, I will post the new samples on a new thread "Conventional analog capture workflow vs alternative methods - comparison"