Ok thank you so much for the effort trying to improve the quality of the capture.
However it might be interesting in terms of the purpose of this thread if we could conclude whether the problem is from the source , camera and tapes or is subsequently created during capture by the device. Maybe if I put a sample of the same capture in DV directly via firewire (using the A / D internal D8 camera decoder ) and other capture from a older tape when the camera was still new (79-80) and where the probability of the camcorder being in best condition was bigger.
In this case at least we could try to conclude whether the device is acceptable, able to extract the tape potential and not altering for the worse the tapes contents, as it could be important for people wanting to acquire a capture device and not being able to buy one of the old legacy recommended devices that are increasingly difficult to find , I would even say impossible to find ,namely in Europe . If you think such’s samples could give definitive clues I will post them.
+ Reply to Thread
Results 61 to 90 of 154
As far as broken lines and aliasing, I'm not sure that you've demonstrated anything more than the problems with interlaced content in general. Do you have an example of interlaced video where a shaking camera is zooming in along a diagonal line without these issues?
Here is a 480i frame I generated by downscaling a 1080p60 camcorder video and interlacing it. Lines are similarly broken.
I purposely gimped QTGMC here by inputting the PNG rather than the video. If given the generated video, its motion compensation allows it to fill everything in better because the movement is smaller and the video has much less noise than the Video8 example. Hopefully the point is still clear.
Thanks for the input, but your example doesn't deal with what I referred to. No one is rescaling here. I don't see why you're posting a "480i frame generated from a 1080p60". How is this relevant? How about posting a 480i frame from a 480i capture? I'm talking about a different animal and a different process.
- My sister Ann's brother
Both samples below are interlaced. The original AVI capture is unretouched from the very start of a 1992 tape. Other shots were captured but at a different frame size. Anyway, nobody wants to see 400 MB of home video. Yep, there's aliasing, but less of it and pixels don't go wild all over the place. My final BluRay had a lot more processing than shown here, so to keep things even I re-encoded with the same filters used on the mp4 for the USB3 capture. In fact I used fewer filters, only santiag and QTGMC (at "fast" instead of "slow" to give my own funny lines and edges a fighting chance and the benefit of the doubt). No color or levels work, no awarpsharp, chroma shift, maa, sharpener, etc., that I used on the USB3 post.
I chose scenes where my sister did her trombone act with the zoom in "available darkness" interior light and daylight. Has constant motion, low-light noise, out of focus hop and zoom, clothes with stripes and diagonals, a wood plank fence, grass, wrought iron patterns on lawn furniture, even a shot with skinny wood window blinds to show that my own sample ain't perfect. Also used a lower bitrate than on my BluRay archive. I think you'll see that things move smoother, if imperfect.
VCR: Refurbished Panasonic AG-1960 (no longer owned).
capture card: ATI All In Wonder 7500 Radeon AGP in XP SvcPack 2, VirtualDub, cheap Biostar AMD mobo, 2005.
[EDIT]SignVideo PA-100 proc amp for basic capture levels.
The links below are to a pro-level download account. No ads or popups. Just a download window.
Down Home 1992 Sample.avi, 25.5 seconds, unprocessed capture (YUY2 Lagarith), 134 MB: https://www.mediafire.com/?8aohz7h5ktg86oj
Sample_1992_480i.m2t, 49 seconds, 50 MB: https://www.mediafire.com/?l7569rlbmxvdibg
As far as I can tell, most of the disturbances in the USB3 cap are on the original tape.
Last edited by LMotlow; 20th May 2015 at 08:32.- My sister Ann's brother
Anyone know what causes this twitter problem with consumer cameras? Couple of years ago I spent three 8-hour days on the net looking for an answer. I guess you can tell I didn't find one, LOL! Is that a camera shutter problem? Recording circuits? Some kind of line phase or line registration glitch? Sometimes you just get a little of it, sometimes a flood. You get broken lines (lines with serious and visible gaps), lines that are just split or doubled somewhere but with no data missing, or just sawtooth edges.- My sister Ann's brother
Thank you vaporeom800 and LMotlow for the effort done in interpreting the usb3hdcap capture data.
I think that both of you, although having minor differences opinions regarding interpretation of some data, are apparently in agreement that the problems found in the captures are source related ( tape/ camera problems) and not introduced by the capture device itself and that, the capture device is in some way neutral not introducing by itself additional problems , artifacts , whatever.
As I´m not able to find none of these “ wonder” legacy ATI cards ,and I have to choose the best I can get ,I was really interested to find if this device could be my best choice without going for professional stuff like the AJA KONA, BlueFish or similar cards for capturing analog tapes who may have in addition to the excessive price particular problems ( even knowing this device is gamers market oriented , but with retro gamers also in mind ,that´s why the analog conversion was included).
I´m posting the samples I previous referred, a D8 firewire direct capture of the same clip .If the problems are tape/camera related they also must be there.
I tried also with some old capture devices owned (unfortunately not very good quality), Philips Saa7134 perhaps the best of this lot, Dazzle DVC100, pinnacle PC TV usb, Hauppage WinTv 878/9 and it seems to me that the problems are always present independently of the capture device used.
I ´m also posting a video8 sample capture from the camcorder early days when it was still new and a clip with less motion (pan and zoom ) that perhaps will show off less problems.
It would be nice if we can conclude ,if it is or not a satisfactory device for analog tapes captures or if I still need to continue searching for a new one.
Last edited by FLP437; 20th May 2015 at 23:34.
Re: Sample_DV.AVI. As I suspected. Something in your USB3 capture setup treats analog as DV, so you get the DV field treatment that results in those flying pixels and aliasing every time something moves, making for noisy video. Sample_DV plays more smoothly. You still have some line twitter now and then, common with many consumer cameras, but Sample DV plays smoother and is easier to clean up. As DV, it's PC-only playback. So re-encoding to another final delivery format will be necessary but at least you can maintain 25i without problems. You do have typical DV compression artifacts (mosquito noise) but that's easy enough to fix (you can see it more clearly if you watch the date stamp area during play).
Remember in your post-processing that this video is Bottom Field First. You should keep it that way.
First time I've seen that reversal of interlaced field problems. But it works, so go with it.
The other cap is grainy (no huge problem) and there's still a little line twitter (no surprise). I didn't work on it, but it looks useable.- My sister Ann's brother
Only to clarify, the Sample_DV as nothing to do with a usb3hdcap capture, the device is not in the workflow.
It's a raw capture from the digital8 camcorder firewire output ( playing a video8 analog tape) in a pc with firewire input through WinDv.
D8 - firewire - PC WinDv
Yes, you're correct. I didn't word the case clearly. I was comparing source->USB3 with DV play->Firewire.
You're also correct about the legacy AGC hardware. Difficult to find anywhere whether Europe or over here. I can't stop using mine just yet -- I still have about 150 hours of old tapes to transfer. If you happen to find one of those AGC or PCI cards, note that there are separate NTSC and PAL editions. Most owners seem to be in no great rush to get rid of theirs.- My sister Ann's brother
Format : AVI Format/Info : Audio Video Interleave Commercial name : DVCPRO File size : 94.4 MiB Duration : 26s 80ms Overall bit rate mode : Constant Overall bit rate : 30.3 Mbps Writing library : VirtualDub build 35491/release Video ID : 0 Format : DV Commercial name : DVCPRO Codec ID : dvsd Codec ID/Hint : Sony Duration : 26s 80ms Bit rate mode : Constant Bit rate : 24.4 Mbps Encoded bit rate : 28.8 Mbps Width : 720 pixels Height : 576 pixels Display aspect ratio : 4:3 Frame rate mode : Constant Frame rate : 25.000 fps Standard : PAL Color space : YUV Chroma subsampling : 4:2:0 Bit depth : 8 bits Scan type : Interlaced Scan order : Bottom Field First Compression mode : Lossy Bits/(Pixel*Frame) : 2.357 Stream size : 89.5 MiB (95%) Audio ID : 1 Format : PCM Format settings, Endianness : Little Format settings, Sign : Signed Codec ID : 1 Duration : 26s 80ms Bit rate mode : Constant Bit rate : 1 536 Kbps Channel(s) : 2 channels Sampling rate : 48.0 KHz Bit depth : 16 bits Stream size : 4.78 MiB (5%) Alignment : Aligned on interleaves Interleave, duration : 41 ms (1.02 video frame) Interleave, preload duration : 500 ms
Last edited by LMotlow; 21st May 2015 at 10:10.- My sister Ann's brother
But it doesn't fly to pieces with exploding psychotic pixels every time something moves.
Thank you for humoring me with the sample, LMotlow. I'll table my thoughts on it for now, if that's okay, as I'm more interested in discussing this DV sample that you uncharacteristically praised.
Here is one QTGMC'd field from the frame you showed earlier to exemplify broken lines. The portion of the column that you showed at 2x looks the same in both to me, except for the very edge where it meets his arm. The next field doesn't even have that distinction. Are you seeing something different?
On the DV I see mosquito noise, some blocking, associated loss of detail on the metal archway to the right, a difference in chroma placement, and blocky chroma (on the UV-only view). On the Huffyuv I see additional noise, and notably edge noise that isn't visible in the DV since it's replaced with mosquito noise. (To be clear, my comparison was done at 2x, not the 1:1 magnification seen here.)
Here is a different frame where you may agree that the Huffyuv holds up better (the man in the middle).
The "tweaks" are an attempt to match levels:
Sorry that we disagree. Both of the samples in your bottom "hufyuv vs DV" are deinterlaced. That would have to be done anyway to clean up the chroma and some other problems. Try re-interlacing each for DVD or BluRay. If the O.P. wants only 50fps playback, his final delivery format options are limited.
Many of those broken or aliased diagonals will always be there to some extent because of the camera shutter. Some will persist even when deinterlaced.
Last edited by LMotlow; 24th May 2015 at 06:18.- My sister Ann's brother
Yes, they're deinterlaced. The only way they wouldn't be displayed via deinterlacing is if the final DVD/BD was viewed exclusively on a CRT, so I don't get your point.
FLP347: Was DNR turned on when you transferred the DV sample? When the camera moves, the street loses detail in a way that I wouldn't expect from an intraframe codec like DV.
I'm not sure now but I think both tbc and dnr were turned off in the dv capture.However if you need I can make a recapture assuring both are turned off. It would be interesting if someone could make a synthesis including pros and cons regarding this device. Unfortunately I have not the technical background to do it but I can provide any data eventually necessary.
Nah don't worry about a recapture, but thanks for the offer.
One minor con with the USB3HDCAP is that it doesn't capture all lines available for PAL video. You lose 2.5 lines at the top and gain 3 lines of nothingness at the bottom. But your TV7134 does the same thing.
I hate to spoil your day, but CRT's never displayed two complete sets of interlaced scanlines at the same time.- My sister Ann's brother
The point is that reinterlacing the original non-DV capture just gets you right back where you started with the noise.
Now you're sying that DV cap is smoother, which is what already said.
I hate to spoil your day, but CRT's never displayed two complete sets of interlaced scanlines at the same time.
Point is: the USB3 interlaced cap looks like the same old pooh-pooh when it's reinterlaced. That means 50 fps playback, which means no disc formats like DVD, BluRay, or AVCHD. OK for an archive, but what happens when other family members want a family tape but don't have BluRay players? If the O.P. figures he'll never need universal playback, ever, it's not a problem. The DV does deinterlace and reinterlace cleanly and doesn't appear to be an impossible cleanup (what's so great about the other version, anyway, with all that chroma bleed and displacement that looks worse than the DV?). Some of that line twitter, we both know, isn't going away in either case.- My sister Ann's brother
Sony Bdps1200 Wired Streaming Blu-ray Disc Player, Full Hd 1080p Blu-ray Disc Playback (Certified Refurbished)
$46.99 and free shipping!
vaporeon800, could you please help with this points.
As I have for now no other option If I want a HuffYUV lossless capture file, I have to use the usb3hdcap or the philips Saa7134 , if not I have to stay with the previous DV captures I´m trying to see if i can make any small improvements in the captures with the startech device.
Related to the capture pin properties - I frames interval
is the default value (30 ) the best one or a smaller value (15 ) could improve capture results?
As I got some chroma problems in the captures and namely chroma bleeding could the the DNR on in the camcorder help?
I turned off DNR and TBC because I got the impression of a less detailed captured image . But has the device is much less prone to drop frames than the philips ( with this one a external tbc is almost mandatory if not there are a lot of drop frames) perhaps some tweaks could improve the captures.
I don't believe the I Frame Interval and Quality options there actually do anything. You could try cranking the Interval as high as it goes and the Quality as low as it goes to see whether any setting has an effect towards degrading the image.
In my opinion, the camera is DNRing the chroma regardless of the DNR setting being off or on; turning it "on" doesn't seem to improve things further.
I think you're getting the best you can get, so long as you set the Proc Amp appropriately for each capture.
Well the idea was a little crazy but yes changing The I frame Interval works , for worst . I tried only decreasing the I Frame interval for 15 and I got lots of inserted frames. So it seems I´m getting the best I can probably . The TBC and DNR circuits on the camcorder seem to have virtually no effect I´m beginning to think if they are working properly , but I also don't get better results from my video8 deck neither from the Philips Saa7134 capture card. I probably have to accept that these are the best results I can get for now. Perhaps Black magic learns how to make drivers and I will have an alternative choice.
That's surprising. So did the quality change at all?
Well the quality for individual frames didn´t changed a lot but you get hundreds of inserted frames with captures of only some minutes and is likely that sound and video synchronization will suffer I think.
I have now made several 90 min. captures and can confirm the device is not prone at all to drop frames , however with some tapes it does insert frames . I noticed that in this cases the internal camera TBC (theoretical a line tbc ) slightly reduces the number of inserted frames , but only an external TBC avoids completely the inserted frames, even in the more difficult tapes. The inserted frames tend to appear during scene changing when the camera ends a recording and initiates a new one.
I decided to buy one of these to check out for myself. On two different computers I experienced the issue shown in the screenshot when I attempted to capture an S-Video signal.
Even with a stable NTSC signal coming from a DVD player, there are many frames dropped and inserted every second, causing the average frame rate to be in the range of 20-24 fps instead of 29.97 as set. I tried composite too, with the same result.
The USB 3.0 controllers are Renesas & Intel, both supported according to StarTech. I started with the current drivers on StarTech's website, then I uninstalled them and grabbed the Micomsoft pack from TheThrillness' page. No change. And on this AMD machine, I confirmed that I could capture 1080p60.00 from the Intel's HDMI output with 0 drops/inserts.
The only fix I could find turned out to be Capture -> Timing -> "Ignore video timestamps from capture driver". The average rate is now approximately correct, with 0 drops/inserts. I'm running a test capture currently to see whether the captured frames are actually intact. FLP437 never mentioned having to use this option, so... this is odd.
I can confirm that I didn't need any work around to put my usb3hdcap to work namely capturing video8 or VHS through s video. Only framerate in settings, bellow capture tab in virtualdub as a tendency to change value without notice from 25 fps to 50fps and in this situations it indeed insert a lot of frames hundreds and thousands but did not however drop frames. I was obliged before capturing to confirm the frame rate value and change it if necessary to 25, after that everything run normally.
The portables that I have used were both with intel usb3 host controllers and intel CPU ( i5) and used one with windows 7 and miconsoft drivers and the other with windows 8.1 with the standard drivers from startech and the situation was similar both working without problems and without drop frames only with bad quality tapes some frames were inserted, namely when changing scenes but usually no more then 10 or 20 for a 60 or 90 min capture.
For anyone looking to buy this specifically for VHS capture, I would say don’t bother.
I was initially very pleased because it does in fact do a great job correcting horizontal jitter from unstable sources (= "line TBC"). But it's all pointless, because it loses VSYNC at the slightest hiccup, resulting in either a "NO SIGNAL" screen or dropped/inserted frames. A guy on DigitalFAQ said the same thing about the XCAPTURE-1, of which the USB3HDCAP is apparently a clone.
Adding a full-frame TBC like the AVT-8710 to resolve the issue would mean relying on its poorer wiggle correction instead.
Such a shame. It could have been a great option.
I wonder why FLP437 didn't encounter this issue with his VHS captures. Maybe it's more stable in PAL mode?
Attached is a 1-minute sample of a tape played back simultaneously from one VCR to two different capture setups. Over the course of 1 hour, there were 68 drops & 1422 inserts, and the resulting file is 26 frames longer than the content should be. Admittedly the tape I recorded this test pattern onto is garbage, so it's a bit of a torture test, but I get the same "NO SIGNAL" issue with my regular camcorder tapes that don't present issues to other capture devices. It just doesn't pop up as often as this sample.
- Macrovision test (Home Alone): video replaced with "HDCP" message.
- Can capture 2 extra bottom lines of 480i content by forcing 720x576 frame size. Combined with VirtualDub's Cropping filter during capture, allows me to directly capture a 720x482 file without garbage blackness underneath. [Not useful to anyone else, I'm sure.]
- Sharpness: added moiré on Snell & Wilcox zone plate indicates that the default setting of 32 adds ringing. Set to 0 to reduce. [Only tested CVBS input.]
- Horizontal timing is wrong (very-right edge actually appears on the left side of the screen -- see screenshot in post #88). Not an issue for VHS.