Confirmed: Pinnacle 500-USB works with VirtualDub!
 

I just lunched the VirtualDub to see if it recognizes Pinnacle 500-USB device, First selected the device in the device list and got some sort of encrypted looking video with intact audio, but upon switching from overlay to preview, I was able to get a clear picture, I tested both NTSC and PAL tapes using Huffyuv compressor, the output files read 4:2:2 8bit for both NTSC and PAL. Is there any other way to test the quality of this device?

Pinnacle USB Box 500 work only under Win XP SP-3

Lagarith or UtVideo [V.12.2.0]

Resolution 704 x 576 in YUY2

Proper driver on the Software Studio 10.5 .... only the driver can be installed automatically.
Installing NO Studio

Windows 7 Ultimate, Windows recognized it by name and installed the driver, VirtualDub lists it by the exact name in the device list. Yes I do have Studio 19 installed that may have to do something with it but it works cherry. I'm impressed actually by the quality of the caps, I haven't done long caps to check for audio sync, But I will do it in the weekend and post some samples here.

-- merged --

Here is a sample of a recorded S-VHS tape from TV broadcast, I still don't understand why VirtualDub not capturing NTSC in 4/3 format ??
http://www.mediafire.com/download/00...S-VHS_NTSC.avi

If you want regulation DVD for your final output, you captured at the correct frame size (720x480). A frame that's physically 4:3 in size (640x480) isn't acceptable for DVD, or for SD-BluRay for that matter. If you want 640x480, set that frame size in VirtualDub capture. VDub can capture any of several frame sizes, but you have to tell the poor critter what to do or it takes the default, usually via your capture drivers. If you capture at 640x480, you would have to resize to 720x480 or 704x480 for DVD.

You can't resize interlaced or telecined video. The TV ads in your sample are progressive video with hard-coded 3:2 pulldown (telecine). For resizing you have to remove telecine to get 23.976 progressive video. Avisynth's TIVTC plugin is the only good way to remove duplicate and field-combed telecine effects -- anything else, and you'll get low quality results. For standard interlace, use QTGMC. Do NOT use an NLE to do this. None of them do it correctly.

The image doesn't have much noise but looks a tad soft -- then again, maybe it was broadcast that way, and a lot depends on the original recorder and the playback machine.. There are some edge halos and one ad has commonly seen shimmer and line twitter, but that could be the broadcast itself (I see it all the time on TV). Overall it's a very workable capture, better than most, but definition looks a little scanty. I'll give it a run tonight and post some results and ideas.

In VirtualDub, if you want to change the display aspect ratio, right-click on one of the viewing panels and choose an aspect ratio from the pop-up dialog. Decoded AVI doesn't have an "aspect ratio" except its innate frame size (in this case 720x480 is 3:2), but any PC media player lets you adjust the aspect ratio during playback. (I believe Windows Media Player is the exception, but it's not a very good player anyway. Microsoft has ruined it).

I did note that some masking or other processing appears to have been bused. If so, we'd prefer unprocessed samples. Was edge masking used during capture?

No processing has been done except the TBC was ON during the capture, The recording was done on a S-VHS tape as follows:
Media player with cable tuner output thru HDMI as 1080i broadcast to a HDMI to S-Video converter and then to S-Video input of a S-VHS VCR, recorded on a TDK S-VHS tape and played back on the same deck and captured from S-Video output of the VCR with Pinnacle 500-USB and then to VirtualDub using HuffYUV compressor, All the cables used are better shielded and short usually 1ft each, Not that cables matter a lot but it's a hobby and has to be perfected.

Thank you for the additional information. I was in the process of composing some notes about the oddities with this capture and more questions about it, but the reasons are more clear to me now. The video does have what we call an over-processed look to it, having gone through a number of resizers and analog-digital conversions and back again. The low color density gives it an unsaturated look despite chroma bleed, chroma ghosting, and occasional blooming effects. The video responded to telecine removal with inaccuracies such as color corruption and severe multi-frame chroma ghosting at scene changes. It shows signs of being "played" with telecine removal and then poorly re-interlaced (i.e., it looks like sloppy inverse telecine was performed at one point). The video can be conventionally deinterlaced, which should not be possible with true telecine, and had to be followed with conventional decimation of duplicated fields.

This is one way of saying that multiple stages of pre-processing has pretty much "cooked" the original signal and gives an over-processed look before anything gets done to it. One can try correcting the obvious red color casts and the absence of clean whites, but doing so makes the image look even more lacking in color density. Shadow detail doesn't hold together with efforts at improving dynamic range and contrast. One gets a similar effect by decompressing and recompressing jpeg images several times over. There's also subtle chroma blotching and discoloration in what should be neutral areas like whites and especially grays.

The absence of typical tape noise is deceiving at first glance. But effectively what I'm looking at is a multi-generational copy of the original signal. As I worked with the video I became more and more convinced that this looked like a lossless capture that had gone through some lossy processing, with resizing interpolation errors and artifacts, buzzing edges, severe line twitter and shimmer at times, and definition loss from multiple colorspace conversions (HDMI to s-video converter? That's definitely one of the culprits). Subtlety and fine detail are absent except as fuzzy remnants, like video that has been resized several times.

The result looks very "digital". Should look OK to the average viewer. There are already too many interpolation losses, so if you want true 4:3 square-pixel captures you should cap at 640x480. Video processed as you describe won't resize very well. The way they will look on a big TV is strictly up to the player and the TV.

If that's really recorded in S-VHS mode, and not just VHS mode on S-VHS stock, it looks really bad. Sorry to say it.

To convert from the 1080i original, your equipment had to deinterlace the signal to 1080p, scale it to sub-480p ("sub" for 16:9-in-4:3), then reinterlace it. A cheap adapter likely has no provision for film mode detection, so it would just interpolate everything.

I could disagree with that in some respects, notably that HDMI and s-vido conversions are definitely "processing". Whether they caused the ghosting problems seen in the sample or whtehr the VCR is doing it is up for debate, but I'll show examples of frames where the field structure is obviously corrupt.

Below: from the original sample, a 3-frame sequence including a scene change. The three frames are 115, 116, and 117 from the original sample. Frame 116 was originally a telecined scene change containi9ng two fields, the first frame as the top field, the third frame as the bottom field. Frames like this from scene changes is common with both telecined and interlaced video.
http://www.digitalfaq.com/forum/atta...1&d=1459433031

With either inverse telecine or deinterlacing, double-image frames like the middle frame are separated into two full-sized frames. With telecine, duplicates will occur every few frames, so dupes are removed.

Below: frame 116 double-image, top field transposed to a new full-sized frame. top field. This new frame looks fairly normal:
http://www.digitalfaq.com/forum/atta...1&d=1459433272

Below: frame 116 double-image frame, bottom field transposed into a new full-sized frame. Ghosting elements from the previous frame are obvious: ghosting of the red flowers in the upper left, ghosting of bright whites and the woman's shoulder in the lower right, and green, magenta, or other discolorations from other parts of the previous frame. This ghosting persisted for several subsequent frames.
http://www.digitalfaq.com/forum/atta...1&d=1459433505

The two images above were the result of inverse telecine. Unfortunately deinterlacing produced exactly the same frames. With correct telecine or interlacing, these artifacts should not appear when decoding interlaced or telecined frames.

Lest you feel that inverse telecine or deinterlacing caused these affects, here is frame 117 from the original, unchanged, and showing the same artifacts:
http://www.digitalfaq.com/forum/atta...1&d=1459433669

Ghosting wasn't limited to scene changes. There are countless examples of this ghosting that follows telecined frames elsewhere, without scene changes.

Below: frame 791, unprocessed original, cropped to upper right-hand corner. Arrows point to ghosting from background objects in the previous frame, and double-image that shouldn't be seen on progressive frames:
http://www.digitalfaq.com/forum/atta...1&d=1459433854

The same frame area, inverse telecined:
http://www.digitalfaq.com/forum/atta...1&d=1459433918

The same frame area, deinterlaced:
http://www.digitalfaq.com/forum/atta...1&d=1459433951

All three versions have the same artifacts. You might want to re-examine your recording, converting, resampling, and capturing process. Also, on many frames I see distortion during motion. If you want to see examples, look at frames 10, 11, 12, and 1004 and 1005 in the original sample for examples of distorted fingers and hands.

BTW, some of that motion distortion is obviously due to camera blur, but it seems excessive. Maybe it was broadcast that way. Chroma ghosting is a different matter. Was DNR turned on in the VCR?

Thank you guys for the accurate details, As you can probably see that I'm just a learner, I do really appreciate learning from the pros like you guys, I forgot to mention that the broadcast was 16/9, I set the media player to pan and scan to cut off the sides of the wide screen, And I meant by no processing is no software processing on my side, I probably should have used a different material, Commercials usually are heavily processed at the studios.
And yes as I mentioned before the TBC was ON all the time, I will post another clip and try to start from a clean source, So far I think it's a great device.

Here is another S-Video capture from a cleaner 1080i broadcast, I just posted the S-Video capture but I've included the original 1080i broadcast file, Watching the S-Video feed on my computer and playing the HuffYUV S-Video capture file couldn't see any difference, So besides the driver that probably require a Pinnacle software to be installed on the computer even if it is not used, this device has potential.

http://www.mediafire.com/download/j7...00-USB_CAP.zip

Thanks for the samples, but you might not get much feedback from those who won't download a .75GB video.

Both the original HD mpg and the capture are pretty sharp. Oversaturation, low gamma, crushed darks, and hard highlight clipping are apparently part of the original broadcast -- the engineers' token to viewers who enjoy the plastic digital look and usually watch TV in volcano mode. Some of the bright colors are at the edge of blooming, especially red. As far as levels and color go this style is pretty popular, if also indicative of why film would be superior to digital video's more restricted contrast range. It would look a bit more lifelike in downscaling if the color matrix had been properly converted from Rec709 to Rec601 for SD video. Downscaling produced obvious aliasing, stairstepping, line twitter, wiggling, and moire on fine lines and edges. This could be avoided with proper resizing with low-pass filters, but some of it appears to be in the original -- common grunge with HD broadcasts these days. It's also a bad idea to resize interlaced video while still interlaced.

Meanwhile you might have noticed the following. Below, I had to reduce the frame size from 1920x1080 to save a little forum bandwidth and make the images the same height, but here is frame 185 of the Original Video (it's a digital original, not "film"). You see the frame and the entire original scene here at 16x9:

Original 1920x1080 @16x9:
http://www.digitalfaq.com/forum/atta...1&d=1459776074

Here is frame 185 as 720x480 in the capture at its frame aspect ratio of 1.5:1:
http://www.digitalfaq.com/forum/atta...1&d=1459776187

As you can see, the capture is cropped on each side. Of course you wouldn't display as 720x480 with the wrong aspect ratio anyway, as the tires are stretched and the guys look a little too heavy-set.

Here is the same 720x480 frame at 4:3, with objects looking the way they're supposed to:
http://www.digitalfaq.com/forum/atta...1&d=1459776330

If you display captured frame 185 at 16x9, it's unnaturally stretched and the scene is still cropped:
http://www.digitalfaq.com/forum/atta...1&d=1459776396

Obviously the cropped capture looks correct only at 4:3. Is this what you had in mind for your final output?

Sorry for the big file size, I was trying to shove as many different scenes as I could, As for picture cropping, I set the media player to output as pan & scan so I can get 4/3 aspect ratio, The purpose of this is to test the performance of the capture device in an attempt to recapture my analog tapes that has been previously captured in DV.
Yes I want to output 4/3 aspect ratio but because the pixel shape in NTSC is not square it always gives me a horizontally slightly stretched frame, With PAL I don't have a problem. And I don't want to change the resolution in case I need to produce a popular format such as DVD later on based on your suggestion above.
This brings me to a question, I have a multi standard HP 16/10 1080p monitor with all kind of inputs, Composite, S-Video, Component, HDMI, and Display Port, When using composite or S-Video in NTSC mode I get the stretched picture problem with tiny black bars on the sides It has no setting to change it to 4/3 aspect ratio, In PAL it displays perfect 4/3 frame with black bars on the sides I even measured the frame with a tape measure as the monitor doesn't overscan, Is there a way to display NTSC frame properly? I mean like a hardware device that is inserted in the stream to correct the aspect ratio? Note: This question related to video display only.

The capture device seems to look the same as an ATI USB. Doin OK there.

What are you encoding with? 720x480 encoded to a 4:3 DAR displays as 4:3, not stretched, whether NTSC or PAL.

I unde4rstand that, but 720x480 is encoded to DVD as either 4:3 or 16:9. DVD and standard definition BluRay can't have any other aspect ratios.

PC media media players can be adjusted for several aspect ratios. So can the viewing panels in VirtualDub. And I guess you know that 16/10 is wider than 16:9. PAL and NTSC don't have the same anamorphic pixel ratios (PAR), so without a properly encoded aspect ratio they won't display alike. But I don't get this. Don't you have media players on your PC? Or are you using your PC monitor as TV for your VCR (expect problems there)?

I don't understand how you're playing videos into your monitor. I have no aspect ratio problems with my TV or PC monitors. Then again, I don't play a VCR through a PC monitor. I use a TV equipped to display aspect ratios correctly from whatever input. I don't use a PC monitor as a TV. Pan and scan I never use.

Your 4:3 capture has scaling artifacts everywhere, including ripples, aliasing, moire -- I mentioned all those. There is some of that already in the HD MPEG, so it's a nightmare to clean up, but interpolation errors in the capture make it look worse. If you deinterlace frames you'll see objects shifting fame by fame, just slightly, but enough to cause plenty of problems and noise.

The attached 4:3 Capture_downscale_4x3.mp4 shows plenty of noise (look at the ripples in the left-hand lamp post in the opening shot, and twitter in the iron grill. It's in every shot, a little or a lot. As the car passes in the next shot there's aliasing and excessive combing in the car window trim, floorboards, dashboard instruments, front window bottom border, front fender, the front side mirror, the hood ornament, and canopy trim, and nervous twitter in small background objects like the building's windows and the white posts in the ground. It's worse than the original.

There are different ways to clean that stuff. I gave your original MPEG a trial rather than cut off chunks of the image the picture or get into obsolete 1980's pan and scan. It's cleaner than the 4:3 or the original (the lamp post and ironwork are pretty steady) and about 85% of the other artifacts are calmed. The attached Original_downscale_16x9.mp4 is full frame 720x480 @16:9 DAR. It can be tweaked some more -- I ran out of time -- but you can't maintain a cleaner image if you take an 8x8 block of pixels and try to squeeze them down into 3.5x3.5. You can either try to clean it up or live with the noise. The car's front grill is an eyesore but it's a little better here, not as bad as the buzzing original.

I'd say my eyes are less tolerant of artifacts and distortion than yours, but to each his own. :wink2:

Both videos display properly on my PC monitor and on my HD TV through my Oppo Player.

The problem with my monitor is when hooked up straight to the S-Video output of my VCR, But I can live with that as it is for monitoring purposes only, My tapes are analog 4/3, So I don't have to worry about resizing the video.
The artifacts you described are most likely introduced during the conversion from HDMI to S-Video, Like I said I watched the live S-Video stream and the captured HuffYUV file and I couldn't spot any difference so I guess I have the OK to go ahead and start re-capturing the tapes with the 500-USB, I have a problem with one tape with waived edges, So the audio goes up and down and sync track introduced a frame jump, I might be shopping for a full frame TBC, And there nothing I can do to the audio I guess.

Good luck, then.

I still don't understand the point behind HDMI to s-video for 4:3 VHS. Why don't you plug your VCR into the capture device? If you already have a wide screen digital video, or any other digital video, why go through all that? Digital is copied from place to place, not converted\recorded\captured. Or maybe I don't understand your workflow (it's highly possible).

If you have a line tbc in your player or elsewhere in the capture chain, you shouldn't have flagged top borders. Likely need a frame tbc as well to prevent dropped or inserted frames and bad sync.

It's confusing I know, I used the HDMI to S-Video to obtain a clean S-Video signal to test the capture device with, to my believe to test a device you have to feed it a clean signal, If the signal is already bad such as my home VHS-C tapes I wouldn't know if the device is performing correctly, Maybe a pro spot tape signal defects and device lack of performance but not me, I even tried recording a S-VHS tape with the VCR to make a test tape but the quality was inferior to what I posted above, so I opted for a clean S-Video signal to test with.
The tape I mentioned to be captured has waviness is in the tape itself not in the video signal, The edges of the tape are damaged, It was played before by a family member on a VCR with a bad tracking, The sound spikes up as the tape moves along and the frames jump up, That I believe a full frame TBC would take care of the damaged sync track. But it's only one tape and the 1st few minutes only.

That's totally mysterious. Analog via HDMI is converted to digital, then re-converted to analog. There are color and artifact problems associated with that. And a clean HDMI->s-video converter costs $400 or more.

Do whatever you have to do.

What are you recording to S-VHS?

I'm not recording, I was testing only, I have no desire or need to record a HD digital source to a SD analog format other than testing the capture device, The capture device has S-Video input and I have to supply that signal, I don't have test tapes to test with so I had to come up with my own using a HDMI to S-Video converter that supplies me with the needed signal to test the card, I hope I am clear this time.

Ah, there we go. thank you. Clear now, and especially after I backtracked to your earlier posts.

The problem that remains is that your old VHS tapes won't look like those tests, as you've already mentioned with the tape that has flagged borders. To put it another way, tape played directly to capture is the acid test. True, you've checked on problems the capture card might have. I'm afraid the s-video converter and other in-chain pre-processing created problems your capture device probably won't display.

If problems arise during conventional capture, that would be a different story. I'm afraid the s-video business made it seem that something else was amiss.

I'm aware of that: Garbage in garbage out, But at least the same garbage out better than a worse garbage, Actually there is two full size VHS PAL tapes 90min each where recorded with a semi professional shoulder held camcorder with a decent quality, The rest of the VHS-C tapes are recorded on consumer camcorders those are the garbage.

I officially started the recapture process, Here is a small segment of the capture:

Thank you for your new sample. :salute:

Tree.avi is a little bright but the capture appears to be faithful to the source. Its worst enemy is spastic and camera motion and jiggle, which wreaks chaos with sharpness and smooth edges. Even when I deinterlaced with QTGMC, edges were ragged. This is mostly the fault of severe motion, not the player or cappture card -- the latter are doing the best they can with the source.

Color looks accurate but the black level and gamma look rather high. That's easy enough to fix in post-processing. Often it's difficult during capure to control levels with home video, they change so frequently. I made some simple corrections in Avisynth. Color is largely personal, of course. You can tweak however you wish.

Below, the original frame 152 (*borders removed for clarity*) resized to a 4:3 image. The original is a loittle pale and lacks a little of the perception of depth.
http://www.digitalfaq.com/forum/atta...1&d=1459874876

Below, frame 152 with levels and saturation adjusted:
http://www.digitalfaq.com/forum/atta...1&d=1459874995

Both frames are still interlaced, so they're a little fuzzy due to motion. I also noticed very noisy edges, especially in the right-side wall structures whose edges looked very ragged and kept moving in ways that made the edges "buzz".

The attached Tree.mp4 is encoded at 4:3. Some side border pixels were cropped off to get a 704x576 frame (Note that 704x576 is valid for 4:3 anamorphic DVD). There were 10 pixels of bottom-border head switching noise that I cropped off and replaced with 6 black pixels along the bottom and 4 along the top to maintain the original height. The original top border has a broken 2-pixels at the top, which were also replaced with solid lines. A thin uneven top border is common with many home camera transfers. I also made an identical MPEG for DVD, which I haven't posted.

Smoothing continuous camera jiggle take some work. It also requires higher encoded bitrates. The motion makes inherent VHS noise look worse, and when encoded much of the data bits are wasted on that noise.Calming some of the horizontal and vertical shake is often done to make the encoding look smoother and to help temporal denoisers that are based on motion. But all deshakers move the image around in such a way that some of the original pixels from all 4 sides will be cut off. I used VirtualDub's DeShake to relieve some of the jiggle. The settings were mild and came at a cost of about 15% of the original image. With less motion, the cost would hardly be noticeable. With bad motion and strong settings, the image would be very stable but much less of the original image would remain. Another cost with motion interpolation deshakers is a bit of contrast.

The DeShake example looks a little smoother, attached as Tree_Mild_DeShake.mp4 .

Wow the processed videos looked amazing, A little blurrier but definitely looked better than the original. And yes the cameraman is a rooky he shakes a lot and he most likely doesn't know what white balance is, There is a lot of scenes have bluish whites.
Is there a post processing easy to use software that doesn't use keyboard typing and scripts?

Noise and motion smoothers always take something away. But actually it's blurry even when deinterlaced because motion blur and aliasing in the camera don't have clean edges. I guess one could say it's not the detail that's sharp, it's the noise. It's possible to sharpen a bit more but only after defects are cleared, or you'd have a very sharp disaster.

Jiggly cameras, not unusual. We see them all the time. A recent post had camera motion so bad, denoisers became confused and just created artifacts -- and the noise was still there!

Color imbalance is common, even with retail tape. Most people would fix too much blue by adding red, which wouldn't be correct. You'd end up with an imbalance of pink (blue + red = pink). You would either reduce blue or reduce blue contrast, or add yellow (red + green).

Umm, I'm afraid there's no Avisynth equivalent except with big studio gear, although pros do use Avisynth as well. A favorite with us amateurs, too. No M.I.T. degree required, but you do have to learn up on a few tricks. Studio gear is automated to a degree, but instead of scripting you need to know programming -- a bit more difficult. There's a graphical interface for Avisynth (AvsPmod) but, really, I can't stand it. Filter setup is a pain, often doesn't work correctly, and the filters and AvsPmod keep getting updates. And you have to write a script anyway. AvsPmod can't run without a script.

Like many people I don't re-type entire scripts from scratch. Text files full of boilerplate text are kept around. Just copy and paste, and make a few changes if needed.

The scripts can be very basic or very complex. Avisynth exists on many levels. I should add that I also used VirtualDub, especially for the DeShake version.

Avisynth is easier than first appears. Most of us mere mortals learned by example. And don't forget the Getting Started pages in its Help files. The app and most of its popular filters have lots of documentation -- maybe too much, for some of us. Most people tend to use the same batch of filters over and over. Gets rather routine after a while.

I use VirtualDub as often as Avisynth. There are things you can do in VDub that you can't do in Avisynth, and vice versa. We can get into details later. Meanwhile the sample capture doesn't seem to have defects from the player or capture card, given that VHS and webcams/cameras are what they are.

[EDIT] Oops, forgot. The original sample has a greenish stain along the right border. There's a filter for that. Works most of the time and worked OK here (some videos are just impossible, though). You'd need Avisynth for that filter. The filter's name is chubbyrain2. I have no idea how the maker came up with that one.

I will get into avisynth later on after I'm done with the capturing, I have to learn it since I need it. Is there an option to change individual colors in VirtualDub?

Yes.

I didn't change any hues in the sample I posted. I lowered black levels and gamma, and pushed saturation up a tiny notch. When gamma and black levels are too high it can often make color look washed out.

Here is some pics showing the guts of the Pinnacle 500-USB if anyone interested:

Also remember that some broadcasting engineers are untrained monkeys with sticks. :knock:

It took several years for me to understand IVTC, because most of my homemade cartoon sources would not cleanly convert. Chroma ghosting was always an issue. This was 10+ years ago, when I was into the "music video" scene (anime, cartoons mostly), and into creating advanced motion menus for fanmade DVDs. Trying to IVTC those for processing in Premiere was not cooperating. Back then, the best deinterlace methods involved nnedi, pre-QTGMC. (I actually still have the sources for my favorite music vid, and plan to redo it at some point.)

So again, sometimes the source suck before we ever get them. :(

And this: http://www.digitalfaq.com/forum/news...embership.html
You overdid it. Too much green/yellow. Coincidentally, while watching baseball yesterday, I was reading a magazine about editing, and how too many people, even pros, have gotten into a bad habit of overdoing color in the digital age. It's something I heavily agree with.

____

It's funny how this was latreche34's thread, but I've only replied to you sanlyn. :laugh:

I was playing with the Pinnacle 500-USB firewire port to see if I can use it in the future if I ever get a laptop without firewire and I find out that the device acts like a bridge between the computer's USB port and its firewire port, I was able to hookup a DV/HDV deck and was able to test view/capture a tape using several DV and HDV capture software, Not only that but I was able to control the deck from the software like play/stop/ffwd etc all from the USB port.

That's surprising, and could be useful for many people. The only other device known to do this is the Dazzle Hollywood DV Bridge.

So you could capture a DV cassette over USB without additional compression?

The ADC is the SAA7113H, which was a widely-used chip in the early 2000s.

HDV and DV with no additional compression, It just converts Firewire to USB with no external power needed as it draws its power from the USB port, Pretty amazing I almost bought the Pinnacle Moviebox for that purpose which is bulky and requires external power until I tested the 500 and found to be working, Pinnacle never advertised this feature.