How can I confirm that my DVD player correctly displays 16:9 content or not? It displays 16:9 content from my DV camera too wide by about 5%, but 4:3 content is displayed perfectly. I've checked and tried all the player settings and nothing helped.
Is there some test or reference disc I can play on it to check if the problem is with the player or elsewhere in my authoring process?
Alternatively, how does the attached DVD display on your hardware player? My camera has the slightly wider Rec.601 display aspect ratios to support nominal analogue blanking (http://en.wikipedia.org/wiki/Nominal_analogue_blanking). A software player displaying all pixels at exactly 16:9 will make the circle appear slightly too tall, but a hardware player should display it correctly. Except mine doesn't! It's a Sansui DVD 3500, in case it makes any difference.
Thanks in advance,
+ Reply to Thread
Results 1 to 30 of 52
Just my 2 cents.
The dvd is pure 16:9 and has the correct DAR of 1.778 so what is on the disk should display as pure 16:9.
But the Pixel Aspect Ratio is fooked - probably caused by the camera as you have mentioned.
IMO a hardware player should show it the same as a software player.
Others may have direct ideas but I would suggest that you upload a small sample of actual video from the camera ideally with a circle or something else that can test it.
If that's not the problem... the DVD spec refers to the MPEG 2 spec regarding aspect ratios. The MPEG 2 spec says the full frame (720x576 in your case) comprises the 16:9* picture. As you are aware, that conflicts with the rec.601 spec where the inner ~702x576 comprises the 16:9 picture and the rest is padding. If you want to avoid this discrepancy (mostly) crop the frame down to 704x576 before making a DVD. Or stretch the inner 702 width to 720. Or use a Sequence Display Extension to specify the inner 702 is the 16:9 picture.
In my experience, how the DVD player outputs the video varies depending on the output. Analog outputs (at least composite and s-video) follow the rec.601 spec, digital outputs follow the MPEG 2 spec.
I just checked your DVD on my two Blu-ray players and TVs. The players were set to upscale to 1080p to HDMI. Both TVs where set to non-overscan. In both cases the circle rendered about 2 percent taller than it is wide, as expected.
I dug an old DVD player out of storage and connected it to one of the TVs via a composite cable. The player was set to output as 4:3. So the 16:9 DVD was letterboxed. The TV was set to pillarbox a 4:3 input, non-overscan. So the DVD rendered with both letterboxing and pillarboxing, and the circle was perfectly round, as near as I could measure. That fits with my previous experience where the composite output follows rec.601. That same DVD player follows MPEG DAR when upscaling to HDMI, just like the Blu-ray players mentioned above.
* Note that DVD only specifies the DAR. There is no PAR flag in MPEG 2 encoding except when using square pixel (1:1 PAR). The PAR is implied by: SAR = DAR / FAR, clearly spelled out in the MPEG 2 spec (SAR = sample (pixel) aspect ratio, FAR = frame size aspect ratio).
Last edited by jagabo; 28th Feb 2013 at 08:38.
The last paragraph of the section
summarises the issue nicely also (although there's better justification for the value of 704 at
At this point, it then seems my problem is caused by both overscan/nominal analogue blanking (these are seen as the same or different things by different authors), as well as MPEG vs Rec.601. My DVD player is outputting via composite (which should? produce a correct circle as you found), but that's fed to an old CRT TV without overscan control. I recalculated the error and it's just over 2%. That tallies with the percentage of overscan one could expect on my PAL TV of (720-704)/720*100 = 2.2%.
So it seems my options are to either just let it be, or to author explicitly for computer or DVD, as the case may be. If I remember correctly the DVD spec demands 720 horizontal pixels, so for that option I'd have to scale the picture to 702x576 and pad to get to 720x576, right?
Last edited by fvisagie; 1st Mar 2013 at 07:45. Reason: Corrected overscan amount
The PAL DVD spec also allows for 704*576
And once again, the behaviour at the HDMI output (MPEG AR) and the composite/s-video output (rec.601 AR) differ. So either way, one output will be wrong.
I'd really hoped this would work
Spline16Resize(704, 576, 9, 0, -9, 0)
Unless I made some error during the encode, but it doesn't look like it. So it seems it's less unsafe to just stick with 720 pixels .
Upon reflection, it still seems possible to make DVDs display close to perfectly on hardware players by scaling to 704 and then padding by 8 left and right, as mentioned at https://forum.videohelp.com/threads/342423-Upscaling-and-proper-SD-DVDs-(704x480-576i). But then they would look terrible on computer, which brings one back to the situation of explicitly authoring for one or the other, or just letting the 2% be...
Last edited by fvisagie; 1st Mar 2013 at 07:52.
Again, pls post a sample of the raw video from the camera. Only then can we try to solve this. I tried with the original dvd but he AR is alread comprimised.
Jagabo is rarely wrong so he has probably nailed it when he states that sucess will depend on the type of connection. But it is worth giving it a go.
I recommend you completely ignore the issue. Most commercial DVDs do. Nobody can see the difference without measuring the image on screen. If you use a CRT display the AR of the CRT is probably off by more than the difference between rec.601 and MPEG 2 scaling. Scaling from 702x567 to 720x576 is going to reduce image quality.
Last edited by jagabo; 1st Mar 2013 at 08:30.
Well if you are happy for the status quo then who am I to argue.
I just saw a wee challenge in taking your analog source video and experimenting with some ways to have the best of both sides of the coin. Probably would not work but there is no harm in trying.
Below are 2 images: one is a perfect circle and one is an ellipse (circle that is 2% slightly wider than it is high). JUST BY LOOKING AT THE CIRCLES (not counting knowing ahead of time which is which, or by measuring), it is very difficult-to-impossible for most people to tell the difference (if you can, it's probably because they are side-by-side, or because the left got a glitch in it). That's why it really doesn't matter too much for the purposes we've been discussing.
Joke: What's the difference between an egg and a good f@art (against thunder)
punchline: you can not beat a good f@art........
I am now somewhat confused. The specs of that sample video - 4:3 against your OP which was 16:9 - suggest that we have a bog-standard PAL DV file. There is no reason whatsoever why that will not display correctly when authored to dvd.
If I over-simplify the issue then I will retire gracefully. But do check out the sample below. Happy to do a 16:9 sample, if you care to upload one, as well.
Hehe. I posted the 4:3 sample because the 16:9 looks terrible, although they both exhibit the same 720-704 pixel problem. The camera seems to have shot the 16:9 in 4:3 aspect ratio before cropping to 16:9 creating horrible vertical resolution loss. I attach the 16:9 though, seeing you've been so kind and helpful .
However, unless I'm overlooking something, the sample you posted proves jagabo's and Scott's point. On my computer the sample renders ~2.5% too high, and with my DVD->composite out->analogue TV it renders ~2.2% too wide.
I'll leave the jokes to the politicians from now.
Below is the 16:9 dvd from your sample. The only thing that throws me is that you claim the AR to be 1.82:1 whereas both the DV and the dvd are consistant at 1.42:1.
Where, or how, do you get that 1.82:1 from ?
I do not have any 16:9 DV to compare but the 4:3 sample was identical to one on my HDD. If that is so then any pure 4:3 or 16:9 dvd will display what you are seeing except that I can not see it but then I do not have the original paper to compare against.
Display Aspect Ratio = Pixel Aspect Ratio * Storage Aspect Ratio (using the terminology from http://en.wikipedia.org/wiki/Pixel_aspect_ratio#Confusion_with_Display_Aspect_Ratio)
For "4:3" clips: DAR = 59/54 * 720/576 = 1.366 (i.e. the DAR should actually be 1.366)
For "16:9" clips: DAR = 118/81 * 720/576 = 1.821
Thanks for having gone to so much trouble with this!
Who am I to argue with Wiki. But.....
720*1.42 = 1024(rounded) * 576 the SAR of the DV and the DVD. If you do not believe me then just run both the vob and the DV through mediainfo which reports the 1.42 PAR and 1.78 DAR. Add to that any pure 16:9 dvd will report 1.78 on the tin.
It's actually called a compass that draws a circle. A protactor is used to measure an angle. But it is many moons ago when I studied geometry.
I just checked my old Canon Optura Pi DV camcorder by recording a round object, transfering to a computer via firewire, then measuring the resulting video. It appears to use rec.601 scaling (though the picture fills the entire frame). To make the round object appear round upon display the 720x480 frame has to be resized to 654x480. So the PAR is 10:11. MediaInfo says 4:3 DAR and 0.889 PAR, which is wrong for this camcorder.
Last edited by jagabo; 4th Mar 2013 at 11:32.
MediaInfo is assuming the full frame holds the 4:3 DAR. I haven't seen the DV spec myself, but as I understand it, DV uses rec.601 scaling. So MediaInfo appears to be incorrect for both PAL and NTSC 4:3 DV. But I wouldn't take that as the final word until someone can quote the DV spec.
720x576 PAR 59:54 -> 704x576 PAR 12:11 I guess, i.e.
That makes sense if you reason as follows: the "4:3" Rec.601 720x576 source holds a symmetrical picture in 720 horizontal pixels (if displayed at 1.366 granted), the analogue DVD->TV conversion presents the picture in ~704 pixels, and for that to be symmetrical, the full Rec.601 width of 720 must be resized to 704. That's how it works here at least, see attached.
I'd be very much interested to hear how that comes out on a digital output, jagabo?
Even if this comes out symmetrically on a digital output, there are still some caveats. This comes out too narrow on PC (because we started off with a Rec.601 DAR 1.366 picture now displayed at 1.333), and there's loss of horizontal resolution as jagabo points out above.
edit: It's clear that the analogue DVD->TV conversion still stretches, then blanks the 704-wide picture. If it displayed all horizontal pixels as encoded, the picture would have come out as on the computer, i.e. ~2.5% too narrow. Therefore, there's double loss of horizontal resolution: resizing from 720 to 704, and then the analogue conversion throwing away another ~8+8 pixels. 32 in total from 720, hm...
Last edited by fvisagie; 5th Mar 2013 at 02:58.
IMO (and opinions vary!) I think you're giving him bad advice jagabo. You should just encode 704x576 and be done with it.
Scaling it to 720x576 is silly. Any decent image is going to be unnecessarily compromised by that small scaling. Whereas 704x576 keeps the original pixels without scaling, and will display correctly on all kinds of outputs - both ITU and MPEG agree that it's exactly 4x3 (or 16x9) to within one pixel.
I recall the thread about 704x576 only being valid for 4x3, and not 16x9 - but across the entire internet there's only one reference to one old DVD player not playing it properly. I wonder if there's a single example of that one model of DVD player still working anywhere in the world today? Certainly all modern DVD players cope with it, because 704x576 support has been mandated for 4x3 and 16x9 MPEG-2 broadcasts across Europe for well over a decade - and silicon vendors don't put one MPEG-2 decoder into DVD players, and a different one into TVs and STBs.
I agree 704x576 is close enough, a legal frame size for DVD, and avoids picture degradation from scaling. I said so in post #3. In that later post I was telling him the proper way if he wants a 720x576 frame.
Regarding whether 704x576 is a valid frame size for 16:9 DVD I can't say. I don't have the DVD spec and I've never seen an authoritative post on the issue. But I've never seen a commercial 16:9 DVD with a 704x576 (or 704x480) frame size.
Last edited by jagabo; 14th Mar 2013 at 05:46.