That may have been your impression Nigel, but 1440 simply cannot show fine detail, my images show this, it doesn't matter whether it is BBC pre or after the encoder change or ITV (1) HD we are talking about. If people could get hold of some of the 1920 transmissions from BBC Scandinavia I am sure these would show the same effect (on material that is filmed in 1920, if it is filmed in 1440 obviously there won't be any difference). I did a comparison on a pre encoder change recording with the 9.6 encoder transmission of Joanna Lumley's Northern Lights- to be quite honest there was hardly any difference at all, this old vs new encoder is really a red herring. It may be that on some situations the new BBC HD encoder does not increase bitrate when it should, i.e. with smoke or mist or shadow contrasts, which I have observed at 4-6Mbps, will lead to compression artifacts and unrealistic appearance, which may not have happened with the old encoder fixed at 16Mbps.

It doesn't matter if you consider it an 'impression' or not, the old BBC HD transmissions were a better picture and contained more detail than all others.

Presumably you are a newcomer to HD?, and never saw them - the 'wow factor' was huge, and there's none at all now.

No, I have had HD for about 3 years now, I saw the old BBC HD encoder at 16Mbps, and still have recordings using that encoder, as I explained in my addition to my post, I did a comparison of Joanna Lumley's Northern Lights on the old 16Mbps encoder and compared to the new encoder when it was fixed at 9.6Mbps, there was no difference at all, there was absolutely not more detail in one or the other. A drop in perceived quality was due to increased artifacts when there was fast movement or complicated imagery, this was largely fixed when the channel switched to VBR, although as I explained above there are still situations where the encoder should increase the bitrate but it doesn't which will lead to artifacts and an unrealistic appearance. The simple fact is 1440 x 1080 has 518,400 LESS pixels of information than 1920x1080, 1440 cannot magic this extra information back, it is gone, as many of my examples show, where vertical patterning completely disappears.

We too had examples of previous recordings, and played them back side by side on identical TV's - the detail and quality was massively better on the old recordings.

Are you denying the old BBC HD pictures weren't the best that were available?.

Well that would be pretty hard to judge as other channels didn't transmit the same programmes, so what are people meant to compare?

A perception that one might be better than the other?

I have seen several programmes on 'other channels' that have fine detail in certain scenes that I can't remember seeing on BBC HD. But they weren't the same programmes, so I can't say whether they or the BBC HD were the "best available", full stop.

Are you saying you compared recordings of the same programme under the old encoder and the new encoder, either at 9.6 or VBR? If not, and you are saying you just played two recordings under the old encoder and they looked good, that may well be, but unless you can show the same programmes looked bad under the new encoder I don't think that is really valid I'm afraid. Some recordings may certainly have looked worse under the new encoder at 9.6, where there was fast movement or complicated scenes, but I think VBR solves most of these problems. A good example of where the new encoder when fixed at 9.6 produced obvious artifacts which no longer occur as the bitrate can increase, is the BBC HD ident with the rotating 3D diamond, this collapsed into blocks previously but is now clear throughout the animation.


It would be good if we could get hold of some of the BBC HD Scandinavia content as that is broadcast in 1920, but we would only see difference on 1920 recorded content, I think quite a bit of the older stuff, possibly still some studio content is filmed in 1440 so 1920 would give no advantage in that case.

I guess I must be missing something - why does an engineer from an STB manufacturer that will handle 1920 perfectly well have a 'massive interest in defending 1440'? It's not as if I can buy an alternative product that will give me access to the same programs in 1920 is it?


So are the two ITV 1 HD streams identical in every aspect (bit rate, encoder settings, etc) on a frame by frame basis except for ITV1 HD London downsampling to 1440?

Because virtually all FTA HD receivable by Humax Freesat or Freeview HD receivers is 1440, it's conceivable that some may not be happy with that, and either not invest in a HD receiver at all and just rent blu rays (as I know some people do), or choose to subscribe to Sky to get Full 1920 HD when they might have gone with a Freeview or Freesat product, when they thought that was proper HD.


It is not known what encoder they use, but they do use the same bitrate 10.2Mbps, it is known 1920 generally requires a higher bitrate for the same amount of clean artifact free image than 1440 due to the increased number of pixels, but ITV1 HD Granada, Central and STV HD use a slightly inadequate bitrate for the resolution- I think the EBU recommend 13 or 14Mbps for 1920 CBR so you can see more compression artifacts on ITV1 HD Granada. Please look at my images and read the descriptions and at Graham's BBC HD test card- you can clearly see the lack of horizontal resolution often results in complete loss of some detail, such as vertical lines in clothing or the lines in the test card, and when it doesn't cause complete absence of detail it blurs other detail reducing its 3D lifelike quality, i.e. in image 4 1440 London vs 1920 Granada, the vertical lines in the French flag completely disappear in the London version, Jodie Kidd's earrings and Julian's glasses, and both their hair in the 1920 version is noticeably sharper and brighter and more lifelike as a consequence.

I doubt it, the three streams on the Sky transponder STV-HD, ITV1-HD Granada and ITV1-HD Central West presumably will be stat-muxed an in consequence have vbr depending on content

I have no vested interest either way, I am explaining why your test is not significant to the argument.

Yes, we had examples of the BBC HD Preview channel recorded - so could compare them with EXACTLY the same programming side by side on identical TV sets. It's the advantage of working for a TV shop - we have multiple HD boxes and multiple TV sets available

The original BBC HD pictures were the best of all of the HD channels, then over night quality plummeted to be the worst. The pictures were soft, out of focus looking, no definition, and with far more artifacts.

The later 'improvement' made it a little better, but it's still soft and lacks definition, a 'poor mans' HD

There's an opportunity to directly compare this week with the ATP tour finals showing the same feed on both Sky Sports HD and BBC HD.

Just about to post that, & the Sky picture is better than the BBC HD picture.

They are not stat muxed, all fixed bitrate at 10.2Mbps.


Of course you have an interest as I have already stated, virtually all the HD channels accessible by your boxes are 1440 bar one (save for Luxury Life & STV HD in non Freesat mode). My tests would not be significant if I and others couldn't tell the difference between the 1920 and 1440 versions from a normal seating distance, as we can, they surely are significant.


Yes, it is well known the 9.6Mbps rate was highly inadequate, I don't think I can accept there was such a massive difference between the 16Mbps and new encoder under VBR though unless someone can produce precisely the same scenes or programmes to compare.


I wouldn't be surprised at all, basically anything that is not a long distance landscape shot with things clearly in focus and you can tell the difference 1920 makes.

So all the .png files are 1920x1080 even though London transmitted as 1440? Why are the London pictures larger files?

The ITV1 HD London pngs were output directly from DVBViewer, whereas the ITV1 HD Granada pngs were output from MPC-HC as MPC-HC will only output 1440 screenshots as 1440x1080, I would have had to put them through another process of resizing otherwise. It was also easier to match the precise frame in MPC-HC as it has frame skip which DVBViewer does not have. Any file size difference is therefore down to the compression setting difference, this will have no visual impact as png is lossless. The same decoder and deinterlace setting was used in both programmes & I did test other decoders and also made 1440 screenshots using MPC-HC and resized to 1920 png to make sure it made no difference to the missing detail in the 1440 samples.

Fist of all the mpeg decoders upscale 1440 x 1080 to 1920 x 1080 before you can grab an image. (they respect the transmitted pixel aspect ratio information)

Secondally because jpg and other picture compression techniques produce files that are smaller with larger areas of similar colour. jpg additionally is lossy compression algorithm that loses data it thinks you won't notice.

Simplistically imagine an image with 255 adjacent pixels all the same colour. Assuming 24 bit colour you would need 3 bytes one for each value of Red, Green and Blue x 255= 765 Bytes.

The same information can be transmitted losslessly in just 4 bytes.
3 for the colour information and 1 for the number of repeats

So how does a line of 1440 separate pixels turn into 1920?

The same way a SD 720 x 576 pixels turn into 1920 x 1080 to display on a full HD TV.

Basically the video processor looks at adjacent pixels and invents the intermediate pixels. The better it does this the better the picture. Horizontal interpolation is much easier than vertical.

Many thanks to jzee and everyone else here for this interesting discussion.

From my viewpoint, viewing on a 42" Panasonic 720p plasma screen with content from a Sky HD box, it is clear that the picture quality of BBC content is noticeably degraded from it's pre August 2009 position.

I have three Sky HD boxes, two of which have content from 2008 and early 2009; the pre August 2009 PQ is markedly better providing the Wow we were promised from the BBC.

I find it rather disingenuous of the BBC to repeatedly refer to 'up to five times more detail than ordinary standard definition TV' then broadcast at 1440.

I would concur with others comments about better content and improving SD quality, (could we get rid of all the shopping, sex, +1 channels from terrestrial), but would also like our publicly funded national broadcaster to set the standard as far as technical excellence is concerned.

Sadly the BBC appear to restrict their transmissions to the limited bandwidth available via Freeview, rather than utilising the increased availability of satellite.

Restore the Wow and Ditch the DOG!

john

Not strictly true.

The 1st episode of Lost Land of the Tiger recently blew me away for its jump in quality. Either there was a enocder experiment or the production was exceptional on this episode and compensated as it really was a step up from anything seen for a long time.

Its just a pity episodes 2 & 3 dropped back to a lower level with every subsequent episode. Episode 1 sublime, 2 good in parts, 3 quite poor. Almost makes me wonder if the BBC did an encoding experiment on it and dropped the encoder per episode to see if anyone would notice.


As an engineer, I don't think Bob has anything to gain from this. After all he doesn't make the final decisions on resolution nor chip purchasing. Thats a boardroom decision. Hopefully the new version of the HDR will see a 1920 chip. I think the problem in justifying putting it in a product at the moment anyway stems from the fact that the BBC don't seem to have plans to transmit 1920 at the moment odd test apart.

Also, I have to wonder is 1440 a limitation of the decoding chip or the firmware? If firmware then its upgradeable anyway.

@jzee,

I'm amazed anyone would discount the difference between your screen shots. The difference is obvious. The 1920 versions are pin-sharp, the 1440 version are quite blurred - certainly not as soft as upscaled SD, but not really HD.

However, I think there's something wrong with your screen caps, or something very wrong with ITV1 which isn't intrinsic to 1440.

Try this:
1.Take the 1920 version.
2 Resize it to 1440x1080.
3 Resize it back to 1920x1080.

This gives you a version that's been scaled to 1440, and then re-scaled back to 1920 for display. Which should be pretty similar to the 1440 version you captured. It's not!. You'll find that the captured 1440 is far softer than the 1440 version you just created.

Obviously the 1440 version you created hasn't been through MPEG encoding at 1440, but if you take a look at the differences between your own 1440 and the captured 1440, I think you'll realise that MPEG encoding has nothing to do with it. Either ITV, or your screen capturing, is messing it up.

1440 doesn't have to look that bad.

Comparing the 1440 you generated to the captured 1920, of course the captured 1920 looks better - but it's not such a big difference.


I think bob_cat is wrong to dismiss this comparison out of hand, but he's obviously right that different silicon will perform the 1440>1920 upscaling in slightly different ways, and maintain more or less sharpness. The scaling in my photo software introduced slight jaggies, but I know if I'd used some edge-directed interpolation algorithm this would have been avoided.

Cheers,
David.

P.S. It's interesting that broadcasters have made choices about broadcast standards which mean the quality of reception equipment becomes important. MPEG decoding is fully standard - a STB/chip either does it correctly, or not. Broadcasting 1920x1080p50 would leave little for STB and display processing to do other than follow the standard. But good deinterlacing comes at a price. Good scaling some at a price. BBC HD needs deinterlacing and scaling!

Who said the Humax HDR can't decode 1920x1080?! Did you just make that up, or have I missed something?!

Cheers,
David.

I presume he just made it up?.

I wouldn't be much use as a box if it couldn't accept the transmission standard it's designed for

Given these would have identical pictures for large parts of the daily schedule, there wouldn't be any advantage in stat-muxing..