what should the hd200 do for native resolution, non standard formats?

[cat6man]

i love the native resolution feature that takes 480i, 480p, 720p, 1080i and 1080p inputs and outputs the same format, permitting a more powerful video processor downstream (outboard or in a modern tv set) to do a better job
processing the video than the quite mediocre processing done by the sigma chipset.

my question is, what should the hd200 do in native mode for non-standard resolutions? i was watching an AVI file and found that the hd200 was sending out 1080p which looked lousy on a big screen.
the info said the video was mp4 @24hz with something like 6xx by 4xx resolution. when i forced a lower output resolution, the picture (routed through a vp50pro video processor outputting 1080p) was dramatically better.

so what should native mode do with non-standard format inputs?
since the objective of native mode is to do as little (ideally none) processing
as possible, at minimum it should never do any deinterlacing if native mode is set.......i.e. it should not provide a progressive output...............as for resolution, should it scale to 480i or 1080i? if the post processor is going to convert to 1080 at some point, for example, is it better to let the sigma chip go to 1080i directly, or let the sigma take it to 480i and let the video processor do a 2nd conversion to 1080i? it is not clear to me, other than
it should never (in native mode) convert to progressive.

any thoughts?

[TwistedMelon]

I've already outlined this problem to SageTV in quite some detail a couple of months ago.

The latest betas include some improved choices when making mode switches, but the ultimate way to fix this will be with preference tables.

Resulution ranges and properties for the source are specified and linked to specific output resolutions.

This enables Sage to provide defaults and allows customers to tweak the experience. As you've noticed with your source video, Sage is defaulting to the highest mode available when it can't find an exact match. This is the incorrect behavior I've also seen and outlined to them originally.

This behavior makes native resolution switching useless. The primary purpose of the feature is to maintain the best quality image. The way to do that is to maximize on the abilities of the hardware a customer has in their system. Customers who are serious about video will have video scalers either externally or integrated into higher end projectors and screens.

For these scalers to be able to do their thing, they need to receive an input that's as close to the original content as possible. If Sage is pre-scaling content out to 1080 then it's doing most of the work and the external (possibly high-end) scaler is left with nothing to do.

The output resolution needs to be the lowest possible resolution that will contain the dimensions of the source video. You might want to make some exceptions and that's why I suggest the use of tables and being able to specify source to destination mappings, including ranges and properties such as refresh.

It is a MUCH lesser evil for Sage to convert to progressive than it is for it to scale up the resolution. You can set up more output resolutions as long as your scaler is capable of accepting them. So you won't necessarily be limited to the four or five default modes. Some scalers can detect that content has been made progressive and undo it, such as products from Anchor Bay.

[cat6man]

hi bruno,

i agree with much of what you said but not all.........not sure why we have different views on some of this.

for example, i found scaling the "crummy" AVI video to 1080i was much better than letting hd200 scale to 1080p, which implicates the de-interleaver (known to be weak).

in general, i've found the sigma scaling to not be terrible (though not great) but the de-interleaving to be horrible (though my vp50pro has a PrEP mode that unravels poor de-interleaving)

i find native resolution very very useful for most of my videos, since they are primarily 480i, 720p or 1080i recordings made with silicondust hdhr tuners
or hauppauge hdpvr recordings off my STB

for the few that are oddball, they are typically torrent downloads of tv shows that i missed somehow and typically very over compressed to begin with.......like the one reported above

i'm quite happy with native resolution as it now stands for 95% of my recordings, but think it would be better if the default for oddball resolutions went to 480i instead of the highest 1080p...............in my case, i'd probably be happy with an option to choose one the supported output resolutions as the oddball case default (and 1080p is not it!)

cheers

[briands]

If the content is interlaced, you MUST de-interlace it before you scale it or you will get terrible results. There may be an exception if the output will be interlaced at the same frame rate as the source. The scaling will still be less than the best but will not have the same degree of artifacts.

[valnar]

Originally Posted by briands

If the content is interlaced, you MUST de-interlace it before you scale it or you will get terrible results. There may be an exception if the output will be interlaced at the same frame rate as the source. The scaling will still be less than the best but will not have the same degree of artifacts.

I always wondered about that pecking order - deinterlace first, then scale second. I mean... I could have looked it up many times before, but just never got around to it.

'Good to know.

[stanger89]

Originally Posted by cat6man

i love the native resolution feature that takes 480i, 480p, 720p, 1080i and 1080p inputs and outputs the same format, permitting a more powerful video processor downstream (outboard or in a modern tv set) to do a better job
processing the video than the quite mediocre processing done by the sigma chipset.

my question is, what should the hd200 do in native mode for non-standard resolutions? i was watching an AVI file and found that the hd200 was sending out 1080p which looked lousy on a big screen.

What it "should" do is generally a rather tough question, the reason being there are generally only a limited subset of resolutions/timings supported for transmission of video, eg:
480i60
480p60
720p60
1080i60
1080psF (sorta rarely supported)
1080p60
1080p24

While "home brew" video can be basically of arbitrary dimensions/timings.

the info said the video was mp4 @24hz with something like 6xx by 4xx resolution. when i forced a lower output resolution, the picture (routed through a vp50pro video processor outputting 1080p) was dramatically better.

However despite those arbitrary timings, the situation isn't really that bad. Almost all resolutions/timings are based of one of the above listed resolutions. So ideally "native" output should probably determine the "right" output format, where "right" is most likely the next larger resolution. For example your 6xx by 4xx video would "fit" into 480p60, so likely that should be used.

so what should native mode do with non-standard format inputs?
since the objective of native mode is to do as little (ideally none) processing
as possible, at minimum it should never do any deinterlacing if native mode is set.......i.e. it should not provide a progressive output...............as for resolution, should it scale to 480i or 1080i?

Well obviously 480i should be output as 480i and 1080i as 1080i. Fortunately with interlaced it's relatively simple as the only way you're likely to find interlaced video is in one of those two formats/timings.

if the post processor is going to convert to 1080 at some point, for example, is it better to let the sigma chip go to 1080i directly, or let the sigma take it to 480i and let the video processor do a 2nd conversion to 1080i? it is not clear to me, other than
it should never (in native mode) convert to progressive.

I'm not sure I follow that. If it's 480i you want it output as 480i, if it's 1080i you want it output as 1080i. You never want the extender reducing the resolution.

[cat6man]

@stanger

i agree you never want to go lower rez........my point is that are 2 places in my system to do this since i have a video processor after the hd200

assume i want to get to 1080p at the output.

should i let the hd200 produce 480i, then let the vid proc take it to 1080p
or
should i let the hd200 produce 1080i and let the vid proc deinterlace to 1080p

an analogy is a mobile-mobile cell phone call with tandem vocoders.
with tandem vocoders, mobile one codes to a low bit rate (imagine mp3), this is converted back to pcm (imangine WAV) which is transmitted to the other mobiles location, where the WAV is converted to a low bit rate (2nd mp3 coding) sent to the mobile, then converted back to pcm (WAV) which is what you hear.

this is how the cellular network operated for years and going through 2 stages of low bit rate coding was not pretty on the ears...........smarter systems later transmitted the low bit rate coded voice through the network to the receiving mobile, where it was converted back to voice with only
one net low bit rate encoding.

it obviously also depends on the quality of the codecs involved and the sigma chip is not as good as my dvdo processor..........assuming they were equal in scaling ability, the answer would clearly be to go to 1080i directly at the sigma chip and let the vid proc do the deinterleaving........with different quality of encoding, the answer is not so clear

[cat6man]

Originally Posted by briands

If the content is interlaced, you MUST de-interlace it before you scale it or you will get terrible results. There may be an exception if the output will be interlaced at the same frame rate as the source. The scaling will still be less than the best but will not have the same degree of artifacts.

the case i most care about is what you call the exception............if the source
is 60hz and not frame rate change is needed, i believe proper deinterlacing is more important (but i could be convinced otherwise if you can point me to a reference)...................when the frame rate changes, i probably agree with your ordering

in my experience at the same frame rate, the sigma chip is not too bad at scaling and s*cks at deinterlacing.........now this
may be due to basic bad implementation and not theoretical in nature

your mileage may vary........if it does, i'd be curious as to why

cheers

[stanger89]

Originally Posted by cat6man

@stanger

i agree you never want to go lower rez........my point is that are 2 places in my system to do this since i have a video processor after the hd200

assume i want to get to 1080p at the output.

should i let the hd200 produce 480i, then let the vid proc take it to 1080p
or
should i let the hd200 produce 1080i and let the vid proc deinterlace to 1080p

I can't think of one good reason you'd want the extender to scale to 1080i for the VP to convert to 1080p. If you've got a 480i file, let the extender output 480i and the VP deinterlace/scale it. You definitely wouldn't want to have the extender deinterlace/scale 480i, at least not if you've got something like a VP50 sitting there.

it obviously also depends on the quality of the codecs involved and the sigma chip is not as good as my dvdo processor..........assuming they were equal in scaling ability, the answer would clearly be to go to 1080i directly at the sigma chip and let the vid proc do the deinterleaving........with different quality of encoding, the answer is not so clear

I think you're missing something though, you can't "go to 1080i directly" without deinterlacing (assuming an interlaced source). If you've got a 480i signal, it's impossible to scale it to 1080i without deinterlacing. So if you have the Sigma chip "go to 1080i" then it will deinterlace the 480i to 480p, scale it to 1080p, and then interlace that back to 1080i, which would have to be again deinterlaced by the VP. Basically there's absolutely no point converting from one interlaced resolution to another for transmission.

I'm not saying you can't do that, but the VP won't do you any good if you feed it 1080i. Basically the only choice is if you want 480i/1080i sources deinterlaced in the extender or an external VP.

Originally Posted by cat6man

the case i most care about is what you call the exception............if the source
is 60hz and not frame rate change is needed, i believe proper deinterlacing is more important (but i could be convinced otherwise if you can point me to a reference)...................when the frame rate changes, i probably agree with your ordering

in my experience at the same frame rate, the sigma chip is not too bad at scaling and s*cks at deinterlacing.........now this
may be due to basic bad implementation and not theoretical in nature

The biggest problem with the sigma chip seems to be that it's just a flag-reading deinterlacer. So if your content is flagged right, it's not bad. But if your content is flagged wrong, or changes a lot, then the sigma chip has a very hard time doing anything other than rather bad.

[Taddeusz]

Originally Posted by stanger89

If you've got a 480i signal, it's impossible to scale it to 1080i without deinterlacing.

I wouldn't necessarily say that. In my mind I can see how it could be done. Not sure how the TV networks are doing it. But I can definitely see how it could be done. I think the key would be to do the scaling field by field and make sure the 240 lines are mapped and scaled identically to 540 lines from field to field.

Again, I can see how it could theoretically be done but I don't know how to turn that into something practical.

[stanger89]

I mean yes, you can resize/scale each field individually, but you get completely horrible results, eg:



http://www.100fps.com/

The problem is with interlaced content, each field can come from a different actual picture. Look at it like this, say we've got a 6x4 interlacedsource image:

AAAAAA
BBBBBB
CCCCCC
DDDDDD

Lines a and c are from one field and b and d from the other. So lets scale this whole image to 12x8, scaling fields at a time. We end up with two new fields (small letters are derived from scaling):

AaAaAaAaAaAa
aaaaaaaaaaaa
CcCcCcCcCcCc
cccccccccccc

BbBbBbBbBbBb
bbbbbbbbbbbb
DdDdDdDdDdDd
dddddddddddd

So now if you just interlieve these as you would normal "8i" fields:


As you can see the actual order of the fields in relation to the image is completely messed up.

There is of course another option, you could "smartly" interlieve them:

AaAaAaAaAaAa
aaaaaaaaaaaa
BbBbBbBbBbBb
bbbbbbbbbbbb
CcCcCcCcCcCc
cccccccccccc
DdDdDdDdDdDd
dddddddddddd

But this would likely result in Jaggies because each field was resized without "knowledge" of the other. This two would be suboptimal, though probably less offensive.

[Taddeusz]

I didn't say it would be easy. It would probably be very computationally intensive to ensure that the fields line up properly between the original and output videos.

[stanger89]

At the very least I don't see how that would have better results than deinterlacing first. You're dealing with very low spacial resolutions if you're scaling fields only.

[TwistedMelon]

Stanger and I seem to concur. The heavy lifting should be done externally, which is the whole point of native resolution switching. It's goal is to as best as possible, source match, giving your external processor a signal as close to the original as possible.

Any processing done by the HD200, which includes scaling and deinterlacing, 3:2 pulldown, etc... Is likely only going to serve to lessen the quality of the final output produced by a good external scaler.

The current native mode switching implementation will often choose an inappropriate output resolution.

With the previous RC, the output resolution would be your maximum available resolution if the movie dimensions didn't match exactly one of your output resolutions. So for instance you would potentially get 1080p or i out of the HD200 when playing back a 640x320 movie file.

Let's not confuse this thread with discussion about how Sage should handle scaling and deinterlacing when it's being used as a scaler. Again, the point of the native resolution switching is to allow the use of an external scaler/processor, so we want Sage doing as little as possible.

[jkohn]

I don't really see how it's much of an issue. Anything that doesn't match one of the standard resolutions is almost certainly poorly-encoded garbage, and isn't going to look good regardless of where it gets scaled.

The problem isn't so much with the HD-200 scaling, scaling is easy with today's hardware/software. It's deinterlacing that really matters, and I have to wonder how many interlaced files you're going to find that don't match a standard resolution.

Everything I play on the HD200 is either 720p or 1080i, so those are the only two resolutions I have enabled. My only real complaint is constant resolution-switching when going in and out of the UI, but I don't think there's really a good solution for that.

[TwistedMelon]

Originally Posted by jkohn

I don't really see how it's much of an issue. Anything that doesn't match one of the standard resolutions is almost certainly poorly-encoded garbage, and isn't going to look good regardless of where it gets scaled.

I hate to be blunt, but that comment is total garbage. If you don't understand, then don't participate in the discussion, but don't call other people's content and by extension their suggestions, garbage, when you clearly demonstrate you haven't the slightest bit of knowledge about the subject.

It's an issue and it's an important one for people with varied (and high quality) video collections and equipment.