AVS Forum banner
Status
Not open for further replies.
1 - 20 of 30 Posts

·
Registered
Joined
·
47 Posts
Discussion Starter · #1 ·
Hi,

Consider the situation I am in..A prog scan dvd palyer (rp56) driving an LT150. With this setup I am pretty much by passing the supposedly bad deinterlacer in the LT150. Now LT150 is doing the scaling to it's native resolution.


I keep hearing in the forum that an HTPC would still improve the picture quality by bypassing the internal scaler of the projector which is again presumed to be inferior. Now forgetting all the other benefits from an HTPC..from a pure PQ perspective for a dvd..can somebody explain the technical reasons for this? In the limited knowledge I have on scaling..it should not be very complex (not as complex as de interlacing)..Mostly some sort of matrix multiplication. What leads people to believe that this operation is blotched by the internal scaler?


Or is the scaling operation much more complex than I thought?
 

·
Registered
Joined
·
79 Posts
Scaling is a very simple operation. I think these external scalers are expensive because there is a small marketplace for them. It is certainly not difficult to write the code to scale an image. The only issue is the performance necessary to scale the image in real time requires a fairly fast processor. There are DSP chips available for a reasonable price that do this very well. I'm not sure how the HTPC software is doing this but it could be directly by the PC processor or using the GPU on the graphics chip. All newer 3D graphics chips have the capability of scaling images using bilinear interpolation.


--sdc
 

·
Registered
Joined
·
47 Posts
Discussion Starter · #3 ·
In my experiment with LT150 here's what I found..


1) S-video into LT150 vs 480p (from rp56) into LT150.

There is a noticeable difference ..although not earth shattering or jaw dropping. This kind of proved to me that the deinterlacer of rp56 is infact better than LT150..or it could be that the deinterlacing without the intermediate d/a conversions is the reason for this improvement.


Either case I am convinced about the benefits of bypassing the de interlacer of LT150.


Now the issue I am trying to understand is how much more improvement can be done by using an HTPC to scale the picture. If you ask me now I would be very sceptical of ANY improvement.


So what is the improvement people are realising going from rp-56 to an HTPC (again..from a dvd playback viewpoint).


Or is it that the deinterlacers used by HTPC are better than the sage/faroudja chip
 

·
Registered
Joined
·
6,225 Posts
OK, I'll take a crack at it.


With a progressive DVD player you get an analog output in the 480p video mode. Since this does not match the native projector resolution it must be scaled. The steps are:


1) Digital DVD data is de-interlaced in the DVD player, and scaled digitally if the source is anamorphic and then letterboxed or pillarboxed if the source aspect ratio does not match the output aspect ratio - all of these steps are in the digital domain, with minimal degradation, although only very expensive DVD players approach HTPC performance levels here.


2) Digital to analog conversion occurs to drive the analog component video outputs of the player, and picture quality is lost. In either the DVD player or HTPC video board, this step is not too bad - very high quality converters are used.


3) In the projector analog to digital conversion occurs, and picture quality is lost.


4) Projector digital scaling occurs, and significant scaling artifacts are introduced, unless you have a really expensive projector with the Faroudja scaler chip.


5) Partial digital to analog conversion occurs in the digital projector driver circuitry (some of which is analog), and some picture quality is lost. If your projector is an analog CRT display, total digital to analog conversion and a little more quality is lost.


Contrast this with the HTPC:


1) Digital DVD data is de-interlaced and scaled in the digital domain, with about equal performance with the most expensive DVD players. However, the scaler produces not the standard 480p but the exact native projector resolution.


2) Digital to analog conversion occurs in the very high quality HTPC video board, and the RGBHV (VGA) output quality is even better than the progressive component video signal with it's embedded sync signals.


3) The native resolution bypasses the internal projector scaling, and the A/D and D/A conversions that accompany such. Thus no picture quality degradation in the relatively simple projector circuitry due to scaling.


4) Partial digital to analog conversion occurs in the digital projector driver circuitry (some of which is analog), and some picture quality is lost. If your projector is an analog CRT display, total digital to analog conversion and a little more quality is lost.


The last step is usually the same no matter what the projector, unless you have a DVI input on your projector. Since such DVI inputs have very limited cable lengths however, the projector interface of choice is typically one of two analog inputs, the RGBHV via (5)BNC or (1)HD-15 (VGA) connectors, or the almost-as-good analog component video via (3)RCA connectors.


In theory, the only time a standalone progressive DVD player would approach HTPC video performance levels would be in the exception case where the projector was a 480p native projector such as the Plus HE-3100, aka "Piano". Of course, it isn't as bright as most projectors, and at "only" 480p, you will have to be carefull about pixel visibility. Furthermore, unless you drive it with RGBHV, it will still be a few percent short of a good VGA-connected projector.


The real advantage to the HTPC is not only does it produce video quality equal to the best DVD player, but it gets the best possible performance from an inexpensive business presentation grade projector, where the video inputs and associated circuitry are extra features, not particularly sophisticated, and definately somewhat compromised quality wise. This advantage begins to disappear once you move up the food chain to very expensive home theater projectors.


Gary


[This message has been edited by Gary McCoy (edited 10-04-2001).]
 

·
Registered
Joined
·
47 Posts
Discussion Starter · #5 ·
Thanks gary,

I guess the real difference is probably going to be the extra d/a a/d conversions( i didn't take this into account).


I am going to borrow a friend's HTPC tonite and try it out and see how much of a difference it makes.


Any suggestions on specific dvd's/scenes where the scaling artifacts would be very apparent?


(Right now i have benhur,bug'slife,chicken run,gladiator,matrix,pitch black,crouching tiger and a few more in my possession)
 

·
Registered
Joined
·
2,091 Posts
Schintam,


I would be very interested in your impressions of the difference. I'm using a Faroudja scaler (scales to 480 or 600p) with my LT150 and see no artifacts, but I have wondered if a HTPC or Faroudja Native Rate 1024x768 scaler would yield an appreciably better picture.


------------------

John
My HT Picts
 

·
Registered
Joined
·
1,679 Posts
let me just dissect these:


Progressive scan player:


1. Standalone progressive players are better at de-interlacing than PC players. PCs are extremly lousy at de-interlacing. If the source is anamorphic you're just outputting it directly with no letterboxing or pillarboxing. You're supposed to let the projector do the anamorphic scaling unless you want to lose resolution.


2. ok


3. ok


4. ok


5. For CRT projectors you'd use a separate scaler that's at least half decent, or input the 480p resolution directly with no scaling whatsoever and no AD/DA loss whatsoever.


HTPC


1. ok


2. ok


3. Duh what? you'd still have an AD/DA conversion - just no scaling.


4. ok



The real difference between the two is really in the inferior scaling algorithm used by the chip in the LT150. AD/DA has nothing to do with it. A HTPC with a good video card (Radeon) has much better scaling and interpolation performance than the internal processor of the LT150. That's it.


------------------

/frode
 

·
Registered
Joined
·
6,225 Posts
Frode,

In actual fact, modern software DVD players have pretty elaborate deinterlacing algorithyms, and are generally considered the best players available. There are tons and tons of posts in the Forum archives comparing players, I'm not making it up. I've been following the topic for years, and even today most posts favor the software player - but you are certainly entitled to the minority opinion if you wish.

External scalers, no matter how expensive, still must scale in the digital domain, so there is an inherant A/D and D/A conversion involved. You correctly point out that analog CRT projectors have multiscan capability, so that sometimes scaling is not necessary as the signal is within the scan range of the CRT, and can be tweeked to the screen size and shape with simple size and position controls. My first home projector was a CRT, so I'm familiar with their advantages, however, there is no way I would go back after owning a digital projector.


Gary
 

·
Registered
Joined
·
60 Posts
Quote:
In actual fact, modern software DVD players have pretty elaborate deinterlacing algorithyms, and are generally considered the best players available.
The only thing elaborate about software DVD players' de-interlacing is that they listen to flags encoded in the material to indicate when it's playing a film source DVD. They then reconstruct the progressive frames out of that. So, yes, for film source material, software DVD players are excellent.


However, a good standlone deinterlacer will rule any current software DVD player when playing video source DVDs, or DVDs with improperly-encoded flags. This is because standalone deinterlacers use a number of complex and sophisticated algorithms to reassemble interlaced frames, even for video source material. They are not dependent on flags in the source DVD data to tell them when film source data is being played, and so they handle even incorrectly-encoded sources well by doing their own 3:2 pulldown detection. The end result is way better than software DVD players which tend to just do simple bobs and weaves if it's not a film source.


Aaron
 

·
Registered
Joined
·
1,679 Posts
Gary..


Umm.. HTPCs are flag-based players. They don't even have adaptive de-interlacing, and even the Radeon's is so basic that it's almost worthless. For standalones however they are far better in the de-interlacing capacity as you can get cadence-reading, and proper adaptive de-interlacing (though of course not all standalones are created equal). HTPCs are by definition crap at de-interlacing any kind of material not optimally mastered.


I think you also might have misunderstood the original poster. He asked about the progressive standalone+_internal_ scaler vs. HTPC, not external scaler vs. HTPC. The only real difference between internal scaler and HTPC in this instance would be actual scaling in the projector itself as the AD/DA steps are the same.


The reason people prefer HTPCs is due to the superior scaling - not de-interlacing. I assume you've read Secrets comparison of de-interlacing capability? http://www.hometheaterhifi.com/volum...-2-8-2001.html


------------------

/frode
 

·
Registered
Joined
·
47 Posts
Discussion Starter · #11 ·
Wow,

I am going in circles here http://www.avsforum.com/ubb/smile.gif . Now I begin to think that my initial thought of scaling being the ONLY difference was infact correct.

Anyway..DVD/scene Suggestions please for my own shootout...


Before I give my impressions after the comparison..please note that my screen size relative to viewing distance is probably smaller than what most of you have..so please keep this in mind.


Screen size : 100" diagonal 4:3

viewing distance about 13-14~
 

·
Registered
Joined
·
1,679 Posts
schintam,


The further back you sit the less of a difference there will be. However - have you read my HTPC vs. Skyworth (FLI2200) comparison for the LT150? The Skyworth uses the same chip as the RP56 and so should have comparable picture quality. http://www.avsforum.com/ubb/Forum10/HTML/007503.html


------------------

/frode
 

·
Registered
Joined
·
6,225 Posts
Frode, interesting read - I guess I'm a HTPC bigot - I just realized the only video source DVD I own is Akira although I've rented many more - I use CineMaster 2000 and some different DVDGenie tweeks for video source material.


The one comment in the Secrets article I am questioning is the remark that modern CPUs don't have the horsepower for adaptive deinterlacing. Since the minimum CPU spec is typically 500Mhz or below, and modern CPUs much faster, I just don't agree. Maybe it's time we push the software player venders to acquire the missing function.


After all, adaptive deinterlacing appears to work well in DScaler.


The other thing we should do as consumers, is squawk about improperly flagged DVDs!


Gary


[This message has been edited by Gary McCoy (edited 10-04-2001).]
 

·
Registered
Joined
·
1,679 Posts
Gary,


Akira isn't video source - it's a hybrid interlaced/progressive encode. That means it's primarily progressive film, but it has been improperly authored so it contains lots of interlaced sections. That's even worse than pure video in a way as the normal software DVD players on a PC will flip back and forth between bob and weave. I did ream Pioneer out for this - and hopefully they did learn something from my comments. Their next project "Ah! My Goddess" will be the next test case.


dScaler is nice - but it's still not close to what you can get out of a FLI2200. There's also the added limitation of it currently being bound to the inferior hardware of the BT chips and the noise&softness that comes with them. I would love to see dScaler rivaling a FLI2200 based standalone player for video source DVDs, but I can't exactly see that happening today. Until then I'm stuck with the dual solution of using a software DVD player for film source, and the Skyworth for everything else.


------------------

/frode
 

·
Registered
Joined
·
6,225 Posts
Frode,

I think the real limitation in DScaler performance is not the BT hardware but the simple lack of picture information in broadcast signals. Essentially composite video is used to Amplitude Modulate the RF video carrier, then envelope detection in the receiver is used to attempt to recreate the composite video signal. The best it's ever going to get for broadcast analog video is almost as good as direct composite video, until HDTV is universal.


I wasn't suggesting we actually use DScaler in it's present form, rather that perhaps the deinterlacing part of that product might find a home in a software DVD player, specificly for video source material. The source quality would be much higher than broadcast video.


Gary
 

·
Registered
Joined
·
10,201 Posts
Gary, the computational power available in an ASIC design is, even today, vastly higher than is available in a single-processor PC. ASICs have the potential for serious parallel processing that can't even be considered on a PC, and video processing is perfect for parallel implementation. I don't think it's an exaggeration to say that you a chipset could do the 1000 times more computations per second than a PC.


Now the real question is whether or not 1000 times more computations are necessary.


On one hand, you have to consider how much horsepower is under the hood of a Teranex, which is the cream of the crop of software-based video processors.


On the other hand, in the more reasonable price range I think it's reasonably safe to say that the best software-based deinterlacer out there right now is in the Rock+ firmware. Certainly, a sufficiently powerful HTPC could implement the deinterlacing algorithms that the Rock+ uses. The HTPC would also have to implement an MPEG2 decompression engine, too.


Judging from discussions on this forum, most people think that Faroudja's hardware deinterlacer is better than the Rock+ on video material---but the gap is closing. So I would say that we are rapidly approaching the point where HTPCs could match the deinterlacing and scaling quality of, say, a Faroudja + SDI system. (Not that a Faroudja + SDI system exists yet, but hopefully you get my point.)




[This message has been edited by mcg1969 (edited 10-04-2001).]
 

·
Registered
Joined
·
79 Posts
Gary,

You may be right about that the software vendors are not doing as good a job as they could with adaptive de-interlacing. One question I have is do these products perform radically different on P2 vs P3 and P4 processors. The P3 has additional instructions that are very useful in doing image processing and the P4 extends those further. If the software vendors are not taking advantage of that the products would be vastly inferior.


That said the comment made about and ASIC being much faster is true because they can be designed to pipeline a specific processing function. Ultimately a custom piece of hardware can do more in real time than a general purpose processor. The real question though is can a fast P3 or p4 do what is required for good adaptive de-interlacing.


--sdc
 

·
Registered
Joined
·
6,225 Posts
Yes but we have very powerfull ASICs in the PC already, such as the video chipset on the video board. Visualize the critical compute-intensive tasks like motion-adaptive de-interlacing relegated to an ASIC that was optimized for that function. Hardware MPEG decoders are also common examples of powerfull ASICs.


Finally, if we are still banging up against the computational limit, there remains the option of using a dual processor motherboard and a video application designed for that environment.


If someone were to offer such an application to me, it would stimulate me into something I seldom do - another round of HTPC hardware and software upgrades. I just need a good reason to spend the money.


Gary
 

·
Registered
Joined
·
47 Posts
Discussion Starter · #19 ·
Well I finally got to compare my prog scan vs a HTPC (won't take names...one of the pre built ones from a well known HTPC source).

My intention was not to compare an HTPC vs prog scan player but to determine if they make a difference in MY SETUP. So YMMV.

Here's my setup (video)

LT150 + Dalite hi power 60x80 (100" diag) + home made masking ..sitting in a reasonably light controlled basement (almost passes the hand wave test).


seating is 13-15 feet from screen.

panasonic rp-56/sony s330 interlaced player/HTPC (for today).


I wear contacts (1.75 power in each eye)..recently checked prescription http://www.avsforum.com/ubb/smile.gif ..(hey..this does make a diff alright)


and I am an IC designer..(might skew my bias towards hardware solutions http://www.avsforum.com/ubb/smile.gif )

UNfortunately all I had in hand were pretty good transfers..Fith element,Pitch black,Crouching tiger and benhur. I did not want to LOOK for defects so I did not move closer to screen to see any differences.

ANyway first off the difference between the S video interlace and component progressive was clearly visible but not jaw dropping.

The real test:

HTPC took some time to boot up and configure and some minor hiccups later fired up the picture and first impression was really impressive. Now i thought ..if I go back to the rp56 now I will see the difference.

Absolutely NO! I mean there were differences in the picture but from my seating distance the PQ is nearly identical. Infact factoring in the usability perspective I can say the rp56 has a slight edge...


Well looks like I saved some money for now http://www.avsforum.com/ubb/smile.gif


Anybody has any specific suggestions on scenes where I MIGHT see a glaring difference please let me know.


Well, somebody recently mentioned in this forum that for DVD playback the best bang for the buck is LT150 + rp-56 ..I whole hearedly second that.
 

·
Registered
Joined
·
79 Posts
Gary,

The ASICs on the video board will not do much good unless they have been designed to do de-interlacing. It is possible that some boards do have this capability. I am not as familiar with the video end of these boards but will look into it. It makes sense that some capability must exist on boards that can capture NTSC video.


A dual processor will only possibly double the speed of the image processing and that if memory bus conflicts can be avoided. There are could also be potential caching problems for some algorithms. Nothing will slow a computer down faster than cache misses.


--sdc
 
1 - 20 of 30 Posts
Status
Not open for further replies.
Top