Sign in with
Sign up | Sign in
Your question

Best comb filter?

Last response: in Home Theatre
Share
Anonymous
September 7, 2004 6:41:49 AM

Archived from groups: alt.tv.tech.hdtv (More info?)

Is there any one set or brand that has the best comb filter?
Steve Grauman

More about : comb filter

Anonymous
September 7, 2004 8:00:41 AM

Archived from groups: alt.tv.tech.hdtv (More info?)

In article <20040906224149.08576.00000279@mb-m07.aol.com>,
oneactor1@aol.com (Steve Grauman) writes:
> Is there any one set or brand that has the best comb filter?
>
I know that I am not directly answering your question, but I can
give SOME input.

Firstly, I have seen several 3D combs (which is the theoretical best
for NTSC), most that have excellent performance on stationary scenes. One
good thing about 3D comb concept (beyond the obvious seperation of
chroma and luma) is that the diagonal resolution can
be unlimited by the color subcarrier frequency range. (With normal
line combs, you'll see that the comb can provide wide horizontal freq
response, but tends to cause the diagonal response to rolloff
significantly -- essentially in the diagonal equivalent 200TVL range.)

The effect of poor diagonal response is best seen when running through
multiple encode/decode cycles, where the resulting image does measure to
have good horizontal and vertical response, but the picture still looks
oddly missing detail. At one generation, the loss of image detail is
noticeable (in direct comparison) but often doesnt' jump out at you.
The effect tends to look like an oddly slightly sharpened cartoony
looking video, obviously a transient response issue due to some of the
spatial domain signal processing.

Also 2D combs will still allow significant leakage even for static scenes,
but more and more sophisticated schemes can help to incrementally improve
and remove artifacts.

A combo 2D/3D comb, where the best 2D technology is used, while the 3D
scheme is added on will produce an image that does look almost like Y/C,
perhaps will some loss of detail and some artifacts (but far far far
better than simpler schemes.)

The biggest bugaboo that I have seen with 3D combs is the comet-tailing
associated with the time domain filtering (and attempted noise reduction.)
Proper removal of comet-tailing (and intelligent switching between 3D and
2D modes) can look damned good. A good example of a comet-tailing
susceptable 3D comb will be the JVC 30000 DVHS/SVHS deck. It is a fairly
good 3D comb, but the designers didn't seem to adequately deal with the
temporal effects. (This effect is one reason why PAL combs aren't 3D,
it isn't the 50Hz thing, but the 4 cycle phasing instead of two. This makes
the interval over which the image must be unchanging to be so long that
the advantage of a non-motion compensating 3D comb is nil.)

The best CONSUMER, BUILT-IN, but still imperfect 3D comb that I have evaluated
in my own equipment selection might be in the RCA HDTV. The bad news is
that the horizontal pixel rendition is only 640H for NTSC, which throws away
information.

I haven't carefully reviewed my Canopus ADVC 300? yet, but I'd suspect that
the built-in 3D comb is probably the best in my equipment array (I GUESS that
it is either on-par or better than the RCA.) In transcoding some of my
old, noisy LDs, my resulting video quality (after significant video processing)
and using the Canopus device for Composite decoding, the resulting video
looks VERY SIMILAR to a soft DVD, and most NTSC artifacts are essentially
invisible. (The NR on the TMPGENC does much of the work, but the Composite
deocding in the Canopus Analog <---> DV25 box does its' fair share of Composite
video processing.)

So, when reviewing a TV set with a 3D comb, check carefully the motion
rendition and how well the chroma/luma seperation is sustained during
movement.


Even though I haven't answered your question directly, I have tried to
help without giving TOO MUCH stealth bias.

John
September 8, 2004 3:36:25 AM

Archived from groups: alt.tv.tech.hdtv (More info?)

On 07 Sep 2004 02:41:49 GMT, oneactor1@aol.com (Steve Grauman) wrote:

>Is there any one set or brand that has the best comb filter?

Good question, but getting to be moot, for many people, with all their
video sources S-video or component.
Related resources
Anonymous
September 8, 2004 10:39:58 AM

Archived from groups: alt.tv.tech.hdtv (More info?)

>In transcoding some of my
>old, noisy LDs, my resulting video quality (after significant video
>processing)
>and using the Canopus device for Composite decoding, the resulting video
>looks VERY SIMILAR to a soft DVD

It's funny you should mention this, as my primary concern for the comb-filter
is how well it will clean up LD-video. I generally run my DVD player via
component and my DSS/DirecTV reciever via s-video so the comb filter isn't in
use. But I run my LD player via composite output because the comb filter in a
new set is usually far and away better than the one that would be utilized
inside the player if I choose to use the S-Video output. I've had several
people reccomend to me the Crystal Vsision VPS-1, maufactured by Camelot
Technologies. It's an outboard comb filter, a 2D unit which apparently is based
on Phillips hardware. They've said that running the VPS-1 between a LD player's
composite outputs and the TV makes a significant difference in video quality.
However, the VPS is a several year old design and I'm tempted to believe that a
high quality 3D Y/C unit such as what would be found in Sony's 34-XBR960 HDTV
set would be superior. Any thoughts?

>Even though I haven't answered your question directly, I have tried to
>help without giving TOO MUCH stealth bias.

You're being quite helpful, thanks.
Steve Grauman
Anonymous
September 8, 2004 10:40:37 AM

Archived from groups: alt.tv.tech.hdtv (More info?)

>Good question, but getting to be moot, for many people, with all their
>video sources S-video or component.
>

True, I'm only concerned because I still feed a LD player to my TV via
composite, so the comb filter will be getting some use.
Steve Grauman
Anonymous
September 10, 2004 5:12:32 AM

Archived from groups: alt.tv.tech.hdtv (More info?)

In article <20040908023958.25562.00000312@mb-m29.aol.com>,
oneactor1@aol.com (Steve Grauman) writes:
>>In transcoding some of my
>>old, noisy LDs, my resulting video quality (after significant video
>>processing)
>>and using the Canopus device for Composite decoding, the resulting video
>>looks VERY SIMILAR to a soft DVD
>
> It's funny you should mention this, as my primary concern for the comb-filter
> is how well it will clean up LD-video. I generally run my DVD player via
> component and my DSS/DirecTV reciever via s-video so the comb filter isn't in
> use. But I run my LD player via composite output because the comb filter in a
> new set is usually far and away better than the one that would be utilized
> inside the player if I choose to use the S-Video output. I've had several
> people reccomend to me the Crystal Vsision VPS-1, maufactured by Camelot
> Technologies. It's an outboard comb filter, a 2D unit which apparently is based
> on Phillips hardware. They've said that running the VPS-1 between a LD player's
> composite outputs and the TV makes a significant difference in video quality.
> However, the VPS is a several year old design and I'm tempted to believe that a
> high quality 3D Y/C unit such as what would be found in Sony's 34-XBR960 HDTV
> set would be superior. Any thoughts?
>
The Philips design likely has a very good multi-line comb filter. I have
looked at the Philips chips in the past, and some of their latest multi-line
combs (not 3D, but just 2D) look damned good. However, the multi-line comb
isn't as good as a GOOD 3D comb. The loss of diagonal resolution is still
existant, but who cares?

I have to say that the Philips chips are surprisingly good, however. For
applications desiring the very best performance, they don't achieve the best,
and they wouldn't be the best choice. For applications that desire damned
good performance, especially if a pre-encoder filter (for MPEG2) would
be used, then some of the problems with some of the NTSC decoders will
be mitigated.

So, if there are some limitations on the available funds, then the Philips
based solution would be a good second choice.

For my own application, I get INCREDIBLE performance by using a
DPS290 TBC (totally optional, not really needed except for the worst
time base problems), Canopus ADVC300 for conversion to DV25, then converting
to MPEG2 using the TMPGENC software. I use small amounts of NR, using the
3D mode on the Canopus. I use the TBC only for problematic LDs, and it
can provide that 'extra' amount of NR, but causes comet-tailing if overused.
I use the largest amount of NR that I can use without artifacting on the
Canopus ADVC300. Also, I do a little high frequency compensation with the
TMPGENC package.

My end result is almost (well, not quite) indistinguishable from a good
DVD that perhaps has a little too much HF rolloff. Casual viewing
by this very picky video person makes it unlikely to detect that the
video had ever been LD. The result looks mostly like a slightly soft
DVD instead of a very good LD.

The DV25 stage might give some purists a little discomfort, but there
is really little/no loss because of the much greater amount of quality
limitation comes from the composite frequency response limits. DV25
gives 1.5MHz of bandwidth, which is better than any standard NTSC
composite.

John
Anonymous
September 10, 2004 5:12:33 AM

Archived from groups: alt.tv.tech.hdtv (More info?)

>My end result is almost (well, not quite) indistinguishable from a good
>DVD that perhaps has a little too much HF rolloff. Casual viewing
>by this very picky video person makes it unlikely to detect that the
>video had ever been LD. The result looks mostly like a slightly soft
>DVD instead of a very good LD.

Do you have the ability to make DVD copies of your LDs? If so, have you ever?
I'm interested to know how good a LD to DVD-R (or DVD+R, or whatever disc
format you use) really looks when played back on a set-top DVD player.
Steve Grauman
Anonymous
September 10, 2004 5:28:29 AM

Archived from groups: alt.tv.tech.hdtv (More info?)

In article <20040909210939.22988.00000454@mb-m10.aol.com>,
oneactor1@aol.com (Steve Grauman) writes:
>>My end result is almost (well, not quite) indistinguishable from a good
>>DVD that perhaps has a little too much HF rolloff. Casual viewing
>>by this very picky video person makes it unlikely to detect that the
>>video had ever been LD. The result looks mostly like a slightly soft
>>DVD instead of a very good LD.
>
> Do you have the ability to make DVD copies of your LDs? If so, have you ever?
>
Yes.

>
> I'm interested to know how good a LD to DVD-R (or DVD+R, or whatever disc
> format you use) really looks when played back on a set-top DVD player.
>
Given moderately noisy LD source (e.g. Bananarama Venus), with lots of red
chroma noise, and some pattern noise obviously from the source, the result
ends up with rock-solid chroma (well, looks like BetaSP instead of near
perfect DV50, but still damned good.)

So, yes it looks really good on both my JVC and Pioneer DVD set top players.

John
Anonymous
September 10, 2004 7:18:17 AM

Archived from groups: alt.tv.tech.hdtv (More info?)

In article <20040909210939.22988.00000454@mb-m10.aol.com>,
oneactor1@aol.com (Steve Grauman) writes:
>>My end result is almost (well, not quite) indistinguishable from a good
>>DVD that perhaps has a little too much HF rolloff. Casual viewing
>>by this very picky video person makes it unlikely to detect that the
>>video had ever been LD. The result looks mostly like a slightly soft
>>DVD instead of a very good LD.
>
> Do you have the ability to make DVD copies of your LDs?
>
PS:
If there is a way that I could provide a demo that doesn't
take too much space/toolong a download, I'll see if there
is a way.

I'll have to boot NT/2000 to try some experiments.

I haven't looked at the comparisons recently (in the last
few months), but AFAIR, the difference between 'raw' copies of
the LD and 'cleaned-up' results were significant.

Again, using a good 2D comb will still be okay, especially
if you use the noise reduction on TMPGENC or equivalent.

One reason why I can get by using DV25 for my encoding is that
the noise reduction in the Canopus is so good that the bits
aren't wasted too severely. However, providing a noisy signal
to a DV25 (or MPEG2) encoder definitely has negative effects.

My guess is that one reason why DV25 didn't take off for timeshifting
on normal (noisy) NTSC is that the quality loss on DV25 when recording
noisy signals is substantial enough that it might degrade DV25 to a
level that is only slightly better than SVHS (even though DV25 is
still better.)

My absolutely best LD recording setup is like the following:


COMP COMP DV25 DV25 DV25
LD --> DPS290 TBC --> Canopus --> DHR1000 --> P4 PC Disk -->
ADVC300 Archive
(OPT NR) (MIN NR) (OPT TAPE) (COPY TO DISK)

MPG IMG
TMPGENC --> TMPAUTHOR --> Create DVD
(MPEG ENC) (MASTER DVD)

Also, I do the following:

COMP COMP COMP DV25 DV25
LD --> DPS290 TBC --> D9 DV50 --> Canopus --> DHR1000 -->
(OPT NR) Archive ADVC300 Archive
(MIN NR)


DV25 MPG IMG
P4 PC Disk --> TMPGENC --> TMPAUTHOR --> Create DVD
(COPY TO DISK) (MPEG ENC) (MASTER DVD)


1) The DPS290 is only seriously needed if the LD time base
stability is especially poor, needs the brutal NR provided
by the DPS290 or or has drop outs. D9 also HATES unstable
sources, and analog, non-tbc'ed LD isn't

2) The D9 copy is nearly perfect (almost NO difference between
D9 and LD.) Every last noise nit is fairly well reproduced on
D9, and so there is really no noticeable modification of the
signal by D9.

3) Most of my recording setup is only STEREO capable.

4) The Canopus converter can be nearly as brutal as the DPS TBC
noise reduction. It is important to use the lowest levels of
NR and use absolutely minimal (or no amount) of edge enhancement.
Use just enough noise reduction to mitigate the noise and improve
performance of DV25 encoding. The lowest amount of NR is adequate
except for the worst source material. Also, I don't play with
significantly rotted disks.

5) The TMPGENC noise reduction can be used at very high levels without
extreme negative impact. They seem to have done a good job to
avoid comet tailing. Also, use very small amounts of horizontal
and vertical boost. If appropriate, using film modes can also be
helpful (I forget the details.)

6) DV25 copies of the LD material is NOT a faithful reproduction, but
might be significantly improved. This adds a layer of generation
loss and also might be irrecoverably removing useful information
for the future. The D9 copies are quite faithful, and if the format
continues to exist in the future, it will provide incredibly
faithful copies of the original.

GOOD LUCK!!!

John
!