|
|
| Re: VidHD color artifact quality [message #373005 is a reply to message #372974] |
Wed, 29 August 2018 13:32   |
STYNX
Messages: 453
Registered: October 2012
Karma: 0
|
Senior Member |
|
|
On Wednesday, 29 August 2018 03:14:04 UTC+2, John Brooks wrote:
> For those curious about how VidHD handles NTSC artifact colors in HGR and DHGR graphics modes, here is a video capture of several Apple II programs:
>
> Airheart
> Rescue Raiders
> Captain Goognight
> Drol
> Hard Hat Mack
> Mr. Do
> Bard's Tale
> Wizardry
> Dazzle Draw slideshow
> Cavern Creatures
>
> The VidHD board is coming October 2018 for $129. Email vidhd@blueshiftinc..com to join the waiting list.
>
> https://www.youtube.com/watch?v=quX3k9rYZck
> -JB
> @JBrooksBSI
The NTSC Emulation looks really good. It mostly corresponds to my own experience. Here is a grab of the real NTSC output of my Apple IIe.
https://c2.staticflickr.com/2/1878/29411671187_b2ce6284a7_o. png
It was captured at 560x192px resolution with a professional frame grabber with custom scanning frequency to get a near pixel perfect scan. Maybe the contrast and brightness are not perfect but you can compare it with your rendition. I think that your implementation is very good.
-Jonas
|
|
|
|
| Re: VidHD color artifact quality [message #373019 is a reply to message #373005] |
Wed, 29 August 2018 15:49 |
TomCh
Messages: 242
Registered: November 2012
Karma: 0
|
Senior Member |
|
|
Hi John,
I see The Bard's Tale has blended horizontal lines to produce the pink and cyan colours. Is this intrinsic to your video emulation?
What do the other blend combinations look like, eg. Fantavision or Blazing Paddles (or even Karateka)?
Thanks,
Tom
|
|
|
|
| Re: VidHD color artifact quality [message #373020 is a reply to message #373019] |
Wed, 29 August 2018 16:27 |
|
|
Originally posted by: John Brooks
On Wednesday, August 29, 2018 at 12:49:57 PM UTC-7, TomCh wrote:
> Hi John,
>
> I see The Bard's Tale has blended horizontal lines to produce the pink and cyan colours. Is this intrinsic to your video emulation?
>
> What do the other blend combinations look like, eg. Fantavision or Blazing Paddles (or even Karateka)?
>
> Thanks,
> Tom
Is it possible that you are watching the movie at less than 1080p and Youtube is showing a down-sampled image?
When I look at the YouTube video fullscreen, it shows the pink hat as alternating lines of orange and purple.
Similarly, the cyan hat has alternating lines of blue and green.
VidHD is not doing any blending vertically. The only horizontal blending is due to NTSC and Woz overdriving it at 14MHz.
-JB
@JBrooksBSI
|
|
|
|
| Re: VidHD color artifact quality [message #373021 is a reply to message #373005] |
Wed, 29 August 2018 17:11 |
|
|
Originally posted by: John Brooks
On Wednesday, August 29, 2018 at 10:32:08 AM UTC-7, STYNX wrote:
> On Wednesday, 29 August 2018 03:14:04 UTC+2, John Brooks wrote:
>> For those curious about how VidHD handles NTSC artifact colors in HGR and DHGR graphics modes, here is a video capture of several Apple II programs:
>>
>> Airheart
>> Rescue Raiders
>> Captain Goognight
>> Drol
>> Hard Hat Mack
>> Mr. Do
>> Bard's Tale
>> Wizardry
>> Dazzle Draw slideshow
>> Cavern Creatures
>>
>> The VidHD board is coming October 2018 for $129. Email vidhd@blueshiftinc.com to join the waiting list.
>>
>> https://www.youtube.com/watch?v=quX3k9rYZck
>> -JB
>> @JBrooksBSI
>
> The NTSC Emulation looks really good. It mostly corresponds to my own experience. Here is a grab of the real NTSC output of my Apple IIe.
>
> https://c2.staticflickr.com/2/1878/29411671187_b2ce6284a7_o. png
>
> It was captured at 560x192px resolution with a professional frame grabber with custom scanning frequency to get a near pixel perfect scan. Maybe the contrast and brightness are not perfect but you can compare it with your rendition. I think that your implementation is very good.
Thanks Jonas.
There are some unusual artifacts in the IIe capture which are likely coming from the NTSC decoder being used:
1) The orange, green, and blue squares to the left of the compass should be solid blocks, but the top row is darkened, which may be caused by a comb filter. The left edge of each colored block also has a strange ringing or darkening effect inside it.
2) The blue boat hull should be a constant color but has a horizontal pattern.
3) The inner edges above and below the compass 'N' are choppy instead of a smooth constant brightness.
4) The artifact colors around the whites and grays are missing ('N' in compass, and the banner at the wreath base).
You could try connecting your IIe to an old-school CRT TV without a comb filter and see if these issues are from the IIe or the monitor.
Here is another NTSC version of Airheart for comparison:
https://zellyn.github.io/apple2shader/
-JB
@JBrooks
|
|
|
|
|
|
| Re: VidHD color artifact quality [message #373085 is a reply to message #373021] |
Thu, 30 August 2018 10:18 |
STYNX
Messages: 453
Registered: October 2012
Karma: 0
|
Senior Member |
|
|
|
The captured picture was scanned from a frame doubled source. The frame doubler is a professional converter for crt projectors and uses a 12bit signal path per color (rgb) from ntsc. The resulting picture can be viewed on older vga monitors and is interpreted as 640x240 pixel with 120hz (a bit of adjusting is needed to get a centered and fitting image). Since the converter is trying to sync to the color burst and samples at 4 times of this frequency, I can get near pixel-perfect video on a vga display. The grabber was configured to scan the vga-like signal by setting front and back porch to get the resulting 560pixel picture. I could try to scan with 1920x240 pixels at 120hz with a higher pixel-clock but I would have to get the hardware out of storage. I suspect that the frame doubler has some tricks to get a better image on crt projectors which might conflict with the grabber.
|
|
|
|
| Re: VidHD color artifact quality [message #373098 is a reply to message #373036] |
Thu, 30 August 2018 10:26 |
zellyn
Messages: 173
Registered: April 2013
Karma: 0
|
Senior Member |
|
|
On Wednesday, August 29, 2018 at 7:45:26 PM UTC-4, Doug Dingus wrote:
> John,
>
> That shader page is amazing. First thing I've seen that reproduces what my Sony PVM does with an Apple 2 composite signal. Well done!
Huh. Thanks for the reminder that the web page completely lacks any credits or attribution. All glory for the stunning NTSC emulation goes to the author of OpenEmulator, the now sadly disappeared-from-the-internet Marc Ressl.
Source here: https://github.com/zellyn/apple2shader
Zellyn
|
|
|
|
|
|
| Re: VidHD color artifact quality [message #373164 is a reply to message #373020] |
Fri, 31 August 2018 15:05 |
TomCh
Messages: 242
Registered: November 2012
Karma: 0
|
Senior Member |
|
|
On Wednesday, August 29, 2018 at 9:27:26 PM UTC+1, John Brooks wrote:
> Is it possible that you are watching the movie at less than 1080p and Youtube is showing a down-sampled image?
>
Yes, that was it.
> VidHD is not doing any blending vertically. The only horizontal blending is due to NTSC and Woz overdriving it at 14MHz.
>
Do you have any interest in supporting some level of vertical blending?
(probably aka low quality NTSC TV set)
Tom
|
|
|
|
| Re: VidHD color artifact quality [message #373167 is a reply to message #373164] |
Fri, 31 August 2018 18:30 |
|
|
Originally posted by: John Brooks
On Friday, August 31, 2018 at 12:05:14 PM UTC-7, TomCh wrote:
> On Wednesday, August 29, 2018 at 9:27:26 PM UTC+1, John Brooks wrote:
>> Is it possible that you are watching the movie at less than 1080p and Youtube is showing a down-sampled image?
>>
> Yes, that was it.
>
>> VidHD is not doing any blending vertically. The only horizontal blending is due to NTSC and Woz overdriving it at 14MHz.
>>
> Do you have any interest in supporting some level of vertical blending?
> (probably aka low quality NTSC TV set)
Not at the moment. The strict real-time nature of drawing 120 million pixels a second prohibits large-area blurring.
VidHD is generally trying to take advantage of the large 1080p display to show off the Apple II's graphic detail rather than recreating a low quality CRT.
-JB
@JBrooksBSI
|
|
|
|
| Re: VidHD color artifact quality [message #373179 is a reply to message #373164] |
Sat, 01 September 2018 02:24 |
Michael J. Mahon
Messages: 1767
Registered: October 2012
Karma: 0
|
Senior Member |
|
|
TomCh <tomcharlesworth26@gmail.com> wrote:
> On Wednesday, August 29, 2018 at 9:27:26 PM UTC+1, John Brooks wrote:
>> Is it possible that you are watching the movie at less than 1080p and
>> Youtube is showing a down-sampled image?
>>
> Yes, that was it.
>
>> VidHD is not doing any blending vertically. The only horizontal blending
>> is due to NTSC and Woz overdriving it at 14MHz.
>>
> Do you have any interest in supporting some level of vertical blending?
> (probably aka low quality NTSC TV set)
>
> Tom
>
Ha!
But in fairness to older NTSC TV sets, they were expensive, and pretty well
engineered (unlike many inexpensive B&W sets). Any overlap of scan lines
was usually the result of poor adjustment: focus, dynamic focus, regulated
HV, and static and dynamic convergence.
Older color sets were a bear to adjust properly, and if they weren’t
properly degaussed, the settings might shift if you changed the orientation
of the set!
Thank Sony for changing all that. Their Trinitron vertical stripe CRT made
shadow masks obsolete for all practical purposes. With a simple vertical
grid of wires, they greatly simplified the color TV and its adjustments.
BTW, broadcast TV actually had twice as many lines as Apple video, since it
was interlaced—requiring even better line focus.
--
-michael - NadaNet 3.1 and AppleCrate II: http://michaeljmahon.com
|
|
|
|
| Re: VidHD color artifact quality [message #373216 is a reply to message #373179] |
Sun, 02 September 2018 15:30 |
TomCh
Messages: 242
Registered: November 2012
Karma: 0
|
Senior Member |
|
|
On Saturday, September 1, 2018 at 7:24:13 AM UTC+1, Michael J. Mahon wrote:
> TomCh wrote:
>> On Wednesday, August 29, 2018 at 9:27:26 PM UTC+1, John Brooks wrote:
>>> Is it possible that you are watching the movie at less than 1080p and
>>> Youtube is showing a down-sampled image?
>>>
>> Yes, that was it.
>>
>>> VidHD is not doing any blending vertically. The only horizontal blending
>>> is due to NTSC and Woz overdriving it at 14MHz.
>>>
>> Do you have any interest in supporting some level of vertical blending?
>> (probably aka low quality NTSC TV set)
>>
>> Tom
>>
>
> Ha!
>
> But in fairness to older NTSC TV sets, they were expensive, and pretty well
> engineered (unlike many inexpensive B&W sets). Any overlap of scan lines
> was usually the result of poor adjustment: focus, dynamic focus, regulated
> HV, and static and dynamic convergence.
>
> Older color sets were a bear to adjust properly, and if they weren’t
> properly degaussed, the settings might shift if you changed the orientation
> of the set!
>
> Thank Sony for changing all that. Their Trinitron vertical stripe CRT made
> shadow masks obsolete for all practical purposes. With a simple vertical
> grid of wires, they greatly simplified the color TV and its adjustments.
>
> BTW, broadcast TV actually had twice as many lines as Apple video, since it
> was interlaced—requiring even better line focus.
>
> --
> -michael - NadaNet 3.1 and AppleCrate II: http://michaeljmahon.com
I took a few pictures (admittedly not great quality) of titles apparently doing alternate horizontal lines for mixing/blending to make new colours.
The Bard's Tale, Archon, Karateka & Fantavision:
https://github.com/AppleWin/AppleWin/issues/344#issuecomment -417927539
NB. This is from a PAL TV, not NTSC.
It'd be interesting to see the equivalent screenshots from an NTSC TV.
Tom
|
|
|
|
| Re: VidHD color artifact quality [message #373219 is a reply to message #373167] |
Sun, 02 September 2018 16:39 |
TomCh
Messages: 242
Registered: November 2012
Karma: 0
|
Senior Member |
|
|
On Friday, August 31, 2018 at 11:30:59 PM UTC+1, John Brooks wrote:
> On Friday, August 31, 2018 at 12:05:14 PM UTC-7, TomCh wrote:
>> On Wednesday, August 29, 2018 at 9:27:26 PM UTC+1, John Brooks wrote:
>>> Is it possible that you are watching the movie at less than 1080p and Youtube is showing a down-sampled image?
>>>
>> Yes, that was it.
>>
>>> VidHD is not doing any blending vertically. The only horizontal blending is due to NTSC and Woz overdriving it at 14MHz.
>>>
>> Do you have any interest in supporting some level of vertical blending?
>> (probably aka low quality NTSC TV set)
>
> Not at the moment. The strict real-time nature of drawing 120 million pixels a second prohibits large-area blurring.
>
> VidHD is generally trying to take advantage of the large 1080p display to show off the Apple II's graphic detail rather than recreating a low quality CRT.
>
> -JB
> @JBrooksBSI
I watched your KansasFest'18 presentation and understand that you are pretty tight on compute resource.
On a different topic, have you tested with French Touch's "Ansi Story" demo? The final effect (on side-2) where they vertically split the display into DGR / TEXT80 / DGR is a very good test of video timing accuracy. Of all the FT demos, this was the effect that caused AppleWin the most trouble.
Tom
|
|
|
|
| Re: VidHD color artifact quality [message #373226 is a reply to message #373219] |
Sun, 02 September 2018 17:54 |
Michael J. Mahon
Messages: 1767
Registered: October 2012
Karma: 0
|
Senior Member |
|
|
TomCh <tomcharlesworth26@gmail.com> wrote:
> On Friday, August 31, 2018 at 11:30:59 PM UTC+1, John Brooks wrote:
>> On Friday, August 31, 2018 at 12:05:14 PM UTC-7, TomCh wrote:
>>> On Wednesday, August 29, 2018 at 9:27:26 PM UTC+1, John Brooks wrote:
>>>> Is it possible that you are watching the movie at less than 1080p and
>>>> Youtube is showing a down-sampled image?
>>>>
>>> Yes, that was it.
>>>
>>>> VidHD is not doing any blending vertically. The only horizontal
>>>> blending is due to NTSC and Woz overdriving it at 14MHz.
>>>>
>>> Do you have any interest in supporting some level of vertical blending?
>>> (probably aka low quality NTSC TV set)
>>
>> Not at the moment. The strict real-time nature of drawing 120 million
>> pixels a second prohibits large-area blurring.
>>
>> VidHD is generally trying to take advantage of the large 1080p display
>> to show off the Apple II's graphic detail rather than recreating a low quality CRT.
>>
>> -JB
>> @JBrooksBSI
>
> I watched your KansasFest'18 presentation and understand that you are
> pretty tight on compute resource.
>
> On a different topic, have you tested with French Touch's "Ansi Story"
> demo? The final effect (on side-2) where they vertically split the
> display into DGR / TEXT80 / DGR is a very good test of video timing
> accuracy. Of all the FT demos, this was the effect that caused AppleWin the most trouble.
>
> Tom
>
Line-to-line dithering is well established, but the blending should occur
in the eye rather than the display.
--
-michael - NadaNet 3.1 and AppleCrate II: http://michaeljmahon.com
|
|
|
|
|
|
| Re: VidHD color artifact quality [message #373235 is a reply to message #373226] |
Mon, 03 September 2018 04:16 |
TomCh
Messages: 242
Registered: November 2012
Karma: 0
|
Senior Member |
|
|
On Sunday, September 2, 2018 at 10:54:14 PM UTC+1, Michael J. Mahon wrote:
> Line-to-line dithering is well established, but the blending should occur
> in the eye rather than the display.
>
> --
> -michael - NadaNet 3.1 and AppleCrate II: http://michaeljmahon.com
Hi Michael,
Yes: with those 2 close-up pictures I took (Bard's Tale: "pink" & "cyan" hats), then at this proximity to the display you can see the individual scanlines alternating between the 2 colours. But when viewed from a distance (eg.. a handful of inches or more) then as you say, it's blended in the eye.
But the rendering I'm doing (AppleWin) and presumably John is doing (VidHD), is (a) scaled up from 192 lines to 1080 lines (so chunkier); and (b) the resulting image is on a sharper, higher fidelity display.
So perhaps we need to provide a little help to the eye to blend? (ie. a vertical "roll-off" of the colour, so it's not such a hard "edge".)
Or perhaps we need to do better at recreating a simulation of the CRT display (ie. at the sub-pixel level)?
Tom
|
|
|
|
| Re: VidHD color artifact quality [message #373242 is a reply to message #373235] |
Mon, 03 September 2018 12:28 |
Michael J. Mahon
Messages: 1767
Registered: October 2012
Karma: 0
|
Senior Member |
|
|
TomCh <tomcharlesworth26@gmail.com> wrote:
> On Sunday, September 2, 2018 at 10:54:14 PM UTC+1, Michael J. Mahon wrote:
>> Line-to-line dithering is well established, but the blending should occur
>> in the eye rather than the display.
>>
>> --
>> -michael - NadaNet 3.1 and AppleCrate II: http://michaeljmahon.com
>
> Hi Michael,
>
> Yes: with those 2 close-up pictures I took (Bard's Tale: "pink" & "cyan"
> hats), then at this proximity to the display you can see the individual
> scanlines alternating between the 2 colours. But when viewed from a
> distance (eg. a handful of inches or more) then as you say, it's blended in the eye.
>
> But the rendering I'm doing (AppleWin) and presumably John is doing
> (VidHD), is (a) scaled up from 192 lines to 1080 lines (so chunkier); and
> (b) the resulting image is on a sharper, higher fidelity display.
>
> So perhaps we need to provide a little help to the eye to blend? (ie. a
> vertical "roll-off" of the colour, so it's not such a hard "edge".)
> Or perhaps we need to do better at recreating a simulation of the CRT
> display (ie. at the sub-pixel level)?
>
> Tom
>
The eye should blend the lines if the picture subtends approximately the
same angle as with a CRT display, but if the picture is larger to the eye,
then the original lines will be “zoomed in” and the blending will not be as
effective.
This is like any heavily dithered picture, where backing up greatly
improves the appearance. It’s also why SD television looks so bad on a
large display.
I certainly understand the motivation, but I doubt that you’d like the
blurring in areas where dithering is not the intent.
--
-michael - NadaNet 3.1 and AppleCrate II: http://michaeljmahon.com
|
|
|
|
|
|
|
|
|
|
Members
Search
Help
Register
Login
Home
