Improving Madvr HDR to SDR mapping for projector - Page 285 - AVS Forum | Home Theater Discussions And Reviews
Forum Jump: 
 8861Likes
Reply
 
Thread Tools
post #8521 of 9394 Old 01-17-2020, 02:00 PM
Member
 
Join Date: Aug 2015
Posts: 166
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 95 Post(s)
Liked: 68
yes if you use MADVR and therefore the AMD api there is bug where only 709 is outputted instead of 2020 so colours are massively unsaturated, I logged a bug report on MADVR but its not really a bug with MADVR I dont think but i'm hoping MADSHI can put something into MADVR, a switch maybe to force BT2020 or bring back the windows API.

You wont be able to use any tone mapping with MADVR with this issue without a workaround and none of them can be built directly into MADVR, you either have to run a script or an EXE, at the moment i'm tone mapping to SDR instead as my HDR TV only does about 430 nits anyway so i'm not missing much.

If you are unsure at all which card to buy right now for MADVR tone mapping, its not an AMD5700 series IMHO as this has been broken for months.

more info here, best place to discuss this really.

https://forum.doom9.org/showthread.php?t=176013&page=40

bug logged here:

http://bugs.madshi.net/view.php?id=630
Mclingo McKeown is online now  
Sponsored Links
Advertisement
 
post #8522 of 9394 Old 01-17-2020, 03:41 PM
Senior Member
 
Join Date: Jan 2018
Posts: 410
Mentioned: 109 Post(s)
Tagged: 0 Thread(s)
Quoted: 226 Post(s)
Liked: 382
Quote:
Originally Posted by madshi View Post
I'll upload a new test build soon (not today, but hopefully in the next 1-2 days). Are there any fancy new HSTM curves I should include? For now, I'm planning to include those listed in this post:

https://www.avsforum.com/forum/24-di...l#post58893810

I've been doing some testing over the last month and come up with some new suggestions, which you'll hopefully be able to include

My taste has evolved a little in that I’m aiming to be more conservative than I once was. I’m trying to preserve director’s intent (snuggle fairly close to HSTM off) without deviating too much.

So these are my new suggestions:

From top to bottom: Low strength, medium strength, high strength






edit: my strong option (the third row) is very similar to Neo's very high....Neo's very high has a little more contrast on the lower end (i.e. see frame 843, 987 S&M benchmark disc) than my strong curve. My strong curve has a little more shadow detail and highlights in faces (i.e. see Lucy 9119, Casino Royale 96990, Peregrine 52812 etc). YMMV.



and this is very strong (also 100 strength)





My favourite is low.



The very strong is too strong for my preference (maybe those who favour a strong curve would be more approving of it?)


As I'm evolving with my taste and preferring to be more conservative, I did look at Neo’s two options again (the “conservative” option and the “equilibrium” option ), and whilst I think that the conservative option can look good in some scenes and are a fraction more conservative than mine in some respects, I don’t like what it does in other scenes: it can take small pockets of contrast in the image and actually flatten them, so that the net result is that the HSTM setting is actually flatter than HSTM off in certain scenes. (This also applies with the equilibrium curve)


i.e.

Joker:

frame 40593,

off vs NC vs 750 1:2




Look at the buildings, window area, the pockets of contrast in his doctor’s coat, the outline of the contrast around the bed lights on the wall. The top of his mask on his head. Something seems a bit off to me... it loses contrast. It seems flatter than off.



frame 96043





The contrast of the wall/lights in the background. Flattened compared to off.




frame 93055





The contrast of the wall/lights in the background. Flattened compared to off.



(Joker actually had a few other scenes like this although I didn't jot down all the frame numbers as it wasn't officially part of my testing)


The Meg:

frame 96990





The lost contrast that define the contours of the face




Kingsman: The Secret Service:

frame 109322





The lost contrast in the dark side of the face



Miss Peregrine's Home for Peculiar Children:

frame 45275






The slight flattening of the face





LG Chess demo:

frame 804 (this isn't too bad, but there is a bit of flattening in the contours of the face). Didn't include screenshots of this.


So it would be great if you could include these options. Thanks.

Last edited by Fer15; 01-17-2020 at 04:25 PM.
Fer15 is offline  
post #8523 of 9394 Old 01-17-2020, 04:53 PM
AVS Forum Special Member
 
Neo-XP's Avatar
 
Join Date: Jun 2018
Location: Switzerland
Posts: 1,208
Mentioned: 220 Post(s)
Tagged: 0 Thread(s)
Quoted: 928 Post(s)
Liked: 1323
@Fer15 Hi!

If you want to compare to the best one for me, which is not a new one, here it is:

Quote:
Originally Posted by Neo-XP View Post
Neo-XP 1.5, Pure Log, Strength 100

- medium: 750-951-1038-1089-1153-1195-1240-1290-1326-1354-1377-1397-1413-1441-1464-1500
It seems pretty close to your new favorite.
SamuriHL, madshi, Manni01 and 3 others like this.
Neo-XP is offline  
Sponsored Links
Advertisement
 
post #8524 of 9394 Old 01-17-2020, 05:54 PM
AVS Forum Special Member
 
stevenjw's Avatar
 
Join Date: Jul 2001
Location: Palm Coast, Florida, USA
Posts: 3,004
Mentioned: 10 Post(s)
Tagged: 0 Thread(s)
Quoted: 679 Post(s)
Liked: 902
Quote:
Originally Posted by tommarra View Post
I agree JRiver is very powerful but feels iTunish. Plus for some reason I couldn’t get out to use the test binaries that madshi posts on this thread
To update JRiver MC to work, you first have to chose madvr as a render option under video (LAV is handled there as well). Once this is done, run a video which should show *Adding Component at movie start time. This creates the madvr plugin folder under %APPDATA% roaming directory. The eight files from the test zip go there, overwriting the original ones. Best to save off the settings.bin first, then run the reset executable and copy back the settings.bin file. This is from memory, but hope it helps.
tommarra likes this.

...Steve
"Opinions are like orgasms… mine matters most and I really don’t care if you have one or not." ;)
 
My HT gear
stevenjw is offline  
post #8525 of 9394 Old 01-18-2020, 12:33 AM
AVS Forum Special Member
 
madshi's Avatar
 
Join Date: May 2005
Posts: 7,939
Mentioned: 658 Post(s)
Tagged: 0 Thread(s)
Quoted: 2883 Post(s)
Liked: 4358
Quote:
Originally Posted by Fer15 View Post
I've been doing some testing over the last month and come up with some new suggestions, which you'll hopefully be able to include

My taste has evolved a little in that I’m aiming to be more conservative than I once was. I’m trying to preserve director’s intent (snuggle fairly close to HSTM off) without deviating too much.

So these are my new suggestions:

From top to bottom: Low strength, medium strength, high strength

edit: my strong option (the third row) is very similar to Neo's very high....Neo's very high has a little more contrast on the lower end (i.e. see frame 843, 987 S&M benchmark disc) than my strong curve. My strong curve has a little more shadow detail and highlights in faces (i.e. see Lucy 9119, Casino Royale 96990, Peregrine 52812 etc). YMMV.

and this is very strong (also 100 strength)
Thank you! I'll include your new curves.

Quote:
Originally Posted by Neo-XP View Post
If you want to compare to the best one for me, which is not a new one, here it is:

It seems pretty close to your new favorite.
I'm pretty happy to hear that you and Fer15 have new favorites which seem very close to each other. That's always a good sign!

Your preferences have changed back and forth a bit during the last couple of weeks. Could you please give me an updated list of curves that you would like to see in the next build? My preference is not to include too many, but I will of course include all those curves which users liked a lot, and also those curves Fer15 and you like most now (because you two guys are my most active and trusted testers).

If you have a chance, could you give Fer15's and your favorite curves a deeper look? If you like your current favorites better I'll include both your and Fer15's new favorite curves. But maybe you'll like his better (or the same) as yours, if they're pretty close now. That would reduce the number of curves in the next build. But I'll let you decide. Thanks!
madshi is offline  
post #8526 of 9394 Old 01-18-2020, 03:29 AM
AVS Forum Special Member
 
Neo-XP's Avatar
 
Join Date: Jun 2018
Location: Switzerland
Posts: 1,208
Mentioned: 220 Post(s)
Tagged: 0 Thread(s)
Quoted: 928 Post(s)
Liked: 1323
Quote:
Originally Posted by madshi View Post
If you have a chance, could you give Fer15's and your favorite curves a deeper look? If you like your current favorites better I'll include both your and Fer15's new favorite curves. But maybe you'll like his better (or the same) as yours, if they're pretty close now. That would reduce the number of curves in the next build. But I'll let you decide. Thanks!
Comparing quickly with the most famous test scenes, Fer15's new favorite and mine are very very close, and Fer15's is slightly stronger.
To reduce the number of curves, I'll have no problem to only go with Fer15's for the low/conservative choice.

Quote:
Originally Posted by madshi View Post
Could you please give me an updated list of curves that you would like to see in the next build?
Only this one in addition to Fer15's:

(strength 100) 500-721-879-998-1164-1275-1385-1496-1562-1607-1638-1662-1681-1707-1726-1750

I find better than the "5) Neo-XP's 1.0 High" one for a higher strength, but I didn't have the time to compare it to Fer15's new stronger curves yet.

Thanks!
Neo-XP is offline  
post #8527 of 9394 Old 01-18-2020, 03:38 AM
AVS Forum Special Member
 
madshi's Avatar
 
Join Date: May 2005
Posts: 7,939
Mentioned: 658 Post(s)
Tagged: 0 Thread(s)
Quoted: 2883 Post(s)
Liked: 4358
Quote:
Originally Posted by Neo-XP View Post
Comparing quickly with the most famous test scenes, Fer15's new favorite and mine are very very close, and Fer15's is slightly stronger.
To reduce the number of curves, I'll have no problem to only go with Fer15's for the low/conservative choice.

Only this one in addition to Fer15's:

(strength 100) 500-721-879-998-1164-1275-1385-1496-1562-1607-1638-1662-1681-1707-1726-1750

I find better than the "5) Neo-XP's 1.0 High" one for a higher strength, but I didn't have the time to compare it to Fer15's new stronger curves yet.
Great - thanks!

FWIW, I plan to keep all the various curves in the options for several weeks, to give all users plenty of time to compare. It will only be one drop-down-box, after all, and it doesn't hurt too much to have several options in it. However, in the long run I'd like to thin it out to a reasonably small number of curves with nicely descriptive names.
madshi is offline  
post #8528 of 9394 Old 01-18-2020, 04:21 AM
Senior Member
 
Join Date: Jan 2018
Posts: 410
Mentioned: 109 Post(s)
Tagged: 0 Thread(s)
Quoted: 226 Post(s)
Liked: 382
Quote:
Originally Posted by Neo-XP View Post
@Fer15 Hi!

If you want to compare to the best one for me, which is not a new one, here it is:



It seems pretty close to your new favorite.

Interesting!

It seems that it's one that got lost amongst the mix for me: I probably didn't test it or overlooked it at the time tbh. I know I haven't tested all of the suggestions in this thread as there have been so many, and there have been countless revisions (I would also have spent time doing my own tests), so I've lost track. I best recall your very high and most recent 1000-2000 curves.




Quote:
Originally Posted by madshi View Post
Thank you! I'll include your new curves.


I'm pretty happy to hear that you and Fer15 have new favorites which seem very close to each other. That's always a good sign!
Thanks!
Manni01, strike shadow and Neo-XP like this.
Fer15 is offline  
post #8529 of 9394 Old 01-18-2020, 06:11 AM
Senior Member
 
Icaro's Avatar
 
Join Date: Jul 2007
Location: Italy
Posts: 484
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 184 Post(s)
Liked: 94
@madshi

For the new test build, I would still like to keep this:
Neo-XP's 1.0 High: strength 100
500-903-1076-1178-1306-1390-1479-1581-1653-1709-1754-1793-1826-1882-1928-2000
strike shadow and Yves Claeys like this.

Sorry for my bad English
Epson LS10500 - AMD RX580 - LightSpace CMS
Icaro is online now  
post #8530 of 9394 Old 01-18-2020, 07:29 AM
Member
 
stefanelli73's Avatar
 
Join Date: Jan 2008
Posts: 147
Mentioned: 7 Post(s)
Tagged: 0 Thread(s)
Quoted: 57 Post(s)
Liked: 25
Quote:
Originally Posted by Icaro View Post
@madshi

For the new test build, I would still like to keep this:
Neo-XP's 1.0 High: strength 100
500-903-1076-1178-1306-1390-1479-1581-1653-1709-1754-1793-1826-1882-1928-2000

I agree it's perfect except for some flickering on some movie scene
stefanelli73 is online now  
post #8531 of 9394 Old 01-18-2020, 11:14 AM
AVS Forum Special Member
 
stevenjw's Avatar
 
Join Date: Jul 2001
Location: Palm Coast, Florida, USA
Posts: 3,004
Mentioned: 10 Post(s)
Tagged: 0 Thread(s)
Quoted: 679 Post(s)
Liked: 902
I'm happy as long as there are an equal number of curves that help projectors, especially those with low nits, vs. curves that are tested on and look great on flat panels.

...Steve
"Opinions are like orgasms… mine matters most and I really don’t care if you have one or not." ;)
 
My HT gear
stevenjw is offline  
post #8532 of 9394 Old 01-18-2020, 12:00 PM
AVS Forum Special Member
 
Manni01's Avatar
 
Join Date: Sep 2008
Posts: 9,745
Mentioned: 384 Post(s)
Tagged: 0 Thread(s)
Quoted: 5924 Post(s)
Liked: 6503
No need to panic, madshi said that he would put in the upcoming test build all the popular curves (i.e. those we have already selected in the previous testing phase, they will simply be hardcoded).

The latest conservative Neo-XP and Fer15 curves mentioned recently will be in addition to those we have already selected (listed in the post that madshi referenced).

At least that's my understanding.

Then we test all the curves in the upcoming build, and we see which ones should stay and which ones can go.
stevenjw, neo_2009 and Neo-XP like this.

Batch Utility V4.02 May 16 2019 to automate measurements files for madVR with support for BD Folders
JVC Macro feature on Vertex/Vertex2/Integral2/Maestro/Diva
Manni01 is offline  
post #8533 of 9394 Old 01-19-2020, 03:56 AM
AVS Forum Special Member
 
Neo-XP's Avatar
 
Join Date: Jun 2018
Location: Switzerland
Posts: 1,208
Mentioned: 220 Post(s)
Tagged: 0 Thread(s)
Quoted: 928 Post(s)
Liked: 1323
Quote:
Originally Posted by Neo-XP View Post
Only this one in addition to Fer15's:

(strength 100) 500-721-879-998-1164-1275-1385-1496-1562-1607-1638-1662-1681-1707-1726-1750

I find better than the "5) Neo-XP's 1.0 High" one for a higher strength, but I didn't have the time to compare it to Fer15's new stronger curves yet.
To be replaced by this one after learning from Fer15's medium strength curve:

(strength 100) 500-633-728-799-898-965-1031-1098-1137-1164-1183-1197-1208-1224-1236-1250
Manni01 likes this.
Neo-XP is offline  
post #8534 of 9394 Old 01-19-2020, 04:15 AM
AVS Forum Special Member
 
madshi's Avatar
 
Join Date: May 2005
Posts: 7,939
Mentioned: 658 Post(s)
Tagged: 0 Thread(s)
Quoted: 2883 Post(s)
Liked: 4358
Quote:
Originally Posted by Neo-XP View Post
To be replaced by this one after learning from Fer15's medium strength curve:

(strength 100) 500-633-728-799-898-965-1031-1098-1137-1164-1183-1197-1208-1224-1236-1250
Ok, noted, thanks. Where does this new curve fit into Fer15's curves, or how does it compare to them? On a quick check it seems your previous "High" curve (the one this new curve replaces) seemed near to Fer15's new "High" curve (I mean the 2nd strongest of his 4 new curves) curves. Not always the same, though.

Edit: I wonder if your new "High" doesn't fit better into Fer15's curve system than his own "High", because looking at the first number (which seems rather important), his curves go 750 -> 500 -> 600 -> 300, which seems a bit weird, and in my own quick tests sometimes I found his medium curve to be stronger than his high curve. So your new high curve fits better because the first number would then be 750 -> 500 -> 500 -> 300, which seems more logical to me.
Manni01 likes this.

Last edited by madshi; 01-19-2020 at 04:19 AM.
madshi is offline  
post #8535 of 9394 Old 01-19-2020, 04:19 AM
Member
 
Join Date: Feb 2019
Posts: 18
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 13 Post(s)
Liked: 1
60fps playback with MadVR and HDR to SDR mapping

Hello all,

I have trouble playing all 60fps movies when using HDR to SDR mapping. I have recently upgraded to an RTX 2070 Super thinking this problem would go away, but I still get a lot of dropped frames with just those file types. All 24fps files play fine, even with all "trade quality for performance" items unchecked.

When playing the 60fps files (i.e. the new Gemni Man movie), the CPU load goes up to 75% and GPU Load about 80%, so there is still some breathing room to go.

If I select all "trade quality for performance" items, then the 60 fps movie plays without dropped frames.

I have also played with the video decoding options, and DXVA2 seems to allow for the best performance on both MPC-HC and Kodi-DS players.

My HTPC is an i3, so I decided to do further testing and moved the RTX 2070 to my more powerful gaming i7 PC, but the results were the same, so I started suspecting there could be something with MADVR and 60fps files ?

Please, appreciate your thougths.

Thanks in advance !
Cbermejo is online now  
post #8536 of 9394 Old 01-19-2020, 04:40 AM
AVS Forum Special Member
 
Neo-XP's Avatar
 
Join Date: Jun 2018
Location: Switzerland
Posts: 1,208
Mentioned: 220 Post(s)
Tagged: 0 Thread(s)
Quoted: 928 Post(s)
Liked: 1323
Quote:
Originally Posted by madshi View Post
Ok, noted, thanks. Where does this new curve fit into Fer15's curves, or how does it compare to them? On a quick check it seems your previous "High" curve (the one this new curve replaces) seemed near to Fer15's new "High" curve (I mean the 2nd strongest of his 4 new curves) curves. Not always the same, though.

Edit: I wonder if your new "High" doesn't fit better into Fer15's curve system than his own "High", because looking at the first number (which seems rather important), his curves go 750 -> 500 -> 600 -> 300, which seems a bit weird, and in my own quick tests sometimes I found his medium curve to be stronger than his high curve. So your new high curve fits better because the first number would then be 750 -> 500 -> 500 -> 300, which seems more logical to me.
Fer15's curves go like this:

- Low: Hill, min: 750, range: 750, mid:0.25, S: 1, strength: 100
- Medium: Hill, min: 500, range: 500, mid:0.25, S: 1, strength: 100
- High: Hill, min: 600, range: 1200, mid:0.5, S: 1, strength: 100


I didn't test the "High" curve yet, but the "Medium" one was not strong enough compared to the "Low" (often the "Low" was stronger).

So I just adapted it like this to make it "always" stronger:

- Medium: Hill, min: 500, range: 750, mid:0.25, S: 1, strength: 100

And it fits well for me.

My old one was like this:

- Medium: Hill, min: 500, range: 1250, mid:0.25, S: 1, strength: 100

But what I leaned is that higher values on the high end do not necessarily mean stronger effect. The low end seems the most important, then also the range and the curve shape.

Edit: Logically, the higher strengths should be these for me (not tested yet):

- Low: Hill, min: 750, range: 750, mid:0.25, S: 1, strength: 100
- Medium: Hill, min: 500, range: 750, mid:0.25, S: 1, strength: 100
- High: Hill, min: 250, range: 750, mid:0.25, S: 1, strength: 100

- Very High: Hill, min: 0, range: 750, mid:0.25, S: 1, strength: 100
Manni01 and *Mori* like this.

Last edited by Neo-XP; 01-19-2020 at 05:06 AM.
Neo-XP is offline  
post #8537 of 9394 Old 01-19-2020, 05:14 AM
AVS Forum Special Member
 
madshi's Avatar
 
Join Date: May 2005
Posts: 7,939
Mentioned: 658 Post(s)
Tagged: 0 Thread(s)
Quoted: 2883 Post(s)
Liked: 4358
Logically, that makes sense to me. Will be interesting to see what Fer15 thinks about it.

Of course another option would be to lower the low end and keep the high end the same. So basically the range increases.
madshi is offline  
post #8538 of 9394 Old 01-19-2020, 05:22 AM
Senior Member
 
Join Date: Jan 2018
Posts: 410
Mentioned: 109 Post(s)
Tagged: 0 Thread(s)
Quoted: 226 Post(s)
Liked: 382
Quote:
Originally Posted by Neo-XP View Post
Fer15's curves go like this:


I didn't test the "High" curve yet, but the "Medium" one was not strong enough compared to the "Low" (often the "Low" was stronger).



Interesting, I can't recall observing that myself (I'll look again later to double check)

From my testing over the last month, my testing pattern was that I liked to work with whole ratios between the start/end point i.e. 1:2, 1:3, 1:4


With 50 strength 1:8 worked the best and I wouldn't aim to go above 4000 for the endpoint so I had quite a bit of room to work with.

With 100 strength, I couldn't really go over 2000 (or thereabouts) for the endpoint so I had less room to work with hence 1:2, or 1:3 or 1:4


I went back to Hill as I prefer the extra flexibility. I only went to log for my f range as I noticed less posterisation with log, but then Madshi told us to ignore that for now.


With Log, it's easier to work with as you can only decrease the start point. The lower you go, the stronger the contrast of the curve both in the highlights and on the low end (the blacks). If you go too low, you'll get very dark contrast at the expense of shadow detail.


With Hill you have more flexibility, but it's a PITA to work with as there are more variables.


I tested with 0.25, 0.5, and 0.75 for nearly every range.


Generally speaking, at a fixed start/end point, 0.25 will favour darker contrast (including contours) on the low end with slightly more reserved highlights. Increasing it to 0.5 will increase shadow detail and increase the highlights, increasing it to 0.75 will further increase shadow detail and highlights. Although the jump on paper is the same, I generally noticed a slightly bigger impact going from 0.25 to 0.5, than when going from 0.5 to 0.75.


Hill also means that you can have a low start point without crushing shadow detail as you can simply increase the Hill curve.

From my previous selection, low is my favourite followed by medium (both have 0.25 Hill). There also appeared to be a consistency in strength when going from one to the next

I wanted to complete a set with a high recommendation and thought that suggestion with a larger ratio (1:3) and slightly higher Hill (0.5) seemed to be best from the bunch that I looked at for being a reasonably balanced strong curve. I did notice that my medium could sometimes be stronger than my strong curve, which might have had to do with a lower start point, but also the different Hill curves (0.25 (medium) vs 0.5 (strong)

My very strong had the biggest ratio (1:4) and lowest start point. It also had a 0.5 Hill curve, but it's too strong for my liking but decided to throw it out there as others might have preferred it
Manni01 and *Mori* like this.
Fer15 is offline  
post #8539 of 9394 Old 01-19-2020, 05:29 AM
AVS Forum Special Member
 
madshi's Avatar
 
Join Date: May 2005
Posts: 7,939
Mentioned: 658 Post(s)
Tagged: 0 Thread(s)
Quoted: 2883 Post(s)
Liked: 4358
@Fer15 , so you're saying that increasing the Hill curve value from 0.25 to a higher value helps countering the negative effects of using a lower starting value? In that case, starting with Neo's suggestion, maybe something like this would make sense?

Low: Hill, min: 750, range: 750, mid:0.25, S: 1, strength: 100
Medium: Hill, min: 600, range: 750, mid:0.30, S: 1, strength: 100
High: Hill, min: 450, range: 750, mid:0.35, S: 1, strength: 100
Very High: Hill, min: 300, range: 750, mid:0.40, S: 1, strength: 100

Or alternatively:

Low: Hill, min: 750, max: 1500, mid:0.25, S: 1, strength: 100
Medium: Hill, min: 600, max: 1500, mid:0.30, S: 1, strength: 100
High: Hill, min: 450, max: 1500, mid:0.35, S: 1, strength: 100
Very High: Hill, min: 300, max: 1500, mid:0.40, S: 1, strength: 100

In any case, it seems Neo-XP and Fer15 agree on the "Low" curve. So we only need a good way to scale up to higher strength. I'd prefer a logical way (e.g. like I suggested above), to improve the chances that a higher strength curve actually produces a higher visual strength in most cases.

Edit: Your "ratio" approach stops working as the "min" value approaches 0. Because obviously with a min value of 0, the ratio will always be infinite. So I think it might be better to scale up by either keeping the range or the max identical, and just lower the min value?
Manni01 likes this.
madshi is offline  
post #8540 of 9394 Old 01-19-2020, 06:35 AM
Senior Member
 
Join Date: Jan 2018
Posts: 410
Mentioned: 109 Post(s)
Tagged: 0 Thread(s)
Quoted: 226 Post(s)
Liked: 382
Quote:
Originally Posted by madshi View Post
@Fer15 , so you're saying that increasing the Hill curve value from 0.25 to a higher value helps countering the negative effects of using a lower starting value? In that case, starting with Neo's suggestion, maybe something like this would make sense?

Low: Hill, min: 750, range: 750, mid:0.25, S: 1, strength: 100
Medium: Hill, min: 600, range: 750, mid:0.30, S: 1, strength: 100
High: Hill, min: 450, range: 750, mid:0.35, S: 1, strength: 100
Very High: Hill, min: 300, range: 750, mid:0.40, S: 1, strength: 100

Or alternatively:

Low: Hill, min: 750, max: 1500, mid:0.25, S: 1, strength: 100
Medium: Hill, min: 600, max: 1500, mid:0.30, S: 1, strength: 100
High: Hill, min: 450, max: 1500, mid:0.35, S: 1, strength: 100
Very High: Hill, min: 300, max: 1500, mid:0.40, S: 1, strength: 100

In any case, it seems Neo-XP and Fer15 agree on the "Low" curve. So we only need a good way to scale up to higher strength. I'd prefer a logical way (e.g. like I suggested above), to improve the chances that a higher strength curve actually produces a higher visual strength in most cases.

Edit: Your "ratio" approach stops working as the "min" value approaches 0. Because obviously with a min value of 0, the ratio will always be infinite. So I think it might be better to scale up by either keeping the range or the max identical, and just lower the min value?
Yeah, increasing the Hill curve appears to increase shadow detail at a fixed point. Lowering the start point appears to increase contrast on the low end at a fixed point, so you can tinker with the Hill curve to help balance things out a bit.



On paper, those suggestions may work on a more consistent logical basis (but maybe the Hill increments would need to be a bit bigger), but ultimately I would need to see them in action. Which I'm happy to look at (along with Neo's) but it would be preferable to look at in a drop-down menu as it's much more slower process manually having to key in the HSTM inputs, open the files, enter then frames numbers, go through the scenes, then rinse and repeat. It would be easier and quicker going through the images in one sitting.

With regards the ratio system, I never went below 500 for a 1:2 ratio and didn't see the need to go below 300 for my very strong curve (where I decided to increase the ratio 1:4). It's true that it would break once you go very low.

I'll recheck my medium (500 1:2) with 0.25 Hill vs 0.5 Hill later just to be sure as I know I liked 0.5 Hill as well.
Manni01 and Neo-XP like this.

Last edited by Fer15; 01-19-2020 at 07:07 AM.
Fer15 is offline  
post #8541 of 9394 Old 01-19-2020, 09:54 AM
Senior Member
 
Icaro's Avatar
 
Join Date: Jul 2007
Location: Italy
Posts: 484
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 184 Post(s)
Liked: 94
Quote:
Originally Posted by Neo-XP View Post
To be replaced by this one after learning from Fer15's medium strength curve:

(strength 100) 500-633-728-799-898-965-1031-1098-1137-1164-1183-1197-1208-1224-1236-1250
After trying your last two favorites,
I still prefer "5) Neo-XP's 1.0 High"
Better contrast, and generally more precise, especially on highlights

Edit:
This yours last favorite is better with some movies
+1

Sorry for my bad English
Epson LS10500 - AMD RX580 - LightSpace CMS

Last edited by Icaro; 01-19-2020 at 12:44 PM.
Icaro is online now  
post #8542 of 9394 Old 01-19-2020, 01:31 PM
AVS Forum Special Member
 
Neo-XP's Avatar
 
Join Date: Jun 2018
Location: Switzerland
Posts: 1,208
Mentioned: 220 Post(s)
Tagged: 0 Thread(s)
Quoted: 928 Post(s)
Liked: 1323
Quote:
Originally Posted by Icaro View Post
After trying your last two favorites,
I still prefer "5) Neo-XP's 1.0 High"
Better contrast, and generally more precise, especially on highlights

Edit:
This yours last favorite is better with some movies
+1
If you take for instance frame 45275 of Miss Peregrine's Home for Peculiar Children, "5) Neo-XP's 1.0 High" is flatter than HSTM Off.

One way to avoid this is to lower the start point, without modifying the end point or the curve type, for instance:

Quote:
Originally Posted by Neo-XP View Post
Neo-XP 1.5, Pure Log, Strength 100

- very low: 1000-1134-1192-1226-1269-1297-1326-1360-1384-1403-1418-1431-1442-1462-1476-1500
- low: 875-1043-1115-1157-1211-1246-1283-1325-1355-1379-1398-1414-1428-1451-1470-1500
- medium: 750-951-1038-1089-1153-1195-1240-1290-1326-1354-1377-1397-1413-1441-1464-1500
- high: 625-860-961-1020-1095-1144-1196-1256-1298-1330-1357-1379-1399-1431-1458-1500
- very high: 500-769-884-952-1037-1093-1153-1221-1269-1306-1336-1362-1384-1422-1452-1500
- extreme: 250-586-730-815-922-992-1066-1151-1211-1257-1295-1328-1355-1402-1440-1500
- insane: 0-403-576-678-806-890-979-1081-1153-1209-1254-1293-1326-1382-1428-1500
And as you noticed, this one is less detailed/precise on highlights than "5) Neo-XP's 1.0 High":

(strength 100) 500-633-728-799-898-965-1031-1098-1137-1164-1183-1197-1208-1224-1236-1250

So I discard it.
Neo-XP is offline  
post #8543 of 9394 Old 01-19-2020, 01:38 PM
Senior Member
 
Icaro's Avatar
 
Join Date: Jul 2007
Location: Italy
Posts: 484
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 184 Post(s)
Liked: 94
Sorry Neo, I don't understand,
in the end what is your favorite currently?
Thanks

Now I'm watching "Dawn of the Planet of the Apes" (2014), with this yours last curve:


(strength 100) 500-633-728-799-898-965-1031-1098-1137-1164-1183-1197-1208-1224-1236-1250

with very, very high detail.
Same thing before with "Jumanji"
So at the moment this has become my favorite

Sorry for my bad English
Epson LS10500 - AMD RX580 - LightSpace CMS

Last edited by Icaro; 01-19-2020 at 02:13 PM.
Icaro is online now  
post #8544 of 9394 Old 01-19-2020, 02:34 PM
AVS Forum Special Member
 
Neo-XP's Avatar
 
Join Date: Jun 2018
Location: Switzerland
Posts: 1,208
Mentioned: 220 Post(s)
Tagged: 0 Thread(s)
Quoted: 928 Post(s)
Liked: 1323
Quote:
Originally Posted by Icaro View Post
in the end what is your favorite currently?
The ones in bold below:

Neo-XP, Pure Log, Strength 100

- very low: 1000-1134-1192-1226-1269-1297-1326-1360-1384-1403-1418-1431-1442-1462-1476-1500
- low: 875-1043-1115-1157-1211-1246-1283-1325-1355-1379-1398-1414-1428-1451-1470-1500
- medium: 750-951-1038-1089-1153-1195-1240-1290-1326-1354-1377-1397-1413-1441-1464-1500
- high: 625-860-961-1020-1095-1144-1196-1256-1298-1330-1357-1379-1399-1431-1458-1500
- very high: 500-769-884-952-1037-1093-1153-1221-1269-1306-1336-1362-1384-1422-1452-1500
- extreme: 250-586-730-815-922-992-1066-1151-1211-1257-1295-1328-1355-1402-1440-1500
- insane: 0-403-576-678-806-890-979-1081-1153-1209-1254-1293-1326-1382-1428-1500

Fer15, Hill 0.25, Strength 100

- very low: 1000-1089-1152-1199-1266-1310-1354-1398-1425-1443-1455-1465-1472-1483-1490-1500
- low: 875-986-1065-1124-1207-1262-1318-1373-1406-1428-1444-1456-1465-1479-1488-1500
- medium: 750-883-978-1049-1148-1215-1281-1348-1387-1414-1433-1447-1458-1474-1486-1500
- high: 625-780-891-974-1090-1167-1245-1322-1369-1400-1422-1438-1451-1470-1483-1500
- very high: 500-677-804-898-1031-1120-1208-1297-1350-1385-1411-1430-1444-1466-1481-1500
- extreme: 250-471-629-748-914-1025-1135-1246-1312-1357-1388-1412-1431-1457-1476-1500
- insane: 0-266-455-598-797-930-1062-1195-1275-1328-1366-1395-1417-1449-1471-1500

But you can try the higher strengths if you want more
SamuriHL, Manni01, *Mori* and 1 others like this.

Last edited by Neo-XP; 01-19-2020 at 02:39 PM.
Neo-XP is offline  
post #8545 of 9394 Old 01-19-2020, 02:40 PM
AVS Forum Special Member
 
tommarra's Avatar
 
Join Date: Aug 2013
Location: USA
Posts: 1,019
Mentioned: 11 Post(s)
Tagged: 0 Thread(s)
Quoted: 1016 Post(s)
Liked: 597
Quote:
Originally Posted by Icaro View Post
Sorry Neo, I don't understand,

in the end what is your favorite currently?

Thanks



Now I'm watching "Dawn of the Planet of the Apes" (2014), with this yours last curve:





(strength 100) 500-633-728-799-898-965-1031-1098-1137-1164-1183-1197-1208-1224-1236-1250



with very, very high detail.

Same thing before with "Jumanji"

So at the moment this has become my favorite


What is your display device. My understanding is that Neo is on a TV, Manni is on a projector, and not sure about Fer15.

I am looking to try HSTM out with these settings if you are on a projector as well





Sent from my iPhone using Tapatalk
noob00224 likes this.

Gear: Fronts: B&W 804D3 | Center: B&W HTM2D3 | Surrounds: B&W 684S2 | Backs: B&W 684S2 | Heights: B&W DM601S3 | Subs: 3x PSA S1500 Receiver & Amps: Arcam AV40 + Monoprice Monolith 7x200 + Outlaw 7000x | Projector: JVC NX7 | Screen: 126 inch 16:9 Stewart Studiotek 130 G4 | Source: HTPC with Nvidia 2070, Kodi + External Player: MPC-HC with MadVR (for Blu-ray rips), Nvidia Shield (for streaming content)
tommarra is online now  
post #8546 of 9394 Old 01-20-2020, 01:02 AM
AVS Forum Special Member
 
madshi's Avatar
 
Join Date: May 2005
Posts: 7,939
Mentioned: 658 Post(s)
Tagged: 0 Thread(s)
Quoted: 2883 Post(s)
Liked: 4358
Quote:
Originally Posted by Neo-XP View Post
The ones in bold below:

Neo-XP, Pure Log, Strength 100

- very low: 1000-1134-1192-1226-1269-1297-1326-1360-1384-1403-1418-1431-1442-1462-1476-1500
- low: 875-1043-1115-1157-1211-1246-1283-1325-1355-1379-1398-1414-1428-1451-1470-1500
- medium: 750-951-1038-1089-1153-1195-1240-1290-1326-1354-1377-1397-1413-1441-1464-1500
- high: 625-860-961-1020-1095-1144-1196-1256-1298-1330-1357-1379-1399-1431-1458-1500
- very high: 500-769-884-952-1037-1093-1153-1221-1269-1306-1336-1362-1384-1422-1452-1500
- extreme: 250-586-730-815-922-992-1066-1151-1211-1257-1295-1328-1355-1402-1440-1500
- insane: 0-403-576-678-806-890-979-1081-1153-1209-1254-1293-1326-1382-1428-1500

Fer15, Hill 0.25, Strength 100

- very low: 1000-1089-1152-1199-1266-1310-1354-1398-1425-1443-1455-1465-1472-1483-1490-1500
- low: 875-986-1065-1124-1207-1262-1318-1373-1406-1428-1444-1456-1465-1479-1488-1500
- medium: 750-883-978-1049-1148-1215-1281-1348-1387-1414-1433-1447-1458-1474-1486-1500
- high: 625-780-891-974-1090-1167-1245-1322-1369-1400-1422-1438-1451-1470-1483-1500
- very high: 500-677-804-898-1031-1120-1208-1297-1350-1385-1411-1430-1444-1466-1481-1500
- extreme: 250-471-629-748-914-1025-1135-1246-1312-1357-1388-1412-1431-1457-1476-1500
- insane: 0-266-455-598-797-930-1062-1195-1275-1328-1366-1395-1417-1449-1471-1500

But you can try the higher strengths if you want more
So you kept the end point identical now, instead of keeping the range identical. I think that might make more sense. Did you try both?

Also, have you had a chance to quickly check if increasing the Hill value from 0.25 to something higher helps, as you lower the starting point? E.g. if you compare Hill "extreme" with 0.25 vs 0.35, do you have a preference for either 0.25 or 0.35?

Finally, would you like me to add e.g. medium, very high and extreme for both Hill and Log in the next test build?
madshi is offline  
post #8547 of 9394 Old 01-20-2020, 05:47 AM
Advanced Member
 
ddgdl's Avatar
 
Join Date: Mar 2007
Posts: 686
Mentioned: 6 Post(s)
Tagged: 0 Thread(s)
Quoted: 379 Post(s)
Liked: 360
Quote:
Originally Posted by madshi View Post
So you kept the end point identical now, instead of keeping the range identical. I think that might make more sense. Did you try both?

Also, have you had a chance to quickly check if increasing the Hill value from 0.25 to something higher helps, as you lower the starting point? E.g. if you compare Hill "extreme" with 0.25 vs 0.35, do you have a preference for either 0.25 or 0.35?

Finally, would you like me to add e.g. medium, very high and extreme for both Hill and Log in the next test build?
I found it helpful to see some of the extremes during testing, but only if the flickering and other associated problems with extreme settings was also fixed, I think.
ddgdl is offline  
post #8548 of 9394 Old 01-20-2020, 07:00 AM
Senior Member
 
fingersdlp's Avatar
 
Join Date: Oct 2002
Location: Omaha NE
Posts: 466
Mentioned: 2 Post(s)
Tagged: 0 Thread(s)
Quoted: 184 Post(s)
Liked: 98
I have just got around to playing with the HTSM. I experience micro-stutters on some content. The opening of Blade Runner 2049 Sony Logo Pan/Resize shows it well. No stutter with HTSM off. When it is on I do not see any missed frames or display glitches in the stats and the render times don't seem out of line - just see a stutter on the pans. Is this a known issue? My card is a 1070 so maybe I just need more? I also only have an I5 for CPU. I am building a new HTPC so all is not lost if it is just a horsepower issue. I like the results otherwise.

Any feedback appreciated.

Thanks
fingersdlp is offline  
post #8549 of 9394 Old 01-20-2020, 09:50 AM
Senior Member
 
Icaro's Avatar
 
Join Date: Jul 2007
Location: Italy
Posts: 484
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 184 Post(s)
Liked: 94
Quote:
Originally Posted by Neo-XP View Post
The ones in bold below:

Neo-XP, Pure Log, Strength 100

- very low: 1000-1134-1192-1226-1269-1297-1326-1360-1384-1403-1418-1431-1442-1462-1476-1500
- low: 875-1043-1115-1157-1211-1246-1283-1325-1355-1379-1398-1414-1428-1451-1470-1500
- medium: 750-951-1038-1089-1153-1195-1240-1290-1326-1354-1377-1397-1413-1441-1464-1500
- high: 625-860-961-1020-1095-1144-1196-1256-1298-1330-1357-1379-1399-1431-1458-1500
- very high: 500-769-884-952-1037-1093-1153-1221-1269-1306-1336-1362-1384-1422-1452-1500
- extreme: 250-586-730-815-922-992-1066-1151-1211-1257-1295-1328-1355-1402-1440-1500
- insane: 0-403-576-678-806-890-979-1081-1153-1209-1254-1293-1326-1382-1428-1500

Fer15, Hill 0.25, Strength 100

- very low: 1000-1089-1152-1199-1266-1310-1354-1398-1425-1443-1455-1465-1472-1483-1490-1500
- low: 875-986-1065-1124-1207-1262-1318-1373-1406-1428-1444-1456-1465-1479-1488-1500
- medium: 750-883-978-1049-1148-1215-1281-1348-1387-1414-1433-1447-1458-1474-1486-1500
- high: 625-780-891-974-1090-1167-1245-1322-1369-1400-1422-1438-1451-1470-1483-1500
- very high: 500-677-804-898-1031-1120-1208-1297-1350-1385-1411-1430-1444-1466-1481-1500
- extreme: 250-471-629-748-914-1025-1135-1246-1312-1357-1388-1412-1431-1457-1476-1500
- insane: 0-266-455-598-797-930-1062-1195-1275-1328-1366-1395-1417-1449-1471-1500

But you can try the higher strengths if you want more
Of the two, I prefer the Fer15:

Fer15, Hill 0.25, Strength 100
- medium: 750-883-978-1049-1148-1215-1281-1348-1387-1414-1433-1447-1458-1474-1486-1500

More dynamic, more contrast and three-dimensionality.

Sorry for my bad English
Epson LS10500 - AMD RX580 - LightSpace CMS
Icaro is online now  
post #8550 of 9394 Old 01-20-2020, 09:53 AM
Member
 
Join Date: Nov 2017
Location: Stockholm, Sweden
Posts: 106
Mentioned: 2 Post(s)
Tagged: 0 Thread(s)
Quoted: 76 Post(s)
Liked: 28
Quote:
Originally Posted by fingersdlp View Post
I have just got around to playing with the HTSM. I experience micro-stutters on some content.



Any feedback appreciated.



Thanks
Try changing the Vertical sync to "ON" in the Manage 3D settings in NVIDIA control panel. That completely solved the micro stutter I had in Fullscreen Windowed mode.

You can also try and change "frames presented in advanced" to 1, that works for me as well, no stuttering at all.

Sent from my SM-N975F using Tapatalk
fingersdlp likes this.
arcspin is offline  
Sponsored Links
Advertisement
 
Reply Digital Hi-End Projectors - $3,000+ USD MSRP

Tags
dynamic tone mapping , hdr , madvr , sdr , ton mapping

Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page


Forum Jump: 

Posting Rules  
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off