Amlogic Super Scaler - works for me!

Revisited the Amlogic Super Scaler and for my current setup and my viewing situation, I have found that activating the super scaler does improve the sharpness of the TV picture. In particular, I find the “long distance” scenes to be noticeably sharper, such as a distant shot of a forest where the greenery appears sharper etc. I am viewing from around 7 feet away. So far, I have not noticed any negatives when viewing from a normal distance.

My set up is a 43 inch budget Toshiba 4k HDR TV (gives a very acceptable picture), CoreELEC Nexus (latest alpha) running on a S905x2 box (x96 Max) with 4GB ram and 32GB internal storage. GUI is set to 1080p/60Hz. I have all the available viewing resolutions whitelisted, so for example, in my case the 720p/25Hz videos are output to the TV at 1280x720p resolution at 50Hz. It’s my understanding that with this setup, my TV is doing the upscale from either 720p or 1080p to 4k. Does anybody know how the super scaler operates in this scenario? Is it a sharpening algorithm that is applied to the video before the TV upscales it, or does it apply the sharpening after the TV has upscaled? Any info appreciated as I could not find much info on the feature.

Just to remind anybody who wishes to try out the Amlogic super scaler for themselves, create or modify an “autostart.sh” file and place it in the “.config” folder. The source code to add is:-

echo 1 > /sys/module/amvideo/parameters/super_scaler

I understand that some people may not like the sharpening effect, but it’s easy to try out. I did also test my animation videos and other titles such as Avatar, and they all looked fine.

If I remember correctly the super scaler lead to a strange cartoon look. If you like that have fun.

1 Like

Gave it a try and in general I could find no difference except hen it came to fast motion such as when watching todays F1 Qualifying, where there did seem to be an improvement when tracking the high speed cars on track.

No downside though so far.

Gave it a crack, no idea if theres a way to confirm it’s on. May have looked slightly different but hard to tell. Might go back to standard.

If you can ssh into your box and run:

cat /sys/module/amvideo/parameters/super_scaler

and enter. If it returns a Y, then it is activated AFAIK.

I remember seeing that complaint in the past, however, so far I have not seen anything unusual, just a slightly sharper picture.

I only use CoreELEC to play back locally stored SDR videos from my USB HDD. I don’t know if that makes a difference as far as the cartoon effect is concerned.

I did search and find this post:- Odroid N2 test builds - Single Board Computers / Odroid N2/N2Plus/C4 - CoreELEC Forums

I see nothing like this on my setup, images are very natural. If my pictures looked like this, I would definitely disable it. A reply to that post mentions a variable /sys/module/amvideo/parameters/sr0_sr1_refresh and setting this to 0. Mine is currently set to 1 and I have not altered it from the default CE setting.

There is a sample and it was quite visible. If it works for your use case great.

I can’t believe how bad that image was of Alan Partridge! I have never seen anything like that and I’ve been using CoreELEC, and before that, LibreELEC for over 5 years.

I did find a 3-second sample in that thread called “bn10_copy.mkv” and that plays back fine with super scaler turned on. I could not find any other sample.

It seems that in your case the super scaler upsclales from SD to HD and the rest HD to 4k is done by the TV.
Maybe the cartoon look appears if the super scaler upsclales from SD to 4k by itself.
One thing I can say for sure - upscaling from SD to 4k that is done by my TV looks ugly and both other options (SD → 4k by CoreELEC and SD → HD by CoreELEC + HD → 4k by TV) are acceptable for me.

I have no SD videos. All videos are either 720p (HD Ready) or 1080p (Full HD). AFAIK SD is 480p. I agree though that the TV does the heavy lifting to get the native image upscaled from 720p or 1080p to 4k, as I have whitelisted all the available TV resolutions. It must be that the amlogic super scaler applies some extra enhancements on the final image after it has been upscaled by the TV, or it applies it before the image is upscaled by the TV. Either way, it definitely is doing something.

If I disable the super scaler, I can see a difference (not big, but it is there) when viewing long distance shots in particular - ie when you see a landscape against the horizon or a long distance view of a forest etc. There is a slight enhancement on edges that is not there when the scaler is disabled. The edges just look sharper. For facial closeups etc, I can’t really see a difference. However, my TV is “only” 43 inches. It would be interesting to see what the comparison looked like on a 65 inch TV.

The box outputs a 720P signal to the TV, it can’t (and doesn’t) do any processing after your TV upscales the signal to 4K.

Yes, I don’t know why I thought that amlogic would perform some processing after the image is passed on to the TV. So, yes, the processing must take place before. I just revisited some scenes where I can see a (small) difference and I can definitely see a very slight improvement. From some very brief tests, it appears the effect is more pronounced on native 720p videos than on 1080p videos. As I said, all resolutions are whitelisted and I can verify with the ‘o’ on the keyboard that the amlogic display resolution is 720p for 720p videos and 1080p for 1080p videos.

Of course, it’s possible that I may be “wishing” that there is an improvement and maybe there actually is none at all. Still, I suppose that does not matter, as I have convinced myself there is a slight sharpening effect occurring.

In my honest opinion, there is absolutely no way the eye/brain can make out small differences in picture quality unless they are literally side by side. You can make yourself believe there is or isn’t, but unless side by side it’s all guesswork.

Sie by side is always great but if you have a specific video and can mark out definitive areas for comparison then you can define metrics to make comparison viable, especially from repeated viewing.

I agree. I saw the difference by repeatedly playing back a specific 5 seconds of video with the amlogic scaler enabled and disabled. I was able to repeat playback very quickly, as I had a PC that was “ssh’ed” into CoreELEC and was able to “echo” the scaler value from 0 to 1 very quickly and replay that specific section of video on the TV. I repeated this about 5 times and the total time taken was less than a minute.
I’m 99% certain there is a subtle difference in sharpness (only noticeable in certain sections of video). That’s why I left the sharpening enabled as there has been no downside in leaving it enabled - ymmv.

My upscaling comparison: [link]

Thanks very much for that link. I downloaded the images and I’ve been able to use a free utility called “video-compare” ( GitHub - pixop/video-compare at 20200724). It allows a rudimentary (but effective) side-by-side slider comparison of png files. By dragging the slider back and forth, I can see the differences. These differences are small to non-existent at normal viewing distance but proves, to me, that there is something happening under the hood. As said in prior posts, I see no downside to turning on the amlogic superscaler.

I would rank them:-

  1. Pixop AI Super Resolution image
  2. Shield 2019 AI Low
  3. CoreELEC + SuperScaler and the Dune Real Vision 4k about tied, then the rest in no particular order.

I found by dragging the slider quickly to the left and right shows up the differences clearly. Comparing the two CoreELEC images, there is a sharper picture on the SuperScaler image. I also see the SuperScaler image is slightly sharper than Shield 2019 Enhanced. It does appear that the AI algorithms are the best, but I think amlogic superscaler fares well.

Does anybody know of a better comparison tool with a slider??

1 Like

I wonder if there is a chance to add sharpener to CE?
Those android boxes running Ugoos ATV have a setting for that.

You can do it by looking at my earlier post.

  1. Create file “autostart.sh” and place in “.config” folder.
  2. Add the following line to “autostart.sh” - “echo 1 > /sys/module/amvideo/parameters/super_scaler” (without quotes)
  3. Reboot

Done that with no visible effect on Matrix.
Should I use Nexus Alpha for that?