Page 1 of 1

Acer VG270UP strobe crosstalk tweaking

Posted: 15 Dec 2018, 22:01
by elgreco
Bought a new monitor (the acer VG270UP), has the AUO panel with 1440p ips and 144hz. I bought it because it had backlight strobing and was cheaper then ULMB variants.

I made a picture of the performance while strobing.

Image
https://i.imgur.com/ZBTsjKq.jpg

Is this any good? I see some crosstalk on top and bottom. I tried to increase total vertical in CRU, but no difference in crosstalk.


Overdrive is maxed out at extreme, with some overshoot. Overdrive normal was to slow.

Re: Acer VG270UP strobe crosstalk tweaking

Posted: 15 Dec 2018, 22:14
by Chief Blur Buster
The JPG you uploaded is fairly over-compressed but looking closer, I was able to see.

This looks average for IPS 1440p for a third-party implementation not made by a big organization (non-NVIDIA, non-Samsung). Top/bottom extreme crosstalk is quite normal.

How many percent did you increase VT? Did you try lowering refresh rate first?
Generally, you need to increase VT by 25%-50% to see human-noticeable difference in crosstalk, and usually the only way to increase VT by such big margins is to do it during lower refresh rates (e.g. 100Hz). Large VT's massively increase the bandwidth requirements of the video cable when we're trying to use large blanking intervals GPU-side (BenQ/ZOWIE/LG technique) instead of monitor-motherboard-side (NVIDIA/ULMB technique). Mathematically, large VT's needs to be multiple milliseconds long. For example, a Vertical Total of 1900 at 1440p means 1440/1900ths of a refresh cycle is spent refreshing and (1900-1440)/1900ths of a refresh cycle is spent in VBI between refresh cycles. But VT1900 is impossible at 144Hz, and requires dramatic lowering of Hz to work. VT1500 and VT1600 will not be worthwhile investments in trying to reduce strobe crosstalk, you must try to VT 50% bigger to see big improvement, and 25% bigger to see slight improvement.

Image

The "bigger-crosstalk-free-zone" effect is successfully observed on BenQ/ZOWIE (especially at 144Hz) and was observed to an extent on LG 240Hz.

There is no guarantee large VT tricks work on ASUS monitors but try (A) testing 100Hz strobed default, then (B) switch back to 144Hz and then derive a large-VT 100Hz strobed using the Appendix A instructions of http://www.blurbusters.com/crosstalk#appendixa to modify a 144Hz small-VT mode into a 100Hz large-VT mode while keeping pixel clock (Dot Clock) unchanged... You may get the crosstalk-free-zone-widening effect roughly similar to the above if your monitor scanout velocity stays at its full max-Hz velocity during the lower-Hz operation.

Also, in addition to the above, you can also help reduce crosstalk by narrowing dynamic range a smidgen (5% or so) -- increasing digital black levels and reducing digital white levels (e.g. via NVIDIA Control Panel) since the fulldark / fullbright colors are the ones that generate the most problematic ghosting and maximum strobe crosstalk. (That's why LightBoost had crappy colors... it intentionally reduced dynamic range). Many of us prefer the colorful full dynamic range, but we do have the option to intentionally reduce the dynamic range via various adjustments to try to reduce ghosting/crosstalk further.

Re: Acer VG270UP strobe crosstalk tweaking

Posted: 16 Dec 2018, 05:21
by elgreco
Chief Blur Buster wrote:The JPG you uploaded is fairly over-compressed but looking closer, I was able to see.

This looks average for IPS 1440p for a third-party implementation not made by a big organization (non-NVIDIA, non-Samsung). Top/bottom extreme crosstalk is quite normal.

How many percent did you increase VT? Did you try lowering refresh rate first?
Generally, you need to increase VT by 25%-50% to see human-noticeable difference in crosstalk, and usually the only way to increase VT by such big margins is to do it during lower refresh rates (e.g. 100Hz). Large VT's massively increase the bandwidth requirements of the video cable when we're trying to use large blanking intervals GPU-side (BenQ/ZOWIE/LG technique) instead of monitor-motherboard-side (NVIDIA/ULMB technique). Mathematically, large VT's needs to be multiple milliseconds long. For example, a Vertical Total of 1900 at 1440p means 1440/1900ths of a refresh cycle is spent refreshing and (1900-1440)/1900ths of a refresh cycle is spent in VBI between refresh cycles. But VT1900 is impossible at 144Hz, and requires dramatic lowering of Hz to work. VT1500 and VT1600 will not be worthwhile investments in trying to reduce strobe crosstalk, you must try to VT 50% bigger to see big improvement, and 25% bigger to see slight improvement.

Image

The "bigger-crosstalk-free-zone" effect is successfully observed on BenQ/ZOWIE (especially at 144Hz) and was observed to an extent on LG 240Hz.

There is no guarantee large VT tricks work on ASUS monitors but try (A) testing 100Hz strobed default, then (B) switch back to 144Hz and then derive a large-VT 100Hz strobed using the Appendix A instructions of http://www.blurbusters.com/crosstalk#appendixa to modify a 144Hz small-VT mode into a 100Hz large-VT mode while keeping pixel clock (Dot Clock) unchanged... You may get the crosstalk-free-zone-widening effect roughly similar to the above if your monitor scanout velocity stays at its full max-Hz velocity during the lower-Hz operation.

Also, in addition to the above, you can also help reduce crosstalk by narrowing dynamic range a smidgen (5% or so) -- increasing digital black levels and reducing digital white levels (e.g. via NVIDIA Control Panel) since the fulldark / fullbright colors are the ones that generate the most problematic ghosting and maximum strobe crosstalk. (That's why LightBoost had crappy colors... it intentionally reduced dynamic range). Many of us prefer the colorful full dynamic range, but we do have the option to intentionally reduce the dynamic range via various adjustments to try to reduce ghosting/crosstalk further.
Yeah sorrry for that, sending it to pc and uploading compressed it...

Yeah I tried a VT1825 on 120hz which was the max for the panel. Didn't work. On 100hz I tried VT2200 max but that also didn't work. 100Hz does also have more noticable crosstalk. I tried narrowing the dynamic range, but no succes either :(

Anyways. Is this how all non ULMB perform? I am doubting if I maybe should have bought the ULMB (g-sync) version.

Re: Acer VG270UP strobe crosstalk tweaking

Posted: 16 Dec 2018, 05:25
by Chief Blur Buster
elgreco wrote:Yeah I tried a VT1825 on 120hz which was the max for the panel. Didn't work. On 100hz I tried VT2200 max but that also didn't work. 100Hz does also have more noticable crosstalk. I tried narrowing the dynamic range, but no succes either :(
Did you try the exact-dotclock-number-preserving trick? That's pretty critical to keep the horizontal and vertical numbers exactly the same as before, and only increase one vertical number (Vertical Total) while decreasing the refresh rate. The horizontal numbers, the vertical numbers (except Vertical Porch) must remain exactly the same, as does Pixel Clock (dotclock) exactly the same.

So usually this should converge into only one unique VT number that is often a unique unusual number (like 987 or 1493 or 2186 or whatever). Locking the other numbers allows a 10x bigger success rate than trying random vertical total numbers.

This is much easier to do in ToastyX CRU than with NVIDIA Control Panel.

Re: Acer VG270UP strobe crosstalk tweaking

Posted: 16 Dec 2018, 06:28
by elgreco
Chief Blur Buster wrote:
elgreco wrote:Yeah I tried a VT1825 on 120hz which was the max for the panel. Didn't work. On 100hz I tried VT2200 max but that also didn't work. 100Hz does also have more noticable crosstalk. I tried narrowing the dynamic range, but no succes either :(
Did you try the exact-dotclock-number-preserving trick? That's pretty critical to keep the horizontal and vertical numbers exactly the same as before, and only increase one vertical number (Vertical Total) while decreasing the refresh rate. The horizontal numbers, the vertical numbers (except Vertical Porch) must remain exactly the same, as does Pixel Clock (dotclock) exactly the same.

So usually this should converge into only one unique VT number that is often a unique unusual number (like 987 or 1493 or 2186 or whatever). Locking the other numbers allows a 10x bigger success rate than trying random vertical total numbers.

This is much easier to do in ToastyX CRU than with NVIDIA Control Panel.
No I just tried random VT.. :oops: How can I use an exact doctclock number in CRU? And is dot clock = pixel clock? Because it says max 600MHz on my panel. and 580.08MHz on 144hz refreshrate.


"Lock the dotclock (put radio button on Dotclock). Make sure dotclock DOES NOT change."
I can't find where I should do this in CRU by ToastyX

Re: Acer VG270UP strobe crosstalk tweaking

Posted: 16 Dec 2018, 10:25
by RealNC
elgreco wrote:How can I use an exact doctclock number in CRU? And is dot clock = pixel clock?
Yes. Dot clock means pixel clock.
"Lock the dotclock (put radio button on Dotclock). Make sure dotclock DOES NOT change."
I can't find where I should do this in CRU by ToastyX
Image

Re: Acer VG270UP strobe crosstalk tweaking

Posted: 16 Dec 2018, 12:02
by elgreco
RealNC wrote:
elgreco wrote:How can I use an exact doctclock number in CRU? And is dot clock = pixel clock?
Yes. Dot clock means pixel clock.
"Lock the dotclock (put radio button on Dotclock). Make sure dotclock DOES NOT change."
I can't find where I should do this in CRU by ToastyX
Image
Ok, then I understood correctly. I tried to to use a fixed pixel clock and everything, doesnt really work. The lower the refreshrate the more slightly more strobe crosstalk I get. So no room for improvements on this panel :P

Re: Acer VG270UP strobe crosstalk tweaking

Posted: 16 Dec 2018, 16:53
by Chief Blur Buster
I see, that's really too bad. Usually it's vice versa (less crosstalk when you do this trick)

Try one more time, and instead increasing Total instead of increasing Back Porch.

The worsening crosstalk may be the strobe phase being pushed into the active visible refresh cycle, so if only we had access to a strobe phase adjustment (strobe delay adjustment). The monitor's internal strobe flash timing may synchronize to the beginning of Blanking (after the porches), so increasing one of the two numbers may cause shifts in strobe crosstalk in monitors with unadjustable strobe phase. Kind of like a crude strobe phase adjustment via balancing between Back Porch (standalone setting) and Front Porch (embedded via increasing/decreasing in Total).

The way monitor manufacturers program their monitors sometimes works against us.

(BTW, for monitor manufacturers reading this, strobe tuning and validation services are available; contact me at mark[at]blurbusters.com ...)

Re: Acer VG270UP strobe crosstalk tweaking

Posted: 18 Dec 2018, 09:17
by elgreco
Chief Blur Buster wrote:I see, that's really too bad. Usually it's vice versa (less crosstalk when you do this trick)

Try one more time, and instead increasing Total instead of increasing Back Porch.

The worsening crosstalk may be the strobe phase being pushed into the active visible refresh cycle, so if only we had access to a strobe phase adjustment (strobe delay adjustment). The monitor's internal strobe flash timing may synchronize to the beginning of Blanking (after the porches), so increasing one of the two numbers may cause shifts in strobe crosstalk in monitors with unadjustable strobe phase. Kind of like a crude strobe phase adjustment via balancing between Back Porch (standalone setting) and Front Porch (embedded via increasing/decreasing in Total).

The way monitor manufacturers program their monitors sometimes works against us.

(BTW, for monitor manufacturers reading this, strobe tuning and validation services are available; contact me at mark[at]blurbusters.com ...)
Yeah I tried that too, but with no succes. I do noticed 85hz has way better strobing. Near perfect strobing. To bad it's to low hz, feels so sluggy.

Anyway thanks for helping :) Love your work! :D

Re: Acer VG270UP strobe crosstalk tweaking

Posted: 18 Dec 2018, 16:45
by Chief Blur Buster
Not surprising, strobed 85Hz is much more pleasant in motion clarity and probably what you should use when input lag isn't problematic. Try testing it with fast-motion racing games, space simulators, WWII dogfighting simulators, and other situations where the extra few milliseconds of response isn't as critical as during FPS / CS:GO / Fortnite / PUBG.

Strobed 240Hz is better latency-wise if you want CS:GO or PUBG though, but it's mainly found in BenQ and LG monitors. But nearly not as clean strobe quality as 85Hz.