Page 4 of 5

Re: Benq XL2411 - Can't get rid of ghosting

Posted: 16 Feb 2017, 13:25
by Falkentyne
Yes that looks right. If you set the strobe duty to 003 or lower, there will be no crosstalk at the top but the image will be very dim.

Duty and phase are VERY easy to understand.

Duty is simply persistence.
A strobe lasts for a refresh cycle. So at 120hz that's 1/120th of a second. I know you can understand that.
A strobe is an ON+OFF backlight cycle.

So that's 0.083 ms per "point" of strobe duty" at 120hz. (8.3 divided by 100, because 120hz is 8.3 MILLISECONDS). But remember a strobe is an ON +OFF CYCLE. The ON time is much much shorter than the off time. The TOTAL Time is 8.3 milliseconds for both on and off. So if you were using a "1.0 millisecond" persistence, the on time would be 1.0 milliseconds and the OFF time would be 7.3 milliseconds. Because 7.3 + 1.0 = 8.3. Simple math.

The problem is, the 0.083ms per POINT of strobe duty is assuming NO VT tweaks are in use. So on a fresh config without a VT Tweak. strobe duty=012 would be 1.0ms of persistence because 12 x 0.083= 0.996=1 ms (close enough).

VT tweaks however cause the strobe timings to "Revert" to 60hz timings. which means the timing is based on 16.7ms, or 0.167ms per point of strobe duty. Why? I don't know. But if you check the monitor OSD, you will see it reports "60hz" at 120hz. That's why it reports 60hz. I assume this is a failsafe by the firmware when a Vertical Total is used that doesn't belong to an established refresh rate. So therefore, Strobe Duty 006 is equal to 1.0ms with a VT tweak, because 0.167 x 6=1 ms.

Again unless you're just bad at math, you should be able to understand this.

Strobe PHASE is simply at what "POINT" in a frame (the 8.3ms frame) the strobe begins at. Changing strobe phase moves the start of the strobe from the "beginning" of the current refresh cycle (strobe phase=100) to the "END" of the current refresh cycle, in effect adding 1 frame of input lag (Strobe Phase=000). When using a VT Tweak, the strobe phase is COMPRESSED into a LOWER effective range. Remember everything is based on a 60hz baseline and other refresh rates are "interpolated" by the firmware so that strobe phase 0-100 still works. HOWEVER when using a VT weak, this interpolation gets reverted back to the base 60hz values, so that 0-100 doesn't work anymore at 120hz with a VT tweak. Since 120hz is double 60hz..........where do you think the "backlight shuts off" at?

(remember strobe phase starts at 000, NOT 001).

You got it.

50.

That's why the strobe phase range at 120hz with a VT Tweak is 0-49.

(Without a VT tweak it would be 0-100).

Re: Benq XL2411 - Can't get rid of ghosting

Posted: 16 Feb 2017, 13:49
by Helban
Thank you Falkentyne, this makes perfect sense, thanks for the explanation!

When it comes to the on screen service window that currently shows 012 duty, is that correct? or it just shows 012 but the utility I've set to 2 ms does the job and the duty wont change from 012 on the service window no matter what i set?
So persistence = duty and crosstalk = phase?

So using the utility I've set duty to 002 (2ms) and phase to 000 (0) ? But it wont reflect in the service window?

And when looking at the excel sheet you showed, does it mean i should set duty(persistence in the utility) to 1 and phase (crosstalk in the utility) to 49 to get the best results? Or did i misunderstand it yet again? haha Dont know what i can do with the excel sheet really...

To make it simple:

In a step-by-step instructions: Is there anything more I can change, do, try, make, in any of the settings to improve it any more? Or is this the best as it gets?

Thank you soo much for all the help!

Best regards,
Adam

Re: Benq XL2411 - Can't get rid of ghosting

Posted: 16 Feb 2017, 15:33
by Falkentyne
The utility is based on 0.167ms (60hz) strobe persistence widths. So 12 x 0.167 is 2.0ms These persistence widths happen whenever the Vertical Total is set outside of specification for an established refresh rate, or at 60hz.

Default persistence widths per refresh rate
60hz= 0.167 (16.7 / 100).
85hz= 0.117 (11.7 / 100)
100hz=0.100 (10 / 100)
120hz=0.083 (8.3 / 100)
144hz=0.069 (6.9 / 100)

Using a VT tweak changes all of the persistence widths to the 60hz values regardless of refresh rate.
This also causes the strobe phase to be 'compressed' by the percentage increase of the CURRENT refresh rate, compared to 60hz.

The utility is NOT designed to be used WITHOUT a VT tweak active, because then the persistence widths are the refresh rate divided by 100, and the strobe phase isn't compressed anymore and ranges from 0-100.

Re: Benq XL2411 - Can't get rid of ghosting

Posted: 16 Feb 2017, 20:58
by Dakkon
Thanks Falkentyne, you've been a great help! I'm happy with the monitor now. I'll put up with the ghosting on consoles/older PC games, at least it's no worse than my old monitor.

So if I understand right with the explanation above, with VT tweaks active the utility is more accurate than the service menu? Or do you have to just take into account the maths above when changing it in service menu? If that makes sense.

Also do you just not use 144hz at all? As it would be annoying to change brightness each time. It's just some games like overwatch are fine at 144hz, and I feel I notice the increased FPS, which sounds crazy for such a small increase but I do.

Re: Benq XL2411 - Can't get rid of ghosting

Posted: 16 Feb 2017, 21:44
by Falkentyne
The service menu is fine. The utility is just a front end for the service menu. And the utility is CAPPED at strobe phase 046 or 047, I forgot. It wont go higher.

Re: Benq XL2411 - Can't get rid of ghosting

Posted: 17 Feb 2017, 12:12
by Chief Blur Buster
"Changing strobe phase moves the start of the strobe from the "beginning" of the current refresh cycle (strobe phase=100) to the "END" of the current refresh cycle, in effect adding 1 frame of input lag (Strobe Phase=000)."
Close, but not always quite that simple. Assuming the backlight cut-out bug at high strobe phase numbers didn't exist, you're also bringing the _previous_ strobe flash length back, so suddenly you jump from 1 frame of input lag to 0 frame of lag. So 0 is essentially in theory exactly equal to 100 in latency, assuming it wraps around properly (assuming 100 is a full exact 8.3ms delay in strobe flash) -- because if you pushed the strobe flash length 1 full frame ahead, you've also brought the _previous_ strobe flash to exactly where it would be if it was a "0". Making "0" and "100" identical in look & input lag -- it is a wraparound effect. However, there are bugs in some monitors (like my V2 of XL2720Z) that causes the backlight to black out before the wraparound.

It's mostly a matter of _where_ on the screen you want the least screen.

In theory, excluding the fudge factor that blanking intervals provide, the average input lag of *ANY* strobe phase is all identical, if you measured the input lag of _every_ pixel on the display and then _averaged_ it all out. So at the end of the day, strobe phase doesn't matter in input lag -- if blanking intervals were practically zero. But in the real world, we can use bigger blanking intervals (or faster scanout logic built into monitors, like LightBoost which does its own equivalent of "large vertical total" internally, bypassing Windows) that gives more flexibility in strobe phase adjustments before things become bothersome (double-image effects).

It'd be much easier to explain using images.
But you reach situations like:
--> In one setting you may have 1ms lag at top edge, 4ms in center, and 7ms at bottom edge of display
--> In different strobe phase setting, you may have 6ms lag at top edge, 1ms lag at center, and 4ms lag at bottom edge.

This non-linearity occurs when strobing occurs during the middle of scanout. So, you see, strobe phase never really increases/decreases input lag, it simply reduces input lag of one part of display, and increases input of a different part of the display.

This is due to the asymmetry of LCD scanout versus the globalness of the flash. (8.3ms scanout cycle, or rather 7.8ms scanout cycle + 0.5ms blanking interval, depending on size of vertical total; at VT1350 you got a 3:1 scanout:blanking ratio, which means ~6ms scanout + ~2ms blanking interval).

Anyway, back on topic... Crosshairs is usually in the center of the screen, so sometimes people want the strobe phase to be calibrated to the point where crosshairs has the least lag. That's sort of how BENQ monitors ship Blur Reduction by default, but it has a great cost: huge amounts of strobe crosstalk (double-image effect during fast motion) just underneath the middle of the screen. Lots of us complained to BENQ, and they fixed with the V2 firmware, by providing us with a choice of adjustability.

Now, that said -- with large blanking intervals (e.g. VT1350), there's more flexibility in where you want to time the strobe flash, so you can have a slightly earlier strobe before you begin to get bothered by strobe crosstalk.

So, if we're aiming for lower input lag AND high-quality strobing, you want the largest possible Vertical Total *and* you strobe-flash as early as possible after the scanout. "early" is relative (Einstein) ... the strobe phase numbers do not necessarily correspond to minimum lag. If the monitor had proper phase wraparound (not my monitor, the backlight cuts out at larger phase settings) wraparound point might occur at "38" during a mode, or even at "94" etc. In reality, phase is relative to the blanking interval, and there's the 1ms GtG to contend with, so ideal strobe phase is variable depending on the ratio of LCD GtG time versus the timings to delay the beginning of the active visible refresh (e.g. back porch timing)

Easiest way to do this is to run a motion test (TestUFO) while adjusting the service menu. Get the smallest number you possibly can get, without being bothered by the strobe crosstalk (double image effect) if you're adjusting for good full screen motion clarity.

Re: Benq XL2411 - Can't get rid of ghosting

Posted: 17 Feb 2017, 12:56
by Chief Blur Buster
Better understanding the numbers in ToastyX
For those who don't understand the concept of the numbers in ToastyX....

To help you understand what those numbers in ToastyX means: A calendar
Graphics cards transmit pixels from the computer in this sequence: Back Porch then Active then Front Porch then Sync Interval. (First horizontally, then vertically, calendar-style). Let's visualize:
-- A calendar has 5 rows of weeks.
-- A calendar has 7 days in one week.
-- Horizontal Back Porch is like Sunday (left edge of calendar) = 1 pixel Back Porch)
-- Horizontal Front Porch is like Saturday (right edge of calendar) = 1 pixel Front Porch)
-- Vertical Back Porch is like the first week of a calendar (1st week) = 1 pixel Back Porch)
-- Vertical Front Porch is like the last week of a calendar (5th week) = 1 pixel Front Porch)
-- Horizontal synchronization interval would be like a hidden extra days between Saturday and Sunday
-- Vertical synchronization interval would be like a hidden extra weeks below the 5th week
-- Active pixels would be Monday through Friday, of the 2nd through 4th weeks.
(A visible resolution of 5 pixels wide by 3 pixels tall) since 1st and 5th weeks are vertical porches, and the weekends are horizontal porches.

There's many other ways to visualize this, but this is one of the easiest ways to visualize what the custom timings / ToastyX numbers do to a video signal.

Now "days" are output sequentially, one at a time, left to right, top to bottom. That's how pixels travel over a video cable from a computer to monitor. The hidden days (synchronization) are also transmitted too, at the same rate. It takes a finite amount of time to transmit one pixel. Instead of taking 1 day to transmit a single day of a calendar, a graphics output can send hundreds of millions of pixels per second through the cable to the monitor (Common numbers like 300 Mhz dotclock, aka 300,000,000 pixels a second, for 1080p @ 120Hz) including the invisible porch/sync pixels.

Basically, VT1350 is like a calendar with 1350 weeks (Each of which is 2080 days long), and each 'day' (pixel) is transmitted one at a time, sequentially, over the period of the refresh cycle. Bigger dotclock, pixels get transmitted faster. And if you've got lots of extra padding at the bottom (as in VT1350) the 'weeks' (pixel rows) are transmitted faster (sooner) -- aka accelerated scanout.

Anyway, the sequentialness of all (one-day-at-a-time, one-pixel-at-a-time) of this versus the globalness of the strobe backlight (everything gets flashed into visibility all at once) is why you have various kinds of weird lag interactions for different parts of the screen.

High speed video of scanout versus strobe flash is also found at http://www.blurbusters.com/lightboost/video

Re: Benq XL2411 - Can't get rid of ghosting

Posted: 17 Feb 2017, 13:05
by Chief Blur Buster
Falkentyne wrote:That's why the strobe phase range at 120hz with a VT Tweak is 0-49.
(Without a VT tweak it would be 0-100).
There are many variables, but assuming the strobe phase range is exactly equal to a refresh cycle, and there are no bugs in wraparound logic -- you got to consider time of scanout versus time of blanking interval. VT1350 is a ratio of 1080:270, or a ratio of 4:1, so 80% of strobe phase is during middle of scanout and 20% of strobe phase is during the blanking interval. So in theory, with a VT1080 (not possible on 1080p) 0 through 100 would map to VT1350 range of 0 through 80.

But that's not what it behaves in the real world, since we've got both the GtG transition to consider (1ms GtG average) -- and also behaviourial quirks of strobe phase adjustments where the backlight (incorrectly) cuts out during high strobe phases, rather than wrapping around, and how the cutout occurs earlier with longer strobe flash lengths.

So there's no hard and set range rule like "0 through 50", as that actually varies with shorter/longer strobe duty cycle (strobe flash length) and how big a vertical total you manage to have.

Re: Benq XL2411 - Can't get rid of ghosting

Posted: 17 Feb 2017, 13:32
by Falkentyne
Chief do you still have your old Benqs?

Do you remember that when you are not using a VT Tweak, even at a strobe phase of 100, you can use the full range of strobe duty (0-30)? However when you do use a VT Tweak (Note: see below--I'm talking about any VT change that changes far from the default identifier for that refresh rate, not necessarily 1350/1354 or 1500), not only does the strobe phase (maximum) get cut by a percentage of the refresh rate increase over 60hz, but in addition, it limits the maximum strobe duty you can use also. Masterotaku mentioned this to me in a steam chat. He was wondering why when the strobe phase gets cut in half when any VT change is used (or more at 125hz--max strobe phase is 047 ---048 shuts off the backlight), why the formula seems to change to "Backlight shutoff point - Strobe Phase = Strobe duty maximum".

Example: 50 - 49 = 1.

The issue chief is that the monitor seems to be based on a 60hz default guideline. This could very well explain why the VT range of 1497-1502 doesn't work at 60hz (Out of range!) but DOES work at 61hz.

Did you know you can make the backlight shut off even with a VT 1110 tweak at 120hz? Yes, just 1110--something in that range--not 1117, 1117 is for 75hz).
If you set a VT of 1110 at 120hz, the backlight will shut off at strobe phase 050.

And the OSD will report "60hz" again.

Basically, it seems that whenever the VT is changed UP or Down, to any VT That doesn't belong to an established refresh rate, the monitor will report "60hz", the strobe pulse widths get changed to 0.167ms, and the strobe phase range gets "compressed" because the refresh rate isn't 60hz.

Re: Benq XL2411 - Can't get rid of ghosting

Posted: 19 Feb 2017, 13:02
by Chief Blur Buster
Now I understand. Yes, you're right. I now recall realizing the Strobe Utility scaling is somewhat off during some ToastyX timings. You're right about that one.

I recall it was initially calibrated for XL2720Z V2 non-vertical-total 120Hz (which behaves similiarly to 60Hz) but things suddenly vary when you use large vertical totals (VT1350), as you described. I will eventually have to re-calibrate the scaling, or keep it as-is, as Strobe Utility only works on older BENQ's.