Blur Busters Forums

TFTCentral has a great performance evaluation of the BenQ XL2720Z here:

http://www.tftcentral.co.uk/reviews/benq_xl2720z.htm

Now, I'm interested in this monitor for use in research in cognitive science, where we're still using CRT monitors for fear of the poor temporal performance of LCDs. Until now? It strikes me that from performance data produced by TFT Central, the BenQ XL2720Z might actually have better temporal performance than a CRT. That is, if set up with:
- blur reduction on
- instant mode on
- AMA off
- 120Hz (could use 144Hz, but researchers are used to 120Hz because 6 frames then take a nice round 50ms)

The display should have the performance described by the attached figure, which attempts to show what happens as you try to tell the display to go from either all-black to all-white (solid lines) or from dark-grey to light-grey (dotted lines), then back again.
The horizontal axis represents time and the vertical axis represents luminance at two locations on the screen; the top half shows luminance-through-time of a pixel near the top of the screen while the bottom half shows luminance-through-time of a pixel near the bottom of the screen. The lines (dotted and solid) show what the luminance would be if the backlight were on the whole time, but the vertical white lines show the periods during which the backlight is actually on.

So, my understanding is that typically a dark-grey to light-grey transition takes longer to actually complete (with AMA off) than a black-to-white transition, hence the shallower slopes for the dotted line. Additionally, with a traditional display the bottom pixels start their transitions later than the top pixels. So is my understanding of the BenQ's strobing backlight correct that it strobes the entire screen for the last 2ms (ish, if you're using 120Hz) of the refresh period, thus catching both the top and bottom and both transition types at a point in their transition curve at which everything has fully transitioned? If so, then this would mean that this display is better than a CRT because a CRT still has a difference in the times at which the bottom of the display is updated relative to the top of the display.

I also have some blue lines and text representing my understanding of the display lag to be expected from this display when using vertical sync. That is, with instant mode on and blur reduction *off*, the TFT Central data suggests a 1ms (ish) display lag, but with blur reduction *on* this should increase to 7ms. Does this make sense? (FYI, I think researchers would be OK with 7ms so long as it's a constant lag)

Thoughts? Is my understanding of the mechanics here awry in any way?

Hm, one misunderstanding/misrepresentation a colleague just pointed out is that my plot of the grey-to-grey transition used the average g2g measure from TFTCentral (and even then I misprepresented it as being faster than a refresh when it's 7.5ms in their 144Hz test), whereas the worst-case might be more important to consider. In their 144Hz test, the worst-case result they get is 14.5ms, which is more than two refresh cycles, so the transition would indeed be "visible" during the backlight strobe. ("visible" in quotes because it might not be perceivable by humans)

Good questions.

The diagram are pretty accurate, and the 7ms is pretty close to spot-on though you need to add 1-2ms extra latency for the cable latency and the GtG pixel transition settlement waiting time period (extra pause at end of refresh, to accomodate GtG lag, before flashing backlight). As you can see at the high speed videos of LCD (Those are high speed videos of http://www.testufo.com/flicker running in full screeen mode). The GtG occurs at different parts of the LCD at different parts of the refresh cycle. But with a strobe backlight, the majority of the GtG occurs in total darkness, massively improving the predictability of the illumination of an image on an LCD. So your diagram is correct.

During 120Hz BENQ Blur Reduction, my photodiode oscilloscope measures a constant 9.5ms latency from Directd3D Present() to illumination of photodiode, with BENQ Blur Reduction enabled. Yes, same 9.5ms latency for both top and bottom edges. All strobe backlight displays are global displays, unlike CRTs which are top-to-bottom sequential scan. This is an advantage for certain kinds of vision research purposes where global presentation is favoured. The 9.5ms figure includes DVI latency, and the GtG waiting period (waiting for pixels to finish transitioning before flashing the strobe backlight). It takes 1/120sec to scan the image (~8.3ms), then you've got a bit over an extra millisecond after.

And since GtG occurs in total darkness, and becomes visible to human eye only on strobe flash, you've turned gradual GtG into more instantaneous presentation.

Also, don't forget the Crosstalk setting in Blur Busters Strobe Utility, which is essentially strobe timing phase. During 1080p@120Hz (135KHz scanrate: i.e. refresh speed of 135,000 pixel rows per second) adjusting the Crosstalk setting in Blur Busters Strobe Utility will add about ~1/135,000th of a second extra latency, for every extra scanline that the crosstalk is moved downwards (i.e. downwards by 1 pixel). So if you move the crosstalk ghosting zone downwards by about 240 pixels, you have about 240/135,000ths of a second extra lag. But otherwise, the lag is constant for a specific resolution, refresh rate & "Strobe Utility" crosstalk setting. Thus, you should recalibrate input lag measurements everytime you change Custom Resolution settings & Strobe Utility settings. (e.g. 1080p with VT1125 may have different lag behaviors than 1080p with VT1350 due to the accelerated scanout and larger pauses between refreshes, causing ideal strobe phase to be affected)

If you use custom resolutions with large vertical totals of 1350 or 1500 scanlines per refresh cycle including blanking interval lines (for brevity, I call them the "VT1350 tweak" or "VT1500 tweak") and a slightly lower refresh rate (e.g. 85Hz, 100Hz) to squeeze more GtG into large blanking intervals between refreshes, this will eliminate a lot of crosstalk (double image effecte caused by incomplete GtG leaking between refreshes at the top/bottom edges of the screen). The Z-Series is capable of syncing to video signals with large blanking intervals lasting up to ~4-5ms (at lower refresh rates), which is more than enough to hide 1ms GtG plus the vast majority of GtG error/overshoot. At 120Hz, there's room for a ~1.6ms blanking interval via the VT1350 tweak. Getting familiar with the strobe crosstalk behavior (aka incompletely finished GtG transitions, leaking between refreshes, resulting in double-image effects during fast motion) will be important to see if it's an error factor in your tests. If avoiding strobe crosstalk is important, you could present your visual stimuli in approximately the center 3/4ths of the screen, to avoid error factors caused by strobe crosstalk.

Note that "Persistence" (Strobe length) varies a bit based on the Vertical Total tweak, the millisecond scale in Strobe Utility is calibrated very accurately (error <0.1ms) in Strobe Utility during the VT1350 tweak. The strobe flash timing is the same for the leading edge of the strobe flash, so lag is not affected by the "Persistence" setting; which just affects the timing of the trailing edge of the strobe flash.

Make sure your Z-Series has V2 firmware, for compatibility with Strobe Utility.
You can also get BENQ Z-Series monitors with V2 firmware preinstalled.

P.S. I consider this discussion worthy of the "Area 51" forum!

mike.lwrnc, you might want to check out this review: http://display-corner.epfl.ch/index.php/BenQ_XL2411Z.
It is still work in progress though.