flood wrote:3. persistence: CRTs have very low persistence but then there is also a soft trail. this is a mild annoyance for bright objects moving on a dark background. i don't know how the best strobed LCDs compare.
The best strobed LCDs (the top 5%-10%) can easily beat many medium-persistence CRTs such as the Sony FW900
The difference is
-- Phosphor ghosting on CRT
-- Strobe crosstalk on LCD
Some CRTs have more phosphor ghosting, while some LCDs have less strobe crosstalk. The venn diagram already overlaps.
It's hard to spend the money to buy enough monitors (CRT+LCD) to figure out where the venn diagram overlaps. But I am Blur Busters, that means I've seen enough to witness the venn diagram.
As Advanced Crosstalk FAQ
indicates, the best LCDs can be literally ~1/100th the crosstalk of the worst LCDs. Unfortunately, most people can't afford to buy 10 monitors to find which one has the best strobing.
However, strobing often degrades color quality due to the way many monitors need to slightly overclock its scanout (to cram GtG into VBI) *and* to stay away from overdrive-problematic fulldark primaries and fullbright primaries (you need ovedrive voltage undershoot headroom below black, and overdrive voltage overshoot headroom above white, to eliminate crosstalk for extreme bright/dark colors). Also, TN panels are prone to LCD inversion artifacts (checkerboard patterns) and strobing amplifies that.
For fans of strobing, what I absolutely love about the upcoming strobing on 240Hz 1ms IPS panels is
(A) IPS is immune to inversion artifacts for the most part
(B) IPS is now approaching 1ms, which finally "crams the GtG elephant into the tiny VBI drinking straw" -- Most of GtG-time is within VBI-time, needed to eliminate majority of strobe crosstalk.
(C) IPS is now 240Hz, which is plenty of hertzroom above comfortable strobing (~100-144Hz), hertzroom needed to eliminate majority of strobe crosstalk.
(D) IPS backlights tend to be incredibly bright, which helps strobing
(E) IPS colors are more immune to color degradation during strobing.
(F) IPS with "LightBoost 10%" motion clarity, without nearly as many drawbacks
Overvoltage-boosted strobing that milks the Talbot-Plateau Law to cram more photons in ever-shorter time periods of a strobe backlight LCD to reduce motion blur -- means the LCDs can gain CRT motion clarity attributes as LED backlight technology improves. The Valve Index virtual reality LCD is able to achieve 0.33ms LCD persistence (less than one-third of LightBoost 10%) while being extremely bright -- far brighter than LightBoost 10% or ULMB Pulse Width 10. That's why LCD strobe backlights can have less motion blur than low-persistence OLEDs, that's part of why HTC/Oculus switched to LCD instead of OLED. It's easier to push engineering closer to the Talbot-Plateau law (double brightness flash in half time = equal brightness), due to the fact outsourced light is easier to make brighter than direct pixel light. This may change when MicroLEDs come (which will be brighter than OLEDs), it's simply "bright + low persistence" is extremely hard (how strobe backlights darken the image), but you can simply use more powerful LEDs + heatsinked backlights. (Remember, LED is getting brighter and is now used in stadiums). Which makes brighter low-persistence progressively easier to achieve; and technologically also helps reduce strobe crosstalk, by keeping the flash length within the VBI too.
Strobe tuning and overdrive tuning will be critically important for IPS strobing, but what I'm seeing on my desk (I have a 240Hz 1ms IPS already, not ready to mention the brand just yet). I am liking what I am seeing ... It's not as zero-crosstalk as LightBoost, but you're getting the top 10%-best strobing *AND* the vaunted IPS quality. You still lose lumens, but at least it's only a drop literally from ~400 lumens to roughly ~150-lumens (ish) -- not a problem for many people who's been used to the poor ~50lumens of "LightBoost 10%".
You still have more strobing lag than CRT, due to the requirement of scanout-in-dark and GtG-complete-in-dark before the flash -- but it's minimal and fine for motion-clarity-priority situations rather than latency-priority situations; and you can still turn off strobing to get the same IPS lag as an average 1ms TN panel.
Little known to users, but there's also the async between scanout versus VSYNC OFF (frameslice streaming into the scanout) so vertical latency gradient mechanics feel different on CRT (sequential-scan) versus strobing (global flash), but this can become irrelevant if you use low-latency VSYNC ON methods to minimize microstutter on CRT or LCD. (In fact, low-latency VSYNC ON feels better with strobe LCDs than with CRTs, because VSYNC ON is global, and strobe is global = zero latency gradient = no latency differentials between top/center/bottom for synchronized global frame delivery mechanisms).
Most screens scanout top to bottom, see http://www.blurbusters.com/scanout
for high speed videos of TN LCD, IPS LCD, OLED, etc.
Common latency gradient mechanics, Present()-to-Photons
VSYNC ON + CRT = latency(top) < latency(bottom)
VSYNC OFF + CRT = latency(top) ~= latency(bottom)*
VSYNC ON + LCD strobe = latency(top) = latency(bottom)
VSYNC OFF + LCD strobe = latency(top) > latency(bottom)**
*Note1: Individual frameslices of VSYNC OFF are latency subgradients themselves, as a global fixed gametime, streamed into the progressive cable scanout in realtime. Lag is always lowest immediately below a tearline, and highest immediately above a tearline, and are modifiers of [+0...+frametime] to pre-existing pixel latency, as a linear latency gradient from top thru bottom of frameslices, and frameslices can overlap VBI and/or wraparound to the next refresh cycle. Given sufficiently high framerate and random frameslice placements (random tearlines), average latency equallizes across the full surface of the CRT.
**Note2: You have Note1 equalization of latency on the cable, but you've got the mandatory scanout-in-dark followed by the global flash. As a result, the freshest pixels are near bottom edge of screen, so when measuring multi-pass average latency of a single pixel, the panel's latency gradient is inverted
As refresh rates go higher (or gets eliminated altogether), the latency range in the latency gradient falls down, as the latency gradient is usually capped to a range of refreshtime (1/Hz) .... That said, assuming framebuffer backpressure is kept in check (e.g. VR-style shallow-queue framepacing techniques for ultralowlatency 90Hz or 120hz VSYNC ON equivalent), perfect high-Hz strobe feels beautifully consistent latencyfeel. That said, for a 100Hz CRT VSYNC ON, if a person is using VSYNC ON with a CRT, most of those people don't care enough about the 10ms latency differential between top/bottom. On the other hand, some of us are simultaneously picky about latency AND latency consistency, in addition to zero-stutter, zero-blur. So, understanding latency consistency along the screen surface plane, is a key element to a "Blur Busters Einstein" brain matter (I currently teach this stuff; in the training/classroom stuff I sometimes do at manufacturers/vendors -- services.blurbusters.com
-- from time to time, a vendor flies me in for a few days or a week)
Absolute lag is lowest with VSYNC OFF + CRT though, but if latency consistency and glassfloor Present()-to-Photons latency is desired for all 2 million pixels simultaneously, it currently favours strobe LCD + VSYNC ON. The problem is getting a really good strobed LCD without the artifact tradeoffs...
Either way, back on topic...
There are preferences between IPS (IPS Glow) versus TN (TN viewing angle) versus VA (VA dark ghosts) but with the major speedup of IPS pixel response, it's easier to "have-cake-and-eat-it-too". Almost.
If you're buying the best strobed 120Hz cost-no-object within 6 months, three words: "240Hz 1ms IPS". Yes, hertzroom, but you need to buy hertzroom if you want good strobing at lower Hz.
We are going to have an exciting announcement soon. Keep tuned.
</Advanced Technical Post>