Lagless HD Screen?

[Arino]

Hi,

I'm looking for a new full HD screen I can use for Xbox360 and PS3.

It should be (almost) lagless (like only 1 frame input lag) and as big as possible (I know most lagless ones are only 24" or 27", but there are exceptions).

Any suggestions? Btw I'm living in Europe (NL).

Thx!

[Fudoh]

55" Sony W8. 2014 (!) model ending in "B". About 880 EUR. Less than one frame of lag.

All the 2015 models are considerably slower or - if you're looking at monitors - considerably smaller.

[Ed Oscuro]

http://www.displaylag.com/display-database/

Search through the "Display" field for HDTV. There's a bunch listed at 17ms, which is one frame of lag. Of course, this isn't all models, and the specific one Fudoh mentioned isn't listed. The other 2014 W8 series came in at 22-23ms, though.

From what I understand, there's still no good 30"+ monitors out there for console gaming. I did see an announcement recently about a 1080p FreeSync display coming, but it'll also be somewhat small (24" or 27"). If you're thinking about next-gen gaming, FreeSync might be a good choice - if the consoles ever support it.

[Fudoh]

The other 2014 W8 series came in at 22-23ms, though.

A reading like that on Leo's handheld tester translates to considerably less than a frame of lag using a classic CRT/LCD camera comparison.

[Edward_Tz]

Really like my 50W700B. Not sure which model that translates too in Europe though. W705 maybe?

[Xyga]

Yeah you just add that '5' for many euro models, so it's W705B. Or a '6' for the gray~metal frame variants.

The 2014 Sony W7 W8 W9 are still the best though, especially the W8 Fudoh pointed at; best bang-buck ever for a gaming and movies competent large set.
There's 50" size and a 42" as well, but the latter is a dumbed-down version.

Those are almost all out of stock everywhere I look. *dramatic music*

For websites there's also rtings.com who use the Leo Bodnar (crudely, only the middle bar so it's above reality but gives an idea).
hdtvtest.co.uk as well but they publish reviews less often.

[Arino]

Thank you guys!

http://www.displaylag.com/display-database/

Search through the "Display" field for HDTV. There's a bunch listed at 17ms, which is one frame of lag. Of course, this isn't all models, and the specific one Fudoh mentioned isn't listed. The other 2014 W8 series came in at 22-23ms, though.

From what I understand, there's still no good 30"+ monitors out there for console gaming. I did see an announcement recently about a 1080p FreeSync display coming, but it'll also be somewhat small (24" or 27"). If you're thinking about next-gen gaming, FreeSync might be a good choice - if the consoles ever support it.

About that list, is the "Lag" data actually about input lag? Or is it the response time?

[Arino]

Also I recently got that one here for free:



Is it any good? I can't test it at the moment because I'm moving from Germany to the Netherlands very soon, so all my stuff is in boxes. I've watched TV on it once and didn't really like the picture I think. Also heard Grundig was not a good brand.

[Xyga]

About that list, is the "Lag" data actually about input lag? Or is it the response time?

Both.

Actually rather than 'input lag' we should say 'signal processing lag' or something; the time the display takes to handle, eventually upscale, and render an image.
Then there's the characteristics of the display (is it just sample and hold or full-frame flickering) and the response time element (things related to the liquid crystal panel itself and how it performs in the display).
The 'total lag' is a sum to begin, an not everyone agrees on the exact formula yet.
The LB lag tester is a 'dumb' machine limited by its capabilities and design, I don't get exactly what it doesn't do right in every situation but basically for someone who wants to know only about the 'processing lag' then most websites using it will show you only the flashing bar of their choice (the LB works displaying 3 white bars on which you apply the tester's sensor itself), or an average of the three figures of the said bars (this is what displaylag does), so depending on the figure we're given and the method used it is necessary to substract a number of milliseconds to get a more accurate 'signal processing lag' figure.

Now there are websites like tftcentral adding half the average measured pixel response times to the signal processing lag, because they think it does count in the 'total percievable lag' or something.

What I like to believe is that when I see a website like hdtvtest giving us both high speed camera (usually the lowest) and LB tester measurements (always significantly higher), is that for a human the reality is somewhere in-between these two figures, a bit like what tftcentral seem to be thinking with their method, also pcmonitors.info as it seems.
- The high speed camera says 6ms ?
- The LB tester says 16ms ? (from the middle bar or average of the three)
- in-between = 11ms

Maybe I'm wrong but I don't think that's extremely important because I can't sense only a few milliseconds difference delay, a full frame I can (with direct side-by-side comparison), not 5ms or so.
Therefore when I read a review with LB measurements I think 'okay so the real lag is a bit lower than that' and that's accurate-enough for me.

[Ed Oscuro]

Also I recently got that one here for free:



Is it any good? I can't test it at the moment because I'm moving from Germany to the Netherlands very soon, so all my stuff is in boxes. I've watched TV on it once and didn't really like the picture I think. Also heard Grundig was not a good brand.

I think you answered your own question!

Lag and blur are often considered different things (the response time part of lag has similar basic causes to screen blur) but either can make gaming less enjoyable than it should be.

Panels in the 30"+ range haven't been produced in good quality and massive quantities for a long time, if ever. There might be some opportunity for 30"+ panels to break into the PC market where people want bigger screens than 27" but so far the 32" screens have been lousy.

The other 2014 W8 series came in at 22-23ms, though.

A reading like that on Leo's handheld tester translates to considerably less than a frame of lag using a classic CRT/LCD camera comparison.

That's useless information for the end user, who can't distinguish between latency caused by input lag, and that caused by pixel response time. The 22ms time is much more accurate to the person's experience.

Besides, the old camera method is unreliable and obsolete. Maybe the Bodnar is biased towards certain display types, but a slow LCD is still slow.

[Fudoh]

That's bull, Ed, sorry to say. Most people still don't even understand how to interpretate the readings they get using Leo's device.

[Edward_Tz]

Besides, the old camera method is unreliable and obsolete. Maybe the Bodnar is biased towards certain display types, but a slow LCD is still slow.

What's unreliable or obsolete about the camera method?

[Fudoh]

nothing. As long you average yours readings over enough snapshots (I usually do 30+) it perfectly represents what you really see.

[Ed Oscuro]

Can't agree one bit.

The camera method is obsolete - you can't compare its figures with those on lag databases because everybody gave it up. It's nearly worthless these days and it's a disservice to consumers to throw these understated numbers at them that they cannot use in comparisons.

It's not repeatable. Every display type has the potential to draw lines in a different pattern, the camera's settings matter, the quality of the reference display matters (and with faster LCD screens coming out, many older CRTs with slow decaying phosphors simply won't cut it for lag testing), the moment in the refresh pattern the picture is taken matters, and where the stopwatch image is on the screen matters. Sure, I would say that somebody who has taken 30+ images likely has a sufficiently large sample to draw a conclusion, but there is no guarantee that other camera test numbers, especially those from other sources, will be compatible.

In comparison, the Bodnar tester throws three white patches on the screen and you measure each, and can interpret them as needed (most people just average them). The timing issues are all irrelevant with the Bodnar tester, which constantly measures panel illumination directly. The only thing that matters is that the screen's temperature is stabilized (and comparable to those of other screens), but that was true of the camera method too.

The only thing the camera method has going for it is that it can show that different displays can have different pixel response and characteristics (blur, overdrive artifacts, different response for different colors) so you could get some trailing artifacts that muddle an image. The Bodnar's accuracy is also reduced slightly by some backlight flickering. However, this would lead one to expect the camera method to show slower times, but it consistently understates latency. Additionally, if blur or latency performance of different colors needs to be tested, then it's better to use the camera in conjunction with a blur test (like the UFO test) than any simple lag measurement.

Finally, I don't see the sense of stating that a display which has a total latency figure of 22ms is "really" under 1 frame. It is obviously not under one frame. If the Bodnar tester's photodiode doesn't notice the state change until 22ms, your eye won't see an image faster either. Only plasmas have been thought to fool the Bodnar (but that might have been due to testing methodology).

These points and some others were discussed here, not to mention other places as well.

[Edward_Tz]

The thread you link says the Bodnar test isn't all that accurate. All the sites that use the Badnar tester have some small paragraph saying their measurements might not be accurate.

Seems like any missing information about the picture test would be cleared up with meta data.

[Fudoh]

Can't agree one bit.

it's ok. We don't have to agree.

Finally, I don't see the sense of stating that a display which has a total latency figure of 22ms is "really" under 1 frame.

well, what people interests after all is how much slower the display in question is to a CRT without any digital processing. So, AT LEAST you gotta subtract the readings you get with Leo's on a such a CRT. If you do that you got a correction factor of -8.5ms and that still lets you neglect all the points like required brightness for registration and other (possibly subjective) arguments.

[Xyga]

You know Fudoh there hasn't been any very clear explanation about the LB's quirks anywehre but in some lost thread deep in the hardware section, iirc for one I thought I read it was off by 7ms from the middle bar for instance, the conversation was kind of obscure and I didn't get the whole thing, for instance how you calculate that error factor now ?

As blurbusters say the first bar reading appear close to camera reading so if that's it, you will have a crowd of people saying "whatever we stare at the middle of the screen most of the time so why should the top reading be the relevant one ?" and what about pixel response indeed as most serious articles explain like on tftcentral and pcmonitors it doesn't seem like BS to me or are we in the middle of some sort of misunderstanding ?

[Ed Oscuro]

well, what people interests after all is how much slower the display in question is to a CRT

No, this isn't true. A CRT with slow phosphor transitions (i.e. bright -> black transition on many CRTs is very slow) is not only not a reference for comparison to fast LCD panels, it's also not an ideal. With camera testing, the speed of the comparison CRT (if indeed the tester used a CRT!) phosphors is yet another unknown. There are some unknowns with the Bodnar, but most of those variables remain unchanged with the Bodnar - if they don't, only then do you need to resort to collecting many data points.

Almost nobody cares about CRTs as a baseline - even amongst gamers - they want to know actually how laggy it is. One direct latency measurement comes closer to realizing this wish than a comparison against unknown equipment.

without any digital processing.

This has absolutely nothing to do with the disparity between photodiode and camera testing, and you still have to account for the difference in testing.

So, AT LEAST you gotta subtract the readings you get with Leo's on a such a CRT.

Nope.

The disparity seems to be mainly due to wishful reading of SMTT numbers, which in live footage might be so blurry as to be nearly indecipherable. The Bodnar tester appears to only take a reading when the measured light value passes a certain threshhold.

We can argue about whether Bodnar's photodiode is too strict in determining at what point to take a reading off a gray-to-gray transition, but it is nevertheless likely a very good forward-looking measurement that should benefit faster screens, unlike camera testing-derived numbers which actually reveal nothing about the subjective readability of the image, especially in the absence of the original images.

Put another way, Chief Blur Buster says:

.......Estimating GtG from a photo is different from estimating GtG from electronic lag testing; SMTT numbers are very visible in photo tests when GtG is less than 50% complete, so when GtG (e.g. 8ms) is a significant error factor, latency measurement error can be quite huge.

In short, camera testing sucks and nobody should rely on it alone since there are too many uncontrolled variables.

Of course, there are many useful things that can be gleaned from camera tests, but subjective image quality is hard to capture even with a camera which will happily adjust exposure and pull up details that the eye would not see clearly. And unfortunately, camera-based lag testing combines the worst of unrepeatable experiments with unavoidably subjective readings.

You know Fudoh there hasn't been any very clear explanation about the LB's quirks anywehre

Some of them are well documented at the Blur Busters thread I posted.

There are some hidden variables, but it's obvious that most of them are being held constant, and others remain constant among most display types. Even though we don't know everything about its operation, it still appears to be a reliable baseline.

Of course I would always suggest taking multiple readings and collecting more data than one thinks is needed - estimated running time, temperature come to mind - but generally speaking people have not seen many fluctuations over many readings with the Bodnar tester, and what is known about its basic methodology certainly doesn't suggest there are many uncontrolled variables.

[Edward_Tz]

Almost nobody cares about CRTs as a baseline - even amongst gamers - they want to know actually how laggy it is. One direct latency measurement comes closer to realizing this wish than a comparison against unknown equipment.

Personally, I'm concerned with how much slower a display is from a CRT. So if the Bodnar test is to be taken as the "direct latency" then it's safe to assume that if you want to know how much lag a TV has over a CRT you can just subract about 8ms from the Badnar reading.

[Ed Oscuro]

I knew that comment would cause trouble. When I say "almost nobody" I (obviously) mean amongst the population that is making lag measurements, not the Shmups Forum regulars. In fact, I know that many people who browse this Forum are interested in LCD TV quality, but not so much in using CRTs as a baseline. It just happens that CRTs represent "the way it's meant to be played," so interesting from a fidelity / preservation standpoint. CRTs add in some confusion because phosphors are nearly instantly responsive to energy input, but then light output decays when the signal is decreased. In other words, a CRT might be fast on the incoming signal, but adds some confusion after.

So, can you just subtract 8ms? Maybe, but...

This older Anandtech article page gives a pretty good introduction to sources of input lag. Pay attention to scanout. Right off the bat, it's assumed that the Bodnar starts to send a frame and starts the timer, and then the display starts to receive the display. So, halfway down a 1/60th second frame (on a 60Hz display) you're at 8.3333... seconds into the timing cycle, if the display doesn't latch the full frame before displaying it (like many plasma panels do). This doesn't strike me as being substantially different than what would happen with a CRT - of course, if they follow a similar scanning convention as CRTs. Of course, LCDs and other flat panels do not have to follow the line-by-line scanning convention, which throws a monkey wrench in the works.

There is some information on how different displays can update differently at AVS Forum. It is a fancier way of saying what Leo Bodnar has stated - plasma in general must have no less than 16ms latency (for 60Hz), and the timer is apparently set up that halfway down a frame the timer will be no less than 8.3333... ms.

Anyhow, to elaborate based on a bit of conjecture and fact:

With the camera method and enough samples, there is no bias against awful panels that are blurry and hard to read, so long as they are legible enough that a tester can guess what the statement is. But here is one obvious methodological problem: One tester might demand legibility of the new digits, whereas another person might count a photograph at a particular time based on guesswork about the panel state and the previous frame. Additionally, some methods don't seem to use a gray-to-gray method, like SMTT reportedly does yellow to blue transitions. Not only does this make it impossible to verify the manufacturer's g2g response time, but it may be a too-lenient method for real use. Unfortunately, due to the inherent inaccuracy of the setup and the still-present response time, we can't use a shorter transition time to try and bypass response time to get the "real" input lag (minus response time), because it's still there - and different panel types can have different response characteristics all the way across the spectrum.

And as I said earlier, the tester takes all those images and usually discards them all, so it's impossible to verify what is actually being measured from the final number.

With the Bodnar tester, it's assumed that it reads pixel brightness past some threshhold - but as Chief Blur Buster states, it's not known what threshhold this is. However, if it is a high threshhold, i.e. 50% or even 100% pixel brightness, then this is biased towards higher-quality screens with lower response time. So the tester is inflexible, but overall I think this is the right choice, even tradeoff if that's what it seems to be, to make: Repeatability and higher quality panels would be my choice.

[Sixfortyfive]

The Leo Bodnar device's measurements are simple to interpret once you understand that a 60Hz CRT clocks in at 7ms to 8ms with that method instead of zero.

[Edward_Tz]

Are we interested in blur or lag here? Because there's separate tests for both.

If the Bodnar tester only cares for a bright clear image then it can hardly be called accurate.

[Ed Oscuro]

Are we interested in blur or lag here? Because there's separate tests for both.

If the Bodnar tester only cares for a bright clear image then it can hardly be called accurate.

The point of display lag tests is to try and provide one single number to characterize a display, but of course display quality always plays a role in how quickly a new frame is displayed on the screen, and in turn on how that can be interpreted. And this is a lot more complicated than a simple number - and some of that complication gets in the way of understanding the essential characteristics of a display. Ultimately, a choice has to be made about how to weigh or consider the information coming from a lag test. It's totally unjustified to say that the Bodnar "only cares for a bright clear image" because it is still using that to read data.

In fact I might be a bit hard on the Bodnar. Take for example two displays which we want to compare:
One has slow-responding pixels, and no other sources of latency
Another has fast-responding pixels, and no other sources of latency

A person attempting to interpret a photo test might have more difficulty interpreting the first display, but so long as its pixels have started to transition enough to be visible at the same time when the second display's pixels are completely transitioned, they would assign the same latency value. If - it's not clear it happens this way - the Bodnar credits the second display's faster pixel transitions to full brightness as indicating a faster display, I'd say it is not misleading us - because that second display really is faster at doing what the Bodnar tester is actually testing. The photo test attempts to be agnostic about this issue, but that can be misleading.

We all must agree that the Bodnar is usually quite precise - defined as the consistency of results. On the other hand, photo tests are undisputably not. Taking multiple photos will not give a photo test better results than the Bodnar, as the Bodnar tester can always be used multiple times as well, and we already begin with a greater degree of confidence in its results.

As for accuracy - defined as closeness to the true value - both methods must choose when to interpret visual data as signifying a new frame. Without further elaboration, "photo test" could mean any number of things, ranging from quite terrible to better, but probably still not as good as the Bodnar tester. With old stopwatch photo methods, it is quite possible to get confusion about the state of any image based on how long an old pixel transition remains, and also based on whether the image being looked at represents new pixels or old ones. SMTT improves the situation considerably because it stretches the stopwatch readings up and down the screen, and also (in later versions, of which screenshots are not common) by putting the columns on both sides of the screen, which should help reduce overlapping (but it doesn't guarantee its elimination or perfectly readable screens, nor does it guarantee that you're reading at the same brightness, which causes uncertainty especially in older stopwatch methods). It also helped fix or at least reduce a lot of problems caused by naive users thinking that "cloned" displays were really in-sync when they were not; with SMTT you're instead instructed to read the most up-to-date numbers, no matter where they are on each screen. But still the author admits that there may be problems due to the software and hardware it is running on.

The Bodnar tester, by comparison, does force everybody to use the same pattern all the time. This may not always be the most fitting, but since it uses a g2g transition with white patches it is covering the most important transitions, instead of having some weird transitions like the yellow and blue of (old?) SMTT.

SMTT isn't available or updated anymore, which does put a cramp in its use and suitability for new systems.

Ideally, the best lag test would characterize a number of relevant points in displaying an image:

- When the image starts drawing, even at low brightness
- When the image brightness reaches certain threshholds (50% brightness, full brightness)
- The quality of pixel decay, to ensure that old information is not delaying or obscuring parts of the new frame

You'd also have to characterize this for different colors and different transitions, as well.

The Bodnar tester simply throws away much of this as a complicating factor that'd be hard to interpret correctly - an assumption I feel is justified when attempting to boil things down to a simple number. In a perfect world we could use the Bodnar tester's methodology to test for 144Hz latency, for off-color transitions, and for different brightness threshholds - but instead it sticks with a simple and repeatable method. Photo tests don't do things better here, since they add additional steps involving the reproduction and subjective interpretation of image data, rather than the couldn't-be-simpler synchronized clock & scanout -> photodiode reading method of the Bodnar tester.

[Edward_Tz]

The more I read about the Bodnar tester the more I prefer the photo tests. At least photo tests have a control. Bodnar you have nothing but a magic box giving you numbers.

[Ed Oscuro]

As opposed to the black boxes of subjective interpretation leading to numbers that aren't directly comparable?

LOL.

So if the closed box nature of the Bodnar is a problem - it'd be good to know what the threshhold is, but there's no reason to expect it's doing something not kosher - you can always look into the Blur Buster arduino-based lag tester design. There you could tweak it to read whatever it is you think a lag tester is supposed to be reading.

[Edward_Tz]

Blur makes them not directly comparable?

If that's the problem then the picture test with a CRT should be the best thing for you since you have that as your control with zero blur.

[Xyga]

For the lulz, that just popped on [H] http://hardforum.com/showthread.php?t=1853884&page=257

[Sixfortyfive]

Blur makes them not directly comparable?

If that's the problem then the picture test with a CRT should be the best thing for you since you have that as your control with zero blur.

Not every CRT, camera, or timer program is made equal. If you want a good control test, you'd be better off with the LB tester, as that functions the same against every screen you test it on.

[Galgomite]

I believe the P-series 4k Vizio sets all have a gaming mode that's around 17ms through a particular input. The sky's the limit on screen size there.

[Ed Oscuro]

Blur makes them not directly comparable?

If that's the problem then the picture test with a CRT should be the best thing for you since you have that as your control with zero blur.

I already answered this question; I don't want to keep replying if you aren't going to read. Whether it qualifies as a control isn't the issue. Whether it's a good enough control after the appearance of other testing methods is at issue! A CRT screen has unknown phosphor decay time, even if you are using a SMTT style method and using the picture of the rolling scanline to figure out what's the most recent data (assuming you are fortunate enough to get an up-to-date timestamp to read on both displays). In many cases the CRT will not be a perfect mirror of the LCD screen. The PC hardware timer and the OS pose additional potential sources of error.

In the SMTT link I gave above, the developer of that program (which is the best photo comparison tool I know of) says that a photodiode + oscilloscope test is the best. The Bodnar tester takes that method and makes it a one-button press design, and the appearance of that device might be a good part of the reason why SMTT hasn't been updated since 2012.

There's no doubt that there are a few sources of error for the Bodnar tester, but they're known to be small for most screens - both in comparison with known figures and in direct comparisons with SMTT at the time the Bodnar tester first appeared. I already went over some sources of error above, like I said, if you want to consider that instead of just throwing bombs into the conversation.

I also wanted to mention the 8.5 (or 8.3333...ms) issue. So yeah, in theory and in reality the Bodnar tester's mid-screen reading should be no less than roughly this number.

In practice, nobody ever needs to jump through this hoop unless they pose the question "what is this like compared to a CRT?" I think that since we have these databases of accurate and directly comparable numbers, we ought to stick with them instead of immediately throwing out a number they won't see listed anywhere. That's confusing and doesn't help the average reader use the online sources. I suppose this isn't a problem if you're an oracle, though.