Leo's exciting hand held LAG TEST device - out now !

The place for all discussion on gaming hardware
User avatar
marqs
Posts: 1034
Joined: Sat Dec 15, 2012 12:11 pm
Location: Finland

Re: Leo's exciting hand held LAG TEST device - out now !

Post by marqs »

Four Wude wrote:Hello everyone, I run Display Lag and I wanted to gather your feedback. As of right now I've personally graded 148 displays and continue to grade as many as I can. I received an e-mail from marqs and I figured I'd chime into this community.

First, I'd like to note something about that Sony you guys are having issues with. Here in the US, my friend has a Sony TV from the same/similar model line as the one that's causing issues for you guys, and it exhibited the same behavior that you're experiencing with 0ms on the middle bar. It was the only HDTV I have graded (even among Sony's 2012 and 2013 models) that caused such an issue. I'm not sure what is causing it, therefore I chose not to report the display in my database as it seemed sketchy. So don't be too quick to think your tester is broken, for some reason that specific Sony model line acts weird with the tester.

I'd also like to gather opinions on the differences between the top, middle, and bottom bars. As some of you know, all ratings in my database are calculated from the bottom bar. I decided to adopt the same standard that AVForums use as it was the first authority website to use the Lag Tester in their reviews; they use the bottom bar. I was also recommended by John Beeson (Leo Bodnar's associate) to use the highest ratings I can get as a reference (which are typically found on the bottom bar).

There has been some discussion about using the top bar instead of the bottom, based on what marqs emailed me. Unfortunately, I only started recording all 3 ratings on the last 55 displays inserted into my database (which leaves a good 93 displays with only bottom bar results).

Please advise, and thanks for the support! I don't intend to mislead anyone and want everyone to be on the same page. If you have any questions, feel free to ask.

- Adeel
Thanks for joining the forum!

We've been trying to figure out an accurate way of measuring the latency here. I think we should ask Leo himself about when exactly the counter is started, which would hopefully clear some confusion. If the counter was really started after the complete frame (showing the bars) was sent to display, it would mean that almost all flat displays (except the previously mentioned Sonys) would have at least 1 frame of latency. I'm not ruling that possibilty out yet, but I find it hard to believe (and which would be quite depressing if true :( )
Fudoh wrote:I would be really interested in understanding the technical background, why some displays would show differences in the readings when using different bars, while other displays will give the same results no matter which bar is used.
Yeah, the refresh logic of flat panels is an interesting topic in itself.
Fudoh wrote:I can understand a certain difference (in the single ms range). That's just the time the display needs to draw the whole screen, so you get subtle brightness differences which show in the readings, but how can a display show a whole frame of difference between the top and bottom bar ? Imagine the input signal not being a static screen like it is now, but a horizontally scrolling game. A time gap of up to 16ms between the top and the bottom of the screen would cause serious tearing - to a degree where the display would not be useable anymore for any kind of gaming...
Good point. However, we all probably agree that this happens on a CRT where the input controls the ray directly, yet there is no visible tearing - right? How I see this is possible, is that the tearing between 2 consecutive scanlines only lasts the draw duration of one scanline (~33us at 480p@60Hz), after which it moves one scanline down. In contrast, tearing in e.g. a non-vsynced PC game lasts the duration of the whole frame. It'd great if someone with a high-speed camera could verify all this. I also thought earlier that all flat panels would draw the frame instantly, but the various results indicate that it's not always this way (which is a good thing - at least from latency perspective).
User avatar
BuckoA51
Posts: 3358
Joined: Sat Oct 02, 2010 10:08 am
Location: Ireland
Contact:

Re: Leo's exciting hand held LAG TEST device - out now !

Post by BuckoA51 »

First, I'd like to note something about that Sony you guys are having issues with.
Just to clarify, do you mean the KDL-xxEX403 series or the KDL-xxW4000 series?

Also, would you consider taking submissions for your lag database from other users?
OSSC Forums - http://www.videogameperfection.com/forums
Please check the Wiki before posting about Morph, OSSC, XRGB Mini or XRGB3 - http://junkerhq.net/xrgb/index.php/Main_Page
User avatar
Ed Oscuro
Posts: 18654
Joined: Thu Dec 08, 2005 4:13 pm
Location: uoıʇɐɹnƃıɟuoɔ ɯǝʇsʎs

Re: Leo's exciting hand held LAG TEST device - out now !

Post by Ed Oscuro »

Are people having problems with getting readings from just the one model, or the one type (chassic, panel type inclusive)? If it's as widespread as it seemed then I have to say I don't have much confidence in the unit. If it is one model, I still wonder if the methodology isn't good enough because there are other methods which seem to work quite well in measuring display lag, which aren't known to suffer from any system-based effects. There might be some systematic issues to work out in some other testing types but they do promise to give consistent results regardless of the panel quirks.
User avatar
BuckoA51
Posts: 3358
Joined: Sat Oct 02, 2010 10:08 am
Location: Ireland
Contact:

Re: Leo's exciting hand held LAG TEST device - out now !

Post by BuckoA51 »

Managed to get some readings from the KDL-32W4000. As for what was throwing off the tester for the past few days, I have no idea. I seem to get best results in a dimly lit room with the contrast on the TV turned up a little.

Top bar - 4.2ms
Middle bar - 11.3ms
bottom bar - Fluctuates between 19 and 22ms

One heck of a discrepancy...

In light of this, I recommend when playing Shmups, to always fly near the top of the screen whenever possible.
OSSC Forums - http://www.videogameperfection.com/forums
Please check the Wiki before posting about Morph, OSSC, XRGB Mini or XRGB3 - http://junkerhq.net/xrgb/index.php/Main_Page
User avatar
marqs
Posts: 1034
Joined: Sat Dec 15, 2012 12:11 pm
Location: Finland

Re: Leo's exciting hand held LAG TEST device - out now !

Post by marqs »

I got an answer from Leo about the counter, which confirmed my earlier assumptions. The counter is started when the center of the top bar is transmitted to the display. To measure display latency in regard to the 60Hz input signal, one has to subtract 7.5ms (mid bar) or 15ms (bottom bar) from the corresponding result.

The bottom-bar reading as-is includes signal latency, which depends on the input device. Any framebuffer-based and vsynced inputs (practically all modern console games) have almost one-frame signal latency at the bottom of the picture, as BuckoA51 commented. However, I wouldn't count that in the display latency.
User avatar
BuckoA51
Posts: 3358
Joined: Sat Oct 02, 2010 10:08 am
Location: Ireland
Contact:

Re: Leo's exciting hand held LAG TEST device - out now !

Post by BuckoA51 »

So you're telling us the true "input lag", as in the amount of time between a frame being sent and being displayed, is the top bar and the rest is added as signal latency from the tester?

I still don't fully understand this.
OSSC Forums - http://www.videogameperfection.com/forums
Please check the Wiki before posting about Morph, OSSC, XRGB Mini or XRGB3 - http://junkerhq.net/xrgb/index.php/Main_Page
User avatar
Ed Oscuro
Posts: 18654
Joined: Thu Dec 08, 2005 4:13 pm
Location: uoıʇɐɹnƃıɟuoɔ ɯǝʇsʎs

Re: Leo's exciting hand held LAG TEST device - out now !

Post by Ed Oscuro »

This makes sense to me. Even a (highly serial) digital signal will have some time elapse during the sending of a signal, so if it is being timed from the beginning of that signal send, one cannot expect it to be done any faster, and while it might be nice to have displays that hurry up and burst-transmit each frame for faster display on low-latency displays, currently it is still difficult enough just to produce enough bandwidth to get frames out in time for higher resolution displays.

To make sense of this one only needs to think about how the various portions of a CRT screen never appear "out of sync" despite only one line being drawn at any time. Although the scanning line effect is pronounced and notable in many situations, one doesn't get the feeling that there is significant lag down the screen - even though that portion is in a very real sense "out of date" compared to the top. And of course the signal is understood to be sent continuously.
User avatar
marqs
Posts: 1034
Joined: Sat Dec 15, 2012 12:11 pm
Location: Finland

Re: Leo's exciting hand held LAG TEST device - out now !

Post by marqs »

BuckoA51 wrote:So you're telling us the true "input lag", as in the amount of time between a frame being sent and being displayed, is the top bar and the rest is added as signal latency from the tester?

I still don't fully understand this.
I'd call "input lag" the whole time between user input (e.g. a button press) to a related action/change on the screen. Since this includes various factors (e.g. button poll interval, source processing latency, display latency), it's harder to quantify and measure. The whole term "input lag" is not unambigious, since it's not always clear what is referred as the "input". Display lag on the other hand is easier to define and measure (at least in my opinion), since you have a clearly defined input signal (for display) and output.

Perhaps signal latency is not the best term, but I meant the time where your controller input does not affect the outgoing video signal. This is always a concern in a vsynced game, because the framebuffer can be updated only at pre-defined points, e.g. every 16.6ms with 60Hz output. Once the frame is in the framebuffer, no matter what your controller input is, the outgoing video signal is unaffected until the next vertical blanking interval. This is where the almost 1-frame latency at the bottom of the picture originates, because it takes that time to transmit the video (where the bottom-right dot/pixel is sent and received last). Compare this to game running without vsync, which is free to update the framebuffer at any time. This time the controller input can theoretically change the video signal at any time (assuming the polling and game are fast enough), resulting in no added latency - but with the price of tearlines.

BTW, I built an updated version of VGA lag tester for AVR microcontroller a couple of days ago. I try to finish with instructions and schematics tonight, and put them out for those who are interested testing with analog sources.
User avatar
Ed Oscuro
Posts: 18654
Joined: Thu Dec 08, 2005 4:13 pm
Location: uoıʇɐɹnƃıɟuoɔ ɯǝʇsʎs

Re: Leo's exciting hand held LAG TEST device - out now !

Post by Ed Oscuro »

In most games today there seems little enough call for scanline effects anymore (any distortions or similar tricks could be achieved in other ways with more control now) and tearing is a significant enough disincentive in 3D games especially.
User avatar
BuckoA51
Posts: 3358
Joined: Sat Oct 02, 2010 10:08 am
Location: Ireland
Contact:

Re: Leo's exciting hand held LAG TEST device - out now !

Post by BuckoA51 »

So if I've understood this right, the reason that some TV's show this discrepancy and others don't is that some TV's always have at least a 1 frame buffer which then evens things out so to speak, whereas the faster ones do not?
OSSC Forums - http://www.videogameperfection.com/forums
Please check the Wiki before posting about Morph, OSSC, XRGB Mini or XRGB3 - http://junkerhq.net/xrgb/index.php/Main_Page
User avatar
marqs
Posts: 1034
Joined: Sat Dec 15, 2012 12:11 pm
Location: Finland

Re: Leo's exciting hand held LAG TEST device - out now !

Post by marqs »

BuckoA51 wrote:So if I've understood this right, the reason that some TV's show this discrepancy and others don't is that some TV's always have at least a 1 frame buffer which then evens things out so to speak, whereas the faster ones do not?
Yes, it is possible for a display to even things out by using a frame buffer if it can refresh the screen faster than the input signal. The in-display frame buffer is actually required if the internal refresh rate differs from the signal refresh rate. You can think e.g. an ideal 600Hz plasma fed with a 60Hz input: it takes 16.6ms to fill the buffer with a single input frame, after which the whole screen is updated almost immediately. The screen is then updated 9 more times with the same buffer contents (no interpolation), after which the next input frame is ready to be read from a secondary frame buffer (can't use the original buffer as it'd contain data from both input frames while read).

Note that the difference is in how fast buffer is read and output to the screen. A simple display may also have a frame buffer, of which contents are output at the same rate as the input signal, resulting in constant signal-to-display latency without evening out the difference between top and bottom portion on the screen. The display discrepancies indeed cause some confusion, but that's small compared to the analysis of whole input latency chain, including the console/PC framebuffer I mentoined in the earlier post :shock: .
marqs wrote:BTW, I built an updated version of VGA lag tester for AVR microcontroller a couple of days ago. I try to finish with instructions and schematics tonight, and put them out for those who are interested testing with analog sources.
Got that finally in a publishable condition - here. Please comment if you think something is missing or I've got something wrong.
NWrain
Posts: 242
Joined: Wed Oct 10, 2012 3:50 am
Location: NTSC-US

Re: Leo's exciting hand held LAG TEST device - out now !

Post by NWrain »

Okay, I'm not sure I understand this 100%. For example, if a monitor using the lag tester measured 2ms on the top bar and 20ms on the bottom bar, does it mean that it is taking 20ms to complete one frame 16.6ms?
User avatar
Artemio
Posts: 648
Joined: Tue Jun 09, 2009 12:55 am
Location: Mexico
Contact:

Re: Leo's exciting hand held LAG TEST device - out now !

Post by Artemio »

marqs wrote:BTW, I built an updated version of VGA lag tester for AVR microcontroller a couple of days ago. I try to finish with instructions and schematics tonight, and put them out for those who are interested testing with analog sources.
Got that finally in a publishable condition - here. Please comment if you think something is missing or I've got something wrong.[/quote]

Quite interesting. Before the HDMI lag tester was released I was working in a flawed idea using the dreamcast 240p test suite, by using the DC microphone and FFT I did implement a routine that measures the lag. In case the system lags the audio to match the video, it works with great accuracy for such a setup (1/4th of a frame if I remember correctly). I was developing the same thing using the DC video camera, just for fun.. but it is too slow and not well supported in kallistiOS yet.. and since Leo´s tester was released I didn't go on with that.

However after reading your page and project, I was wondering if this could be implemented in the DC, and in that way measure a full cycle of input lag generated by a game console in either VGA or RGB... through a scaler or directly. It would be a matter of figuring out a way to connect the protodiode to the DC bus... which might not be as easily done.

Anyway, awesome work, thanks for sharing it =)
User avatar
marqs
Posts: 1034
Joined: Sat Dec 15, 2012 12:11 pm
Location: Finland

Re: Leo's exciting hand held LAG TEST device - out now !

Post by marqs »

Artemio wrote:Quite interesting. Before the HDMI lag tester was released I was working in a flawed idea using the dreamcast 240p test suite, by using the DC microphone and FFT I did implement a routine that measures the lag. In case the system lags the audio to match the video, it works with great accuracy for such a setup (1/4th of a frame if I remember correctly). I was developing the same thing using the DC video camera, just for fun.. but it is too slow and not well supported in kallistiOS yet.. and since Leo´s tester was released I didn't go on with that.

However after reading your page and project, I was wondering if this could be implemented in the DC, and in that way measure a full cycle of input lag generated by a game console in either VGA or RGB... through a scaler or directly. It would be a matter of figuring out a way to connect the protodiode to the DC bus... which might not be as easily done.
Sounds definitely possible. Perhaps the photodiode circuit could be connected directly to the interrupt pin of G2 connector. Dunno how hard that'd be to setup and use on SW side. Anyway, DC would be quite ideal console for latency testing as it can output various video modes (480p, 480i/576i, 240p/288p).
User avatar
Artemio
Posts: 648
Joined: Tue Jun 09, 2009 12:55 am
Location: Mexico
Contact:

Re: Leo's exciting hand held LAG TEST device - out now !

Post by Artemio »

marqs wrote:
However after reading your page and project, I was wondering if this could be implemented in the DC, and in that way measure a full cycle of input lag generated by a game console in either VGA or RGB... through a scaler or directly. It would be a matter of figuring out a way to connect the protodiode to the DC bus... which might not be as easily done.
Sounds definitely possible. Perhaps the photodiode circuit could be connected directly to the interrupt pin of G2 connector. Dunno how hard that'd be to setup and use on SW side. Anyway, DC would be quite ideal console for latency testing as it can output various video modes (480p, 480i/576i, 240p/288p).[/quote]


One other option would be modding a DC controller to have a photoresistor in place of one of the triggers http://segaretro.org/Controller_%28Dreamcast%29
User avatar
marqs
Posts: 1034
Joined: Sat Dec 15, 2012 12:11 pm
Location: Finland

Re: Leo's exciting hand held LAG TEST device - out now !

Post by marqs »

Artemio wrote:One other option would be modding a DC controller to have a photoresistor in place of one of the triggers http://segaretro.org/Controller_%28Dreamcast%29
Yeah, that'd work too if the controllers can be read often enough. Do you know if a controller poll in SW actually triggers a read of buttons/axis, or does it return a latched value which is updated once per frame etc?
User avatar
Artemio
Posts: 648
Joined: Tue Jun 09, 2009 12:55 am
Location: Mexico
Contact:

Re: Leo's exciting hand held LAG TEST device - out now !

Post by Artemio »

marqs wrote:Yeah, that'd work too if the controllers can be read often enough. Do you know if a controller poll in SW actually triggers a read of buttons/axis, or does it return a latched value which is updated once per frame etc?
I believe it is the second case, good enough for a 1 frame precision measurement. I've never tested reading the port several times within a frame and comparing those... But that is something easy to code a test for.
lostinblue
Posts: 26
Joined: Sun May 19, 2013 12:17 pm

Re: Leo's exciting hand held LAG TEST device - out now !

Post by lostinblue »

Can someone test GBS8220 lag via leo bodnar method?

I understand it is 33 ms via the quick camera method, but there seems to be a huge discrepancy between both of them; seeing tv's like Panasonic ST50's rated 16 ms last year via the older method are 47ms on it.

I ask because XRGB-mini appears as lagging 24 ms here, so I'm betting GBS8220 should be closer to 50/60 ms via this method or so?

It's a very frequent entry point board, so knowing how it compares is somewhat important in establishing a difference between entry and prosumer offerings; up till now people would say 33 ms is a lot and that XRGB-mini introduced "virtually no lag".


Just idle curiosity on my account though.
User avatar
marqs
Posts: 1034
Joined: Sat Dec 15, 2012 12:11 pm
Location: Finland

Re: Leo's exciting hand held LAG TEST device - out now !

Post by marqs »

lostinblue wrote:Can someone test GBS8220 lag via leo bodnar method?

I understand it is 33 ms via the quick camera method, but there seems to be a huge discrepancy between both of them; seeing tv's like Panasonic ST50's rated 16 ms last year via the older method are 47ms on it.
Are you sure that you're not mixing up ST50 & ST60? Panasonic outsourced the video processor in this year's ST60, which apparently has more lag than the previous ST models.

33ms sounds quite much for a device which doesn't even offer digital output (though it does the conversion in digital domain).
User avatar
Fudoh
Posts: 13015
Joined: Mon Mar 06, 2006 3:29 am
Location: Germany
Contact:

Re: Leo's exciting hand held LAG TEST device - out now !

Post by Fudoh »

GBS8220 uses an adaptive deinterlacer for 480i signals, so it buffers 2 frames, hence the 33ms (or so) delay.
User avatar
marqs
Posts: 1034
Joined: Sat Dec 15, 2012 12:11 pm
Location: Finland

Re: Leo's exciting hand held LAG TEST device - out now !

Post by marqs »

NWrain wrote:Okay, I'm not sure I understand this 100%. For example, if a monitor using the lag tester measured 2ms on the top bar and 20ms on the bottom bar, does it mean that it is taking 20ms to complete one frame 16.6ms?
Yes, alhough that should be possible only with 50Hz input signal, not with 60Hz the tester is using.

I drew 2 example timing diagrams based on my measurements and previous talk of display latency. Hope that somewhat helps in understanding the issue and display differences:
Eizo FS2333 latency diagram
Panasonic VT30 latency diagram.
Leo's tester (and mine) has the reference (zero) point close to the first red marker. The colored markers on the lower timeline can be though as the top and bottom bar triggering points.
lostinblue
Posts: 26
Joined: Sun May 19, 2013 12:17 pm

Re: Leo's exciting hand held LAG TEST device - out now !

Post by lostinblue »

marqs wrote:Are you sure that you're not mixing up ST50 & ST60? Panasonic outsourced the video processor in this year's ST60, which apparently has more lag than the previous ST models.
No, they retested them with Leo Bodnar and results came back at 47 ms.

ST60 fares way worse; 74.5ms on Leo Bodnar. ST60 is infact 47 to 62 ms via quick camera method, though.

Yes, they're using mediatek on ST60, but it's still weird because S60 should have the same innards and has way less lag. And X60 are almost as silky smooth as X50's.
Fudoh wrote:GBS8220 uses an adaptive deinterlacer for 480i signals, so it buffers 2 frames, hence the 33ms (or so) delay.
I see, that's unchanged then.

I've seem numbers being very bloated up by this new method so I was wondering.
User avatar
Fudoh
Posts: 13015
Joined: Mon Mar 06, 2006 3:29 am
Location: Germany
Contact:

Re: Leo's exciting hand held LAG TEST device - out now !

Post by Fudoh »

I've seem numbers being very bloated up by this new method so I was wondering.
I've been wondering the same. Possibly a problem with the way the stop-watch software is running. I've been doing my lag tests with Artemio's 240p suite running on an Everdrive on a Nomad with stock-panel. The whole setup has been calibrated to reflect the Nomad's panel lag (by running the Nomad against a CRT). My results have always been +/-2ms the same as with Leo's device.
User avatar
parodius
Posts: 720
Joined: Wed Jan 26, 2005 5:54 am
Location: Singapore

Re: Leo's exciting hand held LAG TEST device - out now !

Post by parodius »

Seems it's back in stock.

What are the pros & cons of ordering a 1080p or 720p configuration ?
Do most screens have different lag for those 2 configurations ?
My sales thread : 2020/07/20..MASTER.VER.
User avatar
akumajo
Posts: 542
Joined: Thu Sep 04, 2008 12:37 am

Re: Leo's exciting hand held LAG TEST device - out now !

Post by akumajo »

look at KDL W905A / KDL W805A LED TV's from SONY, they are 20ms and big :)

newest plasma's you can forget, too much processing ...
User avatar
marqs
Posts: 1034
Joined: Sat Dec 15, 2012 12:11 pm
Location: Finland

Re: Leo's exciting hand held LAG TEST device - out now !

Post by marqs »

akumajo wrote:look at KDL W905A / KDL W805A LED TV's from SONY, they are 20ms and big :)

newest plasma's you can forget, too much processing ...
I wonder if you have to use Game mode to get that low, and if that hurts PQ.

F5500 (Samsung) and ZT/VT/GT 6X (Panasonic) plasmas still seem to maintain 30-40ms range from earlier models.
User avatar
Fudoh
Posts: 13015
Joined: Mon Mar 06, 2006 3:29 am
Location: Germany
Contact:

Re: Leo's exciting hand held LAG TEST device - out now !

Post by Fudoh »

I wonder if you have to use Game mode to get that low, and if that hurts PQ
yes, you have and not it does not. But it disables the motionflow settings and this means that the motion resolutions drops to something around 400 lines (compared to 1080 on a plasma set).
User avatar
slk486
Posts: 400
Joined: Mon Feb 28, 2011 11:42 pm
Location: Denmark

Leo's exciting hand held LAG TEST device - out now !

Post by slk486 »

akumajo wrote:look at KDL W905A / KDL W805A LED TV's from SONY, they are 20ms and big :)

newest plasma's you can forget, too much processing ...

The W9's are tested to 8 and 10-16 from what I've seen.
User avatar
brentsg
Posts: 2303
Joined: Sun Jul 12, 2009 9:01 pm
Location: St. Louis, MO USA

Re: Leo's exciting hand held LAG TEST device - out now !

Post by brentsg »

I received my lag tester today so I'm working my way thru devices.

My Pioneer Kuro display (PDP-5010FD) is measuring at 53.8ms regardless of whether game mode is on or off.

I tested an Alienware AW2310, which is a 120Hz gaming monitor. It's measuring at no more than 2ms regardless of what I do. This sounds unrealistic, but I can't get any other result.

If I can dig up another HDMI cable I'll test another monitor I've got handy.
Breaking news: Dodonpachi Developer Cave Releases Hello Kitty Game
User avatar
marqs
Posts: 1034
Joined: Sat Dec 15, 2012 12:11 pm
Location: Finland

Re: Leo's exciting hand held LAG TEST device - out now !

Post by marqs »

2ms signal-to-display latency is perfectly realistic for a gaming monitor. That should result to similar readings as with Eizo FS2333 shown in the end of page 3 of this thread (2.2ms, 9.3ms, 16.6ms).
Post Reply