Processing Delay of ATEM TV HD: "6 lines" in milliseconds?

Questions about ATEM Switchers, Camera Converter and everything live!
  • Author
  • Message
Offline

Daniel Binder

  • Posts: 179
  • Joined: Fri Jan 30, 2015 9:54 am

Processing Delay of ATEM TV HD: "6 lines" in milliseconds?

PostSat Dec 16, 2017 5:55 pm

I noticed, that the tech specs of the ATEM TV HD mention a processing delay of "6 lines" (compared to 1 line for the ATEM Production 4K and "<2 lines" for the ATEM Production Studio 4K 1M/E)

Could someone clarify for me, what exactly "lines" would mean here and how this would translate into milliseconds? Are we talking about "lines" as in 1080 lines in a full HD frame? (meaning anything less than 1080 lines would be less than 1 frame delay? and 6 lines would be somewhere around 0.3ms?

Or are lines basically fields or half-frames in an interlaced signal?
as in 1080i50 - 6 "half-frames" = 3 frames = 120ms delay?


Further, does anyone know, wether the actual processing delay depends on the use of DVE, DSKs and such?
I found older discussion of the original ATEM series saying multiview +1 frame, DVE +1 frame.

I'm trying to find out if the ATEM TV HD is in any way suitable for a live setup where on-location projection would be needed too. 120ms would be definitely too much.

Thanks for any input on this,

cheers,

Daniel

(edited)
Offline
User avatar

Thomas Seewald

  • Posts: 1165
  • Joined: Fri Aug 24, 2012 11:57 am

Re: Processing Delay of ATEM TV HD: "6 lines" in millisecond

PostSat Dec 16, 2017 7:44 pm

Lines are neither frames, nor half-frames -really lines.
Be sure, your cameras and the projector causes much more latency than the ATEM. Did you ever has seen any procetion without noticeable delay ?
thos-berlin - Thomas S e e w a l d
Offline

Daniel Binder

  • Posts: 179
  • Joined: Fri Jan 30, 2015 9:54 am

Re: Processing Delay of ATEM TV HD: "6 lines" in millisecond

PostSat Dec 16, 2017 8:05 pm

thanks for the reply. so the 2 lines or 6 lines delay is negligible for the intended purpose (less than 1 ms), correct?

I'm fully aware of the latency the cameras (in my case about 50ms already) and projectors are causing. that's why I really don't need the switcher to add significantly more to it. I really don't want to end up with a very much noticeable 200ms delay or more.
I know, I can get away with 2-3 frames total, but not 6-8 frames :)

cheers,

Daniel
Offline
User avatar

Xtreemtec

  • Posts: 5387
  • Joined: Wed Jan 02, 2013 11:48 am
  • Location: The Netherlands

Re: Processing Delay of ATEM TV HD: "6 lines" in millisecond

PostMon Dec 18, 2017 1:57 pm

Were did you find the lines??

As it always have been 1 frame delay for all Atem mixers, (in ideal situation, if sources are all genlocked.)
If you have sources that are not genlocked it can have a almost full frame extra for the Framebuffer.)

And if you use a DVE you have to add 1 frame extra latency.

This has been always the thumb rule of delay in the Atem. So i really wonder were the new info about lines come from.. ;)

So in general,
1 frame is 20ms (@ 50FPS)
1 frame is 16,7ms (@ 60FPS)
1 frame is 40ms (@ 25FPS)
1 frame is 33ms (@ 30FPS)

So general video delay in the Atem mixer is between 20 / 60ms depending on your setup.

To note. A camera, depending on model will take about 3 frames latency.. And a monitor, projector or ledwall can go up to 8 frames is worst case scenario.. ;)
Daniel Wittenaar .:: Xtreemtec Media Productions ::. -= www.xtreemtec.nl =-
4K OBV Trailer, ATEM TVS HD, 4M/E Broadcast Studio 4K, Constelation 8K, Hyperdeck Studio 12G, Ursa Broadcast 4K, 4K fiber converters with Sony Control
Offline

Asgeir Hustad

  • Posts: 333
  • Joined: Tue Sep 26, 2017 10:13 am
  • Location: Norway

Re: Processing Delay of ATEM TV HD: "6 lines" in millisecond

PostMon Dec 18, 2017 2:35 pm

Xtreemtec wrote:Were did you find the lines??

As it always have been 1 frame delay for all Atem mixers, (in ideal situation, if sources are all genlocked.)
If you have sources that are not genlocked it can have a almost full frame extra for the Framebuffer.)

And if you use a DVE you have to add 1 frame extra latency.

This has been always the thumb rule of delay in the Atem. So i really wonder were the new info about lines come from.. ;)


Are you sure that there is 1 frame delay even when genlocked? That's not my information, as far as I know, there is up to 1 frame delay for frame sync if not genlocked, some lines for general mixing, one more frame if DVE is enabled, and then the rest is determined by cameras and monitor/projectors.

The official specs also say <2 Lines Processing Delay on the newer mixers. I've gotten my delay-chain all the way down to 30ms from camera to projected image when using only 1 camera, ie. no framesyncs involved, at 720p50. This is 1 1/2-ish frames, including camera and projector!
Offline
User avatar

Thomas Seewald

  • Posts: 1165
  • Joined: Fri Aug 24, 2012 11:57 am

Re: Processing Delay of ATEM TV HD: "6 lines" in millisecond

PostMon Dec 18, 2017 3:56 pm

I remember too, that use of genlocked souces causes some lines and non genlocked sources up to 1 frame delay. But in the manual I found nothing ....

It makes no sense to have a delay of one frame when genlocked...
thos-berlin - Thomas S e e w a l d
Offline
User avatar

Xtreemtec

  • Posts: 5387
  • Joined: Wed Jan 02, 2013 11:48 am
  • Location: The Netherlands

Re: Processing Delay of ATEM TV HD: "6 lines" in millisecond

PostMon Dec 18, 2017 4:16 pm

Well could be a few lines.. But it needs processing. And you can genlock it. But you will not know if you have it genlock running. In your camera you need to set pre or post genlock to sync up to the atem.

Problem is the Atem will not tell you if you have it perfect sync or running 20 lines 2 lines or 2 pixel before or after the internal frame buffer. :lol: :lol:

This means that you can not tell if you are using the minimum delay or are just 2 pixels away from using a complete 1 frame framebuffer..... As the framebuffer is always on. You will always have a picture..
In the old days if you were not sync you would not have a picture at all or have garbage on the screen.

So determening acurate sync is difficult. Also depends on a lot of factors.. If you have fiber converters and a router between it. The sync settings are different then direct cable connections.. All because of reclockers in every SDI device that take just a tiny little bit of time to straighten up the signal.. But if you have enough of them in a line it will trow off your sync. ;)

BUT now a fun fact. So take the Example of 50Fps. That is 1 frame is 20ms. Audio travels with 360 meters per second. There are 50 frames in a second. So we can say 360/50 = 7,2 meters per frame.
Audio travels 7,2 meters per frame!

That means that if you have 3 frames delay in your video. And you are sitting 21,6 meters away from the screen you hear and see in perfect sync. If you are at 7 meters from the screen you hear the audio 60ms earlier then video.. if you are sitting at 43 meters your audio will be 60ms behind the video.

If your venue is long enough (Stadiums, big churches, etc etc.. ) And you sit at the far end of the audio setup. 86m = 120ms delay 143m = 200ms delay.
Audio delay stacks will be placed at such big events.. All delayed with processors to be sure you don't create a delay and so introduce an echo...
Now the most fun fact were every AV company will go wrong.. They place extra screens so the spectators can see it better. But they forget to delay the video.... :lol: :lol:
The result is a completely OFF lipsink..
Daniel Wittenaar .:: Xtreemtec Media Productions ::. -= www.xtreemtec.nl =-
4K OBV Trailer, ATEM TVS HD, 4M/E Broadcast Studio 4K, Constelation 8K, Hyperdeck Studio 12G, Ursa Broadcast 4K, 4K fiber converters with Sony Control
Offline

Gary Adams

Blackmagic Design

  • Posts: 1383
  • Joined: Sat Aug 25, 2012 6:14 am

Re: Processing Delay of ATEM TV HD: "6 lines" in millisecond

PostMon Dec 18, 2017 4:58 pm

This is always an interesting topic to me. The length of one line of video is 15 microseconds here in the US so line delays are not noticeable. While the processing delay of an ATEM ranges from 1-6 lines depending on model and mode, it's important to understand that sources that are not genlocked will vary in time through the ATEM since they are not locked together. This time will slowly vary from lines to a frame (max). To prevent this variance and indeed lock the latency through any ATEM, the source must be locked to the same reference as the ATEM. AND (this is important), the timing of that locked source must be a few lines in advance of the ATEM to be ahead of the processing. If the vertical of the source reaches the ATEM later than the ATEM vertical, the ATEM (or any switcher) must delay that image until the next vertical. So it would be true that a genlocked source could have a frame delay through the switcher if the timing is not very slightly advanced. Many genlocking sources (like our studio type cameras) provide the ability to adjust the advance timing for this purpose. There are sync generators that provide this ability as well. While the ATEMs do not provide indication of lock or delay, I find a simple verification is to feed the video signal to one of the Smartview Duo monitors and the ATEM output to the other. It is very easy to see if they are the same or delayed by a frame by eye or using a phone camera. They could be any pair of identical monitors for this test. Hope this helps.

Regards, Gary
Gary Adams
Blackmagic Design
Offline

rogersuski

  • Posts: 234
  • Joined: Wed Aug 29, 2012 12:37 am

Re: Processing Delay of ATEM TV HD: "6 lines" in millisecond

PostMon Dec 18, 2017 5:26 pm

Gary, would it be *technically* possible for the ATEM to provide feedback of "frame buffer in use" or "ext lock" for various sources? That would be a great benefit, and consider it a feature request.
Offline

Gary Adams

Blackmagic Design

  • Posts: 1383
  • Joined: Sat Aug 25, 2012 6:14 am

Re: Processing Delay of ATEM TV HD: "6 lines" in millisecond

PostMon Dec 18, 2017 5:37 pm

We put that in as a feature request. But we cannot make any promises.

Regards, Gary
Gary Adams
Blackmagic Design
Offline

Daniel Binder

  • Posts: 179
  • Joined: Fri Jan 30, 2015 9:54 am

Re: Processing Delay of ATEM TV HD: "6 lines" in millisecond

PostWed Dec 20, 2017 7:35 pm

Thanks a lot for the clarification.

Happy holidays, everyone!

Daniel
Offline
User avatar

Xtreemtec

  • Posts: 5387
  • Joined: Wed Jan 02, 2013 11:48 am
  • Location: The Netherlands

Re: Processing Delay of ATEM TV HD: "6 lines" in millisecond

PostThu Dec 21, 2017 9:41 am

rogersuski wrote:Gary, would it be *technically* possible for the ATEM to provide feedback of "frame buffer in use" or "ext lock" for various sources? That would be a great benefit, and consider it a feature request.

+100 for this.. Because this gives the user a good feedback in how far the framebuffer is in use. ;)
Daniel Wittenaar .:: Xtreemtec Media Productions ::. -= www.xtreemtec.nl =-
4K OBV Trailer, ATEM TVS HD, 4M/E Broadcast Studio 4K, Constelation 8K, Hyperdeck Studio 12G, Ursa Broadcast 4K, 4K fiber converters with Sony Control
Offline

strulala

  • Posts: 2
  • Joined: Mon Aug 03, 2020 8:29 am
  • Real Name: Leopold Strobl

Re: Processing Delay of ATEM TV HD: "6 lines" in millisecond

PostMon Aug 10, 2020 5:49 pm

Gary Adams wrote:We put that in as a feature request. But we cannot make any promises.

Regards, Gary


Are there any news here, Gary?
Offline

Gary Adams

Blackmagic Design

  • Posts: 1383
  • Joined: Sat Aug 25, 2012 6:14 am

Re: Processing Delay of ATEM TV HD: "6 lines" in millisecond

PostMon Aug 10, 2020 10:54 pm

The request has been noted, but of course, we cannot speak for the future. Stay safe.

I'll use this space to say that all ATEMs have about 4-10 lines of processing delay. This means with a genlocked input that is timed roughly greater than 10 lines in advance of the ATEM will have a 10 line latency. This is a very short period of time. While not scientific you can calculate this time by figuring 59.94 frames per second times 1080 lines per frame. Then 1/x or invert and you will get about 15 microseconds per line. 10 of those would be 150 microseconds. I would consider this latency pretty low. If the signal is not advanced by the processing delay, the ATEM will need to delay it until the next frame. This of course will vary if the signal is not genlocked. So the latency of all ATEM (without special modes like DVE) will vary between about 10 lines to a frame. That also is pretty low latency. What we generally see of course is the build up of delays caused by cameras, scalers, converters and monitors in addition to the 10 lines to a frame. I hope this helps.

Regards, Gary
Gary Adams
Blackmagic Design

Return to Live Production

Who is online

Users browsing this forum: No registered users and 33 guests