ATEM TVS and 2M/E delay

We have had an ATEM TVS the past year, and on a few productions where we have recorded, we have noticed there's a delay of about 5-8 frames between audio and video.

I just want to make sure I understand why this is happening, and what we can do about it.

- Cameras connected directly to TVS
- Program HD-SDI goes to an AJA KiPro
- Audio from a mixer goes to the AJA KiPro

I have yet to try, so I'm not sure. Would it help to feed the TVS with a digital audio signal, so it's embedded in the program HD-SDI to the KiPro?

Yes/No, and why? How can this be fixed.


I'm concerned we might have the same issue on the 2 M/E (which finally came last week). I haven't had the chance to do thorough testing with it yet as I'm out of town.


Have anyone with a 1 M/E or 2 M/E had the same issues?
The event we're doing will be projected live in the arena, but we're recording for archival and possibly promo work for next years event. Delay to the projector is of course crucial too. I Was happily informed the projectors will have SDI inputs.

But 5-8 Frames can't be only by ATEM ! How did you measured the delay ? Can you desribe the whole setup ? Are there any conveters used ?

Well Delay will always be there! In live situations avoid close-ups on somebody who is talking.

For your audio recording you should delay externally. or correct in POST. At this moment it is not possible to delay audio in the mixer (would be very nice)


On the 2M/E you can try to genlock your camera's and your mixer to minimise the frame buffers inside of the mixer.

But still you will have always have a few frames delay.

- Camera (1 to 5 frames to capture and proces the image) Depending on camera
- (if a converter used > "HDMI to SDI few lines of the frame" "Component to sdi 1 frame")
- Mixer 2 frames i guess (framebuffer + processing)
- Beamer (scaler inside the beamer 1 to 20 frames depending on model and processing needed)

I have seen a broadcast companie that displayed people clapping onto the beat of the music in a rock concert. Where they placed a special frame buffer module to 1 camera to delay the video so much the claps were off 1 complete clap / beat. So that they stayed in sync with the real beat.

It has beed said a thousand times in this forum:

The ATEM has a FIXED delay of exactly 1 frame from input to output unless you use the DVE effect, which adds another frame but only for the content of the DVE. So if you use the DVE effect to create a picture in picture effect, only the DVE picture will have 2 frames of delay. The other one has 1 frame.

If you genlock your cameras to the ATEM you wont even have 1 frame - you will just have 1 line of delay.

If you experience more delay, it is related to the rest of your gear like cameras, converters and everything else that is inside your signal chain. It might be that your cameras are really "slow" ones and have an delay of about 6 frames which in addition to the ATEM 1 frame delay would cause 7 frames end to end.

Buy a DEQ2496 from Behringer (or something similar) and delay your audio to match picture and sound. You will find this kind of setup in every digital studio or similar setup. If you have to feed a projector in the same venue, you will have to check which gear produces this delay. Maybe you'll have to buy new cameras. But the interesting part would be the question how you measured the delay? Even any digital monitor has asome amount of delay. Some have more than 4 frames!


Cheers

pro.

Thanks guys.

To clarify;
We are using two JVC HM-series (790 with genlock and 700 without genlock), and two XL H1 with genlock.
We have never used any converters between the camera and switcher.

Delay measurement have never been performed in a technical manner. We once had to use a SDI to Component converter to feed a projector, and I bet that added to the delay right there, but that doesn't tell us why or how it can add such amounts of delay on the recording.


I kind of had a hope that the 2M/E would perform better than the TVS. We didn't get it just because of a mixed feeling on the TVS. The amount of inputs, AUX and clean feed vs. price was the big winner.


Before the ATEM, we had a Brickhouse Callisto R SD-SDI mixer, which, if my memory is correct, had a maximum of 2 frames delay when we checked the KiPro recording in FCP.

The measurement was basically done by counting frames in the timeline. Audio vs. video. For some events it would be super nice to just produce, record, burn it on a Blu-Ray and sell it in a booth just outside the venue. Adding delay to that would be a headache we'd love to be without.


I'll check about the Behringer unit.

Here is table for ECHOLAB ATEM switcher .. I think , BMD ATEMs may be different .. But there lot of interesting infos there .. May be usefull for someone ..

http://atemuser.com/sites/default/files ... gement.pdf

I know, that in ALEX-Berlin (open TV channel Berlin - Germany) TV studio the audio is delayey by the audio mixer before recorded by AJA keypro. But I don't know the delaytime.

EDIT: No ATEM-Mixer !

Maybe a SDI embedded audio would be better. For trying, a simple ans cheep analog to SPDIF converter also works with ATEMs AES/EBU input.

Did you made any measures with other SDI video sources combined with analog audio ? Maybe it is a problem of mixing digital video with analog audio signals to be recorded....

To calculate delay in audio processor, just divide your number of delayed frame by the number of frame per second you are using multiplicated by 1000. (DelayedFrame/FPS*1000) = milliseconds in audio processor. Examples:

Delayed frames | Frames per second | Milliseconds | % of a second
1 59,94 16,7 1,7%
2 59,94 33,4 3,3%
3 59,94 50,1 5,0%
4 59,94 66,7 6,7%
5 59,94 83,4 8,3%
6 59,94 100,1 10,0%
8 59,94 133,5 13,3%


Delayed frames | Frames per second | Milliseconds | % of a second
1 50 20,0 2,0%
2 50 40,0 4,0%
3 50 60,0 6,0%
4 50 80,0 8,0%
5 50 100,0 10,0%
6 50 120,0 12,0%
8 50 160,0 16,0%

Thank you guys.

We haven't had a chance to do the delay tests we were hoping to do. We've spent a ton of time rebuilding our production racks.

We have two racks. Each of them has a 17" monitor for multi view, one for each M/E. Under the 17" we a have a rack mountable set of two 7" monitors which earlier was used for camera reference.

Now we may have them show whatever we want. AUX/Clean feed, additional sources that won't fit on multi view and so on.


One of the racks will be the production rack, with the 2M/E switcher, and everything we need of cabling and equipment for genlock, timecode and tally. So basically, it'll be the base unit, with two racks to accompany with more monitoring and recording to our KiPro units.


As for audio delay equipment, turns out we have a 1U whatever it's branded, that will do just what we need it to.

Really looking forward to get going with this rig this weekend. I'll make sure we take some photos.

We also use the GY-HM790, genlocked to the TVS.

We have tested two scenarios:

1. Run mics into camera, so the audio is embedded over HD-SDI into the TVS. Doing this, there is absolutely no delay. Though, this is not ideal, because controlling levels is a PITA.

2. Run mics into an analog mixer. Analog out of the mixer into our AD (to convert it to AES), and then into the TVS. We have found that we have to set the delay in our AD to around 70ms.

The weekend is over, and 6 productions in total were done. It was conference style and concerts produced for big screens next to the stage. We also made recordings for archive/promo down the road.


Unfortunately we didn't get the chance to test with the audio delay unit, but I'm sure it's the missing link in order to get audio in sync with video.

For the productions we had set it up like this.

atemsetup.png (9.84 KiB) Viewed 6731 times

Forgot to mention we mixed audio from the mixer with sound from one of the cameras on the recording through the ATEM software.

We also had other sources such as a computer for videos (DVI Extender to SDI - AMD is a real PITA!), and another computer with key+fill HD-SDI for texting. Lower thirds were made in Photoshop and uploaded to the mixer.


We had one camera in the chain that doesn't have genlock. The JVC HM700. Unfortunately we just had to get a camera at the time. Quite annoying there was a new camera with genlock on the horizon just a few months later.


Would that one camera without genlock add more delay to the whole process through the ATEM, or would it be some more delay for that camera only?


I'm happy to say the 2M/E style interface on the Mac looked really sexy. And a bit intimidating for those who doesn't know what it does.

We used the 1M/E panel to control the mixer. In some occasions we would make an independent mix for the big screens, and one for the recording. Swapping between the two M/E's proved to be effortless.


We did all sorts of stuff. Keyframed graphics on upstream key, lower thirds, song texts, stingers and all AUX buses were used in some way for either recording or monitoring.

One of the camera operators had a Teradek Bolt, so we got to test wireless transission of HD-SDI. It worked really good when it had link. It had no visible delay compared to the cameras on cable.

Hi Ola Haldor Voll,

the Teradek Cube devices typically have a delay of 300ms which correspond to a delay of roughly 7 frames in PAL land (40ms - 1 frame). The Teradek Bolt has almost non existant delay (I think I read somewhere it has 1ms). Can you tell something about the maximum distance from transmitter to receiver?

Regarding the genlock question:
If you just add 1 single source to the ATEM that is not genlocked I think it acts like nothing is genlocked. Genlock makes only sense if you can genlock all sources. But you can easily test this by recording a superimposed timecode from your camera from a monitor and compare the "in display" timecode with the timecode of the clip. Please note, that even the monitor can add some delay. As I said earlier the ATEM has a delay of 1 frame with unsynced cameras.

Cheers


pro.

Sorry that is not correct.

One camera free-running in a genlocked system will engage the framesync on that input and cause that camera only to be delayed by (upto) one frame.

The genlocked sources will still pass with just one line of delay.

Sorry that is not correct.

One camera free-running in a genlocked system will engage the framesync on that input and cause that camera only to be delayed by (upto) one frame.

The genlocked sources will still pass with just one line of delay.

Exactly. This is why it's important to have cameras genlocked, but for other sources, such as a media player (animated lower thirds, bugs, graphics, etc.), it's not really important. 1-frame delay on those sources isn't a big deal.

Thank you, that's exactly the answer I was hoping for.
We didn't see much delay from any of the cameras, and I suspect the most delay came from the projection itself.
There was very little delay in general, and no complaints from anyone about that.


The next step is to check whether interlaced or progressive is better for big screen projections.
I thought the projected image looked de-interlaced/progressive in the way dancers or close ups on hands and talking heads looked, thus I thought it didn't resemble what we were producing.

And tickle me to death if you will, for asking a stupid question... I've heard or read somewhere that there is more delay when the cameras are set to progressive than interlaced. I find it hard to believe, but it would be nice to hear your thoughts on it.

To be precise:
The delay for an ungenlocked source (camera etc.) is at least one frame up to two frames (continously sliding delay within this window).
The delay for genlocked (and correctly timed/aligned) sources is 1 line.
The delay for genlocked but not correctly timed sources is 1 frame.

We have entered 2 (these boxes are mono, so one for left and one for right)
Behringer Shark FBQ100 at the end of our audio workflow and have them connected to our ATEM 1 M/E.
We gave the audio a delay of 60mseconds and all audio from camera's and microphones are in sync with our video.

We gave the audio a delay of 60mseconds and all audio from camera's and microphones are in sync with our video.

If you are running 50 Hz you are fine.
60ms = 1,5 Frames. You have a delay window of +/-20 ms wich is inside the non visible range.

@Uli,

Thanks!