How to reduce latency?

DLVid · Jul 10, 2021

Hi all, as we get closer to going back on location I thought I'd do some testing of my TC2E to check its latency to see if it will work for Imag. My testing method is as follows: I have a BMD Ursa Broadcast camera connected to the switcher via SDI. I have a black burst generator connected to the switcher and camera. The camera is set to the same format and frame rate as each session. I have a laptop running a millisecond timer. I am shooting the laptop display with the camera and outputting that signal on my critical monitor. To find latency, I take video of both screens, find a clean frame of both images and do some simple math. The results were surprising. Genlocking the system resulted in significantly higher latency. I then went on to test adjusting the camera lines of timing and the switcher H & V position. The camera lines had an effect but not in the right direction. Switcher H & V had no effect. Below is a table of the results. I'd really like the system to be around 60 - 80ms; thoughts? Your feedback is welcome. Anyone using a TC2E for Imag out there?

SESSION		SOURCE TIME	OUTPUT TIME	CAMERA LINES	LATENCY (ms)

720P/59.94	FREE RUN	26.011	25.889	0	122
	GENLOCKED	36.318	36.126	0	192

1080P/29.97	FREE RUN	19.965	19.775	0	190
	GENLOCKED	17.089	16.831	0	258

1080P/59.94	FREE RUN	42.234	42.109	0	125
	GENLOCKED	10.368	10.236	0	132

				CAMERA LINES	LATENCY (ms)
1080P/29.97	FREE RUN	19.965	19.775	0	190
	GENLOCKED	7.007	6.75	0	257
		48.799	48.542	1	257
		58.783	58.528	2	255
		2.269	1.95	3	319
		14.591	14.332	4	259
		38.622	38.366	5	256
		57.723	57.469	-1	254
		19.004	18.688	-2	316
		4.895	4.574	-3	321
		49.695	49.374	-4	321
		23.392	23.199	-5	193

		28.799	28.543	10	256
		2.141	1.887	-10	254

				SWITCHER H POS
		13.182	12.926	20%	256
		16.796	16.54	50%	256
		15.355	15.1	80%	255
		4.446	4.19	-20%	256
		5.951	5.694	-50%	257
		56.574	56.318	-80%	256

				SWITCHER V POS
		17.149	16.894	20%	255
		5.599	5.341	50%	258
		53.726	53.47	80%	256
		54.559	54.302	-20%	257
		45.791	45.535	-50%	256
		55.292	55.053	-80%	239

kanep · Jul 10, 2021

Genlock works by delaying signals, it not surprising this would have an effect, it is what genlock does. The main question is, does the output of the TriCaster need to be timed to something else?

Also the critical viewing display likely has a native resolution. When signals are set to it that do not match, it has be adapted which will add some latency, so that can account for some of the timing differences when working with different resolutions.

What you might try is genlock all of the cameras, but not the TriCaster. Then turn off frame syncs on all cameras.

The other aspect is how much of the timing is in the camera (because it isn't zero millisecond from lens to output) and how much is in the TriCaster. You might try is playback of a video clip in the DDR showing timecode, loop back into an input. Then put the DDR source and input into a M/E side by side and do a frame grab.

DLVid · Jul 11, 2021

DLVid said:
Hi all, as we get closer to going back on location I thought I'd do some testing of my TC2E to check its latency to see if it will work for Imag. My testing method is as follows: I have a BMD Ursa Broadcast camera connected to the switcher via SDI. I have a black burst generator connected to the switcher and camera. The camera is set to the same format and frame rate as each session. I have a laptop running a millisecond timer. I am shooting the laptop display with the camera and outputting that signal on my critical monitor. To find latency, I take video of both screens, find a clean frame of both images and do some simple math. The results were surprising. Genlocking the system resulted in significantly higher latency. I then went on to test adjusting the camera lines of timing and the switcher H & V position. The camera lines had an effect but not in the right direction. Switcher H & V had no effect. Below is a table of the results. I'd really like the system to be around 60 - 80ms; thoughts? Your feedback is welcome. Anyone using a TC2E for Imag out there?

SESSION SOURCE TIME OUTPUT TIME CAMERA LINES LATENCY (ms)
720P/59.94 FREE RUN 26.011 25.889 0 122
GENLOCKED 36.318 36.126 0 192
1080P/29.97 FREE RUN 19.965 19.775 0 190
GENLOCKED 17.089 16.831 0 258
1080P/59.94 FREE RUN 42.234 42.109 0 125
GENLOCKED 10.368 10.236 0 132
CAMERA LINES LATENCY (ms)
1080P/29.97 FREE RUN 19.965 19.775 0 190
GENLOCKED 7.007 6.75 0 257
48.799 48.542 1 257
58.783 58.528 2 255
2.269 1.95 3 319
14.591 14.332 4 259
38.622 38.366 5 256
57.723 57.469 -1 254
19.004 18.688 -2 316
4.895 4.574 -3 321
49.695 49.374 -4 321
23.392 23.199 -5 193
28.799 28.543 10 256
2.141 1.887 -10 254
SWITCHER H POS
13.182 12.926 20% 256
16.796 16.54 50% 256
15.355 15.1 80% 255
4.446 4.19 -20% 256
5.951 5.694 -50% 257
56.574 56.318 -80% 256
SWITCHER V POS
17.149 16.894 20% 255
5.599 5.341 50% 258
53.726 53.47 80% 256
54.559 54.302 -20% 257
45.791 45.535 -50% 256
55.292 55.053 -80% 239

Good idea. I recorded the millisecond timer in DDR 1 and played it back in an M/E with split screen with DDR 1 on one side and mix 4 (DDR 1) re-entered to input 1.

SESSION		SOURCE TIME	OUTPUT TIME	CAMERA LINES	LATENCY (ms)	DDR 1 RE-ENTRY TEST
1080P/29.97	FREE RUN	2.398	2.334	N/A	64
	GENLOCKED	49.664	49.559	N/A	105

DLVid · Jul 11, 2021

I did some additional research on the Blackmagic Ursa Broadcast cameras and learned that they have 60ms of latency. This tracks with the testing I've done.

How to reduce latency?

DLVid

Member

kanep

NewTek SE

DLVid

Member

DLVid

Member