Resolume Forum

I'm looking to put together a system that can ingest (4) 1080p video feeds via HDSDI and output (1) 4k 12G SDI and (3) 1080p on HDSDI. Latency on that video input is a big deal. It will be live in the venue, so delayed video will get ugly fast. I would like to stay in the pro video world, so HDMI and Display port are not super attractive. Thanks in advance for your input!

Datapath sdi4 input with aja HA5-12G and aja AJA HA5-4K. Gives you 4sdi inputs, one 12G and 4SDI 3G.

Going to be the most pro and lowest latency, you also get sync on your 3G outs now. Faster then using blackmagic cards on the output, which don't sync the outputs.

Drazkers,

Thanks for the information. Is there any benefit to using a single card? I'm looking at the AJA Corvid 88. I've done some testing with a Blackmagic Decklink Quad 2 in an external Sonnet enclosure connected via Thunderbolt 3 (Mac). I'm ingesting video in one input, resizing it, positioning it on screen and outputting that from one of the outputs on that same card. My total system latency is at about 7 frames in this scenario. This includes a video switcher, router, LED wall processor too and I know these pieces of hardware are all introducing some of that latency. Some of my reading suggests that AJA products may reduce the amount of latency. I would like to be able to shave off 3 frames somehow. Thanks for you help!

If you use magewell capture cards and the output is done by the graph, you will have 3 frames at the output.

If the output is through a card (SDI) you will have 4 or 5 frames.
The output from the video card is slower than the graphics card because the signal has to be re-encoded and sent to the PCI lane.

Each team you put in is a minimum of 1 frame.

4k signals thru AJA hardware are not supported in Resolume.

For 4k signals you're best bet is probably a Decklink 4k Extreme 12G should you need to stick with SDI.

If you are going to output to LED walls you could also look at (framesynchronized, not tested with Resolume) Quadro GPUs. You will not have any readback/conversion delay getting frames back from the GPU outputting it on SDI. That'll shave a few frames of your in-out latency.
If SDI out is just for being able to monitor in a control room or taking stuff fullscreen I've had good experience with separate DVI -> SDI converters that are synchronized (used that on touchscreens with other software to get latency low).

I don't have numbers on hand regarding in/out latencies per board. In theory we have more control on AJA and should to have a smaller buffer but I haven't done any practical real-world performance analysis.

PVD123 wrote: ↑Wed May 19, 2021 16:48 I'm looking to put together a system that can ingest (4) 1080p video feeds via HDSDI and output (1) 4k 12G SDI and (3) 1080p on HDSDI. Latency on that video input is a big deal. It will be live in the venue, so delayed video will get ugly fast. I would like to stay in the pro video world, so HDMI and Display port are not super attractive. Thanks in advance for your input!

Have studied this problem in another context, unrelated to Resolume. Minimizing latency in IMAG is not difficult. It does require effort to measure the glass-to-SDI output latency in your video cameras. Many years ago, the “industrial” grade of cameras from the likes of Sony, Panasonic, etc. had 3-4 frames of latency! And, cinema cameras like the RED ONE had up to 5-6 frames of latency! It was bonkers! To get 1-frame of latency, for IMAG applications, from Sony or Panasonic required spending big bucks on their higher-end “broadcast” grade cameras, rather than the mid-grade “industrial” cameras. Blackmagic upset the business strategy of the big two with the URSA Broadcast camera, which does feature a 1080i59.94 output with 1-frame of latency. 1080i sounded bad until I realized that broadcast vision mixers are optimized for the lowest delay with 1080i, as are many, many projectors. So, the combination of a 1080i output from a camera, into something like an Analog Way Ascender 16 (or whatever), 1080i, into Barco projectors, 1080i, has basically <2 frame delay. Assuming all the cameras are genlocked. It’s <2 frames. End to end. Incredible.

To achieve that in software is hard. Here is why. A typical SDI input board on a computer is going to have 60ms (i.e. - nearly four frames) of input delay. Then, if you want to use SDI out, you’re going to add even more delay. I don’t remember the exact numbers for that. I think usually about another 60ms (i.e. - nearly four frames). So, it’s hard.

And, that’s on top of your cameras! And, in addition to your projectors or video wall processors!

For lowest delay, this requires using genlocked low-latency cameras and a zero-latency vision mixer.

Period.