I don’t understand most of your questions, so maybe try to elaborate? 

You definitely need camera tracking in large volumes as well, even if you don’t use Set extension. 

Multi cam is definitely supported. You should start with reading the documentation to get a better understanding of workflows: https://aximmetry.com/learn/virtual-production-workflow/preparation-of-the-production-environment-phase-i/led-wall-production/introduction-to-led-wall-production/

Hi,

1.
Tracking devices must be used with LED walls. Even if you don't use digital expansion, the picture on the LED wall must be rendered from the perspective (position) of the camera. The only exception is if your camera is stationary, doesn't move or changes zoom (PTZ with fixed zoom). Then you probably don't need tracking as the parallax won't change.

If you don't use digital expansion, then you can record your show on your camera or connect your camera directly to the broadcasting system. Because Aximmetry won't do modifications to the camera's picture.
Nevertheless, you probably want to connect the camera to Aximmetry, at least initially, as it makes things easier to set up.

2.
Aximmetry rendered objects connected to the AR Overlay pin will always block people:

Things rendered on the LED Wall will always be blocked by characters (talents in the real world).
This is normal, this is a limitation of LED Walls.
If you really want to have objects that move in front and behind characters, then you can use Aximmetry's flow editor to set up a system that switches between rendering an object in front (AR) or behind (LED wall). And have somewhat smooth transitions between these.
Depending on what you want to achieve exactly, this can be extremely hard or somewhat easy.

Also, because only one Unreal can run on one computer, you won't be able to use Unreal to render objects into the AR Overlay pin. However, with two machines you can render the AR Overlay with Unreal: https://aximmetry.com/learn/virtual-production-workflow/preparation-of-the-production-environment-phase-i/led-wall-production/using-led-walls-and-ar-together/#using-two-machines 

3.
You can connect up to 8 cameras to the Aximmetry LED Wall compound. And you can seamlessly transition/switch between these cameras. When the Picture Delays are correctly set.
If you want to actually shoot with more than one camera simultaneously while using the same LED Wall, that is not really possible. Note that the LED Wall will only be able to show the virtual world from the perspective of one camera. So if 2 cameras are looking at the same pixels of the LED Wall simultaneously, they will have to see different things for the same pixels, which the LED Wall won't be able to do.

4.
There is always a delay. Not just the delay of the signal of your camera to your computer, but the delay of the rendered image from your computer to the LED Wall. And the delay of the tracking system. You should try to connect all these devices directly to each other with the least delay possible. Also, in Aximmetry you should try to set as little In-to-out latency as possible: https://aximmetry.com/learn/virtual-production-workflow/preparation-of-the-production-environment-phase-i/inner-workings-of-aximmetry/in-to-out-latency/#in-to-out-latency-1
In the case of digital expansion, you can use a multi-machine setup to reduce the effects of delay, especially if you have moving objects in your virtual scene. If you do, the control machine will re-render the digital expansion with more up-to-date (accurate) tracking data.

Note, the delay of the signal of your camera to your computer only matters if you use digital expansion.

Warmest regards,

Eifert@Aximmetry,Thank you for your very professional answer.

Which deepened my understanding of the working principles of LED walls and green screens. If the size of the LED wall is large enough, my understanding is that the purpose of tracking the camera is only to make the computer render only within the camera's field of view, otherwise a large LED screen may have large pixels.

@Lee555: No, that’s not accurate. Aside from only rendering high resolution for the inner frustum (the cameras field of view) you also render the scene at the correct perspective from what the camera sees. Without that you wouldn’t get any parallax in the environment and it would look very fake. This is the key point of using real-time rendering for LED walls, since otherwise we could just use pre-rendered (or live footage) video loops. 

Hi,

Here is a GIF demonstrating what @TwentyStudios said about parallax:

It is visible on the side walls of the studio that they stretch depending on the position of the camera (parallax effect).

You can see this preview in Aximmetry by turning on the STUDIO view and setting the Preview Mode to Final:

And here is with Fill blurred, while Frustum is clear. And in Preview 2 is what you would see from the camera's perspective (Frustum):

Actually, if you have a PTZ camera, a camera that only rotates but doesn't move or zooms, then you probably won't need tracking as the parallax won't change.

Warmest regards,

Yes, that's exactly what I understand. Thank you. I will provide a process demonstration of Aximmetry LED wall in the future

By the way, there is another situation where when I record with a camera, I don't need to use software to record. Do I not need to connect the camera's video to the software, just link the tracking device and drive the scene to move

Hi,

Yes, if you don't use digital expansion, you don't need to connect the video to Aximmetry.
However, connecting the video would be very useful when setting up the LED Wall positions in Aximmetry. Or confirming that the LED Walls have correct positions.

Warmest regards,

I started to understand the working principle of LED walls. The display content of an LED wall is only related to the distance from the camera to the LED wall. It's not related to the focal length of the lens, am I correct in understanding this?

Because I am currently experiencing issues with cone delay, which may be caused by tracking device delay, theoretically I can reduce the focal length of the virtual camera lens to make the cone larger to avoid this problem.

Hi,

Focal length (zoom) changes the size of the Frustum on the LED Wall:

Also, correct focal length values are needed for digital expansion positioning and rendering.

Warmest regards,