How many cameras can I use in Connect?

Disclaimer – This is the testing carried out by Ian Justin, this is not carried out by Adobe in a lab.  This testing consists of Ian Justin, 6 meeting room instances, 6 cameras and monitoring and capturing the bandwidth display in the meeting room.  This is not intended to be a scientific test, it is however, representative of a typical user experience in my view.

There is technically no limit to the number of cameras that can be used in Connect, the largest number of live feeds I have used personally is just under 50 cameras as part of a proof of concept for a client, and very distracting.

The number of cameras you can display in Connect depends on your end-user bandwidth, what other things you are doing with the bandwidth (VoIP?), the quality of the camera feeds you will accept and whether you are hoping to enhance the experience of your attendees or just trying to prove a point.

The number of cameras you should use in a good presentation is probably one if you want to engage your audience, two cameras if you want to show interactions, maybe five to six cameras if you have a panel answering questions and ten or more cameras if you want to create the best opportunity to have at least one unhelpful distraction happening in your session at any one time.

The caveat here is that some schools or training classes aim to have the feel of participants in a classroom and the trainer or teacher will want to see the audience.  This is possible, but be aware that engaging your audience with compelling content, quizzes and interactions is likely a better way to keep their attention vs. postage-stamp video feeds.

The real practical issue is to determine just how valuable multiple cameras might be in a meeting room or classroom.  The real technical issue here is the bandwidth available to view these camera feeds at the end-user side.

So to the technical side – I carried out some live testing to see how the bandwidth taken by cameras increased with the number of cameras and with the quality of the feeds.  As you should be aware, you can use the ‘Room Bandwidth’ feature in Connect to constrain the meeting to operate in nominally LAN, DSL or Modem bandwidths and that was the primary option used to manage quality in these tests.

Changing the room bandwidth will lower the resolution and frame rate (fps) of the cameras and as such lower the bandwidth taken overall.

The summary of the results, badly presented I grant you, is shown in Figure 1 below.  In the following notes we can dissect some of the numbers.

Figure 1

The tests were carried out at the bandwidth settings on the left side of the spreadsheet as follows:

  • LAN – 480p and up to 15fps
  • DSL – 240p and up to 10fps
  • Modem – 240p and up to 10fps
  • Modem – 120p and up to 4fps
These settings are changed from the ‘Preferences’ pane in your Meeting Room.  The ‘Room Bandwidth’ setting will alter a number of quality settings in your room, we are only concerned with the camera/video here as we were not screen-sharing or using VoIP.
Note – In the preference pane I took the video one ‘notch’ down from the highest setting.  In my experience, this highest setting at ‘LAN’ bandwidth is entirely too bandwidth intensive and the difference between the highest setting and the ‘one-notch-down’ setting should only be the difference between 20fps and 15fps as shown in Figure 2.  To ensure a good meeting experience, I personally *never* have the video set to the highest quality at LAN room bandwidth.  The majority of these tests were carried out at ‘one-notch-down’ for that reason.

The last of the modem settings (120p and up to 4fps) had the video quality slider taken all the way tot he right to get the most  bandwidth-efficient video possible.

Figure 2
The first tests were carried out with 6 instances of a Connect meeting room and by turning on one camera and recording the average bandwidth, then a second camera and so on until the 6 cameras were all running.  Then the bandwidth setting was changed and the test process was carried out again.
As the initial tests were carried out using the ‘grid’ layout (Figure 3) in the Camera Pod, by way of comparison I also ran the tests with all 6 cameras running in the new Connect 9.2 ‘Filmstrip’ mode (Figure 4) as this is intended to reduce bandwidth.
Figure 3

Figure 4

Some finding from the tests and the spreadsheet.
How much bandwidth does a camera use in Connect?
The average bandwidth taken by one camera in my tests for each bandwidth and quality setting is shown below:

  • LAN – 480p and up to 15fps
    • 260 kbits/s
  • DSL – 240p and up to 10fps
    • 160 kbits/s
  • Modem – 240p and up to 10fps
    • 78 kbits/s
  • Modem – 120p and up to 4fps
    • 18 kbits/s
Does the growth of the bandwidth used grow with cameras in an exponential or linear fashion?
Although it is very possible that with more cameras the changes would be exponential, it does seems to be pretty much linear in my tests, so within the margin of error in my tests. it seems reasonable to assume that multiplying the bandwidth by the number of cameras gives a reasonable indication of the overall bandwidth taken.
Does Filmstrip mode help with the bandwidth at all?
This mode does reduce the bandwidth vs. the traditional camera grid a little in my tests, but not a lot.  My numbers seemed to bear out that with 6 cameras the bandwidth was reduced by around 20-25%.  It is very possible that with more cameras the improvement would be higher as there are more cameras ‘off of the main stage’.
How much bandwidth would 10/20/50 cameras take with these findings?
This is an extrapolation of the numbers from 6 cameras, but it does give some understanding of the logistical challenges of multiple cameras.  However you view the accuracy of these numbers, this does put into perspective how the quality of the video feed, the bandwidth setting of the room and the number of cameras are related to the ultimate bandwidth used or required:
  • LAN – 480p and up to 15fps
    • 10 Cameras – 2,599 kbits/s (2.6 mbits/s)
    • 20 Cameras – 5,197 kbits/s (5.2 mbits/s)
    • 50 Cameras – 12,993 kbits/s (12.9 mbits/s)
  • DSL – 240p and up to 10fps
    • 10 Cameras – 1,629 kbits/s (1.6 mbits/s)
    • 20 Cameras –  3,257kbits/s (3.2 mbits/s)
    • 50 Cameras – 8,143 kbits/s (8.1 mbits/s)
  • Modem – 240p and up to 10fps
    • 10 Cameras – 776 kbits/s
    • 20 Cameras – 1,522 kbits/s (1.5 mbits/s)
    • 50 Cameras – 3,879 kbits/s (3.9 mbits/s)
  • Modem – 120p and up to 4fps
    • 10 Cameras – 181 kbits/s
    • 20 Cameras – 362 kbits/s
    • 50 Cameras – 905 kbits/s
What is the typical capacity of an end-user connection?
This is almost impossible to answer, but I personally try to manage my meetings to a 1 mbit/s limit, meaning I monitor bandwidth and drop quality when I see things get in the high 900 kbits/s.  the average connection speed in the US is around 8 mbits/s but as most internet users will tell you it is not often up there consistently.

It is also important to remember that if you are in a corporate environment with multiple people attending from an office, you are growing the required bandwidth in leaps and bounds as the signals pass out of your office to the servers, then back in again for each user in the case of streaming video.

Important – Remember that in this scenario, there is no VoIP, no pre-recorded video and not screen-sharing and these would all be fighting for the bandwidth as well.  VoIP should always take priority so the camera feeds would potentially suffer.