LOS ANGELES—I’m sure you’ve heard about how everything is going IP e.g., Internet of everything. Even live (real-time) video signals are going IP. Sure, SMPTE 2022 has been around for a while and gee, IP switches that can do 10 or 25 Gbps are available, with 40 Gbps just around the corner, so why not? Isn’t this the smart choice?
So I thought, let’s compare the two technologies. For now, I’ll frame this discussion based on 3 Gbps SDI and 10 Gbps IP switches. First, I’d like to describe what a traditional SDI router is designed to do:
- Deliver serial digital input signals, one per input port to one or more output ports;
- Provide consistent delay and performance, no matter how many active signals are being routed and switched;
- Provide deterministic switch times.
Now what is an IP switch designed to do:
- Manage multiple input and output ports;
- Keep track of all routing table entries (network status);
- Process or sort IP data packets and deliver to appropriate next routes.
So it seems these are two devices have differing use cases. Yes, in fact, the technology and protocols for each were developed for different use cases and thus have evolved as a best fit for the applications intended.
So why try to put a square peg in a round hole? Proponents of video-over-IP will say that commercial-off-the-shelf hardware reduces overall costs. They’ll also say that IP technology is the prevalent technology in general use, thus more R&D results in more innovation and future improvements. It’s also available from many large and small manufacturers, and offers flexible design of large switch fabrics.
Not to refute these points—as they are true—the key question is what should we be skeptical of? If major industries, including manufacturing, transportation, telecommunications, commerce and banking, have adopted IP, why not real-time video production?
TALE OF THE TAPE: BANDWIDTH
Let’s tell the tale of the tape. The first tale of the tape I’d like to address is bandwidth. Not just the port bandwidth or the internal processing bandwidth, but the total aggregate bandwidth (TAB) of each solution. For SDI, it’s fairly simple:
TAB = (No. of Input Ports) x (No. of Output Ports) x port bandwidth
As an example, a typical large plant video router is now a 1024 x 1024 3G SDI router with a TAB of 3.15 Pbps. That’s 3.15 million Gbps. This bandwidth is available, on command, finite and deterministic.
Of course, not all of these Gbps are in use all the time. In fact much of the time, the video signals maybe color bars! But when called for, there is never a delay or limit to any input going to any output.
From the IP side, bandwidth is a bit more difficult to calculate. What is typically specified in an IP switch is the “line rate;” that is, the physical interface maximum data bandwidth that can be either input or output from the port or ports. But this doesn’t give us the TAB or throughput of the switch. There are many factors that go into the throughput of an IP switch, such as backplane bandwidth, processing speed, buffering and to an extent, the power supply.
One major IP switch vendor’s specification for a large enterprise IP switch—16 slots, 10 Gbps per slot—can handle a maximum aggregate processing bandwidth of 320 Gbps. Each slot can be equipped with different line cards, but for our discussion, let’s assume a 4-port 10 Gb Ethernet card. So we have the possibility of a 64x64 (4x16x2) video switch, with a TAB of 320 Gbps. This works out to just under 3 Gbps per port, input or output. By the way, this switch occupies 7 RU with a maximum power consumption around 4.5 kW.
For comparison, a SDI 3 Gbps 64x64 router fits in a 4U frame, consumes 0.25kW and has a TAB of 12.3 Tbps.
So far, it seems SDI is a better value in terms of TAB, power and size.
(Continued in “SDI vs. IP: Packet to Packet.”)
