Bob's Adventures in Wireless and Video Headline Animator

Monday, September 6, 2010

Everyone is Jumping on the TDMA bandwagon

TDMA, or Time Division Multiple Access, has been used by HauteSpot Networks to carry video and voice streaming over broadband wireless for over 5 years. It is only in the last few months that we have seen a number of vendors copying our lead and innovation in the market by offering TDMA in their products.

Why does TDMA matter? Early on wireless engineers in the voice business investigated technologies which would reduce jitter in voice streams. Jitter is another way of saying delay variation.

When video or voice is packetized (where the analog stream is converted to digital bits, and then the bits are grouped into packets for transmission on an addressed network) the packets are sent out in a specific order with very specific timing or spacing between the packets. This allows the receiver to decode the received data and play it back using the same timing as when it was sent. If the packets move across the network in an orderly process at the same speed, then they arrive in order and with even spacing and the receiver can play them back as if there was no network in between.

However, in the real world networks are variable. There are many factors that can contribute to delay variation such as collisions, congestion, route variation, forwarding delay by routers or bridges, encoding latency and decoding latency. There are really only a few things you can do to reduce jitter. You can either ignore lost data and fill in the blanks, or you can buffer received data as it comes in. When you have received enough data to smooth out variations in delay, then you can play out from the buffer. We call this jitter buffering.

In the case of dropping data, the issues are obvious. You may have no video or voice, poor quality video or voice, or lots of starts and stops depending on how bad the loss is. In the case of jitter buffering you may see delays ranging from a few milliseconds to minutes. We have all experienced the count down clock on You Tube. This is the jitter buffer at work.

Wi-Fi uses a protocol referred to as CSMA/CA or Carrier Sense Multiple Access/Collision Avoidance. Basically every Wi-Fi device will first listen to the air to see if there is any traffic, then it will send. If two Wi-Fi devices send at the same time, there is a collision. If they detect a collision, then they have to back off a random time period and resend. This scheme works well in short range where the network is heterogeneous. Meaning that the data on the network is bursty and timing is not so important. With voice and video this is not the case. Video and voice cannot wait for retransmission.

TDMA on the other hand is a controlled network where a master unit provides time slot control to all the members of the network. It eliminates collisions by making sure that no two members of the network send at the same time. It also makes sure that the time between slots for each member of the network are evenly spaced. TDMA provides the transport environment that video and voice need to flow smoothly.

HauteSpot Networks realized the issues with retransmission long ago and implemented a variation of TDMA that co-existed with Wi-Fi on unlicensed channels. Our first implementation, the HauteLine protocol, was limited to point to point links. Later we developed the TLP or TDMA Like Protocol, which supported point to multipoint. TLP has been in development and used in HauteSpot routers for the last 4 years. Our competitors only recently figured this out.

But, it is not enough to just implement TDMA. Using unlicensed frequencies means that you need to co-exist with Wi-Fi. Wi-Fi devices will not pay attention to your TDMA network timing. They will still send, even when your network master has not authorized them to do so. So collisions are still possible. Most of our competitors will revert to using collision detection and retransmission using the same methods as Wi-Fi. They will use acknowledgment (ACK) and non-acknowledgment (NAK) frames to indicate if data was received or not. They will wait for acknowledgment before moving on. This will introduce delay variation. So in a crowded environment they will have performance approaching that of Wi-Fi.

Through our years of experience we have learned that there are many other methods of detecting collisions which are much less disruptive to data flow. HauteSpot's TLP protocol uses a sliding window method of retransmission which creates much less jitter. It also eliminates ACK and NAK frames, significantly reducing network overhead and improving efficiency. It also allows HauteSpot to co-exist with Wi-Fi in heavily congested areas and still perform and scale well.

We suppose imitation is the highest form of flattery. But only good imitation really matters. Test your network performance, especially jitter, and you will see what a difference good design makes. It takes time to refine design and prove under real world conditions what works best. Those who are just learning now how to make video and voice work well over wireless broadband have a long way to go.

If you want to learn more about TLP, just ask. We may need to get you under NDA, but we will answer your questions.

No comments:

Post a Comment