The Challenges of Digital Audio (Part 2)
By Alan Ruberg, Systems Architect
This is Part 2 of a 4 part series; here is Part 1.
You would think that, because of accuracy, computer storage and networks would be great for sight and sound. The trouble is that speed and time are largely ignored. For example, a communication stall that would hardly be noticed in a web page load would cause a noticeable sound gap. To overcome this problem, playback is delayed (buffered) until enough data is received to cover the longest anticipated gap plus the time it takes to catch up, assuming the network is fast enough. Sometimes, it’s not. Have you ever tried to make a phone call when the delay (we call it “latency”) is really long or your voice is echoing? If a film is being played, the audio could arrive quite late and cause lip-synch problems. If a skip occurs, the audio becomes permanently early, and if a gap occurs, the audio becomes permanently late. This is called “variable latency” and makes the problem even worse.
It’s not always possible to communicate back to the source, such as in radio or television. The radio station doesn’t stop and retry if someone doesn’t receive the signal properly. It’s up to the radio to do the best with what it gets. The same is done in telephone because long delays cause jokes to fail. In the end, repeating transmissions makes no sense because time and speed are too sensitive to ignore. Very good error correction schemes are frequently used that can increase the transmitted bits by 50-100% even though space is a premium. That leaves accuracy as the last variable.
Fortunately, audio design engineers can use this to trick the human system. The application of these tricks is generally called “concealment.” It’s really no different from a magician bringing attention to one hand while the other hand is pocketing the object that “disappeared” (sorry if I wrecked the illusion).
The example shows what can happen if data is lost when the transmission does little to help the receiver. If the receiver keeps track of time, then a gap is experienced, but the latency remains fixed. If missing data or errors are ignored, then a “glitch” is heard and the audio starts coming too soon. If the data is late, a gap is heard, and the rest of the audio is late. In real life, the transmitter does quite a bit to help the receiver maintain quality. One trick is to send the data out of order, possibly giving preference to “more important” data. This simple example sends every 10th audio sample as a “packet.” The first packet contains samples 1, 11, 21, …, and the second packet contains samples 2, 12, 22, … The size of each packet determines the latency (times 10). The same amount of data is lost as the previous examples, which represents two packets, so two consecutive samples are lost out of every 10. This very simple approach draws a straight line where the samples were lost. It isn’t perfect, but it does a decent job of tricking the ear as long as it doesn’t happen too much. It also conserves time. You have probably experienced “robot voice” on cell phones, blurry blocks on your TV, or skipping on a CD. This is the error correction system doing its best in harsh circumstances. Mostly, you don’t notice anything going wrong. We humans are forgiving on phones, but when it comes to home theater, it needs to sound perfect.
The technology behind WiSA compatibility works much more like TV and radio delivery, instead of like the internet, to reduce latency and improve sound quality while guaranteeing fixed latency. This way, your wireless home theater sound is robust and sounds great. WiSA as an organization has balanced all the trade-offs to help its members deliver pristine, interference-free HD audio. Stay tuned for Part 3 of The Challenges of Digital Audio, (and how WiSA has overcome them).