Coda: What Was That Latency?

And is it low enough for live audio?

If Auracast? is to be acceptable in public venues the transmission delay, or latency, must be low enough. Otherwise the delay between streaming and live audio, the latter of which will still be heard while wearing hearing devices or earbuds, will muddle the sound we hear. Music will become distorted and voices less intelligible.

At the present time, when actual Auracast deployments are still in the future, latency performance remains a concern in some circles. A significant milestone was passed when GN ReSound ran an Auracast trial at the Lincoln Center earlier this month. Writing about that experience, I said “there was no latency effect or obvious artifacts to remind me that I was streaming.” This was with a 50-50 mix of live and Auracast, a condition most likely to reveal issues.

Show me the numbers

Naturally I wanted to know what latency I was actually experiencing, so I measured the Nexia along with the Starkey Edge AI and Sennheiser Momentum TWS 4. All three deliver latencies that should work in this setting with the Sennheiser being the lowest, perhaps showing the possibilities in a situational device with different constraints. Only the Nexia was tried at the Lincoln Center, but stay tuned. Surely there will be further demos where I can experiment with multiple devices.

In my presentation at the ACAud conference in Australia earlier this year, I discussed latency and its measurement in some detail. TL:DR see the short description at the bottom.

Here are the results:

Is it low enough?

Sound is slow, taking about 3mS to travel one meter. In public venues the distance matters. Anyone who has been to a large arena concert has noticed this. A drum beat will be seen on the big screen noticeably before the sound reaches your ears. This works in one’s favor when using Auracast. In fact, latency will have to be increased for distant seating sections in venues like this. This is possible to do.

At the Lincoln Center I sat about 10 rows back from the stage. With a combination of my on-the-spot estimate at the time and a bit of rough math, I’m guessing I was about 12m from the pickup microphones that served as the Auracast input. From the moment the sound hit the mics to when it reached my ears, about 36mS would have elapsed. The apparent latency at my seat was therefore only about 18mS. That is why my perceived experience was so good. This was relatively close to the stage. For someone another eight rows back, the latency would have been about zero.

There are still issues to be resolved for Auracast to become the ultimate audio experience in public settings. These include simplifying the tuning method, especially for those without smartphones, optimizing audio settings for a pleasing sound without a lot of tweaking, and defining proper positioning, spacing, and configuration of transmitters in larger venues where more than one is required. But based on my subjective experience in the Lincoln Center and these measurements, latency is not one of them.

Epilogue: A brief description of latency measurement

To determine latency, I put a hearing device in my manikin colleague Carl van Gogh’s ear and set him in another room. The audio picked up in his ear canal was transmitted back to my office through a very low latency RODE wireless mic. In my office was a metronome clicking away, its sound both transmitted to Carl with Auracast and picked up directly by a second RODE mic. Both were collected by the companion RODE receiver. What little additional latency was added by the RODE mics was cancelled out by using one at both ends. Measuring the spread between live and streamed clicks using Audacity audio software revealed the latency.