How much latency is acceptable in an amplification system for singing?

Question

I teach music in a public elementary school setting. Many teachers (of subjects other than music) have an in-room amplification system. They wear a microphone, the sound of their voice is amplified in a mounted speaker (either in the ceiling or wall-mounted). The idea is that teachers won't tire their voices out so much having to speak loudly, and all students will have an easier time hearing the teacher. Students who have a significant hearing difference carry a transmitter with them, which sends the teacher's voice to their in-ear hearing devices.

I am told that a similar device will be installed in the music room. As I look over the specifications of the system, I expect that there will be some latency (lag) between my singing and the sound of my voice in the speaker. So as I sing and play an acoustic piano, there will be a difference between when I hear the notes I play and when I/the students hear my voice. My question is, how much lag will be noticable or annoying?

EDIT: I have discovered that the system will introduce approx. 25ms of latency. I will update this with my experience once it is installed.

Answer 1

Musicians can discern as little as 5ms (or perhaps 1ms for percussion?). Non-musicians will usually hear two distinct sounds if they are separated by at least 20-40ms, depending on the type of sound. This numbers are the results of studies of the precedence effect (AKA Haas effect).

The consequences of signal delays have been studied and more or less understood for at least 60 years. So sound amplification and transmission device designers are aware of this.

Analog electronic signal paths have such low latency that an amplified voice will come out of a PA speaker before the unamplified voice from the person talking reaches the same PA system through air (in other words, it's faster than the speed of sound).

Digital devices used to sometimes introduce noticeable latency during A/D and D/A conversion, but modern devices are usually much lower latency, under 1-10ms. This is usually not a problem at all.

Two things are more likely to cause problematic latency in modern systems:

1. High latency processing, such as pitch shifting or look-ahead dynamic control. This is less common and I doubt you have to worry about it.
2. Transmission of digital sound over a packet-switched network without a proper QoS configuration. This is also something you almost certainly don't have to worry about. This kind of thing crops up in large venues like stadiums, hotels, and convention centers.

My guessing you don't have to worry about those possibilities is that a public school is extremely unlikely to be able to afford the equipment or expertise to introduce those technologies to the school.

Answer 2

I suppose that would depend on the kind of music, but most amplification systems should not induce an amount of lag that matters. Such kind of lag should only happen with complex DSP chains or unstable transmitting channels, which should not be given in your case. Generally a noticeable lag should also be a problem with speaking, as you’ll hear yourself noticeably after you speak, which would be really hard to work with. So I’m positive that lag won’t be much of a problem for you. Rather you’ll need to test out if the system (which is most likely there for amplifying speech) will give you good acoustics with singing while playing piano.

For your question: I’d say if you are with in 20ms it should be okay. Depending on the music even more. You can simply test this yourself. Record some piano, record some singing and align them. Then try how much you can shift your vocals until it sounds off.