Hacker News Reader

Comment by jancsika | Hacker News Reader

> 1ms round-trip

Ooh, a rare latency claim that actually specifies which type of latency they mean! One millisecond round-trip latency, if true in practice, is quite impressive.

I'm going to continue reading now... :)

View on HN

I've played with Bela [1] which uses the same real time kernel IIUC. It had super stable audio performance down to a 2 sample buffer, which felt like programming hardware. I'm very happy to see more work in this field.

1: https://bela.io/

Bela is fantastic because it pairs good audio hardware with low latency software. Elk looks good, but I'd need something better than a Raspberry Pi in terms of audio IOs to use that full potential.

There's a decent-looking hires 6in/6out hat on the site store, albeit "out of stock".

Bela is indeed fantastic, but the CPU is a bit weak.

I’ve never understood the latency requirement for professional gear, are you able to articulate why there’s a need for lower latency than average consumer gear?

Latency means you can't hear what you're doing. This is especially important for drums and transient sounds because playing with feel will move the beat by around by a few ms. Adding a fixed delay to that makes it more difficult and adds cognitive load. Adding a small variable delay makes it impossible.

No one cares if your pad sounds are a little late, but variable latency absolutely destroys drum grooves, sequenced electronica, and the kind of super-precise playing needed for good classical piano.

You might think most people can't hear the difference. And they can't. But - as with other audio production values - they can feel the difference. Tiny variations in timing are the bedrock of many kinds of musical expression. If they're not there, people notice.

Three things come to mind--

1. If you strum the strings of a guitar and want your laptop to make you sound like Jimmy Hendrix, it had better output sound quickly enough that you perceive it happening "immediately." (I.e., it should sound and feel just like an electric guitar that's hooked up to an amp.)

2. Such a laptop like the one above had better make you sound like Jimmy Hendrix for every moment you are playing the guitar. The website doesn't state it, but they are making a claim to 1ms round-trip latency in the worst case. If your first strum of the guitar convinces you that the resulting sound is happening immediately, then every subsequent strum must also sound like it's happening immediately, too. Users don't merely rant about a sporadically janky digital instrument as if it were your run-of-the-mill Windows 10 UI. They throw it out.

3. There's something I call El Dorado latency, which is always half the latency being claimed and is eternally sought after by users. (Works for system latency, round-trip latency, etc.) Since the unit for measuring audio latency is milliseconds, and since "1" is the last whole number greater than "0", a claim of 1 ms essentially short circuits the search. E.g., if they had claimed "2" then El Dorado latency requires the user to imagine how much better the system would sound at "1". :)

4. Okay, I'm being a butt in #3 above. Round-trip latency in audio seeks to go down to "ultra-low" latency measurements for the same reason gamers want frame-rates to go up to gazillions. In each case it means developers can choose one of the following for their end-users: a) add more hungry, hungry hippos per unit of time, or b) make the same number of hippos hungrier.

Edit: clarifications

Half of 1 millisecond is 0.5 milliseconds.

Just because their published latency is the “smallest whole number” as you seem to be pointing to - doesn’t mean that it can’t be halved. What a strange thing to say.

Ableton measures latency down to the tenth of a millisecond. It’s really not a big deal.

And latencies add up when chaining hardware: 1ms for the effect pedal, 1ms for the amp, 1ms for the audio interface, etc..

what does the "hungry hippos" metaphor/idiom stand for? https://en.wikipedia.org/wiki/Hungry_Hungry_Hippos didn't help

I think it just meant more "stuff"

I thought they were referring to latency vs bandwidth

Different use case - consumer gear is generally only used for listening to recordings, whereas professional gear is often used for real time recording of someone who's playing an instrument while listening to the output.

It only takes a small amount of latency (5 ms or so) to noticeably impact performances by musicians. Latency is also an issue when you want to monitor a live signal along with other recorded audio, like in a DAW. It is not as big a deal with a playback only device.

rule of thumb is that sound travels roughly 1ft per ms, so 5ms is the same as standing 5 extra feet from the speaker. Most musicians don't have much trouble up to around 10-20ms.

Jitter is killer though - if the latency is randomly changing it's really annoying.

Is the speed of sound relevant in most recording scenarios? Assuming headphones and instrument/vocal mics that are within a few inches, then everything is traveling as electrical signals at the speed of light.

Or is this just meant as a visualization of latency in these scenarios versus what it would be like with musicians playing x ft apart from each other in a room?

I recently played on an electronic drum set and it was absolutely fine with headphones. But when we hooked it up to the speakers across the room there was a very noticable delay (~5ms as we calculated) when hitting the pads with the stick.

Since I only used acoustic drums until then I never gave much thought to the idea of having your drums sound delayed. It was quite eye opening.

It absolutely is. Many times when rigging halls for recording live events I've had to insert delay in to sub mixes to match latency to the back of the room. It's standard fair for a live audio tech (a thing I do on the side occasionally)

> Assuming headphones and instrument/vocal mics that are within a few inches

It is relevant, and monitoring via headphones or a floor speaker is a mitigation.

It's also a problem for large orchestras. If everyone else just plays along with what they hear from the people near them, the timing is going to be inconsistent across the orchestra pit even if you have some obvious audible reference like loud percussion. The mitigation is to orient yourself as much as possible around the visual cues from a conductor.

That's a good point, and I'd imagine instruments like a church organ would suffer from this too, though I know in that case there are a bunch of other sources of delay, like the time it takes for the pipes to fill with air.

The latter. The point being that they are trained, comfortable, and proficient in those scenarios and the same translates to a (consistent!) software delay of up to 10ms.

Yes it is true that you can deal with that amount of latency, but 10-20ms is way up in the 'very sloppy' range especially when it comes to recording.

Most musicians use headphones / earplugs in professional settings

which is interesting but not the point?

But it's in part because of the latency due to the distance to loudspeakers, which isn't an issue with headphones

I think the loudspeaker part was a comparison to understand it a little bit better.

"If the system processing your signal introduces a 20ms delay, it would be the same as standing 20feet away the loudspeaker you're using as reference."

Interesting, I wouldn't have expected it to be that low, though in the aircraft space, control harmony can be disrupted with as little as 20ms of delay between control input and control surface movement[0].

Do you have any research you can suggest for that 5ms number?

[0] http://www.klabs.org/history/history_docs/reports/dfbw_tomay... (Keeping the delay at or under 10ms was the recommendation I got from the study P.I.)

Here's one :

https://www.jstor.org/stable/10.1525/mp.2006.24.1.49?seq=1

And another (follow the links to the article, but I thought that particular graph was most informative) :

https://www.researchgate.net/figure/Audio-latency-tolerance-...

It's not that low. Up to 40ms (roundtrip) isn't a problem for solo playing, in my experience. And it's also a matter of training. Church organists learn to play with much larger delays.

I can deal with 10ms roundtrip, but not much more.

One issue is when you get some acoustic sound too (open headphones, bone conductance, singing) and pretty much any delay in the headphone signal causes phase interference.

Yeah, I have to add that for me 40ms is ok when playing a virtual instrument.

Did you ever see people who try to sing with headphones with a delay of around 200ms? That really messes up your performance.

With a virtual instrument you would usually not face the full round trip latency, but rather control latency (MIDI?) and audio output latency (usually 50% of roundtrip). So in the VSTi case that might be closer to 25ms than 40ms in practice. Often the latency jitter is quite dreadful in this case as the control signal is not synchronous to audio.

Press a key on a real piano - instant sound.

At 10ms (and possibly less), you will start to notice the lag and it will affect your performance.

https://www.soundonsound.com/sound-advice/q-how-much-latency...

Nope, you sit 3 or 4 feet from the soundboard in a piano, so the latency is around 3 or 4ms.

The speed of sound in air is ~330m/s so sitting about a metre away will produce ~3ms delay. However the speed of the vibration through human flesh (far denser than air) is about 4 to 5 times faster than the speed in air. You will "hear/feel" the sound transmitted through your body far faster than the sound transmitted through air. This maybe why I'm always seeing this magical "1ms" value.

The 10ms (or 5ms, depending on the source) is a threshold - under which a human perceives the sound as "instant", even though - yes, physically - it is not actually instant.

When playing a software instrument via an external hardware controller (typically via MIDI over USB), even a tiny delay from the musician striking a key or a drum pad to the resulting audio out of the speaker can become very problematic for instruments with a very fast attack, where exact timing matters a lot to the performance. Examples include drums and percussion, pianos, many plucked type of sounds. With slow attack sounds like legato strings, extra latency is generally much less of a problem.

If you're trying to play along with something, synchronizing with the beat, the latency will completely destroy your ability to get into the grove right if what your playing is heard with a delay.

In consumer audio, you produce the sound and you're done. In devices used for musical performance, you produce audio, which then feeds into the brain of a musician, who then makes an decision to change their behavior in response to what they're hearing, which then effects the audio. Those constant round-trips through the brain make it very easy to notice latencies. (They pull you out of "the zone".)

Not sure what you mean by average consumer gear, but when I'm performing, anything higher than a 15ms round-trip latency is starting to have a negative effect on my ability to play tightly. I can do with more latency, but that's the point where it starts being unsettling or more tiresome.

That's necessarily going to be aspirational in at least some respects, though—it's limited by the hardware.

I wonder if there's a way for computers to self-measure round trip. Even if you have to plug an aux cord between the speaker + mic jacks.

My first paper as an undergrad measured MIDI round-trip by tapping the input and output MIDI wires and then looking at how the signals were offset: https://www.kmjn.org/publications/MIDI_ICMC04-abstract.html

For keyboard input, one paper (whose citation I can't seem to find) used a clicky mechanical keyboard, pointed a mic at the computer, and measured the offset between each corresponding pair of keyboard click and sound output.

That is the way to do it. In Ardour it is part of the audio setup.

People have done experiments using high-speed cameras and such.

Reply on news.ycombinator.com