Limiters latency for live use

Synthi

Hi people!
I’ve been doing some research trying to find the most suitable limiter for live use, I mean, having a limiter in the master with no latency.
Been using roughrider in the past, but is too much color and I want some transparent limiting for main dynamics control and protection in my live shows.
For testing I used the AUM built in engine info (press DSP consumption icon), and under node statistics/latency.
This is what I found trying all I have in my AUv3 arsenal:
Plugin: latency in milliseconds
AUM built in peaklimiter: 12ms
AUpeaklimiter: 2ms
WU Aupeaklimiter
Bleass compressor (60:1 ratio) : 5ms
RoughRider3 (1000:1 ratio) : no latency (0 ms)
NuRack (using limiter only): no latency
FAC maxima: 42.7ms !!!
FAC medusa (limiter only) : 10.2 ms
Mixbox (limiter only): 0.1 ms (that’s what I actually use)
Modley (limiter only): no latency
Mazerider : no latency
Bark filter (limiter only) : no latency
Barricade: 2.7ms
MBC (one band 20:1 ratio) : 1ms

I’ve found that almost all dedicated limiters I own are useless for live use, will be great having this database expanded, would like to know the latency numbers for fabfilter and ape soft limiters…

wim

You can either have latency or you can have distortion (color). You can't eliminate both.

The only way to reduce peaks smoothly is to know they're coming ahead of time. The waveform has to be collected, analyzed, and acted on in the plugin before it can be output. Without a lookahead buffer the only way to reduce sound above the threshold is to abruptly clip it.

offbrands

Omitting FACs Maxima, there’s like 2.7ms of latency averaged of these you posted.

Asking with respect to fully understand - What kind of effect does 2.7ms produce have when you’re playing live that you’re trying to avoid?

Or is this a chain effect kind of thing that compounds that I’m not aware of in your signal chain?

Is no latency a requirement of some kind I’m not familiar with when playing live?

I’m genuinely curious! Would be helpful to know what you’re after 🙏🏽

wim

Probably the best bet is the lowest latency limiter as only an insurance policy, but ideally lowering the gain enough going into it so that it never needs to activate.

fwiw, that's basically the same as working with no limiter because the hardware isn't going to output above 0db anyway. It's going to clip just like a limiter. Though the right limiter might do it in a more musically acceptable way.

wim

I believe AUM's measurements depend on the plugin reporting its latency properly. Not all do. The only way to really know is to do some other kind of external measurement.

Synthi

@wim said:
You can either have latency or you can have distortion (color). You can't eliminate both.

The only way to reduce peaks smoothly is to know they're coming ahead of time. The waveform has to be collected, analyzed, and acted on in the plugin before it can be output. Without a lookahead buffer the only way to reduce sound above the threshold is to abruptly clip it.

You are right in general terms, but once you left the surface, In my experience as a mastering engineer you can have transparent limiting with near or no latency. I think would be different ways to do that, DSP researchers have its own recipes including prediction of audio. Also the kind of distortion , what frequencies, phases etc will determine how transparent will sound to our ears.

wim

@Synthi said:

@wim said:
You can either have latency or you can have distortion (color). You can't eliminate both.

The only way to reduce peaks smoothly is to know they're coming ahead of time. The waveform has to be collected, analyzed, and acted on in the plugin before it can be output. Without a lookahead buffer the only way to reduce sound above the threshold is to abruptly clip it.

You are right in general terms, but once you left the surface, In my experience as a mastering engineer you can have transparent limiting with near or no latency. I think would be different ways to do that, DSP researchers have its own recipes including prediction of audio. Also the kind of distortion , what frequencies, phases etc will determine how transparent will sound to our ears.

I'm not gonna argue with a mastering engineer, but I think I'm right in general terms. Yes, certain limiters will mask it better and will sound better to your ears than others, but with no latency, there will be distortion in the digital realm. Analog gear is a different story.

Best of luck in your search, for sure. 👍🏼

(prediction of audio is a good point, and one I was going to muse about. I don't know if we have any limiters that sophisticated on iOS)

Synthi

@offbrands said:
Omitting FACs Maxima, there’s like 2.7ms of latency averaged of these you posted.

Asking with respect to fully understand - What kind of effect does 2.7ms produce have when you’re playing live that you’re trying to avoid?

Or is this a chain effect kind of thing that compounds that I’m not aware of in your signal chain?

Is no latency a requirement of some kind I’m not familiar with when playing live?

I’m genuinely curious! Would be helpful to know what you’re after 🙏🏽

Of course that latency will be added to the actual latency of the converter and the buffer size used. 2.7ms by itself is not much but I’m usually at the limit of the “playable zone” with 128 samples buffer , IME 64 samples is not stable enought for live.
I think it depends of each people and the instrument they play. Some instruments have a very sharp and latency will be most obvious. A couple of times when recording at the studio I needed to set the buffer size to 32 samples because the musician couldn’t concentrate because the delay. I’m fine with 64 samples , 128 is a bit uncomfortable but I can tolerate it.

Synthi

@wim said:
I believe AUM's measurements depend on the plugin reporting its latency properly. Not all do. The only way to really know is to do some other kind of external measurement.

Yes I was thinking about that too! Maybe some plugins are not reporting latency to the host?? That would be a potential problem with phases when using that plugins in parallel processing….

wim

I still don't understand why it's impractical to adjust levels appropriately so that limiting only kicks in on rare exceptions. Are you trying to add have a more squashed sound in addition to just avoiding clipping?

wim

Maybe a dynamics processor with a soft-knee is a better choice. That way you can get softer and less noticeable transition into peak reduction, while still preventing exceeding the threshold. Maybe.

offbrands

@Synthi said:

@offbrands said:
Omitting FACs Maxima, there’s like 2.7ms of latency averaged of these you posted.

Asking with respect to fully understand - What kind of effect does 2.7ms produce have when you’re playing live that you’re trying to avoid?

Or is this a chain effect kind of thing that compounds that I’m not aware of in your signal chain?

Is no latency a requirement of some kind I’m not familiar with when playing live?

I’m genuinely curious! Would be helpful to know what you’re after 🙏🏽

Of course that latency will be added to the actual latency of the converter and the buffer size used. 2.7ms by itself is not much but I’m usually at the limit of the “playable zone” with 128 samples buffer , IME 64 samples is not stable enought for live.
I think it depends of each people and the instrument they play. Some instruments have a very sharp and latency will be most obvious. A couple of times when recording at the studio I needed to set the buffer size to 32 samples because the musician couldn’t concentrate because the delay. I’m fine with 64 samples , 128 is a bit uncomfortable but I can tolerate it.

Thank you for explaining! Green over here but did have an idea that it was more compounded considering signal chains and the like, to the lay person, 2.7 ms is like huh? That should be fine? But in that context it makes way more sense. I hadn't thought about samples either so thank you!

wim

FAC Maxima is indeed whacko tho innit? I ended up checking all my other limiters after checking FAC a few days ago because I thought surely I must have done something wrong.

Interestingly, the reported latency didn't decrease at higher sample rates or lower buffers. I would have thought the lookahead would be in buffers, and so the latency would go down, but it's always the same. Either it's not reporting the latency correctly, or they're going to the trouble to keep the lookahead at a fixed ms by capturing more buffers.

Synthi

@wim said:
I still don't understand why it's impractical to adjust levels appropriately so that limiting only kicks in on rare exceptions. Are you trying to add have a more squashed sound in addition to just avoiding clipping?

@wim said:
Maybe a dynamics processor with a soft-knee is a better choice. That way you can get softer and less noticeable transition into peak reduction, while still preventing exceeding the threshold. Maybe.

Yes win, I do the gain staging the best I can in advance, but since my projects involved, not only AUV3 instruments but also voice, some live percussion , electric violin and several miced instruments is not that easy. The ice on the cake is that most of the performance is improvised so I need to tame and control dynamics as much as possible with the most transparent sound. Of course I’m using compressors too, either in tracks or groups, rarely in the master channel. Also, when mastering sometimes using a couple of limiters in series gave me good results and more transparent and dynamic sound for music with large dynamic range.

Synthi

@offbrands said:

@Synthi said:

@offbrands said:
Omitting FACs Maxima, there’s like 2.7ms of latency averaged of these you posted.

Asking with respect to fully understand - What kind of effect does 2.7ms produce have when you’re playing live that you’re trying to avoid?

Or is this a chain effect kind of thing that compounds that I’m not aware of in your signal chain?

Is no latency a requirement of some kind I’m not familiar with when playing live?

I’m genuinely curious! Would be helpful to know what you’re after 🙏🏽

Of course that latency will be added to the actual latency of the converter and the buffer size used. 2.7ms by itself is not much but I’m usually at the limit of the “playable zone” with 128 samples buffer , IME 64 samples is not stable enought for live.
I think it depends of each people and the instrument they play. Some instruments have a very sharp and latency will be most obvious. A couple of times when recording at the studio I needed to set the buffer size to 32 samples because the musician couldn’t concentrate because the delay. I’m fine with 64 samples , 128 is a bit uncomfortable but I can tolerate it.

Thank you for explaining! Green over here but did have an idea that it was more compounded considering signal chains and the like, to the lay person, 2.7 ms is like huh? That should be fine? But in that context it makes way more sense. I hadn't thought about samples either so thank you!

I remember the early days of digital mixing and the affordable Yamahas. Not sure about the numbers but the first pro mix 01 was like 4ms latency and lot of musicians complained about that because was obvious to them, used to instant analog recording. I think to remember the 03 was a step forward with less than 3ms latency.