Latency issues using I/O 3/4 for FX loops

D.B. Walker

Inspired
Did a little searching ... could not find a thread on this. It may fall into the "You are doing it wrong" but ...

I have loops 3 and 4 setup, and it has introduced latency in my patch. Measuring by ear, I'd say around 8 to 10 ms. Just enough such that I can hear it, and more than enough to feel.

Am I SOL, or ... am I doing it wrong?
 
Measuring by ear? The actual value is about 1/10th amount if both are on at once.
 
Is the amount of latency affected by the number of blocks before or after the loop? I'm running the CPU at about 80% on my kitchen sink patch... I'm sure I would not be able to hear .8ms ... generally I can feel about 5ms and hear around 6ms.

Another contributing factor is my wireless adds 4ms ... (Shure GLDX) ... that on top of the loops pushed it into the hear range.
 
I did extensive research into human auditory perception abilities in grad school and never came across anyone who had enough temporal resolution abilities to perceive sub 10ms auditory gap perception out of 100s of volunteers. I don’t buy that anyone can “measure” 8-10ms latency by ear as that simply is super human.....
 
Is the amount of latency affected by the number of blocks before or after the loop? I'm running the CPU at about 80% on my kitchen sink patch... I'm sure I would not be able to hear .8ms ... generally I can feel about 5ms and hear around 6ms.

Another contributing factor is my wireless adds 4ms ... (Shure GLDX) ... that on top of the loops pushed it into the hear range.
The number of block before or after does not make a difference. Each loop adds about 1/2 ms IIRC. It's certainly not 10 ms. If you are using both loops along with the analog in and out then it's around 3ms. Add 4ms from your wireless and now you're up around 7ms. Don't forget that a lot of monitors now have DSP, A/D and D/A converters in them which adds even more latency. Then there's your audio interface which can add latency.
 
Yeah ... stacking up all the points in the system sounds about right. I'm suing presonus monitors, and UA Apollo, add the wireless and then that extra from the loops and there ya go, getting close to 8ms. The loops are the straw the broke the wireless back ;) In a war between the wireless and the loops, the loops will win out.
 
Do the Send / Return blocks add any additional latency if you want to "artificially" increase how many blocks you can have in your signal?
 
Interesting topic here. I guess if you went out digital then you could reduce a stage of DA conversion
 
No.

If I could sense 1/2 ms I’d be on the cover of magazines.

I was just wondering for curiosity's sake.

I suppose you could test it by splitting your signal and shunting one straight to the output, then shunting one through the Return / Send block and back to the same output and see if you hear any phasing, like this.

1573831200286.png

I can't actually test at the moment because I'm remoted into my home PC where my Axe-Fx is on. I suppose I could setup some synth tone generated Tone Match contraption if I really wanted but I don't really feel like doing that right now, hah. I'll test it when I get home.
 
Last edited:
No.

If I could sense 1/2 ms I’d be on the cover of magazines.


a lot of people can’t sense less than about 40ms, and with some degree of hearing damage it gets even worse. Around 10ms is about as fine of temporal resolution as the human auditory system can respond to.

I think it always helps to put things into perspective by comparing to the distance sound travels in a second, roughly about 1100 feet per second, so about 1 foot per millisecond, right?
If people could really detect single millisecond differences in latency it would be like saying you need to move your amp 3 feet closer because that 3ms extra time it’s taking to reach your ears is affecting your playing. No one can estimate distance down to the foot and no one can detect single digit latency. At least no one that’s ever been tested with objective and quantified measures. It would be superhuman.
 
There are some things to consider. Humans can hear Phase relationships quite easily. Just take your speakers, put a phase switch on one and flip it ... this is the type of thing that introducing latency on one side of a stereo signal can introduce and very low numbers.

When I say I can feel around 5ms ... its true ... me switching on a 5ms delay in my sound causes a mushy, disconnected feeling between my pick and the sound ... its minor at 5ms, but I can feel it.

When I say I can hear 8 - 10ms, that's when I can hear both sources at once. Again, if you take one speaker of a stereo pair and delay one by 8ms you will hear a lot of phase relationships get affected, and a perceived slap delay and shift in the stereo field.

I practice at a low enough volume such that I can hear the acoustic sound of the guitar, its the delay from the acoustic sound to the speakers sound that is bothersome. In a live situation its common to have more legacy introduced by distance and its easy to get used to because you can't (normally) hear the acoustic sound of your instrument.

Cliff made me consider all the things in my chain introducing latency, when I stack them all up vs the acoustic sound of the guitar on my body:
Wireless: 4ms
UA Interface: 2ms
Axe: 2ms
Distance: 8ft (about what ... 2ms?)

It gets into the perceivable range.
 
I did extensive research into human auditory perception abilities in grad school and never came across anyone who had enough temporal resolution abilities to perceive sub 10ms auditory gap perception out of 100s of volunteers. I don’t buy that anyone can “measure” 8-10ms latency by ear as that simply is super human.....
You cant hear it but you can feel it.
 
You cant hear it but you can feel it.
Yeah, exactly... picking starts to get squishy over 7mS - 10mS, even if you can adjust to it, when you go play in front of a real amp it's obvious by comparison. Micing speakers through latency is also a comb filtered mess but thankfully direct monitoring solves it mostly.
 
Old thread being resurrected, but yes, using send/return does add some latency.

Thanks.

I've tested this since that last post of mine and found that doing this does add some lag if you place the Send block to the right of the Return block on the grid. It's not enough of a difference to perceive by itself but it does mess with phase if you blend the final signal with an unbroken left-to-right signal routing on the grid in the same patch.
 
Yeah, a return to the right of the send isn't really a send/return...it's a shunt :).

The latency amount IIRC is one process block, so on the Axe-FX III, that should be 16 samples.
 
Back
Top Bottom