Helix - I/O latency measurements

[Orvillain]

Fwiw, it's 1.8ms (well, 1.791666...).

Yes good shout. Reaper not being thorough enough with its display, as you can see in my screenshots. I'll calculate the true figures and update the posts.

 

[Sascha Franck]

Fwiw, HX Stomp, completely empty preset, no input gate, no input pad: 68 samples @ 44.1kHz, so something like 1.54ms.

 

[Sascha Franck]

Cali IV Rhythm I /w cab (both at Line 6 defaults) adds 15 samples (at 44.1), so that's 0.34ms for that block. Just the amp is adding 9 samples (0.2ms), cab block on its own is adding 9 as well, so in case you're using an amp&cab block you seem to be saving 2-3 samples of latency - not worth even talking about, possibly not even in case you're Steve Vai.
Overall, these are pretty decent figures.

 

[Orvillain]

Cali IV Rhythm I /w cab (both at Line 6 defaults) adds 15 samples (at 44.1), so that's 0.34ms for that block. Just the amp is adding 9 samples (0.2ms), cab block on its own is adding 9 as well, so in case you're using an amp&cab block you seem to be saving 2-3 samples of latency - not worth even talking about, possibly not even in case you're Steve Vai.
Overall, these are pretty decent figures.

Yeah I think the internal processing latency is always more or less going to be minimal and acceptable.

I was most interested in it from a "integrating a bunch of additional pedals" perspective.

 

[Sascha Franck]

I was most interested in it from a "integrating a bunch of additional pedals" perspective.

Similar here. Using a loop seems to pretty much exactly double the device latency (which is obviously making sense as you're dealing with exactly twice the components and processes), unless you're using the mix function (which I haven't tried myself as in my use case it's more about adding dirt and whatever pedals serially).

 

[Boudoir Guitar]

Is it fair to say then.... slap your analog effects into the loops, and keep your digital effects out of them altogether??? Is that a good rule of thumb?

I’d say the better rule of thumb would be use pedals that can be run wet-only in the loops is the better rule of thumb

 

[Vic Winters]

What about the poly effects and the different modes (XFast vs XStable)?

 

[Sascha Franck]

Fwiw, as I was just measuring some stuff, I thought I could do so for some more things on my board:
- Zoom MS-50G, empty patch: 1.1ms
- NUX Amp Academy, amp /w MPT IR: 1ms (actually 0.93)
- Atomic Amplifirebox MkI, amp /w MPT IR: 1.9ms

 

[Sascha Franck]

What about the poly effects and the different modes (XFast vs XStable)?

Poly Pitch, default setting (+7):
- X Stable: 15.1ms
- X Fast: 14.2ms
I actually don't get this. I repeated the tests and while I was using a sharp guitar attack as an input rather than a scientifically properly suited signal, the results seem to somewhat vary from one recording to the next. Also, while playing through Poly Pitch, the differences feel as if they actually were bigger.

 

[DLC86]

Poly Pitch, default setting (+7):
- X Stable: 15.1ms
- X Fast: 14.2ms
I actually don't get this. I repeated the tests and while I was using a sharp guitar attack as an input rather than a scientifically properly suited signal, the results seem to somewhat vary from one recording to the next. Also, while playing through Poly Pitch, the differences feel as if they actually were bigger.

Perfectly normal for a pitch shifter, it probably plays back the note only once it has detected the pitch of the original note, and the speed with which it does this depends on various things (even the pitch of the note itself)

 

[Sascha Franck]

Perfectly normal for a pitch shifter, it probably plays back the note only once it has detected the pitch of the original note, and the speed with which it does this depends on various things (even the pitch of the note itself)

You're describing an intelligent pitch shifter.
Poly Pitch is not that. And it doesn't analyse the input pitch(es). That be almost like a realtime Melodyne DNA. Which Poly Pitch also isn't.

Seriously, you're not trying to tell me that PP is detecting the pitches in a 4 part chord, are you?

 

[Orvillain]

You're describing an intelligent pitch shifter.
Poly Pitch is not that. And it doesn't analyse the input pitch(es). That be almost like a realtime Melodyne DNA. Which Poly Pitch also isn't.

Seriously, you're not trying to tell me that PP is detecting the pitches in a 4 part chord, are you?

It is a poly pitch algorithm. It is doing a lot more than your typical monophonic pitch algorithm.

 

[Sascha Franck]

It is a poly pitch algorithm. It is doing a lot more than your typical monophonic pitch algorithm.

I know.
I'm still calling BS on its latency (!) being dependant on the kind of input signal.
But that should be absolutely trivial to prove (or disprove). Hence, I'll be doing so later on, when I'm back home.

 

[DLC86]

Seriously, you're not trying to tell me that PP is detecting the pitches in a 4 part chord, are you?

Tell me how can it shift a 4 part chord without detecting the pitch of every single note played (no, accelerating/slowing the audio and lenghtening/shortening its duration by a fixed ratio doesn't really work for polyphonic pitch shifting)

EDIT: well, after thinking about it, I think it doesn't really work for mono pitch shifting either without pitch detection, otherwise the lenght of the splices isn't synced with the fundamental frequency, so you'll create beatings and perceive a different frequency mixed in with the signal (kinda like a ring modulator)

 

[Sascha Franck]

Tell me how can it shift a 4 part chord without detecting the pitch of every single note played

Multiplication/Division. That's all.

 

[dk_ace]

Interesting, some of that stuff added up to more latency than I would have expected.

I gigged with the helix using all the loops for a while with a mix of analog and digital pedals. I’m down to just using two now as a stereo pair for the Ventris. I run it 100% wet though so that it isn’t adding any latency.

D

 

[quadgear]

Even pitch shifting one note is going to have large latency due to the wavelength (time) of the low E string unless you are limiting it to octaves up/down


You think you can do better? It's not as straightforward as it appears. Build it before calling BS.

 

[DLC86]

Multiplication/Division. That's all.

Yeah, sure...

 

[Sascha Franck]

Yeah, sure...

Indeed.
How do you think an Octavia/Octafuzz works? Does it analyze the input pitch as well?

 

[Sascha Franck]

Even pitch shifting one note is going to have large latency due to the wavelength (time) of the low E string unless you are limiting it to octaves up/down

Armchair expertise mucho?