Line 6 Helix Stadium Talk

[Sascha Franck]

That time Sascha confused milliseconds for minutes.

Tell that to Samsung's autocorrect.

 

[Orvillain]

See, I have an idea about how pitch shifting works without studying DSP. You're studying DSP and apparently still get things wrong.

I didn't get anything wrong. You just want to take what I said literally, in order to avoid embarrassment. Laughable, stupid, and ignorant.

Tell that to Samsung's autocorrect.

Nah, I'll tell it to Sascha's complete inability to proof read before he posts.

 

[AlbertA]

 

[laxu]

I fully understand what I am posting.


Absolutely nothing. I owe you nothing. I've educated you. It is up to you to research it further if you really want to know how these things work.

If you go back and re-read what I have been saying, you'll understand that there are multiple ways to perform pitch shifting, and not all of them would require 23ms to represent an E0. However, that doesn't change the absolute fact that an E0 is 23ms long - 24.2ms to be exact - for a single cycle.

If you're pitching shifting on a pitch detection basis - whether you use autocorrelation, zero-crossing analysis, FFT peak-tracking - then you absolutely need to have at least one full cycle of the waveform to be able to analyse the pitch. This does not mean that latency would change from pitch to pitch. That simply isn't how it works and belies a gross misunderstanding and ignorance on your part.

If you are using a time-domain granular approach, phase-vocoder shifting, or resampling approaches, then you're avoiding pitch detection altogether. Which means you can reduce the latency. This is what the Whammy does. It is also probably what Helix does.

As I said in a previous post - there are many ways to perform this operation. But there are always trade offs. Latency is always a factor, but some approaches are worse than others. The lower latency approaches have issues with phase reconstruction and sound quality.

Can you improve pitch detection latency simply by deciding you are supporting e.g only standard guitar range or maybe just high pass filtering the signal before detection? If I understand it correctly, this would allow for faster pitch detection but would produce errors if you are using a downtuned guitar, extended range guitar etc.

 

[Orvillain]

Can you improve pitch detection latency simply by deciding you are supporting e.g only standard guitar range or maybe just high pass filtering the signal before detection? If I understand it correctly, this would allow for faster pitch detection but would produce errors if you are using a downtuned guitar, extended range guitar etc.

Pitch detection latency, yes you can save some time by only supporting an expected range, and anything outside of that range gets filtered out.

I'm not an expert, but from my brief reading on the topic:

Autocorrelation approaches - latency is high, because you need 1 or more full cycles. It sounds natural, preserves transients, and is good for monophonic sources. But it needs pitch detection, and is unstable on noisy inputs.

Zero-crossing detection approaches - high latency, but it is a fast and cheap form of pitch detection. Extremely inaccurate though.

FFT peak tracking approaches - Latency is at least as high as the FFT window you use. It is more robust on complex waveforms, but may mistrack harmonics.

Granular approaches - Low latency, low-ish CPU requirements. But sounds grainy and phasey, with smeared transients. Think of the old school AKAI timestretch algorithm.

Waveform simularity - Medium amount of latency. Fewer artifacts. Still has time smearing and very difficult to make work on a range of source material.

Resampling - Low latency. Quality is okay, depending on the implementation. But it can smear transients.

Classic Phase Vocoder - Medium latency. Could be in the region of 5-25ms. Smears transients, has a bit of a robotic sound.

As I understand it, there are also neural network based solutions. But I don't know much about those.

I've implemented the granular approach and the FFT peak tracking approach. I've built a resampler, but haven't used it for pitch shifting exactly. But I had to build something that could manage fractional buffer reads and interpolation in order to make my BBD delay plugin. So turning that into a pitch shifter wouldn't be beyond the wit of man.

 

[Karl Houseknecht]

I'm definitely have issues with Clarity acting like there's a fuzz pedal somewhere in the chain. I cannot figure out what's happening, but it did not happen with my OG Helix.

Okay, I’m not alone. After an evening of playing and dialing in high gain sounds, I returned to my AC30 preset that I’d dialed in for clean, and it was overdriven with nasty, harsh clipping. There is something amiss with this firmware.

 

[Sascha Franck]

I didn't get anything wrong

This is wrong. At the very least, your latency time is dictated by the lowest note you want to support. Because you need at least a full cycle in order to correctly detect the pitch, in order to perform any shifting in the first place.

A low E on bass? 23ms minimum.

This is just bogus. And yes, you *did* generalise it.

 

[Orvillain]

This is just bogus.

Incorrect.

And yes, you *did* generalise it.

I never said otherwise.

Stop fighting. Start learning. Go read that paper I posted.

 

[Sascha Franck]

Go read that paper I posted.

Absolutely zero need. You don't need 23ms to apply pitch shifting to an E1. That's all I need to know to prove your first statement in this discussion is bogus.

 

[Orvillain]

Absolutely zero need. You don't need 23ms to apply pitch shifting to an E1. That's all I need to know to prove your first statement in this discussion is bogus.

See how you're latching onto that one statement, and refusing to see the forest for the trees?

That's you, that is.

 

[Lorne Malvo]

Any chance you guys could take the pissy pitch-shifting argument to Thunderdome ?

The Gear Forum

Guitar Gear Discussion Forum

thegearforum.com

 

[molul]

I've always heard the "you need one full cycle to properly detect pitch", and I never understood why it can't be done in another way (I don't know the technical language, so please bear with me xD): start reading the wave, mark the first direction change, then wait for the next.

Worst case scenario (you start reading the wave right after a change of direction), you've needed a full cycle as always. Best case (you start reading right before a change of direction), half.

Maybe that's already being done, though.

 

[Orvillain]

Any chance you guys could take the pissy pitch-shifting argument to Thunderdome ?

The Gear Forum

Guitar Gear Discussion Forum

thegearforum.com

NEEEEEVVVVEEEERRRRR!!

... I mean, sure thing. Feel free to clean up the last couple of pages.

 

[EOengineer]

If that were true, then your idea of how DSP works wouldn't be reducible to a fucking bumper sticker.

 

[fiftyshadesofslay]

Okay, a few thoughts on the Helix Stadium XL now that I’ve fired it up, before I wade into the pitch shifting debate…

It sounds and feels amazing. Line 6 have absolutely knocked it out of the park for sure. The Agoura models sound great to my ears. I’m normally a JCM 800 user, but the Bogner Blue also sounds incredible.

Feel wise it feels great to play, I don’t know what it is about some modellers but some just feel good. Is it the converters, compression? Dunno. All I do know is this works for me, on par with the Fractal and Kemper units. I always felt a bit detached when playing the OG Helix, QC is the same.

Even the legacy stuff sounds better. I’m sure it’s the same algorithms, but maybe it has better converters now, which translate to a better sound overall? It’s way more 3D and bigger now.

A few minor gripes… I’d definitely prefer an option to use USB to connect to the editor, I’ve got crap Wi-Fi in my music room. You lose sound for a second or so when changing amps to try them out. Doesn’t make much difference to anything but was surprising. Finally, where can I get a decent backpack for this thing? I think the new one looks just fine, but it’s out of stock in the UK and I’ve got gigs coming up!

 

[DG80]

Would suggest not going for the official backpack; every review I've seen has shown it to be pretty poor (certainly not in the same league as the original). I actually cancelled my pre-order for it earlier - I'll find something else until they release v2.

 

[EOengineer]

Would suggest not going for the official backpack; every review I've seen has shown it to be pretty poor (certainly not in the same league as the original). I actually cancelled my pre-order for it earlier - I'll find something else until they release v2.

We must be looking in different places because most reviewers I’ve seen who actually bought and received the Stadium bag have liked it. I’m seeing a lot of negative conjecture from people who don’t own it.

 

[jellodog]

but maybe it has better converters now, which translate to a better sound overall? It’s way more 3D and bigger now.

Could also be the greater dynamic range on the Stadium input? 123 dB vs 126 dB ?

 

[Sascha Franck]

See how you're latching onto that one statement, and refusing to see the forest for the trees?

Nonsense. It's the single statement you've used to enter this discussion and I called it what it is, namely bogus. Only later you decided to come up with whatever "it's not the only way...", so don't accuse me, it's absolutely just you who started with such a (scientifically) stupid generalisation.
And instead of growing a pair you decided to call me names and what not. And it's not for the first time. I generally have a lot of respect for you but in this discussion you acted as if you were an utter douchebag. Perhaps you are.

 

[laxu]

Nonsense. It's the single statement you've used to enter this discussion and I called it what it is, namely bogus. Only later you decided to come up with whatever "it's not the only way...", so don't accuse me, it's absolutely just you who started with such a (scientifically) stupid generalisation.
And instead of growing a pair you decided to call me names and what not. And it's not for the first time. I generally have a lot of respect for you but in this discussion you acted as if you were an utter douchebag. Perhaps you are.

Yet following from the sidelines...you haven't provided any evidence to counter what he said. It basically amounts to "nuh-uh, you're wrong".

Your initial premise was that just more CPU horsepower will allow for faster latency detection. Isn't the whole issue that you have to wait for the full cycle of that note for it to be reliably detected as E0? Thus extra horsepower would do nothing as the system has to wait for that 24.2ms to be able to reliably discern "that's E0!"

You were at any time able to jump in and say "well if you use this other method then it can be faster" but you never did.

Could e.g machine learning based note detection do it faster? Maybe! I have no idea. If you do, please educate us.