High Latency as USB Interface

[mark_gamache]

This has been discussed on other threads, but I would like to start one dedicated to the topic of using the Helix as a low latency interface. I've not been able get < 10ms latency running any sample and/or buffer settings. This has been supported by others reporting similar results.  To get close you have to run 96K/64 samples. Running 96K/64 results in ~10.6ms of latency, but because you have to run 96K/64 samples it puts pressure on the computer I/O and creates buffer under-runs (the random noise you hear if the buffer is not filled). At a more reasonable setting of 48K/64 samples the latency is 13.2ms (all measured by the RTL utility). Note: my cheap Presonus USB interface is less than half that latency at the same settings. 

 

Question #1: Is this due to the latency compensation built-in for each external loop (to account for the analog-to-digital latency on each loop)?

Question #2: Are there any tricks or settings we can use to reduce the latency 

 

I know I can monitor in near real-time via the helix, that's not what I am looking for here. 

 

thanks for any insight people might have,

mark 

(note: I am using a Windows 10 laptop with an i7 processor and 32Gigs Ram)

 

 

[rd2rk]

In this thread:

 

 

I'm over on TGP fighting off the "I don't know but here's my opinion because you're an idiot to want to know" guys.

It's hard, annoying work, but I have every confidence that eventually Digital Igloo (the only person who ACTUALLY KNOWS) will respond with a definitive and official answer to these questions. If you FOLLOW the above thread, that's where I'll be posting it when he does, and we can once and for all dispense with these "It's probably been asked before, but..." threads. Right......

 

Anyhow, I don't mean to be snippy, just trying to drag a bit of organization out of the chaos.....

[mark_gamache]

Thanks for tracking this down. We def should seek a definitive answer as this will continue to come up. 

[SaschaFranck]

23 hours ago, rd2rk said:

So, cutting to the chase, you think it entirely possible that he's getting 5ms RTL glitch-free at 64spls on his 60 euro Behringer and i5 laptop?

 

I'm allowing myself to answer your question from TGP here. The answer is: No. And those 3-5ms the one guy is claiming are an illusion, no less. Maybe some host will report such numbers, but they're wrong.

How would I know? Because I had one of those Behringers over here for a testride (a mate bought it - but sent it back after some tests at his and my place). 64 samples buffersize might work glitchfree in case you don't tax your computer hard, but that still depends on your host as well (with Logic it's trivial - but Logic is still one of the best in terms of balanced low latency performance) and you will likely not enjoy it too much on an i5.

Fwiw, from all I remember, I got around 8ms at 44.1kHz and 64 samples buffersize. Not all that bad for a €60 interface but defenitely not that great, either.

[rd2rk]

27 minutes ago, SaschaFranck said:

 

I'm allowing myself to answer your question from TGP here. The answer is: No. And those 3-5ms the one guy is claiming are an illusion, no less. Maybe some host will report such numbers, but they're wrong.

How would I know? Because I had one of those Behringers over here for a testride (a mate bought it - but sent it back after some tests at his and my place). 64 samples buffersize might work glitchfree in case you don't tax your computer hard, but that still depends on your host as well (with Logic it's trivial - but Logic is still one of the best in terms of balanced low latency performance) and you will likely not enjoy it too much on an i5.

Fwiw, from all I remember, I got around 8ms at 44.1kHz and 64 samples buffersize. Not all that bad for a €60 interface but defenitely not that great, either.

 

The guy who posted those RTL numbers made it clear that he didn't care whether we believed him or not. I don't want to call him a liar, so just color me "skeptical".

The question you're responding to was, to be honest, an attempt to get JR to use his amazing command of technical jargon to do it for me. I'm so evil.....

[SaschaFranck]

1 minute ago, rd2rk said:

The question you're responding to was, to be honest, an attempt to get JR to use his amazing command of technical jargon to do it for me. I'm so evil.....

 

Yeah, I know. Fwiw, Jim really knows his stuff. I know him from pretty much way back in the days when mailing lists were still en vogue.

[rd2rk]

2 minutes ago, SaschaFranck said:

 

Yeah, I know. Fwiw, Jim really knows his stuff. I know him from pretty much way back in the days when mailing lists were still en vogue.

 

I totally agree that he is technically very knowledgeable. and I've learned quite a bit from him in previous threads.

 

In typical TGP fashion, most of the respondents in this thread, including JR, had no business getting involved.

My purpose in starting the thread was clearly stated. It was a simple not very technical question addressed at someone (DI specifically) from Line6.

If DI would respond "Yes, phil_m is right, we're not interested in developing low latency drivers" or "No, the HW isn't really designed for low latency performance", that would be, as I stated, "All she wrote", "End of story". But, being TGP this will probably drag on for 12 more pages if DI doesn't do that, and 20 more pages if he does!

[mark_gamache]

FWIW I think the guy is correct on his reporting of his RTL results. Given how that utility probably works, it might be possible to get a result that is that low if the driver provides extremely low buffer options. We've all had results that are technically possible, but not usable in practice. That (like other interfaces) probably isn't usable until it is set to 32 or 64 samples.  Not sure this advances the understanding of the Helix latency. 

Like others, I want a Line 6 engineer to say it has high latency due to reason X. It might be technical or business related. Maybe it's something that could be fixed, maybe not. 

[SaschaFranck]

5 hours ago, mark_gamache said:

FWIW I think the guy is correct on his reporting of his RTL results. Given how that utility probably works, it might be possible to get a result that is that low if the driver provides extremely low buffer options.

 

Ok. Under OSX there's no way to set the driver lower than to 32 samples (only works in class compliant mode anyway) - so it's perhaps a rather experimental setting achieved through a Windows driver or ASIO4All.

However, as Jim Roseberry pointed out, there's pretty much zero chance the unit would run even remotely properly at anything below 32 samples - and even at that value, it was already causing a lot higher CPU hit under OSX, compared to 64. And even if it was running at lower buffersize values, my tests indicated that there's been a rather high safety buffer at work (very typical for cheaper interfaces), so from one point on, lowering the buffersize in your host results in diminishing returns as that won't have any effect on the safety buffers (and converter latency, too).

In a nutshell, 3-5ms simply isn't realistic, likely not even happening at all on those Behringer interfaces.

[StarF666]

How about putting low-latency drivers as a request on Ideascale? At least if DI says it's not possible/restricted by HX hardware.

[mark_gamache]

3 hours ago, StarF666 said:

How about putting low-latency drivers as a request on Ideascale? At least if DI says it's not possible/restricted by HX hardware.

 

It looks like there already is one...

 

https://line6.ideascale.com/a/dtd/Low-Latency-driver-for-Windows/945091-23508

[gunpointmetal]

i7/32GB doesn't mean much if it's a 3rd Gen or something. Out of curiosity, is this something you're testing just to test, or is there a real-world use for monitoring through the DAW instead of off the hardware? Vocals through the mic input?

[rd2rk]

21 minutes ago, gunpointmetal said:

i7/32GB doesn't mean much if it's a 3rd Gen or something. Out of curiosity, is this something you're testing just to test, or is there a real-world use for monitoring through the DAW instead of off the hardware? Vocals through the mic input?

 

If your Helix is all you ever use, then NO. Direct (hardware) monitoring is obviously (unless you're a total noob) the way to go.

If you occasionally like to mess with VST amp sims, or have specific VST processing needs that you want to hear while playing then YES.

The point of all this is that L6 Helix marketing heavily stresses the AI capability.

The high latency relative to dedicated AIs will continue to surprise new users who expect better.

If better latency is possible, it would be nice if L6 would either make it happen or say why it ain't gonna happen.

An official statement would at least give us something to point at for new users who don't first do a search for "latency" before starting a new thread....

[gunpointmetal]

Good luck! I'm guessing you're not gonna get an answer on whether or not low-latency drivers are coming because even if they were, they wouldn't tell anyone till they were ready, and the use cases where direct monitoring wouldn't be preferable are probably <1%. 

[rd2rk]

1 minute ago, gunpointmetal said:

Good luck! I'm guessing you're not gonna get an answer on whether or not low-latency drivers are coming because even if they were, they wouldn't tell anyone till they were ready, and the use cases where direct monitoring wouldn't be preferable are probably <1%. 

 

I agree that it's probably not high on their priorities list, but either "We're thinking about it", or "It's not happening" would be preferable to the current silence.

Even if the answer is "We're working on it", those of us who've been here a while know better than to expect an ETA, never mind a firm date.

Unless the answer is SPRING!...... :-)

[kylotan]

The very nature of the Helix means it's unlikely to ever be as low latency as a dedicated audio interface. The audio pipeline through the Helix is somewhat like running through VST effects on a computer, and as such any or all of the following can theoretically add latency:

• input processing (e.g. noise gate)
• any audio processing blocks (amp, effect, etc)
• routing a signal path to another signal path (because path 2 can't start processing until it has the data from path 1, etc)
• the global EQ

It's possible that disabling these will give you lower latency. It's also possible that it makes no difference and that the latency is 'baked in' due to the way the DSP is designed. But signal processing in the digital domain is almost never zero latency.

[rd2rk]

14 minutes ago, kylotan said:

The very nature of the Helix means it's unlikely to ever be as low latency as a dedicated audio interface. The audio pipeline through the Helix is somewhat like running through VST effects on a computer, and as such any or all of the following can theoretically add latency:

• input processing (e.g. noise gate)
• any audio processing blocks (amp, effect, etc)
• routing a signal path to another signal path (because path 2 can't start processing until it has the data from path 1, etc)
• the global EQ

It's possible that disabling these will give you lower latency. It's also possible that it makes no difference and that the latency is 'baked in' due to the way the DSP is designed. But signal processing in the digital domain is almost never zero latency.

 

Nor is anybody (in their right mind) asking for zero latency.

Nor does anybody (in their right mind) expect to match the performance of a $3000 dedicated interface.

But seriously - higher latency than a $100 Behringer?

[gunpointmetal]

I'd say driver stability was a higher priority than through-put processing latency, since one would expect that since this is a full-on guitar modeling processing, one would monitor off the hardware for recording purposes (even if the intention was to reamp later). I can't see a practical reason for it, but I also can't see a practical reason to use this big-lollipop floorboard as a dry interface for VST amp sims, or imagine what VST effect would be so revolutionary and different from what is in the box that one would haul a laptop along to do external processing in a live environment. 

[rd2rk]

6 minutes ago, gunpointmetal said:

I'd say driver stability was a higher priority than through-put processing latency, since one would expect that since this is a full-on guitar modeling processing, one would monitor off the hardware for recording purposes (even if the intention was to reamp later). I can't see a practical reason for it, but I also can't see a practical reason to use this big-lollipop floorboard as a dry interface for VST amp sims, or imagine what VST effect would be so revolutionary and different from what is in the box that one would haul a laptop along to do external processing in a live environment. 

 

Putting aside for the moment the price/performance ratio thing. One of the biggest complaints about Helix is the perception among some that the reverbs suck. There are many very high quality reverbs available as VSTs. And yes, people DO use laptops for live performance, for many reasons other than better reverbs or amp sims. People do lots of strange things. Not long ago I saw a guy at an "Art Show" who used a laptop and assorted MIDI controllers for guitar, keys, drums, and complex live looping. For an amp he mic'd his studio monitors!

Go figure! Awesome one man band performance BTW. 

[mark_gamache]

1 hour ago, rd2rk said:

 

Putting aside for the moment the price/performance ratio thing. One of the biggest complaints about Helix is the perception among some that the reverbs suck. There are many very high quality reverbs available as VSTs. And yes, people DO use laptops for live performance, for many reasons other than better reverbs or amp sims. People do lots of strange things. Not long ago I saw a guy at an "Art Show" who used a laptop and assorted MIDI controllers for guitar, keys, drums, and complex live looping. For an amp he mic'd his studio monitors!

Go figure! Awesome one man band performance BTW. 

 

Exactly, try some reverbs like this one:

http://www.zynaptiq.com/adaptiverb/

 

or even the old

SoundToys EchoBoy

https://www.soundtoys.com/product/echoboy/

 

and let me know if you think the Helix is on par. It's not.

 

<rant>

It's a fantastic unit, but it's a walled garden of sorts. Sure you can buy pedals and run them in loops, but if the external pedals are digital that's ridiculous.  In that case, you are running software on a DSP with it's own power / switches patch cables and AD conversion (also adding at least 3.5 ms of latency each loop).  The progress and innovation of guitar pedals has been amazing, but it's not keeping pace with audio plugins (again see the zynaptiq stuff and what that can do).  Keep the plug-in modal in mind as you wait patiently for Line-6 to release a new version to see if your favorite device is included. I want to run the Neural DSP stuff alongside Helix now. I want an EMT plate reverb or and echo from a Struder A800 or a Dimension D Stereo Chorus. If I have to lug a laptop along with the helix no problem, that's easy. As it stands, I pay a heavy latency penalty for trying this. In fact, it's less latency to get a fast interface and run it via the Helix loop (taking the extra AD hit).  

</rant>

 

 

 

 

[SaschaFranck]

2 hours ago, kylotan said:

The very nature of the Helix means it's unlikely to ever be as low latency as a dedicated audio interface. The audio pipeline through the Helix is somewhat like running through VST effects on a computer, and as such any or all of the following can theoretically add latency:

• input processing (e.g. noise gate)
• any audio processing blocks (amp, effect, etc)
• routing a signal path to another signal path (because path 2 can't start processing until it has the data from path 1, etc)
• the global EQ

 

All these are irrelevant for plain audio interface functionality. Using, say, RMEs Totalmix software doesn't affect the latency of their devices, either.

[rd2rk]

5 minutes ago, mark_gamache said:

and let me know if you think the Helix is on par. It's not.

 

 

OPINION noted.

 

5 minutes ago, mark_gamache said:

<rant>

It's a fantastic unit, but it's a walled garden of sorts. Sure you can buy pedals and run them in loops, but if the external pedals are digital that's ridiculous.  In that case, you are running software on a DSP with it's own power / switches patch cables and AD conversion (also adding at least 3.5 ms of latency each loop).  The progress and innovation of guitar pedals has been amazing, but it's not keeping pace with audio plugins (again see the zynaptiq stuff and what that can do).  Keep the plug-in modal in mind as you wait patiently for Line-6 to release a new version to see if your favorite device is included. I want to run the Neural DSP stuff alongside Helix now. I want an EMT plate reverb or and echo from a Struder A800 or a Dimension D Stereo Chorus. If I have to lug a laptop along with the helix no problem, that's easy. As it stands, I pay a heavy latency penalty for trying this. In fact, it's less latency to get a fast interface and run it via the Helix loop (taking the extra AD hit).  

</rant>

 

I want my KIA to be a Ferrari. I just looked out the window....nope, still a KIA. Been waiting since 2006 for the "Italian" update.

 

Seriously, not being a fanboi. It sounds like you've got the wrong device. From what I understand, if you're already licensed, the Quad Cortex will let you run the Neural DSP stuff.

Have you sent in your $200 yet?

 

[gunpointmetal]

I'm just the weirdo that thinks if the delay/verbs are fine for a live audience, I'm cool with it, and I can do whatever I want ITB when I'm recording. I promise, nobody on the other side of the sound system gives a dookie about which reverb is on your guitar signal. I do hope you find an answer to your question, and I also hope L6 doesn't have to spend any time that would have gone into modeling new FX creating low-latency drivers for all six of you that need them for some reason.

[rd2rk]

1 minute ago, gunpointmetal said:

I'm just the weirdo that thinks if the delay/verbs are fine for a live audience, I'm cool with it, and I can do whatever I want ITB when I'm recording. I promise, nobody on the other side of the sound system gives a dookie about which reverb is on your guitar signal. I do hope you find an answer to your question, and I also hope L6 doesn't have to spend any time that would have gone into modeling new FX creating low-latency drivers for all six of you that need them for some reason.

 

Personally, I hope that EVERYBODY gets what they want for Christmas.

BUT....I still think their dev time would be better spent on the Free Pony before they start work on the Unicorn!

[phil_m]

42 minutes ago, mark_gamache said:

 

Exactly, try some reverbs like this one:

http://www.zynaptiq.com/adaptiverb/

 

or even the old

SoundToys EchoBoy

https://www.soundtoys.com/product/echoboy/

 

and let me know if you think the Helix is on par. It's not.

 

<rant>

It's a fantastic unit, but it's a walled garden of sorts. Sure you can buy pedals and run them in loops, but if the external pedals are digital that's ridiculous.  In that case, you are running software on a DSP with it's own power / switches patch cables and AD conversion (also adding at least 3.5 ms of latency each loop).  The progress and innovation of guitar pedals has been amazing, but it's not keeping pace with audio plugins (again see the zynaptiq stuff and what that can do).  Keep the plug-in modal in mind as you wait patiently for Line-6 to release a new version to see if your favorite device is included. I want to run the Neural DSP stuff alongside Helix now. I want an EMT plate reverb or and echo from a Struder A800 or a Dimension D Stereo Chorus. If I have to lug a laptop along with the helix no problem, that's easy. As it stands, I pay a heavy latency penalty for trying this. In fact, it's less latency to get a fast interface and run it via the Helix loop (taking the extra AD hit).  

</rant>

 

"On par" is kind of an odd measure... Do I expect the reverbs from an multi-FX to meet the complexity of something a dedicated reverb plug-in can give you? Not really. All these hardware devices are bound by processing limitations that don't exist with plug-ins. But yet most of them are close enough for most people.

 

As far as using the FX loops, they don't necessarily add 3.5ms of latency. Going in and out of a Helix loops will add like 1.8ms, and if the pedal converts the dry signal, that will add more, but many digital delay and reverb pedals (like the Strymon BigSky) have an analog dry path, so there isn't really any additional latency to worry about. So I don't know... If that extra 1.8ms of latency is a deal-breaker for you than it is, but I suspect most people won't care. I know not every digital effect has that dry path - the Eventide H9, doesn't. But then again, if you're using a dense, lush reverb, you're probably even more unlikely to notice a little extra latency.

 

Everything in any design a compromise in one way or another, and the perfect piece of gear doesn't exist. It's just finding something that works for you. I think things keep on getting better as far as what's available to the average player, though.

[SaschaFranck]

Fwiw, not that I'm happy about the Helix' latency values when used as an interface, but I find them to be pretty much acceptable for anything but software audio monitoring. IOW, playing  virtual instruments really isn't exactly problematic (in this case, interface latency only kicks in once, on the output side of things). Furthermore, I think super low latency operation isn't exactly something one should expect from what basically is a hardware modeler and not a dedicated interface.

 

Anyway, I just wanted to share something that I'm doing all the time (and for a long while already, too). Even if I'm not going for full stop software monitoring, I still use delays and reverbs coming from plugins all the time while tracking. To make this work all you have to do is to mute the dry signal somehow (in Logic I'm using a dedicated track and have it set to "no output") and only monitor the 100% wet returns of whatever reverb/delay busses that you're feeding your input channel into (most likely via sends). Can be done in a number of ways, the only important thing being that you completely mute any dry signal - you're monitoring that through the Helix.

As far as those additional 10ms of latency go, they really don't matter too much on reverbs and delays. I usually have some predelay on most reverbs anyway (to avoid mud), so I can just pull that down a bit, should it really matter.

This really works a treat and I'm using the same approach in my "normal", Helix-free recording environment for almost 2 decades by now, saved me from keeping the old rack FX for monitoring pleasure, even with less powerful computers and/or cheaper interfaces not allowing for full stop software audio monitoring.

 

Fwiw, because this is working so well, my current plans are to purchase one of those nice 2-in-1 convertible laptops somewhen this year (doesn't even need to be a high end thing such as a MS surface, so far I'm looking at some Lenovo Yogas, but I'm not in a hurry, so I'll keep looking around for a while). I will run whatever kind of VST plugin host on it (maybe Gig Performer, maybe Bidule, maybe just Reaper, don't think I need a high end solution) and will finally be able to use things as the glorious MFM2 delay from U-He or Native Instruments Raum plugin or even some IR based reverbs live. This will as well free up a considerable amount of resources on the Helix and with the new switching options apparently coming with 2.9, controlling things from the Helix shouldn't be much of an issue (it shouldn't even be with 2.8). As a bonus, I will have access to HX Edit all the time.

That'll be quite some fun.

[mark_gamache]

Okay, lots of good feedback about this topic. 

 

@gunpointmetal Yes, sometimes we can be too focused on technical stuff when the performance is the main objective. I think the reverbs are not something that jumps out for a live audience after 3 beers. In fact, I am getting ready to start rehearsing for some gigs and might just take the Helix with zero outboard gear as my complete solution.  However, I find playing through some of the higher end reverb / delay plugins gratifying. When you start using plugins the water gets deep quickly and this can work against you with too many options, but I like pushing the capabilities of what stuff can do. Also, Line 6 markets the Helix as an interface, they need to make it work as well as an entry-level interface or explain why it doesn't or can't work. 

 

@SaschaFranck The 100% wet option is a fallback position for me. It does limit you to only using effects like reverb and delay. It's better then nothing. I use BlueCat Audio Patchwork: https://www.bluecataudio.com/Products/Product_PatchWork/ as a VST host (it's really efficient and is Midi programmable). Also I use a Surface 6 Here's a video on optimizing that device (because you brought up MS Surface).

 

@phil_m Yes my latency calculation was for digital pedals that don't have a direct through (for example my H9 or Source Audio distortion). The latency was not the main point. To me, it's silly to have multiple DSP boxes all doing A-D conversion to talk between them over 1950's phone plugs. I know it's what most players do today. 

 

Line 6 markets the Helix as mostly an all-in-one solution. And I think it's great, I want to use it as a platform to build on. I want to use the loops to bring in analog devices and USB to add plugins that being new functionality to the Helix. 

 

Also, the Neural DSP hardware might just be the same as a Helix in terms of being a walled garden (closed platform) https://en.wikipedia.org/wiki/Closed_platform, but I'll withhold judgement until it's in the wild. 

 

Here's something that is interesting (also expensive):  https://en.antelopeaudio.com/products/zen-tour-synergy-core/ (Zen Tour Core) 2ms latency upto 32 channels of audio. Has builtin DSP/FPGA for near zero a lot of effects. 

 

 

 

[kylotan]

13 hours ago, rd2rk said:

 

Nor is anybody (in their right mind) asking for zero latency.

Nor does anybody (in their right mind) expect to match the performance of a $3000 dedicated interface.

But seriously - higher latency than a $100 Behringer?

 

Yes, of course higher latency than a $100 Behringer. Latency isn't something you can just throw money at and it'll go away. The most zero latency part of a signal chain is a cable, which costs very little.

 

Doing stuff in the digital domain takes time. There is no way around it, apart from to do less stuff.

[SaschaFranck]

5 hours ago, mark_gamache said:

Here's something that is interesting (also expensive):  https://en.antelopeaudio.com/products/zen-tour-synergy-core/ (Zen Tour Core) 2ms latency upto 32 channels of audio. Has builtin DSP/FPGA for near zero a lot of effects.

 

Ok, as you bring this up: Right now, pretty much all interfaces delivering substantially lower RTL values (and stable performance) than what we get from the Helix cost a rather significant amount of money. I own a Zoom UAC-2, from all I know, regarding latency, it's still the best in its league, it's the very reason why I bought it (fwiw, 4.5ms at 32 samples, 5.9/64). We're talking the league around 200 bucks here. But it's only stereo I/O. Now, with the Helix, we have an 8 I/O interface under the hood. There's a Zoom equivalent for that, too, the UAC-8. Comes in at around 600 bucks already (ok, it also has 8 mic preamps).

Next thing to up the game in USB-interface-low-latency land would be an RME Babyface Pro already. Now we're at <700 bucks already.

 

So, seriously, what can be realistically expected to be thrown in with the Helix, especially given that a lot of "non-core" hardware elements (and even some core ones) aren't exactly top quality already, either (sadly enough, sometimes it's even the opposite)? I would defenitely not expect anything better than Behringer and the cheaper Forcusrite, Steinberg and what not interfaces. And obviously, the driver needs to be built around the various routing options, which might not be a trivial task (all the cheaper interfaces don't offer anything even close to the Helix' routing options).

 

In a nutshell, while I think it might be possible to somewhat improve the Helix' RTL performance, I don't think there's all too much to gain. Maybe 2ms or so. Which would still not take us into the real low latency range, so it'd still not be possible to use HX processing plus software monitoring, it'd possibly also drag you over the "latency acceptance" cliff once you'd be using any plugins coming with their own latencies (such as the PRS plugins from Waves), etc.

 

Now, I'm not defending Line 6 for the quality of the interface of the Helix, I just think it's unrealistic to expect something noticeably better.

 

[kylotan]

11 hours ago, SaschaFranck said:

 

All these are irrelevant for plain audio interface functionality. Using, say, RMEs Totalmix software doesn't affect the latency of their devices, either.

 

They're not irrelevant if the whole idea of the Helix as an audio interface is that it routes the audio through the DSP system so that you can do stuff to that audio. It might be possible that the dry/DI outputs are lower latency. Or, it might not, depending on how they've set it up.

 

Comparing to Totalmix - in the general case, mixing software is almost zero latency because you can do it on a sample by sample basis. But if you want to start doing frequency-based effects, that means you incur some latency because it's not mathematically possible to reconstruct frequencies from individual samples.

[SaschaFranck]

1 minute ago, kylotan said:

They're not irrelevant if the whole idea of the Helix as an audio interface is that it routes the audio through the DSP system so that you can do stuff to that audio. It might be possible that the dry/DI outputs are lower latency. Or, it might not, depending on how they've set it up.

 

Well, of course you're right, but I was only talking about using it strictly as an interface. Compare it to, say, an Apollo. When you switch off the plugins running on the unit, they're completely out of your way. All routings and what not should be taken care of by the inherent safety buffers, so all processing should be able to take place within these boundaries.

[StarF666]

9 hours ago, SaschaFranck said:

IOW, playing  virtual instruments really isn't exactly problematic (in this case, interface latency only kicks in once, on the output side of things). Furthermore, I think super low latency operation isn't exactly something one should expect from what basically is a hardware modeler and not a dedicated interface.

 

When I bought my Helix I was hoping to use it for exactly that. IMHO even the output latency is way to high to play virtual instruments. They're advertising it as an audio interface. I didn't expect best in market round trip latency but was disappointed to find out I had to buy a dedicated audio interface.

[SaschaFranck]

1 hour ago, StarF666 said:

 

When I bought my Helix I was hoping to use it for exactly that. IMHO even the output latency is way to high to play virtual instruments. They're advertising it as an audio interface. I didn't expect best in market round trip latency but was disappointed to find out I had to buy a dedicated audio interface.

 

Hm, so how would playing virtual instruments with 5ms latency be different from playing guitar through a 5ms RTL chain? I mean, that's what pretty much most folks are getting along with just fine and the Helix does deliver around that value for virtual instruments.

[phil_m]

9 hours ago, StarF666 said:

 

When I bought my Helix I was hoping to use it for exactly that. IMHO even the output latency is way to high to play virtual instruments. They're advertising it as an audio interface. I didn't expect best in market round trip latency but was disappointed to find out I had to buy a dedicated audio interface.

 

What are you using as a controller for your virtual instruments?

[StarF666]

Ist's been a while since I tried this the last time. I use my Nord Electro 5D as a keyboard controller and a NI Maschine Mikro MK3 as a pad controller for drums. As far as I remember I couldn't get 5ms output latency. Have to check if midi input latency could have been the issue. 

[qwerty42]

18 hours ago, SaschaFranck said:

 

Hm, so how would playing virtual instruments with 5ms latency be different from playing guitar through a 5ms RTL chain? I mean, that's what pretty much most folks are getting along with just fine and the Helix does deliver around that value for virtual instruments.

Are you actually getting 5ms round trip latency with the Helix? If so, please tell us how -- I certainly can't get that myself, and I'm not a novice with this stuff. I would love to get 5ms round trip latency out of my Helix as an interface for use with other plugins.

[SaschaFranck]

9 minutes ago, qwerty42 said:

Are you actually getting 5ms round trip latency with the Helix? If so, please tell us how -- I certainly can't get that myself, and I'm not a novice with this stuff. I would love to get 5ms round trip latency out of my Helix as an interface for use with other plugins.

 

No, I'm not. I was just using 5ms as a comparison value that most people seem to get along fine with.

In a nutshell: While I do understand that people aren't fine with 10ms for software audio monitoring (I'm in that camp myself), I don't exactly understand why 5ms would be a too big deal when dealing with virtual instruments (which *is* what we're getting with the Helix). Now, maybe for some keyboarders with a good sense of time, it might be too much already, especially as we'd have to add possible MIDI (or MIDI-via-USB) input latency on the way in, but that's a sort of different issue (and likely the reason why Phil was asking).

 

2 hours ago, StarF666 said:

Ist's been a while since I tried this the last time. I use my Nord Electro 5D as a keyboard controller and a NI Maschine Mikro MK3 as a pad controller for drums. As far as I remember I couldn't get 5ms output latency. Have to check if midi input latency could have been the issue. 

 

I could pretty well imagine how playing pad things like Machine would become a critical thing - even if you can get to 5ms audio output latency (which, fwiw, is not exactly possible to measure properly, so once it comes to those numbers and the division into input and output latency, others than with RTL, we have to believe in the reported numbers), when you add just 2-3ms MIDI input latency, things may start to feel somewhat distracted, especially under headphones.

 

Fwiw, for me, when using headphones, something between 5ms and 7ms seem to be the values when I start to feel things. When using my Zoom UAC, playing any really bone dry clean amp patch, I can usually reliably detect the difference between 32 and 64 samples, hence the difference between 4.5 and 5.9ms of latency (had a mate switching buffersizes a while ago, so it was a blind test). But I can *not* as reliably detect the difference between 5.9 and 8.8ms (the latter being the result of using 128 samples buffersize).

Without headphones, the numbers vary. With a tad of room/ambience on the patch, the numbers vary again. Once overdriven sounds come into play with all their smeared attacks, it's yet again completely different.

 

A little OT, but not really:

In addition, all these are just "feel" things. In raw listening tests without any direct physical or audible references, I don't think I could reliably detect 10ms - and according to some tests I once conducted during a sequencing workshop I was holding at the local music university, very few people can. Fwiw, for these tests I used the most simple drum pattern, kicks on beats 1 and 3, snares on beats 2 and 4. both samples were cut straight at their transient peaks (rather than at zero crossings) and I then applied various amounts of positive or negative delay to the snares (all rendered into audio files so any potential sequencer playback issues could be ruled out). It's been a long time ago already and unfortunately I never grabbed the sort of documenting text files from the universities computer (hasn't been an all too scientific test anyway), all I safely remember was some drummers being quite astonished at their abilities (or lack thereof...) to spot offsets in a significantly better fashion than other instrument players. In fact, the most stunning guy was a keyboard player who would detect 10ms offsets all the time, as long as the BPM would stay the same.

Oh, obviously, there's been no hihats running along, so the delayed snares were rather isolated, the only thing to reference to being the BPM "pulse" that you'd had to kind of internalize. Having said that, I rendered the beats at different tempos, and quite obviously, at very slow tempos, it was a lot tougher to detect any offsets for all of my "contestants". Also, when I was switching between BPMs quickly, such as between 70 BPM and a negative snare delay of 15ms and 130 BPM with a positive delay of 10ms, the results were reliably all over the place, even with the mentioned keyboardist.

In other latency related listening scenarios, yet some other things come into play. With a band it might be optical clues, with stereo files it will be levels (there's quite some interesting reads about the Haas effect around), etc, etc.

Fwiw, when recording vocalists, using *higher* buffersizes for audio monitoring sometimes is preferable as very low latency values will cause irritating phasing issues between their "skull listening" and the phones, whereas higher values may result in a kind of chorus-y doubler effect that might even be pleasant to listen to.

Guess what I'm saying is that there's really lots of implications coming along with various incarnations of latency.

And fwiw, all these are responsible for one of the reasons I went with the Helix, as I can always monitor at more or less fixed latencies while still having complete re-amping freedom - which, unlike with any other modeler, includes using Helix Native, allowing me to get 1:1 what I'm monitoring or just some slight variations. Impossible with any other contender unless you go for realtime re-amping.

[Indianrock2020]

I tried using my Helix today as an interface to my laptop running Jam Kazam for online music.... I used helix USB channels one and two for the guitar and channel 8 for the microphone ...the latency showing in jamkazam was about 12 milliseconds ....this is the local gear related latency, I wasn't connected to anyone on the internet.... my regular audio interface gets about 5.7 milliseconds so unless I'm missing something using the Helix in this way makes no sense.

I verified the audio interface latency that jamkazam shows by running the RTL utility... how would you safely route a patch cable on the helix to run RTL against that?

[SaschaFranck]

17 minutes ago, Indianrock2020 said:

I tried using my Helix today as an interface to my laptop running Jam Kazam for online music.... I used helix USB channels one and two for the guitar and channel 8 for the microphone ...the latency showing in jamkazam was about 12 milliseconds ....this is the local gear related latency, I wasn't connected to anyone on the internet.... my regular audio interface gets about 5.7 milliseconds so unless I'm missing something using the Helix in this way makes no sense.

I verified the audio interface latency that jamkazam shows by running the RTL utility... how would you safely route a patch cable on the helix to run RTL against that?

 

Be happy it's that low. Under macOS, at 64 samples and 44.1 kHz operation, RTL is 16.9ms.

Fwiw, my results posted above in this thread where coming from the days when I was running the Helix through macOS' stock Core Audio protocol. Had to install the audio driver for some reasons ever since and I was really WTF?-ing. Usually, dedicated drivers deliver lower latencies than Core Audio, but not so with the Helix.

Doesn't look as if this is high on L6's priority list.

[Indianrock2020]

1 hour ago, SaschaFranck said:

 

Be happy it's that low. Under macOS, at 64 samples and 44.1 kHz operation, RTL is 16.9ms.

Fwiw, my results posted above in this thread where coming from the days when I was running the Helix through macOS' stock Core Audio protocol. Had to install the audio driver for some reasons ever since and I was really WTF?-ing. Usually, dedicated drivers deliver lower latencies than Core Audio, but not so with the Helix.

Doesn't look as if this is high on L6's priority list.

For RTL did you just run a 1/4 inch cable from 1/4 out to guitar input?