Helix Latency... What is it really?

[Eric_Harbauer]

UPDATED AND CORRECTED!!!

My apoligies for the incorrect numbers (I read the timescale incorrectly on my Oscope screen) In the original post.

 

I'd like to begin by saying I've been very impressed with my helix so far.  I'm really enjoying it, and I am not posting this to bash it in any way whatsoever. But I'm naturally curious, and mostly unhappy with the SPDIF I/O,  and wanted to really understand why, which is what prompted this whole experiment. 

 

With all the connection possibilities I have been experimenting hooking different pieces of gear I own to the various I/Os, with some great, and some not so great results.  One of the biggest questions I've had has been with the digital I/Os.  Is it actually faster to use SPDIF?  Logic would say yes.  No A/D conversion.  It has to be faster right?  Well I haven't been so happy with Helix in that regard.  My 2 notes torpedo Hasn't sounded as good as I think it should through the SPDIF I/O.  So I started digging online.  What I found was a lot of nonsense.  Threads saying that each foot of cable added 1ms of latency (this is wrong by the way) and people claiming that Helix is sub 1ms latency. hmmm    

 

Just a few quick numbers, as tested on my Helix floor with an O scope straight thru on a blank preset. 

 

 

Each and every time you convert A/D -> D/A it's 1.8 ms ...Yes that means each loop you engage adds 1.8 ms

 

 

So to be clear.  guitar in + 4 loops (in series) + 1/4" out = 9ms maximum of latency due to A/d conversion

 

4 cable method should clock in at around 3.6ms  (assuming no other effects in circuit)

 

SPIDIF I/O appears to not run as fast as as the standard I/Os

48k runs slightly faster then 44.1k 88k or 96k

I tested this by looping spdif send and return to itself.

As you add gear latency will increase due to the outboard gear.

 

 

Other interesting notes about the SPDIF I/O  there is no option for sync (internal vs spdif) and it connects to my two notes torpedo and transmits and receives audio even when the two units are set to entirely different sync and sample rate settings.  

 

There's also other spdif weirdness but as I said it's not my intent to bash Helix, so I won't go into them here.

 

On the positive side, adding blocks does not appear to slow the unit down much.  like less than .2ms for an amp.  I haven't thoroughly tested this but a real quick check loading blocks up till I started to run out of DSP and very little  change in latency.   IR Blocks may have more latency associated with them but that is probably more due to the nature of what an IR actually is rather than processing power

 

My takeaways.

 

Use the internal blocks in helix as much as possible.

Minimize A/D conversion.  

Stereo loops (parallel) are 1.8ms for both 

Serial loops are 1.8ms each

Avoid using SPDIF for live playing (it's fine for re-amping etc)

 

 

 

Anyway... I thought I'd share this.

 

 

 

Rock On

✌

 

 

 

[phil_m]

I'm sorry, but your I call BS on your numbers. Something is wrong with your measurements. A round of conversion doesn't take anywhere near 9ms, and it hasn't been that high for, well, decades I imagine. If that were the case digital effects would be all but unusable. It is totally possible to have a round of conversion take under 1ms nowadays. The throughput latency of the Helix is something like 1.5 - 2 ms depending on the complexity of the preset. I'm not sure what is happening with your testing or whatever, but it's not correct.

 

Think of it this way. I can set a delay for 20ms, and I can hear that. A 45ms latency would be like having a slapback echo... There's no way.

[Digital_Igloo]

Yeah, there's something amiss. There could very well be something wrong with your Helix, or perhaps something's tweaky with the test methodology, but full throughput in Helix should be sub-2ms. We've never claimed sub-1ms throughput, but Helix should be faster than other boxes that dedicate a DSP to amps, as we only have one DSP hop—DSP 0 (Path 1) to DSP 1 (Path 2), as opposed to DSP 0 (pre-amp effects) to DSP 1 (amp) back to DSP 0 (post-amp effects and Global EQ). FX Loops will indeed incur additional latency, but that shouldn't be much more than a single millisecond each.

[rzumwalt]

I'm sorry, but your I call BS on your numbers. Something is wrong with your measurements. A round of conversion doesn't take anywhere near 9ms, and it hasn't been that high for, well, decades I imagine. If that were the case digital effects would be all but unusable. It is totally possible to have a round of conversion take under 1ms nowadays. The throughput latency of the Helix is something like 1.5 - 2 ms depending on the complexity of the preset. I'm not sure what is happening with your testing or whatever, but it's not correct.

 

Think of it this way. I can set a delay for 20ms, and I can hear that. A 45ms latency would be like having a slapback echo... There's no way.

 

I immediately had the same thought about the send/return loops. I have a TS9 in Loop 2 and a reberb in loop 3. I very often have one or the other or both in parallel with a direct signal that doesn't use a loop. Even with just 9 ms delay, there would be noticeable phasing or flanging at least. One very subtle setting on the reverb I sometimes use is almost unnoticeable. On that setting, I would probably hear this. I've also recorded to a DAW using the final signal going through 2 loops out of USB channel 1 and the dry direct signal through USB line 7. This would have been an 18 ms delay, and I definitely would have noticed that, as there are double trackers that use less delay.

 

I wonder if the OP's equipment has a delay in the "send" instruction for the original source audio. I would be interested if he could attempt this with some sort of compensation.

[Eric_Harbauer]

Fair enough. It is entirely possible I screwed up my measurements. I may have got the scale or the pulse width wrong. It's the first time I've tried to figure out timescale. I'll check it as soon as I'm back in front of my gear. If that's the case I'll update the numbers in my post.

 

The interesting thing to me was that SPDIF is slower than the the regular send and returns. And the fact that adding blocks had a negligible effect on the latency.

 

I ran 2 simultaneous signals one thru helix and the other was looped back on itself. So essentially I was comparing a to b. My goal being to find out which one was faster. As far as the time measurement went I just read it off the o scope screen which was set to 2 ms per subdivision. I may have gotten the scale wrong. If that's the case the round trip latency should be around 1.8ms per A/D conversion whose does seem more realistic. As I said I'll check it out and update the post

[Thurston9]

FAKE NEWS

[phil_m]

I immediately had the same thought about the send/return loops. I have a TS9 in Loop 2 and a reberb in loop 3. I very often have one or the other or both in parallel with a direct signal that doesn't use a loop. Even with just 9 ms delay, there would be noticeable phasing or flanging at least. One very subtle setting on the reverb I sometimes use is almost unnoticeable. On that setting, I would probably hear this. I've also recorded to a DAW using the final signal going through 2 loops out of USB channel 1 and the dry direct signal through USB line 7. This would have been an 18 ms delay, and I definitely would have noticed that, as there are double trackers that use less delay.

 

I wonder if the OP's equipment has a delay in the "send" instruction for the original source audio. I would be interested if he could attempt this with some sort of compensation.

 

The Helix actually has automatic latency compensation for the type of situation you describe (having an FX loop in parallel path but not the other one). It delays the path without the FX loop a little bit so you don't get phasing/comb filtering.

[Eric_Harbauer]

My Apologies.  I read the timescale wrong.  Oopsies  

 

Anyway I updated the numbers as measured on my Helix.  They seem a lot more reasonable now.  

[rzumwalt]

The Helix actually has automatic latency compensation for the type of situation you describe (having an FX loop in parallel path but not the other one). It delays the path without the FX loop a little bit so you don't get phasing/comb filtering.

Well then, I guess ignore everything I just said. Thanks for the info, I prefer being an informed blowhard to being an ignorant one.

[PeterHamm]

Too much measuring and not enough music-making...

[hal3001]

I just wanted to add that I've been using spdif for recording and monitoring for early a year now and I've never had latency issues. I run my Helux Rack into a UA Apollo 8 via spdif to avoid extra conversion and it works great. Also, you can definitely set your clock sync source in global settings on page 4 of In/Outs. I always have it set to word clock in and have that also connected to the apollo.

[hal3001]

Wait. I just looked at a rear product picture of the floor unit. Did they remove wordclock from the floor unit?

[PeterHamm]

Wait. I just looked at a rear product picture of the floor unit. Did they remove wordclock from the floor unit?

 

 

Word Clock is ONLY in the Rack. Never was in the floor.

 

AES/EBU in is also only in the Rack, as is buffered Guitar Out, and also (unnecessary on floor) connections for all 3 EXPs as well as a switch connection for EXP 1 with a Mission pedal so equipped.

[spikey]

Too much measuring and not enough music-making... 

 

 

My new sig- Thanks Hamm!

[GeorgeVW]

It's also valuable to bear in mind that 1ms in latency is the equivalent of standing a hair less than 1' further away from the sound source (using the average 1125 fps free air speed of sound), so even a 10-12ms total latency is only the equivalent of wandering towards the front of a stage by not even the length of your average guitar cable. 

[aleclee]

It's also valuable to bear in mind that 1ms in latency is the equivalent of standing a hair less than 1' further away from the sound source (using the average 1125 fps free air speed of sound), so even a 10-12ms total latency is only the equivalent of wandering towards the front of a stage by not even the length of your average guitar cable. 

 

 

It's really not because your brain has acoustic cues that help it compensate for time of flight that are absent when the delay is in the box.  1ms isn't gonna make a difference by itself but when the total "in the box latency" (digital wireless + Helix + whatever else is in the chain) gets above 7ms, things start to feel disconnected.  Most folks would notice 10-12 ms of latency in the box where they wouldn't notice 20ms (or even 50) in time of flight.

[GeorgeVW]

It's really not because your brain has acoustic cues that help it compensate for time of flight that are absent when the delay is in the box.  1ms isn't gonna make a difference by itself but when the total "in the box latency" (digital wireless + Helix + whatever else is in the chain) gets above 7ms, things start to feel disconnected.  Most folks would notice 10-12 ms of latency in the box where they wouldn't notice 20ms (or even 50) in time of flight.

 

Let's see that assumption stand up to a double-blind test before we get too attached to it. 

[aleclee]

Let's see that assumption stand up to a double-blind test before we get too attached to it. 

 

You might want to have googled "acoustic latency perception" before referring to my assertion as an assumption. Believe it or not, I wasn't pulling that concept out of the aether.  

https://thecrocoduckspond.wordpress.com/2015/04/02/latency-myths-and-facts-why-echo-perception-is-different/

 

If you're looking for something more quantitative, I will assume that you trust Yamaha's research:

http://www.yamahaproaudio.com/northamerica/en_us/training/selftraining/audio_quality/chapter5/05_absolute_latency/

 

If you want something more academic, you can sift through this: 

The Effects of Latency on Live Sound Monitoring

[GeorgeVW]

Not buying it when it comes to playing guitar onstage with a band. In an environment when your ear is hearing both the original string impulse (even if only very faintly) and an amplified source, yeah, makes sense, and I've experienced it in a studio setting. But in an environment where background noise is masking your string sound, nope — don't buy it. 

[aleclee]

Not buying it when it comes to playing guitar onstage with a band. In an environment when your ear is hearing both the original string impulse (even if only very faintly) and an amplified source, yeah, makes sense, and I've experienced it in a studio setting. But in an environment where background noise is masking your string sound, nope — don't buy it. 

 

Okay, you don't believe research by Yamaha Pro Audio and AES.  They're pretty sketchy organizations when it comes to Pro Audio know-how anyway.  :rolleyes:

 

Here's an easy way to do your own double blind experiment:: dial in a 20ms delay at 100% mix and have someone turn it on and off so you don't know the state.  If the perceptual difference is greater than standing 20' away from your speaker (probably something you have some experience with), you've confirmed my hypothesis.

[GeorgeVW]

Okay, you don't believe research by Yamaha Pro Audio and AES.  They're pretty sketchy organizations when it comes to Pro Audio know-how anyway.  :rolleyes:

 

Here's an easy way to do your own double blind experiment:: dial in a 20ms delay at 100% mix and have someone turn it on and off so you don't know the state.  If the perceptual difference is greater than standing 20' away from your speaker (probably something you have some experience with), you've confirmed my hypothesis.

Not so much "don't believe"(hey, I've got good friends who work in both places) as in believe that it's discardable in real-world applications. Your proposed experiment doesn't test what you think it's testing, though. Can a person detect 20ms of delay? Certainly. That's been well established. But here's the experiment that you need to do. Shot an IR of a hallway from 20 feet away from point A. Place two flat-response, full range speakers (B and C), one 20 feet down that hall, one 20 feet closer. Balance level between the two so that observer at point A gets the same perceived SPL from both. Apply the IR and a 20ms delay (or whatever is necessary so that DSP latency plus delay line are equal to 20ms) to the signal going to the nearest speaker. Switch between the sound sources randomly, rigged so that the person doing the switching doesn't know which position of the switch is which source. Do the test multiple times, shifting the L/R orientation of the speakers for randomization purposes between test sessions. Can the test subject consistently, above the level of pure chance, tell which speaker is actually 20' away and which one is virtually 20' away? That's the experiment that will determine the answer. 

[P1KW]

You have not understood, GeorgeVW.
It's not about latency between 2 sounds.
It is the latency between "your feeling" that you have played a note and your brain listens to it.
12 ms is acceptable for a mediocre musician and unacceptable for a virtuoso.

Cheers!

[datacommando]

9 hours ago, P1KW said:

You have not understood, GeorgeVW.
It's not about latency between 2 sounds.
It is the latency between "your feeling" that you have played a note and your brain listens to it.
12 ms is acceptable for a mediocre musician and unacceptable for a virtuoso.

Cheers!

Nice start, P1KW,

 

First post and you have just resurrected a thread that died almost 3 years to the day.

 

Mind out for zombies!

[karlic]

On 5/4/2017 at 6:42 PM, Eric_Harbauer said:

Fair enough. It is entirely possible I screwed up my measurements. I may have got the scale or the pulse width wrong. It's the first time I've tried to figure out timescale. I'll check it as soon as I'm back in front of my gear. If that's the case I'll update the numbers in my post.

 

The interesting thing to me was that SPDIF is slower than the the regular send and returns. And the fact that adding blocks had a negligible effect on the latency.

 

I ran 2 simultaneous signals one thru helix and the other was looped back on itself. So essentially I was comparing a to b. My goal being to find out which one was faster. As far as the time measurement went I just read it off the o scope screen which was set to 2 ms per subdivision. I may have gotten the scale wrong. If that's the case the round trip latency should be around 1.8ms per A/D conversion whose does seem more realistic. As I said I'll check it out and update the post

Really interesting post Eric and thanks for sharing the results. I remember reading on the Kemper forum that SPDIF was slower than the standard analogue outputs too, so this must be a limitation outside of Helix or Kemper.

 

Did you have a figure for the SPDIF latency? It would good to know how much the difference is.

[datacommando]

12 minutes ago, karlic said:

Really interesting post Eric and thanks for sharing the results. I remember reading on the Kemper forum that SPDIF was slower than the standard analogue outputs too, so this must be a limitation outside of Helix or Kemper.

 

Did you have a figure for the SPDIF latency? It would good to know how much the difference is.

Eric hasn’t visited these pages since September 30 2019...

[PeterHamm]

NECRO THREAD ALERT!

[knoxdaz10]

On 5/6/2017 at 5:02 AM, PeterHamm said:

Too much measuring and not enough music-making...

haha

[datacommando]

Ye Gods, does nobody ever read the dates on these threads?

 

I’m going to put it down to sheer boredom during this virus crisis!

 

Yawn!

[karlic]

At the risk of upsetting some, I did find it interesting and the date didn't bother me at the time. With digital wireless systems adding up to 8ms latency in some cases, it is good know the Helix is low. 

 

I know it's a poor show, but I don't have a guitar or Helix with me during breakfast:-) Such a waste of playing time.

[hideout]

On 5/6/2017 at 5:36 AM, PeterHamm said:

 

 

Word Clock is ONLY in the Rack. Never was in the floor.

 

AES/EBU in is also only in the Rack, as is buffered Guitar Out, and also (unnecessary on floor) connections for all 3 EXPs as well as a switch connection for EXP 1 with a Mission pedal so equipped.

Uhh... care to tell us what that red thing is on the right in the picture?

[rd2rk]

49 minutes ago, hideout said:

Uhh... care to tell us what that red thing is on the right in the picture?

 

AES/EBU L6 Link OUT