Who has the lowest latency Guitar Synth Rig?

Started by Elantric, January 20, 2013, 12:08:19 PM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

Elantric

What would be ideal is if folks could post Stereo Audio files in this thread with the following Characteristics:

1) Use a passive DI box on your normal guitar pickups and feed your Audio interface - record as Left Stereo Channel

2) Connect the output of your Hardware or Software Synth and record as Right Audio Channel

3) Play 16th notes and record

4) Post the resulting Stereo Audio Files


Many of us have post processing tools which can directly measure a comparison of  Left vs Right channel in the Stereo Wave / AIFF / MP3 file - and we can all contribute to
"who has the lowest latency Guitar Synth Rig?
A complete report will follow.
   

PD FX

#1
so here are the rough tests.. there's no music here, just  EAdgbe'
all test with gr30 as sound enginge, local off, no fx.
played without pick, no nails. gk3a.
sorry for the noise in the AXOn recordings, there was no DI available..

arkieboy

Elantric and other contributors ...


Indulge me - I think the most important thing is not latency, but consistency.  Can you make sure you play a fair few notes (>20) on each string so we can do a variance estimation ...


Steve
Main rig: Barden Hexacaster and Brian Moore i2.13 controllers
Boss SY1000/Boss GKC-AD/Boss GM-800/Laney LFR112

Other relevant gear: Line 6 Helix LT, Roland GR-33, Axon AX100 MkII
Oberheim Matrix 6R, Supernova IIR, EMu E5000, Apple Mainstage, Apple Logic, MOTU M4

shawnb

Yes, multiple samples are needed.    And specific strings/notes should be spelled out.
Address the process rather than the outcome.  Then, the outcome becomes more likely.   - Fripp

Elantric

#4
And I realize Shawnb has performed a very similar set of measurements  - which we should not disregard, but learn from.

But, yes lets propose a "VGuitarForums MIDI Guitar Test Specification" - perhaps this thread covers one test element.

I agree, the "Synth rig" that triggers nothing but wrong notes - but extremely fast should be rejected.

Maybe a spreadsheet with "fill in the blanks" on the test configuration, stating:

* Guitar Type
* Guitar Scale Length
* Guitar String Type and Gauge
* Technique (Fingers, Nails, Plectrum (style & gauge)
* Guitar to MIDI Interface Type (13 Pin Pickup, TriplePlay, etc)
* Synth Type (Hardware, Software, Make, Model, Version)
* Chromatic or Mono Mode
* Additional Guitar Synth Tools for Virtual Synth Playback (GTAK4, etc)
* Audio Recorder (standalone, Computer (OS Type and Version), Recording Software & Version.
* Computer Specs (OS, CPU, RAM, Drive, Audio / MIDI Interface type)   
Have Instructions to play across all strings.

The Ultimate would be if a Breakoutbox was used and create six separate stereo "in sync" tracks  - representing all the above, but separate audio file for each string.
And Shot with Live Video, so we can review your technique. (Hammer On, Picked, Pull Off, etc)

PD FX

I wont be able to do such elaborate testing.
It suffies for me to know which apparatus i should use for which job.
If I'm gigging I throw the gr-55 on the floor, it's setup is easy and offers a wide sound palette.
If I want to play piano on the laptop, I use gr-30.
If I want to do highspeed jazz with pick I can use the Axon.
But measuring all my guitars in all my setups, with pick and without pick, would be way too much work.
And Arkieboy is right about consistency, especially in synth only mode latency is not an issue, but consistency is!

drjoness2001

My tests:

http://www.joness.com/gr300/MIDI_SPEED.htm

The fastest, real-world MIDI converter was the Roland GI-20 with a GK-2A equipped Parker Nitefly. Avg 15 ms.

The axon was faster, but for me a little bit glitchy. The GI-20 was slower, but more accurate. Sounds like my piano teacher: you need accuracy before speed.

fokof

#7
I made some tests on the VB99 for latency on that thread:

https://www.vguitarforums.com/smf/index.php?topic=1787.0

It's 2,27ms between "Bass direct" and "mag pickup" and 6,19ms between "Mag pickup" and "COSM Bass"


Pitch to MIDI is wayyyyyy up there , 80-100ms  , not usable for bass.
That's why COSM is a godsent for us bass players.


nix808

I will certainly do the test on my pitch to MIDI software,
however I think I can tell you with great accuracy atm.
On E2, the fastest string it's->
1200 samples buffer, to negate the percussive string attack
44100/1200 = 36.75
1000/36.75 = 27 ms
256 samples soundcard buffer(usually would have clean audio)
44100/256 = 172.65
1000/172.65 =5ms
1 ms MIDI polling via Ruby language

27ms + 5ms + 1ms = 33ms total

Please contact me if you would like to try it 8D
Rock and Rock my boat, as long as you don't rock and rock my boat!hehe

Gaustu

#9
Quote from: drjoness2001 on January 25, 2013, 10:58:23 AM
My tests:

http://www.joness.com/gr300/MIDI_SPEED.htm

The axon was faster, but for me a little bit glitchy. The GI-20 was slower, but more accurate. Sounds like my piano teacher: you need accuracy before speed.

Can you tell me please how Axon was set up during a test?  Did you use 5 pin MIDI out to drive synth or USB virtual port? This is very important and I'll tell why later.

I for sure will post my tests here later when will have a time - comparison of Axon and GR-55
Parker PDF85 modified: Sustainiac, SD TB-4, Piezo Hexpander 13 pin out.
GR-55 mod with 13 din thru, 6 analog outputs D-Sub => Presonus FS Mobile  8 analog inputs.
NI Reactor matchHEX Guitar Framer Ensemble custom.

Gaustu

#10
Quote from: drjoness2001 on January 25, 2013, 10:58:23 AM
My tests:

http://www.joness.com/gr300/MIDI_SPEED.htm
Hi drjoness

I wonder why you decided not to include the tests of Axon in your tables.
And now I'm wondering again looking into these tables because I can't get it how to find this info to be useful for me. As I can understand you did not specify which string/note took part in the test and there are some "average" figures of milliseconds. So I cannot see, for instance, what is going on for low E with Yamaha etc.
Can you explain what means "Overall" column of the table? I can't understand.
And what is the pickup named BX-13-VX (in "Pickup" column)?

P.S.
I would say all is important - consistency as well as latency too. So I will do the comprehensive tests and I like an Elantric's idea of test configuration.
Such a comprehensive information will be the most valuable for those who plans to buy something but not a cat in a cul-de-sac. They will know what to expect with 9X% of probability.
Parker PDF85 modified: Sustainiac, SD TB-4, Piezo Hexpander 13 pin out.
GR-55 mod with 13 din thru, 6 analog outputs D-Sub => Presonus FS Mobile  8 analog inputs.
NI Reactor matchHEX Guitar Framer Ensemble custom.

Gaustu

#11
ok, I have done my tests.

1) Axon_SoftSynth

first part of mp3 file: test for "consistency" through all the strings and fretboard.  Axon drives 6 instances of FM7 loaded to Chainer VSTi.
second part of mp3 file:Axon drives Sytrus soft synth.
third part of mp3 file:the same setup, test for latency. It turned out that latency depends very insignificantly in Axon from high notes to lowest notes. This test wondered me. I expected for lower latency.
Highest notes show 17-19 ms,
lowest notes show 18-22 ms.
Almost the same. But don't harry with conclusions!

2)Axon_SynthModuleHardware
Axon drives GR-30 by MIDI 5pin. 
Highest notes: 13-14 ms,
lowest notes: 17-20 ms  !

3) GR-30(piezo-driven)
This test shows good results with old friend GR-30, but there are some sadness.
Highest notes: 7-9 ms, 
lowest notes: 17-21 ms  (if to pluck staccato!)
If to pluck legato the picture goes much worst on lower strings, especially A and E. This is shown on the next test.

4) GR-30(GK-2A driven)
Lower notes: 30-35 ms
Generally it feels less consistency then with piezo.
Second half of this mp3 file shows problems of Roland MIDI conversion while legato playing.
(This does not depend on the type of hex pickup)
I mean, if play staccato, there is low latency (about 20 ms on lower notes) but if to play inextricably the latency rises drastically to 42 ms. This is what always annoyed me when I played GR-30 with its fast-attack sounds.

5) GR-55(piezo driven) 
Test playing internal PCM sound with fast attack.
Highest notes: 10-14 ms,
lowest notes show 18-22 ms. (if to pluck staccato!)
A bit slower then GR-30 as you see.
The end of this mp3 file shows that if play inextricably (legato) the latency rises drastically to 65 ms!
that's why I never gonna use its PCM fast-attack sounds on lower strings.

Some conclusions of mine.
Axon wins all the way over GRs because of its reliable tracking and it has no such issue that GRs have. The latency it produces does not depends on playing character (staccato or legato).

I found the problem with Axon when it works with USB MIDI out mode. This makes it a bit unstable and "glitchy" as some may find before. I still not sure if it may be solved, and this unstability may be caused by the digital noise through the ground loop when USB is connected. I will see what will happen when ground loop goes away.
If to use MIDI 5 pin out Axon tracks excellently.

My setup was:
* Hollow body guitar
* Guitar Scale Length 650
* Guitar String Gauge - 008, 010, 015, 026, 036, 046
* Technique: Plectrum
* Guitar to MIDI Interface GraphTech Hexpander; Roland  GK2A+G1D
* Synth Type - see the description
* No Chromatic,  Mono Mode
* Audio Recorder: PC; SAWStudio 5.1, Software Audio Console
* Computer Specs Win XP Pro SP3; intel i5 650, 4 Gb RAM, Presonus FSM 

Parker PDF85 modified: Sustainiac, SD TB-4, Piezo Hexpander 13 pin out.
GR-55 mod with 13 din thru, 6 analog outputs D-Sub => Presonus FS Mobile  8 analog inputs.
NI Reactor matchHEX Guitar Framer Ensemble custom.

admin

Now you need to add the Fishman Tripleplay to your MIDI test

In My Experience its fastest using I have - and much faster than my GR-30 regardless if Piezo>GR-30 or GK2A>GR-30 

PD FX

Thanks, Gaustu!
Your results show that the GR30 actually is the fastest (unbiased looking at the milliseconds), this is also my experience. The "sadness" your experiencing with the GR30 is due to the velocitycircuitry: the reason for the sloppiness of the GR30 is this velcocitycircuitry that uses the filtered audio as input (the initial highpitched plucksound is not taken into account, so re-triggering is very bad..)
This is hackable! I hope to get my GR30 hacked soon.
When tapping my GR30 allready has nearly perfect behaviour: see

You use the Axon in monomode, well, I also own an Axon 50 and I wanted to play piano with pedal, that's the reason I didnt want multi-monomode, but I got too many glitches in polymode. now I see you say its the USB that is causing gliches! Very interesting.. I'll test this myself with a separate midi cable.. I'm curious!

Gaustu

Quote from: Guitarpolson on February 04, 2014, 01:53:37 AM

You use the Axon in monomode, well, I also own an Axon 50 and I wanted to play piano with pedal, that's the reason I didnt want multi-monomode, but I got too many glitches in polymode. now I see you say its the USB that is causing gliches! Very interesting.. I'll test this myself with a separate midi cable.. I'm curious!

No, I use Axon in multi-monomode, it called "separate" in settings. Because I need bending and other things to be working fully for each string.
When I use it for software synth/FX control like Reaktor or Guitar Rig its ok. But if to use it to control GR-55 there is a problem, GR-55 doesn't receive CC events in multi-monomode but on one channel only which it set up at MIDI settings of GR-55. I wonder how they could skip there such a thing in its midi design.
Parker PDF85 modified: Sustainiac, SD TB-4, Piezo Hexpander 13 pin out.
GR-55 mod with 13 din thru, 6 analog outputs D-Sub => Presonus FS Mobile  8 analog inputs.
NI Reactor matchHEX Guitar Framer Ensemble custom.

Gaustu

#15
Quote from: Guitarpolson on February 04, 2014, 01:53:37 AM

This is hackable! I hope to get my GR30 hacked soon.
When tapping my GR30 allready has nearly perfect behaviour: see

I can't believe it may be fixed so. But even if this is real, it will be usable for its internal sounds only that bored me so much. If you will use it for driving other synths it will produce bad latency.
Anyway I am intrigued. Please let me know when you'll get it hacked.

Soon I'll get my ordered USB to USB isolator and will see how it helps ground loop fixing.
Parker PDF85 modified: Sustainiac, SD TB-4, Piezo Hexpander 13 pin out.
GR-55 mod with 13 din thru, 6 analog outputs D-Sub => Presonus FS Mobile  8 analog inputs.
NI Reactor matchHEX Guitar Framer Ensemble custom.


DF400

That's some significance latency for the GR55, which I feel somewhat when playing live. I've learned to play slightly ahead to compensate, which results in never feeling "in the pocket." Any actual speed tests on the TRP? I have not been able to use it live due to the latency and wondering if it's my computer processing, of just the unit.

Elantric

As a reminder, the GR-55 is packed with diverse range of possible signal sources feeding the output jack




NORMAL PICKUP = Latency is  under 8 milliseconds) 

MODEL  = Latency is  under 12 milliseconds) 

PCM   = Latency is  under 30 milliseconds for Low E string) 

GUITAR TO MIDI (USB Sofsynth ( in general OSX has lowest latency , but varies with what you are feeding, typically 32 milliseconds for Low E string)  , but can be much higher with a poorly configured Windows computer

GUITAR TO MIDI (5 pin MIDI to Tone module / Synth keyboard / etc))  ( varies with what you are feeding , typically 35 milliseconds for Low E string) 

jonathanf106

Quote from: Elantric on December 31, 2015, 06:35:23 PM
As a reminder, the GR-55 is packed with diverse range of possible signal sources feeding the output jack




NORMAL PICKUP = Latency is  under 8 milliseconds) 

MODEL  = Latency is  under 12 milliseconds) 

PCM   = Latency is  under 30 milliseconds for Low E string) 

GUITAR TO MIDI (USB Sofsynth ( in general OSX has lowest latency , but varies with what you are feeding, typically 32 milliseconds for Low E string)  , but can be much higher with a poorly configured Windows computer

GUITAR TO MIDI (5 pin MIDI to Tone module / Synth keyboard / etc))  ( varies with what you are feeding , typically 35 milliseconds for Low E string) 

So basically the PCM haves the most latency, and yall want to reduce that. Aleast down closer to the latency of the COSM which is practically you dont notice.

What about a fresh set of strings? Light or heavy wound.
As far as the GK settings knowing how much nuisance to use and string sensivity suppose to be aligned with NUA tab aleast that what the instructions says online on the roland website. Then there are other parameters when you edit a PCM tone like dynamics, LFOs, etc. I feel like that too much to be tweaking for something Roland should have got right before the product hit the shelves. But I wonder what settings these guitarists uses during the exhibition of this product because it sounded like they were playing zero latency. Especially with piano patches.

Elantric

#20
Why  not load the Roland UK product demonstrators GR-55 PCM Piano patch and review all settings in Gumtowns GR-55 Floorboard Editor

https://www.vguitarforums.com/smf/index.php?topic=8208.msg58356#msg58356


Quotesomething Roland should have got right before the product hit the shelves

If they did  that then we would never have any guitar synthesizer's to play with. Remember no two guitar players play with the same technique and providing deep level controls to adjust and compensate for that is more valuable.
And read Shawnb's GR-55 Setup doc that goes into detail explaining what the GR-55's parameters like "Nuance" and " Low Velocity Cut" accomplish
https://www.vguitarforums.com/smf/index.php?topic=6972.0

jonathanf106

Quote from: Elantric on July 15, 2016, 07:22:21 AM
Why  not load the Roland UK product demonstrators GR-55 PCM Piano patch and review all settings in Gumtowns GR-55 Floorboard Editor

https://www.vguitarforums.com/smf/index.php?topic=8208.msg58356#msg58356

Oh I can do it. Even though its far from perfect, ill do it Sunday at least that where i be back in town. But we need to find out is what settings these expert users that were promoting the product, what settings they were using because they sounded flawless using the PCM sounds.

StellaHarmony

I don't see this as an issue with the GR-55.  I have had mine 3 years and I have never plugged the pickups in.  Everything I play is COSM, synth patches or a combination.  I play Led Zeppelin, Pink Floyd, Stones, Jefferson Starship, Rush, Yes, 12 string stuff, tunings, it all works great.  I don't do Malmsteen but the opening to "Spirit of the Radio" works fine.  On some synth patches I minimize the attack slope to get a better response.  I also play on 10-38 strings, a holdover from the GR-300 days.  I use my guitar strictly as a controller.  I keep the amp clean with a touch of reverb and let the GR-55 create all the sounds.  My guitar is a late 80s Japanese Kramer focus with a Floyd.  It is a good player, but certainly nuthin fancy.  I just took it in for the first time in 20 years because the Floyd fulcrum studs were loose and it would not keep tune.  Now it's back to normal with a neck tweek and first ever fret touch-up.

Shingles

Quote from: Guitarpolson on February 04, 2014, 01:53:37 AM
Thanks, Gaustu!
Your results show that the GR30 actually is the fastest (unbiased looking at the milliseconds), this is also my experience.



No!

It's not a valid comparison, because there is no GR33 to softsynth, no GR33 to hardware synth.
Sending MIDI externally to soft synths or hardware via MIDI, Axon is faster.

Nik
--------------------------------
Tonelab, VG99, Axon AX100, EDP, Repeater
Godin, PRS, Crafter and Roland guitars
Center Point Stereo Spacestation V3