Thanks for the test! I even appreciate more, to be a MG-30 user. Thank you also again for your first MG-30 test. You were one of the sources on which I based my decision to buy the MG-30 8 month ago. There is a new firmware btw, including a new Princeton amp.
Very interesting video..! I consider that Boss makes very good products. I have never paid attention to latency, but I think your video is pretty interesting for those who need this information. tTanks for taking the time to get this video done!
I truly appreciate your precision, dedication and commitment to a topic, and work in general. Thank you so much for your work, it is very valuable to an amateur musician, such as myself. 🙏
Great video, both this and the 23 modeler latency test. I wouldnt be at all surpriced if the fact that Boss uses their own custom DSP chips unlike the others have a major effect on why they are able to achieve such short latency.
Roland does manufacture their own custom DSP. The GT1000 has one of them, some of their synths have three or more. They must have figured out how to parallelize processing.
Many thanks, Leo. Hope you have time to make a video on the latency of the BOSS GX-100 with a UAFX amp in the FX send/return as you did with the TC X5. Or any other unit you would like to pair a UAFX amp with. Great work for the community.
It's very interesting to see the impact of adding more blocks to the signal chain. I'm also looking forward to a new iteration of the slap fight on thegearpage :)
Hey Leo. Glad to see you finally considered my suggestion :) (Add Quad Cortex latency test). I've also made test and got similar result with you. One more thing need to highlight is what I found is if I use row 3+4, then the latency will increase at least 2 - 3ms. However, I still thank you for such useful video to us as lots of people never pay any attention to "Latency". People learned the "latency" should only happen on audio interfaces, but ignore that any of AD/DA device will bring latency. I would say after saw this video I first time focus on "latency", it's very useful for me when purchasing other accessories (like the wireless system) , I have to pay attention to the latency. Maybe I am not sensitive for within 8ms, but multiple AD/DA device mix together ex. digital wireless (3 - 5ms) --> Modeler ( average 5ms) -> digital console (3 - 5ms) then the latency may come to more than 15ms noticable latency.
Thank you Wilson! Very interesting points...and I totally agree...we have to take into consideration the overall latency introduced by our signal chain, with wireless devices, etc...thank you for pointing it out!
Great analysis Leo. A nice calculation to add would be to determine the trade off between the bit depth and sample rate and the latency between these devices.
Yes. Sorry. In order to get a sense of the trade-off. One would assume that these higher end modes are using dsp (and thus incurring a latency debt) to produce a more complex tone.
I do agree...this was something I said in my last video about the latency of 23 amp modelling pedalboards. But for instance if you add a delay of lets say 100 ms, you may have all the time to make the calculation without adding latency...as the first repetition happen after 100 ms...so maybe some clever solutions can also be studied and found to improve the latency... Thank you!
@@LeoGibsonGtr thank you for the reply Leo. However, I was not referring to the signal chain complexity, but to the complexity of the tone produced by the amp. In other words, what is the interaction between the amp model sophistication (responsiveness, dynamics, etc) and the the bit depth and sample rate and the latency. I see you have a new video that looks at some of these factors, but I wonder what is the specific relationship between bit depth, sample rate, latency and amp model characteristics. I hope this is clearer. Thank you again for your great in depth analyses. Nobody on RU-vid does it to the extent that you do. keep it up.
Thank you Leo for another informative video that includes technical discussion. You ask at the end of the video for other suggestions for a deep dive into other devices. I would like to suggest looking at the input impedance's for multi-effects modellers and how it impacts on tonal output range. Most devices, (mult-effects and pedals) state an input impedance of approx 1 Meg ohm. This high impedance has minimal effect on loading of the guitars passive pickups and no noticeable effect on tonal frequency range. However not all devices have this desired, (for passive pickups) high impedance, case in point being the Line 6 Pod Go which has an impedance of approx 250-300k ohm. This means a loss of tonal range (frequency) in the output. Perhaps you may be interested in doing a deep dive into tonal loss of modellers that have a low input impedance or if in fact many do have a low input impedance? Can this problem be fixed at the expense of adding a buffer pedal between the guitar and modeller? I very much appreciate that you dive into the technical aspects of these devices as well as providing such thorough reviews!
Hi, thank you so much, this is a great subject! I think that Latency, Impedance, dynamic range of the converters and aliasing are some of the most important factors to verify in these modellers... As regards impedance, I need to study the subject and verify if there is a cost effective methodology to measure it...if you know a method, please feel free to post a suggestion here, it would be really appreciated...in the meantime I will also check out what I can do. Thank you!
@@LeoGibsonGtr Thats great that you are considering doing a deep dive into this topic. I am a new (beginner) guitarist as was looking to buy the Pod Go or the GX-100 and while doing research on which was best for my use I came across information on input impedance issues on some multi-effects modellers. I became concerned that perhaps manufacturers are not providing the correct specification and tone loss could result? I found this link provides detail on how the measurements may be done. ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-Ykezs1A7PBE.html I still have not decided which multi-effects unit I will buy and would very much appreciate a deep dive into this subject to see if there is any justification for my concerns in relation to input impedance having an adverse effect on tonal range.
Thanks Leo, I would never have thought that a even a deactivated block would increase latency (and presumably use dsp). I will check out this principle inside of my DAW as well.
You are welcome! I think that deactivated blocks take cpu power in order to allow a very quick switch on of the block itself... I'm not sure...but I think it is reasonable. Thank you!
@@LeoGibsonGtr It can be seen as a "neutral function". Like if you use an EQ but don't change any level. The data will go through it, but won't be changed. Its interesting here how MG-30 is conceptually different here. Instead of dynamically allocating blocks (which add up) it has a fixed chain of 11 blocks. They can be reordered, but they are always in the processing pipeline. The Latency of the MG-30 fx chain adds to a total of somewhere around 2ms. You can use any combination of effects and it will work. But if NUX would want to add a new effect, the DSP power the effect can use is limited to the limit of its block (e.g. EFX). The effect cant take more because that would break the Latency limit of the whole chain. So if they want to have significantly more expensive effects or modeling they will need a new device with more DSP power to allow for more expensive effects. In a device like QC (or Helix) the blocks add up dynamically until the DSP limit is reached. So you cannot create _any_ possible combination of FX. But you could have a new expensive effect added. It might be impossible to be combined with most of the other stuff, but you can put it in. If you want to allow more _combinations_ of effects you need hardware with more DSP power.
Interesting. I am also curious about older gear like Boss GT-3 to GT-8. Some Korg AX series. And maybe Zoom GFX series. It would be fun nostalgia. Thanks Leo. you're the best.
You are welcome and thank you for your suggestions!!! I don't have those devices...but I agree with you...it would be interesting to see the progress between those old modellers and the new ones....
It’s interesting that the quad cortex has this much latency considering most of the main stream reviewer‘s said it feels like it has none at all. I know personally anything over 5 ms is noticeable for me when playing in to a DAW. I guess I will find out either way as mine is being delivered on Tuesday. Thanks for all your hard work.
Depending on the way you're measuring, you might be severely underestimating how much latency you're really experiencing when playing into a DAW, e.g. what you think is 5 ms might actually be 15 - 20 ms. The numbers in the DAW don't reflect full monitoring latency.
Very nice as always Leo.....Thank you mate ;)....As a common rule of thumb, anything up to 5 ms is perfectly fine. ;).....Sound itself has a velocity latency around 340 m/s :)
5ms is 1,7m distance. But if you combine mfx units, add wireless transmitters, digital pedals like Digitech The Drop - you will raise it. So if 5ms is absolutely ok for you, using a transmitter which pushes it to 8ms might be bad. If you modeler had 2ms the final result would be 5ms so still ok. Which means: One modeler would allow you to use a wireless transmitter and the other might not. The distance to the speaker introduces latency of course but: Using an analog amp and monitor speakers 1,7m in front of you would give you acceptable 5ms... and you are used to playing that way. If you replace the amp with a modeler that adds 5ms you end up at 10ms... so it feels like you're playing 3,4m away from your monitor speakers... which might feel strange. Same with putting your ear at 30cm from the speaker (hopefully not to loud ;) ) and then playing a note through the 5ms modeler might feel a bit strange... because the distance to the speaker doesn't fit with the latency of the signal into your ear.. which again might feel strange. So even such small latency values can lead to situations were you can perfectly be able to play, but you feel that something is a bit strange.
@@aptfx yes, true. 5 ms is just a ballpark figure that s an acceptable value. I was referring to equipment only, like playing though a soundcard with virtual amps, or using the GT itself, for recording. What introduces a ton of latency is input monitoring. I stay away from it, as much as I can. many times I can t. LOL. As long as plucking the strings and the feel of the guitar doesn t feel weird, I m fine.
I swear guitarist are some of the most smart musicians on earth Details on how the instrument is build, price and quality is key. So happy to be a guitar player 😊🎸💯
What if we compare the 3.2 millisecond measurement to the 3.3 millisecond measurement (e.g., FM3) and want to know the relative error? If my calculations are correct, the 3.2 millisecond measurement has an approximate error of 3.03% relative to the 3.3 millisecond measurement. This means that if their estimation is correct, yours is accurate to 96.97%. This suggests that the potential error in your measurement is less than negligible. As usual, you nailed it! Great video!
boss latency is almost unbelievable! They are using quality converters for sure and the dsp they use is super optimized. My gx 100 is coming this week, I’ve been out the pedalboards game for 3/4 years (just had a zoom g3x) and been using software sims on stage with a macbook. Been a bit disappointed by neural tone king I bought, sounds amazing but it s so heavy on the cpu I decided to give a try to the boss, a brand I trust. Let s see how they compare, for sure the boss is gonna be latency free
Slightly off subject, I recently upgraded from my a boss katana artist to a mk 2 I have to say the power amp input on the back really does what it says, I can run my GT 100 into with the cabinet models and it doesn’t colour the sound at all. I thought about getting a Headrush FRFR but I’m happy what the katana is doing
@@Matthew_Scan oh snap, sorry. I meant the fx return input, which sometimes people compare to a power amp in since it bypasses the preamp section. But it sounds quite weird to my ears...
@@gusblacksmith9 Yeah noticed a difference between the effect returns on and the power amp in, I was sceptical, but it does do a good job as a powered speaker monitors, I A/B it with my studio monitors and the only difference was the katana added mid range which was a good thing , the monitors have zero mid range I had a Headrush powered speaker on order, i cancelled it and got the katana with the power amp in
Leo, you were very kind to not put the FM3 under a microscope. The 3.3ms of latency could have been a very basic signal chain which might balloon when more effects were added. That would have started a war with Fractal.
Hi, actually I just preferred to keep the video focus on 1 high end unit (QC as there is a lot of interest in this topic), 1 middle bdg e 1 low budget, also considering that these 3 units manage Latency in pretty different ways....I will check out the FM 3 in future videos. I love Fractal products and I have no reason to fight with them...In the video I just wanted to point out that the methodology I have used is right, also confirmed by Fractal data... Thank you for your message!
This is a fantastic comparison. As someone who's very sensitive to latency (I generally start getting annoyed at 5ms. It feels like a super fast slapback delay. 10ms is unbearable) I'm loving this. I can't wait till you do an hx stomp, fm3/9 comparison with the cab blocks on/off. Keep up the good work. Btw, English speaking people would expect to hear "Lio" and not "Leo" when you introduce yourself, they might think you are saying "LEL". (I'm writing the pronunciation in Italian) I understand if you don't want to change how you pronounce your name though ;)
Thank you so much, both for the compliments and for your suggestions! Well, In Italy Leo is pronounced LEO and not LIO...therefore it would be pretty strange for me to call myself LIO...I'm sorry, I can't do it... 😀
Thank you! I'm happy Fractal replied... as it is a constructive and useful discussion, in my opinion... I just hoped that they didn't say my numbers were wrong before understanding the tests and trying themselves...But when basically they confirmed my numbers... I was really happy!
Hi, I have test it with other 23 modellers in this video: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-oU0-_3Hub9o.html With just an amp and a cab in the chain...
Yes, I've seen this test. Thank you so much for it, but there is no Axe fx 3. It is very interesting to compare it with FM3 and find out how it compares with others in terms of latency. Thanks a lot!
Hi, they were placed in the upper row. I'm not 100% sure, but I have read that when you have many blocks, it is better to distribute them among multiple rows. Thank you!
So, what would be a latency value that starts being a problem? I know that this varies person to person, but I'd guess that anything below 10ms is good enough?
Hi, hard to tell...for some of us 10 milliseconds are fine, for other 5 milliseconds are not acceptable. I'm not very sensitive to latency as I have played a lot of midi guitar in the past and I can adapt my playing...it's really subjective. Thank you!
I think you hit the nail on the head with this question and at the same time calling out the "omg I feel 5ms of latency" players. Lol no one can tell 5ms of latency like someone else said here. I would say on the safe side 10 to 15 is kind of normal. I also feel the latency way less through a modeller playing normally than you would monitoring anything through a DAW. Hence don't monitor through a DAW if you're recording, problem solved. As a benchmark, think for example doing the doubler delay effect which petrucci is known for to thicken up rythums, that's around 17ms to 20ms which is clearly an effect.
@@teatime6414 Well, if we take a reasonable approximation of the speed of sound as 343 m/s, we have 0.343 m/ms, or circa 1 foot/ms. So 10 ms will add the distance of ~3.5 m or 10 feet to the real distance to your speaker, IF you're using a speaker. If you're using phones, it's the same as the length of a short cable. I really doubt that this would be a problem. But of course, were ADDING latency to a system, so depending on your setup, it could make a difference.
love my NUX MG30 without a doubt! but it kind of feels like it can't chugg specially with active pickups... which would be your choice if I needee some more hardcore sounds? Latency is a huge Issue for me too, so the other 2 are out! thanks for your work 🤝🏻
Hi, thank you for sharing! Well, it's very hard to make such a suggestion, because an heavy sound that is right for me, maybe wrong for you and vice versa, and actually I don't play too much hardcore tones...I'm sorry...
@@LeoGibsonGtr hi Leo and can be different results with first TAIL off main rig screen and you can off second type TAIL in reverb or delay section in mx5. Because manuel says if "TAIL on" actually mean always open / ready this section even you bypass it.
I'm surprised to see that the ON/OFF status of the blocks has an affect on the latency. It seems like that would make it impossible to have gapless switching in "Scene" and "Stomp" modes, which QC clearly has.
Well, I think in this way the unit allows us to switch on the effect block very quickly. I'm sorry but I've not understood your opinion about the gapless switching. Thank you for your comment
@@LeoGibsonGtr In order to produce a continuous stream of audio, without interruptions, each buffer has to be processed in constant time (or if there are some internal variations, you force it to be constant time by making each packet wait in the buffer until its time, but you must pick a latency number high enough so that the slowest packets still fit inside of that budget). Think what happens when you monitor a guitar through a VST on an old computer and the computer slows down because of another application, or simply set the I/O buffer too low - you get choppy audio. So if turning on a block increases the end-to-end latency, from say, 3 ms to 4 ms, it means that the first packet going through the modified chain will be delayed compared to the previous one by 1 ms, creating a gap in sound. The fact that the presence of blocks on the grid changes latency (regardless of their on/off status) is one of the fundamental reasons why preset changes can't be gapless on QC, but changes via Scene and Stomp modes are gapless, even though they can turn blocks on and off. Now perhaps the gap is there but it's too short for us to hear, but we should be able to see it on the waveform.
Interesting that Kemper has just improved their patch switching latency. I think your previous video has caused a stir in the market and everyone is looking at improving their current and future products latency to compete.
Hi, let me just say that my test is not about the latency of the presets switching, but the latency derived from the ADA conversion and the calculation needed to create the target tone. It's the time that passes from the moment that you pick a sting since the moment that the modellers creates the target tone. Here you can find more info: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-oU0-_3Hub9o.html Thank you!
Latency is certainly an interesting parameter. But anything less than 10ms is likely not detectable by most people. There may be more important characteristics of these modelers to evaluate their applications for you. There may also be an inverse relationship between latency and quality and flexibility of modeled tones due to differences in algorithms and converters. But that would be subjective.
Hi Jim, thank you for sharing...interesting point...yes, there could be a relation...but for some of us latency is pretty important, and well...if you want to check out how these units sound...there are a lot of videos in my channel...😀👍
Hi, I don't think there is a strict rule. Someone put right at the beginning to cut off the incoming signal, someone else after the amp to cut off also the amp noise. I don't use noise gate, as it typically ruins the way I play. Thank you for your message!
Although I am really not a good player, I "feel" latency pretty quickly. 5ms would cause issues for me. So Quad Cortex and Kemper would be not a good choice for me.
Don't play too far from speakers. Sound travel about 3 ms a meter so 1,5 Mt away can cause annoying problem on you playing but can keep safe you ears . It's a good trade of isn't it? 😉
It's a bit sad to see that unit with a price range reaching $2k can't be equipped with beafy CPU enough to handle all the required routing. And huge burst of affordable processors from NUX, Mooer, Hotone etc. proves it very well. I remember like 5-7 years ago small 2-pedal unit was viewed as Zoom-505 territory, but now they sound better than many big units from huge companies.
Latency cannot be noticed up to 15ms. Just play a couple of meter away from the speakers introduce few ms due to the sound that travels on air. The problem I notice with mg30 is that with an empty chain the sound is modified . Looks more compress with more presence and less dynamic. I fear that to be quick in AD conversion NUX loose some sound detail which for me are more important than latency. This is my 2 cents , You may not agree with me and is totally fine. A suggestion for next race if you like. How much AD conversion vs speed impact to the sound quality. I really appreciate you channel and you work
If you test the signal on the AP(audio precision), and bypass all the blocks of MG-30 (remember the pre-eq & post-eq are flat, and choose studio direct out mode), the input and output signal is almost the same including the spectrum.
i always trust you master LEO, you made me purchased the mg300 before though the latency is quite high but i love how the unit sound😃 God bless in Jesus Name