NOTE: Disregard the H9 test. Reason 1: It is not meant to be a polyphonic shifter. If you play a full chord (and not just a quick transient), you'll get all sorts of crackling and artifacts. I should have done more research before including this one. Reason 2: I had a short delay on the preset (not shown in the settings screenshots) which unfairly delayed the signal. I retested without the delay and it's still about 42ms for a half-step down - so still the worst of the bunch, just not 222ms bad lol!. So yeah - ignore the H9 results as this was meant as a polyphonic pitch shifting test. Thanks!

Dude this is the video I was looking for!! Thank you so much

Well done. Thank you for the info. Helped me decide on whether to use a pedal or my AxeFXIII

Interesting data. Thanks for doing this comparison 👍

Amazing video. Thanks.

Thanks for the video! 👍

Great vid, thanks for posting this. I read a study once that found the least human detectable latency was around 9ms. This was a test done using percussion oriented musicians and bass players mostly as they seemed to be the best at perceiving latency according to the testers. I wish I could find the data, it was quite interesting. I have the newer turbo version of the axefx 3. Have you had a chance to compare it with the original version, which I am assuming you used for this test? Again, thanks for taking the time to create and post this. My inner geek is rejoicing.

Great video. This is a great subject. Between the stomp and the variax, which do you feel has a more realistic sound when dropped?

Great test. Dont you mind to test fully loaded fx3 preset?

Just found this. Great test, very informative (especially as I have always been wondering about the Variax's latency). Unfortunately, I seem to be pretty sensitive to latency - and no, it's not because I'm "Mr. In The Pocket" at all. Sure, latency and timing sensivity are sort of related, but they're not always going hand in hand with each other, I know some folks with pretty good timing who are not exactly sensitive to latency and vice verse. Just in case you're wondering, I started measuring device latency almoat around 20 years ago, started a rather huge interface latency thread back then on the LUG (Logic User Group) mailinglist, followed the monster thread @ GS and what not). And after dealing with it for a while I also had some musician friends taking part in some testruns (unfortunately I never exactly documented them as it was just a personal interest, so all that's left is a few draft papers). Anyway, for me it's been most interesting to have someone else conducting ABX blindtests with myself, so I could really find out in which situations latency might become an issue. The results have partialy been obvious (such as latency mattering more through headphones than through whatever monitors), partially interesting and partially astounding. Just as an example, when playing through a clean DI path (no amp sim, no FX, no anything), I can almost 100% reliably detect the difference between the 32 and 64 samples buffer size settings of my UAD-2 interface (resulting in 4.6 vs. 6.1 ms of RTL - physically measured, not calculated). Between 64 and 128 (8.9ms) it works as well, but astonishingly enough between 128 and 256 (14.7ms) it didn't seem *that* much of a difference. In the end, also given some of the test results I got from ABX-ing some friends and some stuff I was just reading about, outside of some only halfrelated situations (such as singers having issues on some kind of latencies when there's phasing between their headphone monitor and their "inner ear", sometimes resulting in them to be better of with larger latency values...), it seems to be that there's a pretty personal "point of no return". Around that very point, we're extremely sensitive but once that point is left behind far enough, things don't seem to matter too much anymore (at least not as much as one would think) until we're really running into timing relevant latency numbers. I guess it's a lot instrument related, too. For example, for a test I forced myself playing through horrible amounts of latency (something like 50ms+) and then alternated (at least I tried...) between listening to the resulting audio and my pick feel. The latter even worked with headphones, hence just concentrating on the feel itself rather than onto what clicky sound might be left when you hit an electric guitar in your home studio environment with some music running through monitors. I'd guess that's possible for most instruments - but it's most certainly no fun because quite obviously, we want to enjoy what's coming out of our rigs. Whatever, my personal takes from all this was trying to eliminate latency as much as possible. One of the measures involved was changing my live monitoring whenever there's an InEar gig, especially after running into some situations when things got really uncomfortable. My modeling based live setup is resulting in overall latency numbers somewhere between 3 and 6 ms (depending on whether I use some devices serially or not), which I'm absolutely fine with. But then, on IEM gigs, the console comes into play. And one thing I've learned the hard way was that you absolutely can't rely on what's advertised or told to you by whatever FOH folks. Yeah, might be true that the minimum latency of whatever digital consoles is just around another 1ms, but you can never bet on the FOH folks (unless you know them well). I had a few gigs when they just slapped whatever compression (or worse: limiting) onto my signal, apparently causing some additional latency, making things anything else but enjoyable. Cheaper digital consoles might be adding a noticeable amount of latency all the time as well. As a result, I am now always bringing a small mixer and only grab the signals of the band from the house monitoring while feeding my own signal myself (usually offers better volume control, too, so that's another benefit). And as a sort of rule of thumb, I usually try to stick to a "no more than two digital devices in series" approach (unless they're offering a dry through function), which is working fine. I'm happily making exceptions for sounds that are more "washed". Doesn't matter too much on those whether the pick attack is straight in my face. But for anything else, that's exactly what I want. Ooops, enough blurb. Thanks for the very informative test again!

Great info. I was extremely disappointed by DropTune pedal, sold it after a week. It was no better than built-in Kemper pitch shifter. You just solved my only reason to buy Axe III instead of FM3. I guess I'll just have to stick to retuning my baritone guitar for lower stuff.

Is there a way to minimize latency to lowest possible with the drop pedal?

Is there an optimal place in a pedal chain for a Digitech Drop box? Or any other pitch changing pedals?

So I saw the new John Petrucci plug in has a transpose section in it, and I checked out the demo. I thought of testing the latency like you did in this video: record microphone of strings, and then a track with pitch shifting. It came out, a measly 2 milliseconds. I was blown away. It's not the best sounding all the time, but it's the fastest pitch shifter I've ever seen yet. You should check it out.

Very eyeopening video for me as I just entered the world of pitch shifting thinking I could finally get rid of 10 different tuned guitars. I have the heelix floor and felt completely out of sync when pitching half step down. It is so annoying I always thought I configured it wrong but seems like its a general problem. I just tested the new omnyss pack in bias fx2 on my ipad pro using focusrite interface. it also comes with a pitch shifter and it definitely has better latency than heelix floor but still noticeable

Great info here. IMO it is all physics, even in a real acoustic environment we are dealing with latency. For digital processing it needs some time for processing unit to process data and output that signal. Esspecially with pitch processing units or apps it needs some buffer as it is as I believe slowing down signal as it receives. It needs to have small chunks of signal to do it one by one. With digital stuff we won't get any better as it needs some time to receive and process data. I guess we can get better few milliseconds here and there but I think if companies tries to go lower it is significantly more expensive to produce such products. I guess that's the reason why some people use analog stuff only including wireless, IDK if analog wireless has some latency basically it should be carried through FM waves. And maybe some analog stuff has latency if there are some chips and conversions going on I think it is far less than digital. We need to learn to live with it somehow adjust playing, or not use it at all.

Im a noob with digital gear. What kind of latency is there if you just want the pedals set up for a “traditional” rig, say, Hx stomp set for OD > Amp > delay > reverb > IR > logic? Would you expect a lot of latency?

would like to see a retest with the H90!

Cool video!! Why no one talk about this specs? Thanx!

That's why I use the Line6 G70 wireless (now discontinued). It has < 2ms latency.

Do you know of any Convincing way to play bass guitar on a regular six string guitar? The way for a person to fundamentally be the bass player in a band with a six string guitar so he can sometimes cover 6 string guitar parts?

I have finished my latest prototype for fast polyphonic octaver (pog-like) pedal. Its based on separate processing of fast transient components, then mixed with the high precision processing of the slow componenets. Its should produce only 10 ms latency on both octave up and octave down output, but not yet tested.

I would guess by now there are chips that if incorporated into a singular device like the digitech drop could pull of much better performance than was is currently on the market. Like somewhere in the 5 or less ms range.. But for whatever reason we keep seeing units or effects built in that just don't cut the mustard..for me at least.

Some perspective: Speed of sound is about 1ms/ft. So adding 20ms of latency in digital is the same latency as standing an extra 20ft from your cab.

is the latenct noticeable on the axe fx

When I use bias fx and have a latency of 3ms it throws me way off so…😬

I think you've made a mistake. The Axe-Fx should not have higher latency at one octave than it does at two.

So, pitch shifting in real time will always have latency, no matter how well technology evolves. The issue with pitch shifting is that it needs a look-ahead time frame to work with data and determine how to process the signal. Imagine predictive text in your phone and how it guesses the word you're typing out, probably coming up with what you were typing after a few letters. To ask for pitch shifting with virtually no latency is like asking predict text to predict what you're typing without typing anything yet. The data used in pitch shifting has to both understand some of the structure of the current signal to determine how to process it smoothly to make it coherent. If you ever tried a daw pitch shifter with a latency control, you'll see the less latency you give it, the less coherent it is, especially when the time is literally shorter than the frequency you're playing. Without that, it would sound like a garbled mess. I honestly think pitch shifting is amazing for what you can do, and how far the technology has come, but one thing I always had trouble with was latency. It's just not super practical for tight rhythm or fast playing

While I appreciate the effort, your tests indicate little of value. Pitch shifting algorithms do not have a fixed time delay (even the most rudimentary that don't detect pitch). The delay time varies between some minimum and some maximum, and is dependent on many factors. This is why many units have parameters such as "lowest note"; to minimize the maximum delay and parameters to minimize the difference between the max and min. When you strike the string, you have no idea if delay is at its minimum, its maximum, or somewhere in between. I suggest that you perform this test on one unit, over and over again, and see if the results vary significantly for the same unit with the same settings. My guess is that they will. It is obvious when playing through a pitch shifter that the delay time is anything but constant. And I believe that your comment that a transient attack will give a different result than a pitch is correct, as different pitches will yield different delay times.