How do I test for latency compensation when using AxeIII as interface?

MackieFX

Experienced
To set latency compensation I used the method in this thread. I came to 597 samples as the compensation needed.
https://forum.fractalaudio.com/threads/how-do-i-send-a-daw-click-through-axe3.177637/

I did a test with normal live recording where I tried as best I could to play as close to a click as possible for about 12 bars. Just one dead-note per click. I did this with both 0 and 597 sample compensation.

What I found was that the recorded audio showed me coming in too early with latency compensation set to 597 samples when compared with 0 compensation.

I am not a robot and cannot play 100% in time so this is about as scientific as I could get. Maybe I have a tendency to come in early and this test merely exaggerated this.

tl:dr
1. Does anyone have experience setting up compensation for both recording and re-amping?
2. Do you need to set compensation for each of these situations?
 
I think, the method in the thread ignores all input and output latency.

I would put IN1 and OUT1 in the grid, but not connect them. Then run a cable from OUT1 to IN1. Pass signal from DAW thru OUT1 and record IN1 to a new track.

You can also use Out2->In1 if you worry about feedback issues.
 
I think, the method in the thread ignores all input and output latency.

I would put IN1 and OUT1 in the grid, but not connect them. Then run a cable from OUT1 to IN1. Pass signal from DAW thru OUT1 and record IN1 to a new track.

You can also use Out2->In1 if you worry about feedback issues.
That works for re-amping, but doesn't replicate IN1-OUT1-DAW situations. Is my assumption that different latency compensations need to be used for re-amping and normal recording?
 
I just did a quick screen recording to show how I do it in Logic. I used my FM3 because my AX3 is packed up in the road case right now but the method is the same on both.

I do not use the same value in all situations as the value returned by the I/O plugin varies from day to day and project to project, so I do this on every new project right before I start tracking guitar/bass.




As I mentioned in the video, on the AX3 you can add the input USB block in parallel to the In 1 block and set the I/O plugin's output to USB 5/6. I have that block permanently on all my AX3 presets which saves me the step of going into the Setup menu and changing the Input 1 Source back and forth from USB to Analog every time I do this test or want to reamp something.
 
Last edited:
I just did a quick screen recording to show how I do it in Logic. I used my FM3 because my AX3 is packed up in the road case right now but the method is the same on both.

I do not use the same value in all situations as the value returned by the I/O plugin varies from day to day and project to project, so I do this on every new project on which I'm tracking guitar.




As I mentioned in the video, on the AX3 you can add the input USB block in parallel to the In 1 block and set the I/O plugin's output to USB 5/6. I have that block permanently on all my AX3 presets which saves me the step of going into the Setup menu and changing the Input 1 Source back and forth from USB to Analog every time.

Wow thanks so much for taking the time to record this - especially so early!

Do I need to do this if im NOT re-amping?
 
Wow thanks so much for taking the time to record this - especially so early!

Do I need to do this if im NOT re-amping?

Yes, this sets the latency compensation for all recorded audio so you need to do it before you start tracking anything through the FM3/FM9/AX3.

This is the single gripe I have with these units. Even cheap units like the Pod Go or Scarlett Solo accurately report their latency so this process is unnecessary with them.
 
Bit confused about all this. Do I need to bother with latency settings if I'm running AXIII via SPDIF to UAD Apollo Twin to DAW via Thunderbolt? All good in that case, right?
 
Bit confused about all this. Do I need to bother with latency settings if I'm running AXIII via SPDIF to UAD Apollo Twin to DAW via Thunderbolt? All good in that case, right?
That's probably ok. But it couldn't hurt to do a loopback test yourself to make sure the latency is being compensated properly.
 
Strictly speaking, there is no input and output latency. There is only buffering latency, and that should be the same regardless of whether the signal source is the A/D converter or usb input. I use the same compensation for all situations.
There is roughly 2ms for AD and DA conversation

That works for re-amping, but doesn't replicate IN1-OUT1-DAW situations. Is my assumption that different latency compensations need to be used for re-amping and normal recording?
When you record you go through the output DA and back in through the input DA. When you reamp, you can keep everything in the digital domain.
 
When you are recording the click test, you are measuring round trip latency (both playback and recording). When you do the test playing live along with the DAW's metronome, you are only getting recording latency. That's why the 597 setting is too high. It's compensating too much and pushing your recorded track too far forward, making it sound like you are hitting too early relative to the DAW click.
 
DO NOT ADJUST RECORDING LATENCY/DELAY IN YOUR DAW! Unless you know you are done tracking and all you plan on doing is re-amping tracks. If you forget to reset your delay to zero when you go back to track more of the song, whatever you just tracked is now early compared to the rest of the tracking. It isn't fun, but just takes a lot of editing to fix. Honestly, I wouldn't even worry about re-amp delay unless you're dealing with a bass DI when re-amping a bass amp you plan to mix with the DI. Truth be told, you should always double check your DI with your bass amp waveform visually if you plan on blending them. Same with blending mic'd guitar cabs; you usually end up time adjusting one of the mics.

There is a perfect solution that doesn't involve DAW or hardware compensation (I don't trust UAD's). Just do a visual check of the two waveforms, and then just slip/move the outboard's recorded waveform back X samples until they are perfectly aligned. This can be really hard to tell if the "re-amp" signal is much different, so I would just figure out the latency beforehand for each outboard device with an easy to distinguished audio source like the single lowest note on a clean bass guitar DI like I did in the picture attached.

In the picture, red and green are different takes because I had to figure out how to record DI with just the FX-III as an audio interface since I feed mine into a ADAT into a Apollo Twin normally. Sorry. The Output of the III is time perfect aligned with the USB direct in it recorded (mine aren't since red and green are different takes). When I re-amped the signals from output 1 and the USB-DI, both re-amps were ~41.5 samples late. If you change the Axe's buffer size, these will change. Just find a buffer size that works for you, and then manually adjust the waveforms in your DAW if you truly feel you need to. Again, I only do this with bass and drums if they sound off.

Screen Shot 2021-10-19 at 2.44.06 AM.png

Since I record into an ADAT into my Apollo, I just keep my buffer size at 256 for Axe-FX III. I Keep Pro Tools at I/O 1024 myself, but I record with low latency mode on. You almost had me second guessing myself about the III's I/O buffer, but even monitoring through my Apollo, I can't feel any input latency like I would if software monitoring through a plugin at lower input buffer.
 
Playing is way too inaccurate. For playing, I did test with metronome: record few minutes of metronome. Play that track, and fiddle with metronome until you hear the recorded and new live metronome perfectly in sync. Then record again, onto track2. Look if they really line up or not.

For reamping, reamp some sharp click and see that you align.
 
Unfortunately Not
That's unfortunate. I'm on the forum to learn about fractal products before hopefully buying one. Between this and the seemingly less common high pitched squealing issue, it seems like they really don't have a handle on USB recording, which would be my main use.
 
That's unfortunate. I'm on the forum to learn about fractal products before hopefully buying one. Between this and the seemingly less common high pitched squealing issue, it seems like they really don't have a handle on USB recording, which would be my main use.
USB recording is fine -- reamping is what requires either setting the latency compensation offset or slipping the waveform after the fact. (This is true with any outboard gear that you "reamp" through)

The problem with setting latency compensation offset for reamping is that -- for some people -- it does not remain constant. I believe this problem is either caused in part by or exacerbated by plugins with delay compensation (PDC). However removing those plugins does not re-stabilize the offset...which is odd to say the least. I have a ticket open with Fractal on this...so crossing my fingers for a solution.

I will say, I have only tested in Reaper, so I can't claim 100% that it's not an issue with certain DAWs. I've tested on two computers, fresh Reaper installs, in many different scenarios and can recreate the problem over and over. I've been tempted to try Cubase (@GlennO mentioned in another therad that it has more flexibility dealing with PDC), but I've been a bit gun shy with the price tag and converting projects over.

For what it's worth, here are a handful of people who I recall mentioning the "variable offset" problem: @sixtystring, @steadystate, @zappafranco, @tla
 
So to close this thread off just 1 final question

- Theres no reason to worry about RECORDING latency and I can just leave the compensation at 0?

- Re-amping / Round trip etc is covered already :)
 
To set latency compensation I used the method in this thread. I came to 597 samples as the compensation needed.
https://forum.fractalaudio.com/threads/how-do-i-send-a-daw-click-through-axe3.177637/

I did a test with normal live recording where I tried as best I could to play as close to a click as possible for about 12 bars. Just one dead-note per click. I did this with both 0 and 597 sample compensation.

What I found was that the recorded audio showed me coming in too early with latency compensation set to 597 samples when compared with 0 compensation.

I am not a robot and cannot play 100% in time so this is about as scientific as I could get. Maybe I have a tendency to come in early and this test merely exaggerated this.

tl:dr
1. Does anyone have experience setting up compensation for both recording and re-amping?
2. Do you need to set compensation for each of these situations?
See this post about latency compensation for details and instructions:

https://forum.fractalaudio.com/threads/latency-compensation-measurement.177851/

Yes, you always need latency compensation, not just when re-amping.
 
Back
Top Bottom