FM3 Internal/Passthrough Latency Measurements - Surprising Results?

[mr_fender]

Are you direct monitoring via the FM3's outputs or trying to monitor via software round trip while tracking? Direct monitoring is the way to go for the lowest latency while recording.

 

[Fredmillard]

if Direct monitoring the track I’m recording, it records out of sync with tracks previously recorded, eg the guitar is out of sync with the drum track, etc. If monitoring the DAW output I’m unable to play the piece with precision because what I’m hearing is delayed from my playing. Either way it’s a nightmare. Have not been able yet to get a reasonable explanation. Would really like to record from the FM3 using its built-in I/O rather than having to connect the FM3’s audio output into the Focusrite’s audio input using the latter’s I/O conversion. Regards. Fred

 

[GlennO]

if Direct monitoring the track I’m recording, it records out of sync with tracks previously recorded, eg the guitar is out of sync with the drum track, etc. If monitoring the DAW output I’m unable to play the piece with precision because what I’m hearing is delayed from my playing. Either way it’s a nightmare. Have not been able yet to get a reasonable explanation. Would really like to record from the FM3 using its built-in I/O rather than having to connect the FM3’s audio output into the Focusrite’s audio input using the latter’s I/O conversion. Regards. Fred

You should be monitoring direct. You'll find information about recording with your FM3 here:
https://forum.fractalaudio.com/threads/axe-fx-for-the-recording-musician.177592/

See this for more details on the latency compensation issue mentioned in that guide:
https://forum.fractalaudio.com/threads/latency-compensation-measurement.177851/

P.S. Your issue is unrelated to the topic of this thread. If you have more questions, you'll probably have better luck getting help if you start a new thread.

 

[laxu]

I am unable to reduce the audio interface latency when using the FM3 to record into a DAW. The pass through audio is instantaneous but when using the FM3 as an interface the latency is unbearable. Ironically if I route the audio out of the FM3 into the audio input of my Focusrite Scarlett 4I4, the latency is unnoticeable. But if I use the FM3 as an audio interface directly into the computer via USB, the latency is significant. I have been told that the issue is with my computer, but I beg to differ. I have also been told that if there is significant internal latency - the real time audio would behave the same. I beg to differ with that theory as well. I have tried every buffer setting to no avail. What is FAS doing to deal with this issue?

This is without doubt an issue with your computer. I have a Scarlett 6i6 2nd gen and have no latency issues on the FM3.

Make sure you have the latest USB drivers installed and check the latency settings of the FM3 when using it with ASIO.

 

[Ugly Bunny]

This thread turned into a train wreck since it was necrobumped - I don't think anyone is on the same page lol! This thread is definitely not about USB recording latency, although that's what it turned into.

In any case, I would love to see the test results as well with the new 5.0 FW and the Virtual Capo, assuming it does get the same treatment the Axe Fx III's VC got (as was mentioned earlier). I did a bunch of latency on gear a few months ago including the Variax, the Axe Fx III, the Digitech Drop - I don't think my tests were as precise as this one, but they still gave a pretty good relative assessment of the latency: I tapped a hot microphone to the strings and measured the difference between when the mic signal was tracked and when the processed guitar signal was tracked (or I tapped the strings with something where the mic, placed right next to where the strings were struck, could pick it up at the same time).

I plan to repeat the tests with 5.0 - but, let's be real: We don't even know for certain there will be Virtual Capo changes, let alone improvements, LET ALONE improved to the Axe Fx III's level. Still, better latency is better and I'll take anything I can get at this point lol!

 

[Fredmillard]

Notwithstanding I’m posting to the wrong thread, I nevertheless thank you for directing me to the article addressing my issue. Although I have not yet performed the fix, I’m glad to know it’s a common issue that Fractal knows about. Regards. Fred

 

[shredhead]

Hi, i just did the same measurements on current firmware (Current FM3 Firmware Version: 4.01) and got identical results, recording in1->out2 or DI gives 2.6 ms of latency in the analog output 2. AMP+CAB gives same 4ms. Is there some specific update (not just firmware) required to fix empty chain latency?

 

[Ugly Bunny]

Hi, i just did the same measurements on current firmware (Current FM3 Firmware Version: 4.01) and got identical results, recording in1->out2 or DI gives 2.6 ms of latency in the analog output 2. AMP+CAB gives same 4ms. Is there some specific update (not just firmware) required to fix empty chain latency?

No, that's correct; I averaged about 3ms of latency as well, so 2.6 or 4ms sounds right. You won't get much less than that. The HX Stomp runs about 2ms, the Eventide H9 about 3ms - it's all in the neighborhood.

 

[Will Chen]

Hi, i just did the same measurements on current firmware (Current FM3 Firmware Version: 4.01) and got identical results, recording in1->out2 or DI gives 2.6 ms of latency in the analog output 2. AMP+CAB gives same 4ms. Is there some specific update (not just firmware) required to fix empty chain latency?

Pretty sure 2.6 is just the time of the AD/DA.

 

[shredhead]

No, that's correct; I averaged about 3ms of latency as well, so 2.6 or 4ms sounds right. You won't get much less than that. The HX Stomp runs about 2ms, the Eventide H9 about 3ms - it's all in the neighborhood.

At the beginning of that thread the guy got the same result in 2020 and was told that 2.6ms is too much and it was getting fixed to 1.7ms by FW update. And it looked like it was fixed in FW 1.06. I am wondering, what happened with the fix

 

[Ugly Bunny]

At the beginning of that thread the guy got the same result in 2020 and was told that 2.6ms is too much and it was getting fixed to 1.7ms by FW update. And it looked like it was fixed in FW 1.06. I am wondering, what happened with the fix

Gotcha. Might have to contact FAS directly about that. I've learned that almost nobody on these forums but myself care about or even notice latency in any form lol. I'd be blown away if it were possible to get 1.7ms on literally ANY digital modeler, but I'll be following this thread in case we learn anything.

 

[shredhead]

Gotcha. Might have to contact FAS directly about that. I've learned that almost nobody on these forums but myself care about or even notice latency in any form lol. I'd be blown away if it were possible to get 1.7ms on literally ANY digital modeler, but I'll be following this thread in case we learn anything.

I thought that Axe FX latency should be less than 2 ms (full latency, with amp and effects). Understandable that FM3 is a bit higher, but still, 4 ms is borderline for my feel, and its 4.7 with effects

 

[smj]

Just curious, does using spdif out reduce latency as it avoids the D->A on the output?

Sean Meredith-Jones

 

[GlennO]

Just curious, does using spdif out reduce latency as it avoids the D->A on the output?

Sean Meredith-Jones

That spdif signal needs to be converted to analog at some point for you to hear it. So, you'll incur some D/A latency one way or another, even if it isn't inside the FM3. (Note that this thread is about A/D/A processing, not digital recording.). Depending on your audio interface, it could be slightly lower though.

 

[smj]

That spdif signal needs to be converted to analog at some point for you to hear it. So, you'll incur that D/A latency one way or another, even if it isn't inside the FM3. (Note that this thread is about A/D/A processing, not digital recording.)

It is recording. I actually use the analog outs of the FM3 into a lynx aurora. That goes AES into my computer. It’s easier to work like that than using the Fm3 USB out. I believe it is possible to convert Spdif to AES, but I won’t consider going down that road if there isn’t much of a payoff.

Sean Meredith-Jones

 

[GlennO]

It is recording. I actually use the analog outs of the FM3 into a lynx aurora. That goes AES into my computer. It’s easier to work like that than using the Fm3 USB out. I believe it is possible to convert Spdif to AES, but I won’t consider going down that road if there isn’t much of a payoff.

Sean Meredith-Jones

That's not digital recording .

On the topic of recording:
a) I'd recommend using a digital connection to your audio interface, regardless of latency, to avoid the superfluous D/A/D conversion.
b) Latency is generally moot anyway when recording since one, in most cases, monitors direct.

To answer your question though: yes, skipping the D/A conversion will reduce the internal latency.

 

[smj]

That's not digital recording .

On the topic of recording:
a) I'd recommend using a digital connection to your audio interface, regardless of latency, to avoid the superfluous D/A/D conversion.
b) Latency is generally moot anyway when recording since one, in most cases, monitors direct.

To answer your question though: yes, skipping the D/A conversion will reduce the internal latency.

I guess I’ll have to do some tests. The assertion was made at the beginning of the thread there is a certain level of internal latency on the FM3 from analog in to analog out. So in my case, there’s an extra conversion going analog out of the fm3 into another digital unit. The caveat being that the Aurora is AES which is a real time audio connection…. so I figured I should be ok.

Fractal mentioned it would be reduced in an update.

I’m wondering if going fm3 usb out right into my computer (which is still a data connection not real-time) is slightly faster than cumulative analog out into the aurora. Thus, eliminating one step of the latency inherent in the FM3.

Yes, diminishing returns… but I have to know!

I take it the AXE3 and FM9 are a bit less latency?

Sean Meredith-Jones

 

[shredhead]

I guess I’ll have to do some tests. The assertion was made at the beginning of the thread there is a certain level of internal latency on the FM3 from analog in to analog out. So in my case, there’s an extra conversion going analog out of the fm3 into another digital unit. The caveat being that the Aurora is AES which is a real time audio connection…. so I figured I should be ok.

Fractal mentioned it would be reduced in an update.

I’m wondering if going fm3 usb out right into my computer (which is still a data connection not real-time) is slightly faster than cumulative analog out into the aurora. Thus, eliminating one step of the latency inherent in the FM3.

Yes, diminishing returns… but I have to know!

I take it the AXE3 and FM9 are a bit less latency?

Sean Meredith-Jones

AD-DA latency is not a big deal compared to USB recording latency. You have to adjust for it manually in DAW anyway, in my case with internal FM3 USB buffer set to 48 I have to do record adjusting in DAW (Cubase) by 9 ms (432 samples) - done by recording metronome in a loop. 0.25 ms from one DA is nothing compared to that

 

[shredhead]

And Axe FX 3 is having the same USB buffer latency as far as I know form people owning it

 

[GlennO]

I guess I’ll have to do some tests. The assertion was made at the beginning of the thread there is a certain level of internal latency on the FM3 from analog in to analog out. So in my case, there’s an extra conversion going analog out of the fm3 into another digital unit. The caveat being that the Aurora is AES which is a real time audio connection…. so I figured I should be ok.

Fractal mentioned it would be reduced in an update.

I’m wondering if going fm3 usb out right into my computer (which is still a data connection not real-time) is slightly faster than cumulative analog out into the aurora. Thus, eliminating one step of the latency inherent in the FM3.

Yes, diminishing returns… but I have to know!

I take it the AXE3 and FM9 are a bit less latency?

Sean Meredith-Jones

It's not diminishing returns. It's irrelevant returns . How are you connecting your Aurora to your computer? Yes, the Axe-FX III will have less internal latency.