FM3 USB I/O sample buffer size - possible improvement?

VanDaven

VanDaven

Member
Using the FM3 as USB for Logic, the minimum usable sample buffer size seems to be 128 samples. While the whole thing is working somehow with 64 samples, clicks and glitches can be found in the recordings.

Fellow FM3 users, what are your findings, and Fractal Audio crew, can the USB performance of the FM3 be optimized so a latency of 64 samples might be possible?

Thanks & Cheers!°°
 
Depends on a lot of factors. Computer, interface, Mac or Windows, OS version, DAW, cables, etc. ... I've used 64 samples in the past successfully (Logic on Mac, USB).
 
I have the same experience. Occasional pops and crackle with 64 samples over USB.
Going through my interface via SPDIF I can go down to 32 samples in Logic without any problems. Also the latency with SPDIF is slightly lower as mentioned in another post in this forum.
 
I've tried a Mac Pro with 32gb ram, 3ghz processor, and 128 or higher (in Logic Pro X, NO other tracks, plugins, etc) was the only thing that could keep the artifacts away. My Presonus 1824c could work easily with 64 samples, and zero 'glitching'.
 
yek said:
Depends on a lot of factors. Computer, interface, Mac or Windows, OS version, DAW, cables, etc. ... I've used 64 samples in the past successfully (Logic on Mac, USB).
Click to expand...
Thank you for your answer. I am on an M1 Mac Mini, as interface I was of course referring to the FM3 directly via USB into Logic Pro X.
From my experience I don't think esoteric USB cables make any difference, but maybe your experience is different?
It's a bit disappointing, I have an Axe-FX III connected to a Mac Pro in my studio and I didn't encounter any glitches even @ 32 samples buffersize.
 
Everytime I see this topic I wonder what's the point in running at such low buffer sizes?
You can monitor with no latency from the FM3 outputs and pretty much any daw can compensate the latency of your recorded tracks.

Unless you're recording some keyboard vsti plugins or need to monitor thru the daw for whatever reason, there's really no point. I have it always set at 2048 samples and I reduce it only in those mentioned rare cases
 
DLC86 said:
Everytime I see this topic I wonder what's the point in running at such low buffer sizes?
You can monitor with no latency from the FM3 outputs and pretty much any daw can compensate the latency of your recorded tracks.

Unless you're recording some keyboard vsti plugins or need to monitor thru the daw for whatever reason, there's really no point. I have it always set at 2048 samples and I reduce it only in those mentioned rare cases
Click to expand...
Do you record much? The FM3 has 'near zero' latency of your guitar signal, but with the USB there is ALWAYS going to be latency. I found this out today while recording Livin' on a Prayer, and was having a heck of a time with the recording not syncing, only to realize that it is much different than the Axe III I used to have, or the Presonus 1824c.
 
mtripp said:
Do you record much? The FM3 has 'near zero' latency of your guitar signal, but with the USB there is ALWAYS going to be latency. I found this out today while recording Livin' on a Prayer, and was having a heck of a time with the recording not syncing, only to realize that it is much different than the Axe III I used to have, or the Presonus 1824c.
Click to expand...
DLC86 is correct. Assuming you're monitoring direct (and you should be), any latency is virtually irrelevant when recording your FM3. Note that you will need to manually adjust the recording delay compensation in your DAW to ensure your tracks sync properly after recording an FM3.
 
GlennO said:
DLC86 is correct. Assuming you're monitoring direct (and you should be), any latency is virtually irrelevant when recording your FM3. Note that you will need to manually adjust the recording delay compensation in your DAW to ensure your tracks sync properly after recording an FM3.
Click to expand...
This, just do that and you can work even with a 375859594938372 samples buffer.

PS: and I should add that this would allow more precise recording than any ultra-low latency cuz you can compensate FM3's internal latency too

PPS: and it should be mandatory when reamping imho
 
Last edited:
I run my signal through Sonarworks Reference 4 in my DAW. I prefer it that way and going 32 samples over spdif has a much better feel than USB with 128 samples. Sure, I could get an external dsp, but I am happy with my setup.
 
GlennO said:
DLC86 is correct. Assuming you're monitoring direct (and you should be), any latency is virtually irrelevant when recording your FM3. Note that you will need to manually adjust the recording delay compensation in your DAW to ensure your tracks sync properly after recording an FM3.
Click to expand...
Why should we need to, when any other interface I have never had the need to do so? Even the Line 6 Helix didn't have concerns such as this...
 
I'll add something (and I'm not trying to say anything negative about FAS)...If a product is advertised as an audio interface, the expectation is that it should perform very similarly to any other audio interface device. There are just too many instances of USB issues with this particular device...
 
mtripp said:
Why should we need to, when any other interface I have never had the need to do so? Even the Line 6 Helix didn't have concerns such as this...
Click to expand...
The fact you haven't felt the need doesn't mean it's not advisable to do that.
Try this (with helix, fm3 or whatever you like):
Load a wav impulse response as a track in your daw, if you zoom in a lot you'll see it as short burst, then reamp that track thru your modeler (any buffer size, even with a blank preset). Do the newly recorded track perfectly line up with the first one? I bet it doesn't
 
GlennO said:
That was a very different topic. That concerned the internal processing latency. This thread is about cpu usage that constrains the buffer size.
Click to expand...
Not to start an argument but the title of this thread contradicts what you are stating here. It seems to me that both threads are discussing the same thing, one just much more in-depth.
 
DLC86 said:
The fact you haven't felt the need doesn't mean it's not advisable to do that.
Try this (with helix, fm3 or whatever you like):
Load a wav impulse response as a track in your daw, if you zoom in a lot you'll see it as short burst, then reamp that track thru your modeler (any buffer size, even with a blank preset). Do the newly recorded track perfectly line up with the first one? I bet it doesn't
Click to expand...
Of course they don't. That's not my point. My point is the FM3 behaves differently. I can max out a LIne 6 Helix cpu, and not have as big a fight with timing as I do with the FM3. I thought I was going insane, that I had lost my sense of timing while recording the talkbox for "Livin' on a Prayer" yesterday. In the end, I had to shift the ENTIRE audio clip over. Unfortunately I didn't even bother paying attention to how much. Listen to the 2nd verse and solo, you can hear it is off. Other than a few notes I flubbed, I'm usually dead on :)

I even mentioned it here:
https://forum.fractalaudio.com/threads/livin-on-a-prayer-talkbox-effect.169502/#post-2036526
 
Last edited:
mtripp said:
Not to start an argument but the title of this thread contradicts what you are stating here. It seems to me that both threads are discussing the same thing, one just much more in-depth.
Click to expand...
Not at all. Processing latency, the topic of the other thread, is unrelated to the cpu load that constrains the buffer size, which is the topic here. Two very different subjects. You are starting an argument, one you will lose ;).
 
mtripp said:
Of course they don't. That's not my point. My point is the FM3 behaves differently. I can max out a LIne 6 Helix cpu, and not have as big a fight with timing as I do with the FM3. I thought I was going insane, that I had lost my sense of timing while recording the talkbox for "Livin' on a Prayer" yesterday. In the end, I had to shift the ENTIRE audio clip over. Unfortunately I didn't even bother paying attention to how much. Listen to the 2nd verse and solo, you can hear it is off. Other than a few notes I flubbed, I'm usually dead on :)
Click to expand...
I even mentioned it in my
GlennO said:
Not at all. Processing latency, the topic of the other thread, is unrelated to the cpu load that constrains the buffer size, which is the topic here. Two very different subjects. You are starting an argument, one you will lose ;).
Click to expand...
Ok buddy, be well.
 
You must log in or register to reply here.
Back
Top Bottom