Latency calculations in Patchwork and Axiom standalone apps

[Tappistry]

Quick question regarding the audio settings menus in Patchwork and Axiom: how is the estimated latency calculated? For instance, running my interface at a 256 sample buffer and 44.1 khz sample rate, the estimated latency in Axiom is 24.8 ms. If Axiom's calculation is correct, I would assume I would be able to audibly notice a 24.8 ms delay of hitting notes of my Stick or guitar, but I don't. Do Axiom and Patchwork double buffer the audio stream by default?

As a point of comparison, the same settings in Cantabile x64 yield an estimated latency of 5.80 ms.

My guess is that somewhere between the calculations of these hosts lies the true latency figure for my system, and is heavily dependent on my interface driver interactions with USB and chipset drivers and the happenstance color of the waxing moon of any given evening.

[Blue Cat Audio]

We are using a third party library for audio I/O, and in some cases it is indeed not very precise to estimate latency (I think it mainly depends on the HW drivers...).

Round trip latency is definitely higher than 5.8 ms though (5.8 ms is just the size of the buffer). By the way, are you running Windows or Mac?

[Tappistry]

Windows 10
Focusrite 2i2 interface

And for what it's worth, it seems that I'm able to set my buffer settings lower in Cantabile and Reaper without audio glitches vs. using the standalone apps, especially with two channels of audio simultaneously being fed into my interface.

[Blue Cat Audio]

It appears the latency is reported by the driver of the audio interface, so I guess there could be an issue with the Focusrite driver here.

Regarding low buffer settings, do you mean that the VST version loaded into these hosts with the same configuration produce less dropouts than the standalone version? I am not sure we can actually trust the buffer sizes & latency here since the reported latency is completely off.

[guatl666]

Hi, I'm quickly jumping into this post because it seems to be almost identical to the issue I am experiencing.
I downloaded a demo version of Axiom and PatchWork (both in their latest version as per today 2019-04-23) onto my Windows 10 64 Bit desktop PC (i7 Quad Core @ 2.6 GHz / 3.1 GhZ max. & 32 GB RAM). My audio interface is an Universal Audio Apollo MKII Twin Duo, connected via USB-C to Thunderbolt 2 adapter.
Under all kinds of applications I can set up a stable "low-latency" environment at around 10 - 15 ms roundtrip time (BIAS AMP, Mercuriall SS-11X, Neural DSP plugins, Guitar Rig 5, etc.). When starting Axiom or PatchWork and setting audio to ASIO and selecting my UAD Twin MKII driver, the settings window displays almost 100 ms latency @ 44.1 kHz and almost 90ms latency @48.0 kHz. That amount of latency makes it impossible to play through with my guitar or bass (althoug what comes out of the speakers sounds great!).
Am I doing anything wrong or is the I/O library incompatible with Thunderbolt drivers (or UA drivers?)

[Blue Cat Audio]

Have you tried reducing the buffer size? Some ASIO drivers will actually require you to set it up using the application installed with the audio interface (it may be the case for UA - I am not 100% sure).

[Tappistry]

Hey look, my old thread!

Since I posted this thread, I upgraded to a Zoom UAC-2 interface. Much better performance overall than my Focusrite.

I'm running the Zoom at 160 samples, 48 khz. Reaper shows 3.3 ms/5.8 ms estimated round trip latency at the top bar in an open project. Axiom and Patchwork show 9.0 ms at the same settings.

These values are still acceptable for jamming without any noticeable lag, but I'm still curious why there's such a large difference?

[Blue Cat Audio]

That's only an estimate, and most of the time the actual latency is smaller that what is displayed, so you should probably not worry about it unless you can hear a delay while playing. Also some DAWs actually show an estimate that is too optimist .

Typically in your case 3.3 ms corresponds to the length of a single buffer, and there are usually several buffers of latency between the input and output. so 6.6 ms is probably more accurate, and I would not be surprised if 9 ms is the actual value, as there is often some double buffering happening in the background.

You may also want to check the size of the buffer set in the application delivered with the driver of your soundcard: sometimes that's the value used by the driver and PatchWork or Axiom's setting is not taken into account by the ASIO driver (you can see it because changing the buffer size in Axiom does not change the estimated latency in this case).