Hi Ralf, On 04.12.16 16:28, Olli wrote: I ran some more tests with jack_iodelay: # Scarlett 18i20 Both connected input and output channels are on the same device. ## 48 kHz, 256 frames, 2 periods total roundtrip latency: 1299,813 frames (27,079ms) extra loopback latency: 531 frames ## 48 kHz, 256 frames, 3 periods total roundtrip latency: 1796,813 frames (37,434ms) extra loopback latency: 772 frames ## 48 kHz, 128 frames, 2 periods total roundtrip latency: 705,813 frames (14,704ms) extra loopback latency: 321 frames ## 48 kHz, 128 frames, 3 periods total roundtrip latency: 834,813 frames (17,392ms) extra loopback latency: 322 frames # Scarlett 18i20 + OctoPre Connected input and output channels are on different devices. ## 48 kHz, 256 frames, 2 periods total roundtrip latency: 1349,806 frames (28,121ms) extra loopback latency: 581 frames ## 48 kHz, 256 frames, 3 periods total roundtrip latency: 1801,806 frames (37,538ms) extra loopback latency: 777 frames ## 48 kHz, 128 frames, 2 periods total roundtrip latency: 713,806 frames (14,871ms) extra loopback latency: 329 frames ## 48 kHz, 128 frames, 3 periods total roundtrip latency: 990,806 frames (26,642ms) extra loopback latency: 478 frames I did not test any higher framerates (e. g. 96 kHz) since my Laptop does not run very stable at them. Regards, Olli

I own a first generation Focusrite Scarlett 18i20. I can't really understand the need of very low latency when recording, because the Scarlett has got an internal mixer, which can monitor the input signal nearly at 0 latency. And I think that the Presonus doesn't have an internal mixer, or at least it's not controllable from a linux machine (the first generation Scarlett mixer is controllable since kernel 3.19 thanks to Robin Gareus' reverse ingeneering work of the protocol). Of course if you use virtual instruments latency is important, but I always used the Scarlett as a simple recorder, so I cannot tell if it works, I usually use a PCIe RME card when using virtual instruments and it works well even with vanilla kernels and threadirqs, I get some xruns when I use some heavy cpu virtual instrument sometimes. Ciao Guido

2016-12-12 13:25 GMT+01:00 Johannes Kroll <j-kroll(a)gmx.de>de>: Monitoring: I meant listening to what you are currently recording along with pre recorded material, and not level monitoring. Hardware monitoring has been removed from the linux driver, I don't know why. Ciao Guido

I use a RME Raydat interface. I managed to get no xruns while recording, and some xruns when using virual instruments like setBfree (which I like very much). I only use vanilla kernels with threadirq and I assign high priority to the card interrupt. The machine is sandy bridge with core i7 processor and 8GB ram. Johannes Kroll wrote: About the Focusrite, Johannes Kroll is right the meter was removed from the kernel driver (I don't know why). I was speaking about the internal mixer, which can mix hardware inputs and computer outputs to the hardware outputs of the card. I use Ardour internal meter to check the input signals, I think the lack of hardware metering is not a problem IMHO. Ciao Guido