I am trying to determine the roundtrip latency of a dusty old Toneport UX8 audio interface I have laying around, and I am getting some unexpected results.

I am testing using CEntrance ASIO Latency Test Utility, output 1 cable plugged into input 1, the channel strip led blinks to confirm that a short audio impulse is indeed sent output to input to measure the actual roundtrip latency.

I am getting 2.75 ms for 128 sample buffer and 5.42 ms for 256 sample buffer, which is unrealistically low, it is about 0.1 ms (converters latency perhaps?) above the minimum theoretical input latency. Seems like the output latency is missing from the timing, plus there should be also additional latency from the USB bus - at least like 2 ms. The results are consistent across tests.

Those results are in sharp contrast to the ones reported by Cubase, which reports 4.8 input / 9.6 output and 7.5 i / 15 o respectively - the input latency alone is higher than the one measures by the latency test utility.

Any idea what is going on here?

enter image description here

  • 1
    My guess is that the Centrance utility is only reporting the latency introduced by the driver itself, as opposed to the whole round-trip latency introduced by all of the components (like the DAC/ADC). Commented May 13, 2016 at 16:51
  • Ya but like he said the reported latency is even smaller than the theoretical minimum latency (256 / 48000 * 1000) = 5.804 ms. So a buffer of 256 should be minimum 5.8ms delay.
    – Scorb
    Commented May 13, 2016 at 22:12
  • @ToddWilcox - the measured latency is definitely incorrect, however I still don't understand what is going on. This application works by timing the generation of an audio impulse, sending it to the output, and listening for it on the input, and measuring the time difference to determine the actual roundtrip latency. It is not "reporting" it is actually "measuring", or at least it is supposed to, seeing how the measured values are exactly the theoretical input latency for a given buffer size.
    – dtech
    Commented May 13, 2016 at 22:49
  • @ScottF - how does that formula work? What I am calculating is 1 second / (48000 s/sec / 128 s/buff), 48k divided by the buffer size gives me buffers per second, and a second divided by buffers per second gives me a buffer's latency. If you make the calculations you get 2.66666... for 128 samples and 5.33333... for 256 samples.
    – dtech
    Commented May 13, 2016 at 22:56
  • n_samples / samples_in_one_second gives you how many seconds n_samples takes to play then x1000 gives you how many milleseconds.
    – Scorb
    Commented May 13, 2016 at 23:34

1 Answer 1


the centrance utility is showing a wrong result. the problem is that in the loopback test you are including the DRY signal (analog in to analog out). you have to mute that. only the signal that comes to input, goes to host (the utility in this case) though ASIO and comes out is what matters. that is the round-trip latency

  • There is no such thing. Nothing goes through unless it is explicitly routed. There is no host involved in the test. It is not how the test works, it generates a pulse, which is sent to the output and intercepted from an input, measuring the time it took from generating through outputting to inputting it.
    – dtech
    Commented Nov 6, 2016 at 0:17

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.