It’s very simple to measure a device’s frequency response using the HP 35670A Dynamic Signal Analyzer. The analyzer has a built in signal generator. This is used to pass a periodic chirp signal through the device under test and then the original signal on Ch1 is compared with the signal output from the device on Ch2 using an FFT. You can do swept sine measurements but it takes a long time to do this with any degree of thoroughness.
Let’s have a look at the PodXT’s measurement again:
The upper graph shows the phase difference between the source signal and the PodXT’s output. At first I thought something strange was happening in the PodXT. Then I thought about it a bit more and ran it past my buddy Martin who works with DSP hardware. I’ll let him explain…
“After having a quick look, I think that’s what I would expect to see. Assuming that the filter inside the box has a fixed latency, then this will produce a linear phase response. Imagine delaying a 10kHz signal by 100µs – this would effectively have a phase shift of 0° because the signal has been delayed by a complete cycle. If a 5kHz signal were delayed by the same 100µs, then this would actually appear as a 180° phase shift because we’ve only delayed by half a cycle. This cycle continues on as frequency increases, producing a repetitive linear sweep in phase. Can you try plotting the phase response on a linear frequency axis? I reckon this will probably make it look like a regular linear pattern.”
He’s not wrong. Here’s closer detail of the phase delay on a linear frequency axis:
OK, so that explains that. We have a certain amount of latency in the PodXT (and all the other DSP multi-fx units tested) – all the frequency content in the test signal is being delayed by the same amount. What’s quite cool is, knowing this, you can calculate the device’s latency. I won’t bore you with my calculation (or spoil your fun if you want to do it yourself) but I make the PodXT’s latency to be about 2.4 ms. Anyone that’s messed around with soundcards and home recording will know that latency of less than 10 ms is basically the target for the delay to be unnoticeable. At 2.4 ms, the PodXT is doing pretty well.