The Short Version: Input delay compensation doesn't seem to work as you might expect after reading the FL 20 Manual. Manual input delay compensation (clock icon by input selector) is doubled for some reason when combined with automatic latency compensation (mixer menu option), but not when used without automatic latency compensation. Latency compensation by itself possibly compensates for both the input latency and the output latency as reported in Status bar in Audio Settings. Or mine is just compensating incorrectly for whatever reason and it's a coincidence that the compensation value is close to my input + output latency.
Background: I've been noticing my guitar takes are too early in songs. When entering a test project file and tapping along on the guitar to a metronome, the waveforms are consistently early. When performing a more objective feedback test to remove the element of player error (holding a headphone to the guitar pickups and recording the FL metronome), the waveforms are very slightly late. I believe that the feedback tests are late because something about the feedback tests introduce some form of real-world delay that's just enough to make the waveforms actually look late instead of early. So I've decided that I need to use tap tests to average out the "true" alignment of my recordings and get rid of that real-world feedback delay that isn't a factor when recording parts normally. That being said, I can still use feedback tests to record objective alignment (no player error variable) relative to other feedback tests. I'm working under the assumption that my playback tracking is in-sync, as recording the FL metronome internally shows perfect waveform alignment on the grid.
Main Issue: To fix my root problem of my guitar recordings being early, I've hypothesized that I need to adjust the Input Delay Compensation, as my input may be over-compensated for and that's why my guitar recordings are early. However before doing this, I wasn't sure if the "manual" input delay compensation option (the clock icon by the input selector) was an offset or a replacement of the "automatic" input latency compensation option (the Disk Recording > Latency Compensation option in the mixer menu). The FL 20 Manual doesn't explicitly say either which way - it does mention the PDC option being an offset to the automatic PDC though which is why I was wondering if that also applied to the input delay options.
Test Setup: So, I decided to run a simple test. Do 4 metronome feedback recordings with every combination of the "manual" and "automatic" input latency compensation options on/off. Without getting into formulas, you can easily calculate how many ms are in a playhead "tick" in the playlist. The ticks are what I used to measure the distance between the metronome waveforms. For the tests where I had the "manual" input delay compensation on, I set it to -5.95 ms (the rough ms value for 1 tick at 96PPQ in a 4/4 song at 105 bpm). I actually ended up doing the measurements of the tests in 960PPQ for better accuracy, but that's fine since the -5.95ms just became an arbitrary interval to measure for instead of trying to get a clean, single tick in my measurements had I done them at 96PPQ.
Test Results:
MIDC = "manual input delay compensation" (the clock icon) and AIDC = "automatic input delay compensation" (the mixer menu option)
Status Bar in Audio Settings reports Input Latency at ~7.3ms and Output+Plugin Latency at ~7.5ms
-Recorded with just AIDC on. Then recorded with both off. In the second recording, the waveforms were ~14.9ms later than the first recording. This is more than I was expecting since I thought AIDC was meant to just make up for the Input Latency as reported in the Status bar: 7.3ms. Note: 14.9 is close to being either double the Input Latency: 14.6 OR the Input Latency + Output+Plugin latency: 14.8.
-Recorded with MIDC on and AIDC off. When compared to the second recording above, both off, the waveform was 5.95ms later. Ok, well there's the MIDC value I entered. So this test works as expected.
-Recorded with both on. When compared to just AIDC being on, it's ~11.9 ms later. Soooo, that's exactly double my -5.95 MIDC value? But why double? When compared to the test of both being off, it's ~3ms ahead which seems unremarkable.
Analysis: I don't really know what to make of it all. Forgetting about the MIDC, the AIDC seems to compensate for more than what I thought when compared to having the option off. Maybe it compensates for both the input AND output latency to essentially move your recorded part to where you heard it back? Of course maybe it's simply being compensated incorrectly and it's just a coincidence that the value is close to what the input + output latency would be. When using MIDC without AIDC, it seems to work as expected - the waveforms are moved by the interval you set. But when using MIDC and AIDC, the waveforms are moved by exactly double the interval you set.
Does anyone have any deeper knowledge of "automatic" Latency Compensation and "manual" Input Delay Compensation that would explain these results. The more I test FL Studio's Playback Tracking and Latency Compensation settings, the more I get the impression that... nothing works.
At least not the way the User Manual seems to convey. Note: Measurements aren't necessarily objective. Decimals were rounded and while 960PPQ provides detailed tick measurements, it still isn't fine enough to land exactly on the waveform's start sometimes. That being said, I'm confident that most measurements are pretty close.