Reply to thread

Message: [QUOTE="Bombastus, post: 91932, member: 2178"] Bertom EQ Curve Analyzer can measure hardware frequency and phase response when its test signal leaves the DAW and returns through an audio interface. The hardware path adds converters, cables, driver buffering, and the processor itself, so setup matters more than it does with a plugin. Use two DAW tracks rather than trying to copy the normal single-track plugin sandwich. One track sends the generated signal to a physical output, while the second receives the hardware return and hosts the analyzer. Keep the routing boring at first. Complexity is where bad measurements breed. [HEADING=2]Build the send and return path deliberately[/HEADING] Put EQ Curve Analyzer on the send track and set that instance as the generator. Route the track to a spare line output on your interface, connect that output to the hardware input, then bring the hardware output back into a spare line input. The return track gets the second analyzer instance. Turn on input monitoring for the return track so the DAW actually processes the incoming signal. Some hosts suspend plugin processing when they think a track has nothing useful to do, which can leave the analyzer sitting there with no valid return. A live monitored input keeps the path active. Check the physical level before chasing the graph. The generator should not clip the interface output, the hardware input, the hardware output, or the interface return. A clean measurement at a sensible level tells you about frequency and phase behavior. A clipped measurement tells you what the overload path does, which is a different experiment. Start with the hardware bypassed if its bypass keeps the signal path active. Better still, run a direct cable from the chosen interface output to the chosen input before inserting the processor. A direct loopback shows what the interface and routing contribute on their own. It also confirms whether [B][URL='https://goldmidi.com/community/threads/bertom-audio-updates-eq-curve-analyzer-to-2-1-0.8488/']the EQ Curve Analyzer 2.1.0 test signal[/URL][/B] is making the round trip cleanly. [HEADING=2]Latency has to match the physical round trip[/HEADING] Hardware cannot return the test signal at sample zero. Audio leaves through the D/A converter, crosses the external device, comes back through the A/D converter, and passes through the host buffers before the analyzer sees it. Bertom therefore needs latency alignment even when the hardware itself is analog. Do not assume the buffer size printed in your DAW equals the delay you need. Actual [B][URL='https://pubs.asha.org/doi/10.1044/2020_JSLHR-19-00419']measured round-trip latency[/URL][/B] includes conversion and hardware delay that a buffer setting alone does not describe. Interface drivers may also report compensation differently from the physical loop you have built. Adjust Bertom’s latency control while watching the phase display. A badly aligned return can produce a strong repeating phase slope that belongs to delay rather than the processor under test. Move the setting until the timing pattern settles and the remaining phase trace follows the hardware instead of the round trip. Run the direct loopback again after changing sample rate, buffer size, interface mode, or routing. Any of those changes can alter timing. Reusing yesterday’s latency value because the cables look identical is an easy way to produce a graph that appears precise while being wrong. [HEADING=2]Separate the hardware response from the interface[/HEADING] The returned curve contains everything between the digital generator and the digital analyzer. Your interface output stage, cable path, hardware processor, return cable, and interface input stage all sit inside that loop. Bertom does not magically know which part of the chain caused a small roll-off or phase shift. A direct loopback gives you a reference for the converter path. If the loopback is essentially flat but the hardware measurement bends at 40 Hz or 16 kHz, the processor becomes the obvious suspect. If both measurements show the same edge behavior, investigate the interface path before blaming the box on the desk. Keep levels identical when comparing bypassed and active states. Analog EQs, compressors, saturators, transformers, and other nonlinear circuits can change behavior with drive level, so moving the send level between captures can change more than loudness. Pick a repeatable operating level and change one control at a time. Stereo hardware deserves two clean passes if the left and right channels matter independently. Channel tolerances, switched components, or calibration can produce small differences that disappear when you stare only at one side. Measure each path with the same routing and level, then judge any difference against the direct-loopback baseline. Noise can also roughen the trace, especially when the return gain is unnecessarily high, or the hardware has a noisy output stage. Fix gain staging before increasing analysis detail. More resolution cannot rescue a measurement whose return path is already contaminated by clipping, unstable monitoring, or a poor signal-to-noise ratio. [/QUOTE]

Name