Reply to thread

Message: [QUOTE="Shamiso, post: 92587, member: 160"] DeltaScope’s Difference spectrum reacts to timing as well as level, while its Magnitude view is designed to ignore phase-only changes. Put the same processor between two linked instances and those graphs can disagree without either one being wrong. The distinction matters more after [B][URL='https://goldmidi.com/community/threads/apollo-view-expands-deltascope-to-windows-and-aax.77962/']Apollo View’s DeltaScope 1.2 update[/URL][/B] made the waveform, spectrum, measurement readout, and Hear Delta output use the same difference signal. Magnitude also changed in 1.2, so reverbs and delays are tracked more usefully for tonal balance, which makes it tempting to treat the two spectrum modes as interchangeable. They are not. A lot of analyzer confusion starts when a residual gets read as a frequency response. Difference tells you what remains after one signal is subtracted from the other. Magnitude is trying to answer a narrower question about level change across frequency. [HEADING=2]Difference spectrum is a residual, not an EQ curve[/HEADING] Take a clean track, duplicate it, and delay one copy by a tiny amount without touching gain. The delayed copy can keep essentially the same magnitude response, yet subtraction leaves plenty behind because corresponding samples no longer arrive at the same instant. Difference sees the mismatch immediately. The weird part is what that residual can look like. A fixed delay creates a phase offset that grows with frequency, so subtraction can produce repeating rises and dips across the spectrum. Someone glancing at the graph could mistake those shapes for EQ moves even though no EQ happened. All-pass filters make the point even cleaner. Their job is to alter phase while keeping magnitude flat, and [B][URL='https://doi.org/10.1109/5.3286']digital all-pass filter behavior[/URL][/B] has been formalized for decades. Run an all-pass between the two DeltaScope inputs and Difference has something real to show, while Magnitude can remain flat. This is why a large Difference trace does not automatically mean a processor changed the tonal balance. It can come from delay, phase rotation, polarity behavior, modulation position, or ordinary gain changes. The residual proves the two signals differ at that moment, not why they differ. [HEADING=2]Magnitude transfer removes phase from the comparison[/HEADING] Magnitude is more useful when you want to see how much a processor raises or lowers energy by frequency. Put DeltaScope before and after an EQ, make a broad boost, and the curve should follow the tonal move rather than reacting to phase rotation as if it were extra gain. A pure delay is the easy sanity check. Move one signal later without changing its level, and Difference should react strongly, while Magnitude should stay flat because the spectral level relationship has not changed. An all-pass filter gives the same basic lesson with frequency-dependent phase instead of one broadband offset. This distinction matters with minimum-phase EQs too. A bell or filter can change magnitude and phase together, so Difference contains both effects mixed into one residual. Magnitude lets you focus on the gain part instead of trying to reverse-engineer an EQ curve from the leftover audio. Level mismatches are less mysterious. Turn the processed path down by 1 dB with no other change and Magnitude can show a roughly flat 1 dB offset, while Difference reports a broadband residual because every sample is now scaled differently. Same two signals, two useful answers. [HEADING=2]Moving processors refuse to hold one permanent curve[/HEADING] Transfer-style measurements get less tidy once the processor changes with time or signal level. A compressor responds to the program hitting its detector, while saturation creates harmonics that did not exist in the input. Chorus, flanging, random modulation, and some reverbs keep moving even if the knobs never move. For those processors, a Magnitude trace is still useful, but it is a snapshot of behavior under the material and conditions you fed it. Calling it the processor’s permanent frequency response would be too strong. A different level, chord, transient, modulation phase, or tail position can produce a different picture. DeltaScope 1.2 specifically improved how Magnitude follows the tonal balance of reverbs and delays, which is useful because those effects spread energy through time rather than behaving like static EQ. You still need to watch the graph across a representative passage instead of freezing one convenient frame and declaring the job solved. The difference can be even more animated on the same effects. A reverb tail contains delayed, filtered, and often decorrelated copies of earlier sound, so subtraction can light up long after the dry transient has passed. Magnitude tells you how the spectral balance is being reshaped, while Difference keeps exposing the time-domain mismatch that created the tail in the first place. [/QUOTE]

Name