Reply to thread

Message: [QUOTE="Bombastus, post: 90915, member: 2178"] A Haas delay sits in the 5 to 35 millisecond window, and every millisecond of that offset turns into comb filtering the moment your mix sums to mono. The notches are not subtle. Engineers regularly report parts that all but disappear in a mono fold-down after sounding enormous in stereo. Most advice on widening a vocal stops at technique. Double the take, pan the copies hard, lift the side channel a decibel or two, then check mono before you export. That list is fine as far as it goes, and it skips the thing that actually decides whether your width survives. The method matters more than the amount. Two widening algorithms can push a stereo meter to the same place and behave completely differently once a phone speaker folds them into one channel. [HEADING=2]Phase rotation and delay do not fail the same way[/HEADING] All-pass decorrelation shifts phase without touching magnitude. Fold it to mono, and you pay roughly one to two decibels of spectral ripple, which is a tonal nudge rather than a hole in the arrangement. Haas is different in kind. It is a short delay, and a delay summed with its own source is a comb filter by definition, so mono playback carves deep notches at regular intervals across the spectrum. Hilbert processing belongs with the first group, because it applies a broadband ninety-degree phase shift rather than a time offset. That is why the three width algorithms in [B][URL='https://goldmidi.com/community/threads/black-rooster-audio-introduced-the-ps-vc1-vocal-chain.76251/']Black Rooster's new vocal chain[/URL][/B] are not three flavors of one idea, even though a single mix control drives all of them. Pitch-shift widening sits lower still. Detuning a copy by a few cents produces comb filtering that moves around over time, so the damage never settles into a fixed tonal shift you could compensate for later. You cannot engineer your way out of the Haas trade, either. Push the delay past about 35 milliseconds to dodge the cancellation, and the ear stops fusing the two arrivals into one sound. What you get instead is a slap echo, which is a different effect doing a different job. [HEADING=2]Correlation meter numbers beat guessing at width[/HEADING] Put a correlation meter on the vocal bus and read it during the loudest phrase, not the intro. Anything from plus 0.3 upward leaves enough shared content between the channels for mono playback to hold together. Between zero and 0.3, you are in caution territory, with noticeable energy loss when the two sides sum. Below zero, the channels are actively cancelling, and the vocal reads as broken in mono. The same caution applies to a mid-side move. A side boost of one to three decibels is safe on most material, but once the side channel carries more energy than the mid, you have built a vocal that only fully exists in stereo. That number is worth more than a width percentage, because it responds to the material rather than to the setting. A breathy verse and a belted chorus can land in different zones through the same preset. Then listen. Thirty seconds of mono playback before export will tell you whether you have [B]width that survives a fold-down[/B] or a vocal that quietly loses its front teeth, and it catches problems no meter flags as such. [HEADING=2]Low end is where the notches actually hurt[/HEADING] Collapse the low frequencies to mono before anything widens them, using a crossover somewhere between 100 and 200 hertz. Your ears barely localize bass, so you give up almost nothing you were hearing in the first place. This matters more on vocals than most guides admit. A male lead's fundamentals sit right in that region, so comb notches from a delay-based widener land on the body of the voice rather than on air and sibilance. Large sound systems make it worse. They sum everything below the crossover point to mono, so the cancellation you dodged on studio monitors arrives anyway, in the range with the least headroom to spare. Signal order decides how much of this you can steer. When the width stage sits at the very end of a fixed chain, as it does in the six-module Professional Series layout, whatever low end got past the filter section is exactly what gets widened, so your high-pass decision is quietly doing two jobs at once. The presets complicate that further. Tight, Close, Far, and Outer are calibrated positions rather than raw parameters, so you are choosing a spatial character, not dialing a delay time you can inspect and correct later. [/QUOTE]

Name