It was a great interview. I summarized it with AI so that people can easily jump to the parts they want to watch.
00:00–01:49 Intro / roles
- Interview with Josh (audio systems engineer at Dirac). He works on signal chains, analysis, tuning and calibration across products (automotive, consumer devices, etc.). Cross-domain learnings carry over.
02:29–03:49 Measurement points: why more points can improve ART
- Even if you keep the measurement “cloud” tight around one seat, adding more points can improve ART.
- ART wants to understand wavefront propagation across space; more measurements provide more spatial info (frequency/pressure variation), improving multi-speaker integration.
03:49–05:09 Does microphone order matter?
- Only the FIRST measurement position is critical (used for coarse time alignment).
- After the first point, the order of remaining positions doesn’t matter; the on-screen sequence is guidance, not a strict requirement.
05:27–05:52 UMIK-1 vs UMIK-2
- From Dirac’s perspective, there’s no meaningful advantage for UMIK-2 for ART calibration.
- Higher sampling rates are not required by the algorithm; calibration files largely account for device differences.
06:14–07:18 Pre-adjustments vs post-adjustments
- Pre-adjustments (sub trim/PEQ/placement fixes) are welcome, especially if there are large issues and the system’s correction gain window is limited (often ~10 dB).
- Post-calibration changes are discouraged because they can break filter integrity more than users expect.
07:18–11:16 Denon/Marantz “Directional Bass” vs ART, and LFE direction
- With ART, subwoofer “support” is calculated to match the main channel target. If subs support Front Left, the perceived direction should stay “Front Left” down into bass.
- LFE is mono by source; ART uses the center channel as a timing reference for LFE, which can make LFE feel like it comes from straight ahead.
- If someone truly hears/feels only one sub playing, something is wrong (possibly a bug/config). All subs should contribute; it can help to listen close to each sub to confirm it is active.
11:42–13:56 Splitting subs into groups and “missing target curve” confusion
- Key distinction: you are optimizing the playback channel target, not “EQing the subwoofer as an instrument.”
- A supporting sub group may not show its own target curve because it is trying to match the target/impulse of the group/channel it supports (e.g., the LFE group). Filters still exist per sub even if the UI doesn’t show a separate target curve.
14:36–16:31 Different hardware (e.g., StormAudio) and filter resources
- Dirac aims for similar core implementation across brands; higher-end processors may allow more cross terms/resources, especially for larger systems.
- Exact filter counts vary; the interview references that some devices show ~98 filters while bigger processors can be far higher.
18:32–19:54 Small speakers can still benefit
- Even bookshelf/small speakers can contribute (often ~70–150 Hz region) and be useful “support” resources.
- Josh mentions using ART on small ~4-inch desk speakers and perceiving a “bigger speaker” effect (e.g., sounding closer to 6-inch), in the sense of capability/weight.
20:18–21:30 “Less space” or “less bass” perception after ART
- Controlling room resonance removes lingering low-frequency energy; this can be perceived as less “space,” but it is mostly removal of unwanted energy.
- With better control, users may be able to raise the target curve a bit for cleaner punch without bloated resonance.
22:00–25:08 Infrasonic toggle / <20 Hz behavior
- ART currently applies a cutoff around 20 Hz. To get <20 Hz back, enable infrasonic pass-through, which is unfiltered.
- Unfiltered pass-through can create a hump depending on phase relationships; visually it may appear, but audibly it is often not a major concern.
- Infrasonic content is typically on LFE, but it can also exist on L/R depending on source; enabling on other groups is optional, not forbidden.
25:35–27:41 Target curve “following dips” (cancellations/nulls)
- In traditional single-speaker correction, boosting into deep nulls wastes headroom and rarely “fixes” the null.
- With ART (multi-speaker co-optimization below ~150 Hz), you generally don’t need to shape the target to follow dips; other speakers can fill gaps due to different positions/response.
28:08–30:51 Support Level parameter: not “volume,” and smearing concerns
- “Support Level” is expressed in dB but it’s a mathematical control input, not a direct loudness/gain knob; the scale is counterintuitive (more negative can mean stronger support/“working harder”).
- If worried about localization/smearing, a better first lever is lowering “F support high” (support upper frequency), because localization is more relevant at higher bass/low-mid than at very low bass.
31:10–33:33 F support high vs F SISO
- F SISO is described as the crossover region between MIMO (ART) and SIMO (traditional Dirac correction).
- Support does not stop abruptly; it rolls off toward F SISO. Keeping F support high and F SISO closer can avoid unintended “roll-off” behavior.
- Josh hints F SISO may be automated/removed in a future update to reduce confusion.
33:41–36:13 Phantom center / no center speaker
- If no center is present, the LFE timing reference can shift (e.g., to Front Left). In many cases localization differences are small.
- For off-center seating/asymmetric rooms, placing the first measurement at the actual listening position can yield a strong phantom center.
38:32–40:18 Center dialogue clarity and correction range
- Case-by-case. Often correction helps; if dialogue clarity needs a lift, adjust target curve around ~1–4 kHz slightly for the center.
- If room acoustics are already good, backing off correction can also sound pleasant. Experiment with presets.
40:18–41:06 Presets / running more filters on Denon/Marantz
- Denon/Marantz may allow multiple “speaker presets,” each with multiple Dirac slots. It can be used to A/B experiments without overwriting the baseline.
41:39–45:28 “Weak bass” after ART: common causes and fixes
- Sometimes “less bass” is simply removal of a huge peak/resonance that was previously perceived as bass.
- If placement is perfectly symmetric and both subs share the same dips, there may be limited ability to fix that specific null.
- With ART’s improved control, increasing the target curve (bass boost) can restore desired weight more cleanly.
46:35–47:52 What’s next / improvements
- Dirac is iterating based on user feedback. UI improvements mentioned: holding Shift to adjust certain parameters across groups, and overriding certain warnings to speed workflow.
- Algorithm improvements are ongoing.
48:25–50:31 Commentary on sub grouping experiments
- Different grouping approaches change which impulse/target the subs try to match (e.g., matching LFE vs matching surround channel impulse if subs are tied into surrounds as “full-range”).
- Splitting groups can be useful to apply different parameter limits based on sub capability (e.g., different F support low), while still applying per-sub filters.
