How to Design Broadcast Studios for Recording

1. Project overview: what, where, who, and why

In late 2024, Sonus Gear Flow was brought in to design and deliver a pair of broadcast studios for Harbor City Public Media (HCPM), a mid-market NPR affiliate expanding into video-first digital programming. The facility was a renovated 7,800 sq ft former retail space in Portland, Maine, with two adjacent on-air rooms, one shared control room, two voiceover booths, and a small “flex” studio meant to pull double duty as a podcast room and remote interview hub.

The client’s goal was specific: build broadcast studios that also behave like recording studios. They wanted true multitrack capture (for post and repackaging), reliable live-to-air operation, and repeatable sound for daily shows with rotating talent. The build needed to be complete before the spring pledge-drive cycle. The delivery window was 14 weeks from lease handoff to first live show.

Stakeholders included the station’s chief engineer, the program director, and an external general contractor. Sonus Gear Flow handled systems design, acoustic design coordination, equipment specification, integration, commissioning, and staff training.

2. Challenges and requirements at the outset

The space looked straightforward until we opened the ceiling: exposed structure, shared HVAC trunks, and a party wall adjacent to a bakery with 4:30 a.m. mixer cycles. HCPM also wanted a glass-forward aesthetic—lots of sightlines—without accepting the typical acoustic compromises that come with it.

Initial requirements were captured as measurable targets:

Noise floor: NC-20 (target) in on-air rooms; NC-25 acceptable in booths.
Isolation: minimum STC 55 between on-air rooms and control; STC 60 preferred between Studio A and the exterior bakery wall.
Recording deliverables: ISO tracks for every mic, plus a clean mix-minus for remote guests, with metadata and consistent file naming.
Operational reliability: failover playout, redundant network paths, and “single-button” routing presets for typical show formats.
Latency management: remote interview return and talent monitoring with minimal distraction; no audible combing when talent hears themselves via open-back headphones.
Timeline constraint: permitting plus construction plus integration in 14 weeks; no “phase 2” budget planned.

The main tension was classic: isolation and HVAC noise control push construction cost and time, while broadcast workflows push speed and repeatability. Add video lighting heat loads and you have a recipe for either a noisy studio or an expensive rebuild later.

3. Approach and methodology chosen

We ran the project using a two-track methodology: (1) architectural/acoustic scope lock by end of week 3, and (2) technical design freeze by end of week 5. That meant we could order long-lead items early (glass, doors, consoles, AoIP switches) while still leaving room to refine signal flow based on staffing and show formats.

For audio transport and routing, we standardized on AoIP (Dante + AES67 where required) with a core switch pair and room-level edge switches. For control surfaces and I/O, we chose a broadcast console ecosystem with strong GPIO, snapshot recall, and clean integration with playout and remote codecs. For recording, we designed a dedicated multitrack capture path independent of the on-air chain so that a missed record button could not jeopardize air.

Acoustic design was coordinated with the GC using a “buildable” specification: resilient isolation assemblies that a commercial crew could execute consistently, rather than boutique details that only succeed with specialty labor.

4. Step-by-step execution narrative

Week 1–2: Discovery, measurement, and show mapping

We started by mapping three show types: daily talk (3–4 hosts), music + talk hybrid, and live pledge segments with phone/VoIP callers. We also measured existing ambient noise during bakery operation and afternoon HVAC load. The worst-case reading in the future Studio A footprint was 46 dBA Leq with distinct low-frequency rumble around 63–80 Hz—unacceptable without isolation and mechanical changes.

A workflow workshop produced a routing matrix: every mic must be available to air, to recording, to talkback, and to remote mix-minus, with predictable default states. This prevented “we’ll figure it out in the control room” later.

Week 3–5: Acoustic scope lock and infrastructure design

We issued acoustic wall and ceiling assemblies with explicit details:

Studio A perimeter: double-stud wall, 2x4 with 1” air gap, two layers 5/8” Type X each side, Green Glue between layers, mineral wool in both stud bays. Target STC 60.
Studio-to-control partitions: staggered stud, two layers 5/8” Type X, mineral wool. Target STC 55.
Ceilings: isolated hat channel on clips with double 5/8” gypsum; sealed perimeters with backer rod and acoustic sealant.
Doors: two sets of STC 50 rated doors for Studio A (sound lock), single STC 45 for Studio B.
Glass: two-pane asymmetrical laminated units (10.8mm + 12.8mm) with 6” air gap in Studio A to control coincidence dips.

HVAC was handled with a separate mechanical addendum: oversized lined ducting, remote fan placement where possible, and duct silencers on supply and return for Studio A. We specified air velocity under 250 fpm at diffusers and required flexible connections at equipment. The GC initially pushed back on duct silencer cost; we held firm because everything else is wasted if the room is noisy.

On the technical side, we pulled conduit and cable paths early. We standardized on shielded Cat6A for AoIP and control, and star-grounded technical power. Dedicated circuits were allocated: two 20A isolated-ground circuits per studio, one for technical racks, one for “floor” power (monitors, chargers) to avoid polluting the audio ground.

Week 6–10: Construction, rough-in, and rack build

As walls went up, we performed mid-construction inspections—specifically looking for the usual isolation killers: back-to-back electrical boxes, unsealed penetrations, and ductwork touching framing. We caught a critical issue in Studio B: a conduit sleeve was hard-coupled between studs and the ceiling grid, effectively bypassing the isolation clips. The fix took two hours then; it would have taken days after drywall.

Parallel to construction, we built the core racks offsite. The rack architecture was split into:

AoIP core: redundant managed switches (primary + secondary), Dante domain segmentation, and an AES67 bridge for a legacy codec path.
Console engine and I/O: dedicated DSP core, stage boxes in each studio, and GPIO interface for on-air lights, cough mutes, and tally.
Recording server: a 1U workstation with mirrored SSDs for active sessions and nightly sync to NAS.
Broadcast playout: two playout nodes with shared storage and a hardware audio fallback.

Week 11–13: Integration, tuning, and commissioning

Once rooms were sealed and HVAC balanced, we did noise checks. Studio A landed at NC-19 during worst-case HVAC and bakery operation—close enough that mic self-noise became the limiting factor. Studio B measured NC-21.

We then tuned the rooms for speech. Rather than chasing a “dead booth” sound, we targeted controlled decay with minimal flutter and strong low-mid management:

Broadband absorption panels (4” mineral wool) at first reflection points and rear walls.
Ceiling clouds above talent positions to reduce overhead reflections to lavs/headworn mics during video shoots.
Corner bass trapping in Studio A to reduce 80–120 Hz buildup from desk surfaces and close mic technique.

Finally, we commissioned the system with day-in-the-life tests: a full show with a remote guest, playout, live reads, and multitrack recording, while forcing a network switch reboot to verify redundancy behavior.

5. Technical decisions and trade-offs made

Console and AoIP ecosystem

HCPM selected a Lawo crystal surface with an AoIP engine, paired with Dante/AES67-compatible stageboxes. The key trade-off was cost versus operational speed. A less expensive analog console would have required more outboard routing, more patching, and less recall—fine for a single-studio facility, but risky with rotating staff and daily format changes.

We used two managed switches (primary/secondary networks) and kept non-audio traffic off the AoIP VLAN. A single converged network would have reduced hardware, but it increases troubleshooting time and makes broadcast reliability dependent on office IT changes.

Microphone strategy: consistent tone and rejection

For host mics, we deployed Shure SM7B on low-profile boom arms in Studio A and Electro-Voice RE20 in Studio B where talent tended to move more. The SM7B choice was deliberate for recording: smoother top end for close speech and less harshness when edited aggressively for podcast distribution. The RE20’s variable-D helped when hosts drifted off-axis during live segments.

We paired both with inline gain stages (Cloudlifter-type) where needed, but we kept the actual mic preamps in the console I/O to avoid adding troubleshooting points. The trade-off was slightly higher budget for quality networked I/O with adequate clean gain.

Monitoring and latency

Talent monitoring was designed around “no surprises.” We provided closed-back headphones (Sony MDR-7506) plus small nearfields (Genelec 8030C) for producers. In Studio A, we added a dedicated low-latency cue bus for talent, with sidetone control. Remote guest return was handled through the console DSP so we could keep round-trip latency perceptually stable even when switching between VoIP and dedicated codecs.

Recording architecture: independent and automatic

The facility’s most consequential decision was separating air chain from record chain. Every mic fed:

a processed on-air path (gentle dynamics + safety limiter), and
a pre-processing ISO path for recording.

Multitrack recording ran on Reaper with a locked template and auto-record scripts triggered by GPIO from the on-air “start” state. Engineers can still manually override, but the default behavior is that the studio records whenever it is live. The trade-off was spending time on scripting and testing, but it eliminated human error in daily operation.

6. Results and outcomes with specific details

The studios went live at the end of week 14, with a soft-launch the prior Friday. In the first 60 days:

Average setup time for the daily show dropped from ~25 minutes in the old facility to 8–10 minutes (mic check, remote line check, routing preset).
Recording reliability: zero missed multitrack captures across 112 recorded segments; two instances of incorrect guest routing were fixed by updating a preset.
Noise performance: Studio A remained under NC-20 even during bakery morning cycles; measured HVAC rumble was reduced by ~18 dB at 63 Hz compared to pre-build readings.
Post-production time decreased by roughly 20% for podcast exports because ISO tracks were cleaner and less EQ was required to fight room tone and HVAC.
On-air consistency improved: host-to-host tonal variation was reduced because mic placement and cue mixes were standardized and the room response was controlled.

Operationally, the snapshot system became the backbone: “Daily Talk,” “Interview + Remote,” “Roundtable,” and “Pledge” presets set routing, mix-minus, headphone levels, and bus processing. Producers stopped “building” shows from scratch and started selecting verified states.

7. Lessons learned and what could be done differently

Two lessons were reinforced the hard way.

First, HVAC coordination must start before walls are framed. We had to revise one return path because the initial duct route conflicted with lighting truss supports for the video grid. The fix was minor but consumed three days and added cost. Next time, we would require a single coordination meeting with mechanical, electrical, lighting, and acoustic scopes on the same reflected ceiling plan before ordering duct silencers and lighting fixtures.

Second, glass placement needs acoustic modeling, not aesthetics-first decisions. The client originally wanted floor-to-ceiling glass on one wall of Studio B. We demonstrated, with simple ray tracing and measurements from comparable builds, that the resulting early reflections would make lav/headworn mic work brittle and force heavier gating. We compromised on a smaller window with angled placement and added absorption opposite the glass. The room sounds better for speech and looks intentional on camera.

If we could change one thing, we’d allocate more time for operator training with failure scenarios: what happens when a remote codec drops, how to swap to VoIP, and how to keep recording uninterrupted. We did training, but the highest value came from running “things go wrong” drills, and we should have scheduled more of them.

8. Takeaways applicable to other projects

Design for recording from day one: insist on pre-processing ISO paths, consistent file naming, and automatic capture. Broadcast workflows are fast; recording workflows must be resilient to human error.
Spend money on silence before spending on gear: NC-20 performance does more for intelligibility than any plugin chain. Duct silencers and proper isolation assemblies are often the highest ROI line items.
Snapshots beat hero engineers: presets, defaults, and routing states reduce operational variance and onboarding time. Document them and lock them down.
Separate networks or at least segment aggressively: AoIP stability depends on predictable traffic. VLANs, QoS, and redundant paths are not optional in live environments.
Make the room work for speech first: control early reflections and low-mid buildup, especially with desks and video lighting. A “pretty” reflective room is a long-term tax on post-production.
Commission with real shows: bench tests won’t reveal the routing edge cases that appear when talent, remote guests, and producers all work at speed.

The HCPM build succeeded because the team treated the studios as recording rooms with broadcast responsibilities, not the other way around. The payoff was measurable: quieter rooms, faster daily operation, and multitrack content that holds up in post and republishing—exactly what modern stations need when every live segment is also tomorrow’s podcast.