We recently ran into a problem that looked simple on paper: connect an Allen & Heath SQ5 to a laptop via USB and record 24 channels at 96kHz into a DAW. Straightforward enough. Except it wasn’t — and what we learned chasing that problem is worth sharing, because if you’re recording multitrack audio on a modern laptop, you’ve probably already hit this problem….  or you will.

The Problem: Dropouts That Shouldn’t Happen

The SQ5’s USB-B connection is rated for up to 32 channels of bidirectional audio. The laptop was fast, the cable was solid, the session settings looked right. And yet — clicks, dropouts, interrupted recordings. Classic symptoms with no obvious cause.

Allen & Heath’s own documentation points to the culprit: USB 2.0 audio devices are not always compatible with USB 3.0 ports. That reads like a minor footnote. It isn’t.

What’s Actually Going On Inside Your Computer

USB 2.0 audio — which is what the SQ5’s USB interface uses — relies on a transfer method called isochronous streaming. This is a time-guaranteed delivery method built into the USB spec, designed specifically for audio and video: data has to arrive on schedule, every time, no exceptions.

USB 2.0 ports managed this using a controller architecture called EHCI (Enhanced Host Controller Interface). USB 3.0 introduced a new architecture called xHCI (eXtensible Host Controller Interface), which consolidated both USB 2.0 and 3.0 under one driver. The problem: xHCI implementations vary wildly between chipset manufacturers, and many handle isochronous scheduling for USB 2.0 devices poorly — causing dropped audio packets that show up as clicks and dropouts in your recordings.

On older laptops, you might find an EHCI controller alongside xHCI and move the interface from xHCI to native EHCI and solve the problem. On modern laptops — including the ASUS we were working with — there is no EHCI controller at all. Every port on the machine, regardless of what it looks like from the outside, runs through Intel xHCI. There’s no fallback. There’s no good EHCI port.

And a USB 2.0 hub won’t help either. A hub is a repeater, not a controller — the data still routes through the same xHCI driver at the top of the chain.

What About Mac?

Mac handles this better in most cases. Apple writes its own USB stack and controls the entire hardware-software chain, so isochronous transfer behavior is far more consistent. But it’s not immune — macOS updates have broken USB audio on multiple occasions, including widespread issues on M1 machines after specific OS releases. And USB hubs cause instability on Mac just like they do on Windows.

Better, but not bulletproof.

The Real Truth: USB is not backward compatible. Plan accordingly

You can safely use any USB2 to USB2 device pair via EHCI stacks, or USB3 to USB3 connections using xHCI, but you cannot use a USB2 device connected to a USB3 port without proper testing of the xHCI stack. If you have a poorly written xHCI stack, the laptop will never be “compatible” with the USB2 audio device. 

Our SQ5 has a Dante card installed and this was an alternative. Dante is an audio-over-IP protocol that moves audio across standard Ethernet rather than USB — no isochronous transfer, no xHCI, no USB at all. For some users, that’s the best solution. Avoid the USB conflict problem from the start. 

A direct DANTE connection setup is simpler than it sounds:

• A single Cat5e cable from the SQ5’s Dante card primary port to the laptop’s Ethernet port
• No switch required — both devices auto-assign IP addresses using link-local addressing and discover each other within seconds
• Dante Controller (free from Audinate) handles channel routing
• Nuendo sees audio through Dante Virtual Soundcard (DVS) running as an ASIO driver

The USB dropout problem disappears entirely.

DVS: Powerful, But Know Its Limits

Dante Virtual Soundcard is convenient and capable, but it’s software running on a general-purpose OS — which means it inherits some of the same limitations as any laptop-based audio solution. A few things to know before you rely on it for critical sessions:

• Channel count is tied to network speed. At 96kHz, standard DVS supports up to 32 channels on a Gigabit connection. A 100Mbps connection drops that to 16. Check your Ethernet adapter — and if you’re using a USB-C to Ethernet adapter, make sure it’s Gigabit rated. Many cheap ones aren’t.
• DPC latency is still a factor on Windows. Background drivers — especially GPU drivers on laptops with NVIDIA discrete graphics — can spike and interrupt audio. Run LatencyMon before a session to identify any offenders.
• Set DVS latency to 10ms for recording sessions. You’re not monitoring through DVS, so the latency doesn’t affect workflow. The higher buffer absorbs transient network hiccups.
• Use static IP addresses. Link-local auto-addressing works, but Windows periodically re-validates those addresses and can briefly drop the interface. Assigning static IPs in the 192.168.0.x range on both the SQ5 Dante card and the laptop NIC removes that variable entirely.

Real-World Test Results — and a Friday Surprise

We ran dual-machine recording — a MacBook Air using USB and a Windows laptop using DVS — and cleared a four-plus hour session without a single error on either machine. The next day on the same setup, the Windows machine dropped mid-session after a couple of hours.

Four hours clean, then a drop, is a diagnostic pattern worth recognizing. Immediate failures point to configuration. Delayed failures point to power management, thermal throttling, background processes, or OS activity. LatencyMon is our friend but wasn’t running.  I suspect the NVIDA chipset anyway. But… the truth is that DVS isn’t a critical infrastructure solution. We’ve experienced buffer drop outs, clock jitter, and real-time processing errors when testing with DVS.  It’s a fine driver for video playback or Zoom returns or sound cue machines. Not a solution as a primary recorder. 

The MacBook Air USB stack proved to be more stable with this specific device, but I still wouldn’t want to trust it as a primary recorder using USB2 to USB3 connections. 

When You Need More Than DVS

For sessions where the recording is irreplaceable — a live performance, an on-location session, anything without a second take — DVS on a laptop is a contingency, not a production-grade primary system.

The step up is the RME Digiface Dante — the only portable, bus-powered USB 3.0 Dante interface currently on the market. It connects via USB 3.0, handles all Dante audio processing in hardware using Audinate’s Brooklyn II chip, and delivers 32 channels at 96kHz to your DAW via RME’s ASIO driver. It completely bypasses the xHCI isochronous problem that started this whole investigation. It also functions as a standalone Dante/MADI converter without a computer attached.

An expensive box — but for multitrack location work at 96kHz where the session can’t be redone, it’s a better choice than DVS. 

The Short Version

If you’re seeing USB audio dropouts on a modern laptop:

• The problem is almost certainly xHCI — not your cable, not your settings, not your interface
• USB 2.0 hubs will not fix it
• Mac is more stable but not immune
• Dante over Ethernet removes the USB problem entirely
• DVS works well when configured correctly — static IPs, Gigabit connection, power management disabled, latency at 10ms
• For mission-critical sessions, dedicated Dante hardware is a good answer, but a dedicated hardware recorder is still best

This is the kind of thing that used to took days of forum diving to piece together. Hopefully this saves someone a Friday afternoon of forum deep dives and Google searching.