Notes on the Signal Path of My Home Listening Setup

I noticed that my Apple Music setup might not be configured correctly, and that became the excuse to write down where the digital audio actually goes and how it gets rewritten along the way.

0. Baseline: my setup

First, let me pin down my own configuration. Everything below assumes it.

What each piece is

Steinberg UR22mk2 — audio interface Steinberg is a Yamaha-owned audio hardware and software company (the makers of Cubase). The UR22mk2 is their entry-level USB audio interface. It is really a recording device, but internally it is just a box that combines

• a DAC (digital-to-analog converter),
• a headphone amp, and
• line outputs

into one unit. Its fundamentals are better than the sound hardware built into a PC, so it is worth using even for playback only. It supports up to 24-bit/192kHz.

To use it from Windows you need the dedicated Yamaha Steinberg USB Driver, and that driver holds one of the two settings discussed below.

Yamaha HS5 — monitor speakers Amplifier built in (powered monitors). Connected directly to the UR22mk2’s MAIN OUT. Fairly flat, with restrained bass. Designed to expose flaws rather than to make things sound pleasant.

Sony MDR-900ST — monitor headphones The de facto standard in Japanese studios. 63Ω. It has a pronounced 4–8kHz region, so it too is tuned for finding flaws.

→ Both the HS5 and the 900ST are monitors. 　 In other words, this setup was built from the start around “accuracy first, 　 enjoyment later.” If the sound feels hard or fatiguing, that is neither a 　 fault nor a misconfiguration — it is working as designed.

1. Digital audio fundamentals

Sound is a change in air pressure — a continuous analog waveform. To handle it on a computer, you measure the height of that waveform at fixed intervals and turn it into a sequence of numbers.

• Sample rate — how many times per second you measure. Unit: Hz
  • 44,100Hz (CD standard) / 48,000Hz (video) / 96kHz / 192kHz
• Bit depth — how many steps each measurement can take
  • 16-bit = 65,536 steps / 24-bit = about 16.77 million steps

In array terms, the sample rate is the number of elements and the bit depth is the precision of each element’s type.

Why 44.1kHz

The sampling theorem (Nyquist–Shannon): you can only record accurately up to just under half the sample rate.

The upper limit of human hearing is about 20kHz → doubling it needs 40kHz → add headroom for the anti-aliasing filter → 44.1kHz.

So 44.1kHz is not a compromise: it is theoretically sufficient to record the audible band. This matters, because it is exactly the area where the intuition “bigger numbers are better” stops applying.

2. My signal path

Apple Music
      │  ← the source's own rate (e.g. 44.1kHz)
      ▼
Windows audio mixer
      │  ← force-converted to the "Default Format" rate
      ▼
Yamaha Steinberg USB Driver
      │  ← may be converted again to the driver's rate
      ▼
DAC inside the UR22mk2
      │  ← digital → analog
      ▼
HS5 / 900ST

There are two independent settings

Here is the problem. The same “sample rate” setting exists in two separate places.

If those two disagree, an unnecessary sample rate conversion (SRC) happens.

3. Why avoid conversion

SRC is the process of interpolating points that do not exist from the existing sample points.

44.1kHz → 48kHz is not an integer ratio (it is 147:160), so the original sample points do not survive as-is — every one of them is recomputed.

Go 44.1 → 48 → 44.1 and that process runs twice. The degradation itself is minuscule. The problem is that you gain nothing by doing it.

In the digital domain, doing nothing is the highest quality. The state you want is called bit-perfect: the original byte stream reaches the DAC unmodified.

4. Exclusive mode, and Apple Music’s limitation

There are mechanisms that bypass the OS mixer and let an app output straight to the device.

Note: it is no coincidence that ASIO is a Steinberg standard — the UR22mk2’s driver naturally supports ASIO. But support lives on the driver side; whether it gets used is up to the app.

This is the crux

Apple Music (Windows) can use neither ASIO nor WASAPI exclusive.

→ It is always in shared mode. Going through the mixer is unavoidable.

→ So the fix is not “bypass it,” but 　 align the rate at every stage so that no conversion effectively happens.

For reference, how other setups behave:

• foobar2000 / JRiver → support ASIO and WASAPI exclusive, follow each track automatically
• Apple Music (macOS) → you have to change Audio MIDI Setup by hand
• iOS → the OS follows each track’s rate automatically (the most sensible one)

5. Apple Music’s quality tiers

The first thing I discovered this time was that Lossless audio was Off. In other words, I had been listening to AAC 256kbps.

• ALAC = Apple Lossless Audio Codec. Lossless compression (same idea as ZIP). After decompression the data is a bit-for-bit exact match of the original.
• Most of the catalog is 16-bit/44.1kHz. Only some material is Hi-Res.
• The warning that “Hi-Res requires an external DAC” → the UR22mk2 is that external DAC. It is boilerplate aimed at people on Bluetooth or built-in speakers, so you can just proceed.

6. What I actually configured

Policy: fix everything at 44.1kHz

Most of the catalog is 44.1kHz, so matching that means zero conversion for the vast majority of tracks.

Fix it at 96kHz and every 44.1kHz track gets upsampled instead. Upsampling adds no information (it only interpolates points that do not exist), so there is no theoretical benefit.

Steps

1. Yamaha Steinberg USB Driver control panel → Sample Rate = 44.1kHz
2. Windows sound settings → UR22mk2 properties → Advanced → Default Format = 24-bit, 44100Hz
3. Turn off “audio enhancements” and “spatial sound”
4. Turn off Apple Music’s equalizer and Sound Check
5. Keep the Windows volume fixed at 100%. Adjust volume with the UR22mk2’s knob

The reason for number 5

Digital volume control eats into bit depth. Dropping to 50% ≒ throwing away about 1 bit of information. The UR22mk2’s knob is in the analog stage, so the digital data stays intact.

UR22mk2-specific notes

• Turn the MIX knob fully to the DAW side. Toward INPUT and the silent noise from the input jacks gets mixed in
• Keep GAIN turned down (for playback only, you are not using the inputs)
• Feed the HS5 from MAIN OUT (TRS balanced). Set the HS5’s own LEVEL to around -10dB and consolidate volume control onto the UR22mk2’s OUTPUT knob

7. Priority of effects (a note to self)

If you rank what actually contributes to sound quality in this setup, it comes out like this.

1. HS5 placement — 20–30cm from the wall, tweeters at ear height, isolation pads
2. Room acoustics — reflections, standing waves
3. Condition of the 900ST’s earpads — once they flatten, the highs stab and the bass leaks out. Replacements cost around 2,000 yen
4. Source quality — AAC 256k or Lossless ← this was Off, and I fixed it
5. Sample rate settings ← what I looked into this time
6. Brand differences between DACs and amps

The higher up the list, the more dominant — the differences are orders of magnitude apart.

Getting number 5 right is about “not leaving an incorrect state in place,” not about expecting a dramatic change. Even the difference between AAC 256kbps and Lossless would be an achievement to reliably identify in this setup.

Do 1 through 3 before buying more gear. That is the conclusion.

8. Glossary

Appendix: restating it in engineering terms

Summed up, here is what is going on.

• Decompress an ALAC file → a sequence of PCM samples (a byte stream)
• “Bit-perfect” is the state where that stream reaches the UR22mk2’s DAC without any modification
• SRC, digital volume control, EQ, and effects are all processes that rewrite that byte stream
• The goal is “hand over the bytes you read, exactly as they are”

It is the same idea as not inserting a needless encoding conversion when you pass a stream through. The Windows mixer is an intermediate layer that always intervenes unless you explicitly route around it, and from Apple Music you cannot route around it. So the fix here is to align the input and output formats so that it has nothing to intervene in.