Back to Resources
How-to

How to Compare Stems Against a Reference Mix

A structured A/B and null-testing workflow. Learn how to build a trustworthy stem sum, why loudness matching is essential, and how to read the residual of a null test.

By Stem CheckerUpdated

Null test

Reference − stem sum
Stem Checker showing a reference mix, the summed stems, and the small residual left after nulling the two

The reference and the stem sum leave only a small residual — what remains is likely master-bus limiting or other mix-bus processing.

Define what the comparison should prove

A reference comparison can answer several different questions: Are all musical parts present? Do the stems begin at the same point? Does their balance resemble the approved mix? Do they reproduce the mix exactly enough to null? These are not the same test, and the expected result depends on how the stems were created.

Important expectation

Stems won’t necessarily null against a mastered reference. Master-bus compression, limiting, clipping, analogue processing, dither, and effects that react to the full mix can change the result even when every stem is correct.

Build a trustworthy stem sum

  1. 1
    Create the session at the correct sample rate. Avoid unnecessary conversion before the test.
  2. 2
    Place every stem at the same origin. Don’t drag individual tracks until the music appears to line up.
  3. 3
    Recreate the intended routing. Respect pan positions, mono/stereo treatment, polarity, automation, and any documented bus structure.
  4. 4
    Set neutral playback conditions. Disable processing that wasn’t part of the original stem print and keep the master path clean.
  5. 5
    Route all stems to a comparison bus. This gives you one signal to switch against the reference.
  6. 6
    Import the approved reference separately. Ensure it starts at exactly the same sample and isn’t routed through the stem bus.

Level-match before judging

The louder signal will usually appear fuller, clearer, and more exciting even when it isn’t more accurate. Match perceived loudness before making tonal or balance judgments. LUFS can provide a starting point, but short-term programme differences and peak limiting mean the final adjustment should also be checked by ear.

ComparisonRecommended level treatmentWhat it reveals
Arrangement checkApproximate loudness matchMissing parts, wrong edits, obvious balance errors
Tonal A/BClose loudness match over the same sectionEQ, width, dynamics and ambience differences
Null testExact gain and polarity relationshipSample-level differences under controlled conditions
Peak checkDo not normalize independentlyUnexpected clipping or limiting differences

Use section-based A/B listening

Loop short, revealing sections rather than switching across an entire song. Good test points include:

  • The first full downbeat
  • A dense chorus
  • A vocal-only moment
  • A transition with effect tails
  • The final decay

Switch often enough to retain auditory memory, but not so quickly that you hear only the switching action.

When a null test is appropriate

A null test inverts the polarity of one signal and sums it with the other. Identical signals cancel. For the result to be meaningful, the reference and stem sum must share the same start sample, sample rate, channel layout, gain, pan law, and processing history. Even a slight sample shift or a tiny gain difference can leave a large residual.

Interpreting the residual

Residual soundLikely interpretation
Almost complete silenceThe stem sum is extremely close to the reference.
Full mix, quieterGain mismatch or incomplete polarity inversion.
Transient clicks or attacksSmall timing offset, different limiter response or edit boundary.
Bass-heavy residualLow-frequency phase, gain or master-bus dynamics difference.
Mostly sides or ambienceStereo width, reverb, modulation or pan difference.
One clear instrumentMissing, duplicated or differently processed stem.
Residual grows through songSample-rate, clock or time-stretch mismatch.

A note on residual

A residual is likely the result of master-bus limiting or other mix-bus processing.

A practical decision hierarchy

First verify that the arrangement is complete. Then compare timing and balance. Next compare tone and dynamics at matched loudness. Only attempt a null when the routing history suggests exact cancellation is realistic.

Engineer checklist

The reference and all stems use the same timeline origin.
The comparison session uses the correct sample rate and channel format.
Pan, automation and routing are recreated where required.
The stem sum and reference are loudness-matched for subjective A/B tests.
Short, revealing sections are looped for comparison.
A null test is used only when exact equivalence is technically possible.
Residual audio is diagnosed rather than treated simply as pass or fail.

QC your stems in seconds — no DAW needed

Stem Checker loads and plays your stems in sample-accurate sync and flags length issues in your exports — mismatches, missing audio, duplicates, clipping, dual mono and more. A faster, more reliable way to review and organise your stems and multitracks.

Save time. Deliver with confidence.

We’ll email a macOS download link. No spam.

Keep reading