I mentioned that my software allows the speaker's cabinet to be centered on the turntable axis instead of the front baffle, and mathematically corrects the result to what CTA-2034 specifies. The question is how large the errors of the centered setup actually are, and how completely they can be removed. I have now built a controlled simulation to test exactly that, and this post summarizes the method and results.
The Problem CTA-2034 defines rotation about the speaker's reference point, which is typically the tweeter on the front baffle. On a physical turntable, this places the entire cabinet behind the platter's center—an impractical arrangement for heavy speakers, and particularly awkward for vertical sweeps where the speaker lies on its side.
The usual practice is to center the cabinet on the platter. The front baffle then sits half the cabinet depth in front of the rotation axis. As a result, the microphone is closer than nominal on-axis and farther away at the sides, meaning each angle is recorded from a slightly different distance and direction than intended. The question is how much this affects the CTA-2034 curves, and whether those errors can be corrected in software.
Method measured physical speaker cannot serve as a reference here, because its true CTA-2034 response isn't known independently of the measurement itself. The software therefore simulates an ideal speaker: a two-way system (6.5" woofer, 1" dome tweeter in a 22 x 38 x 28 cm sealed box, 4th-order LR crossover at 2 kHz, including cabinet edge diffraction, with the mic at 1 m).
Because the speaker is simulated, it can be spun twice:
- Reference Run: Rotated around the front baffle (yielding the exact CTA-2034 standard).
- Centered Run: Rotated around the cabinet's middle (yielding what a physical turntable actually records).
The difference between the two isolates the error of the centered setup with zero external noise.
The centered rig modeled upright for horizontal sweeps, and on its side for vertical sweeps, with the platter's axis passing through the
middle of the cabinet and the mic aimed at that axis.
Results Without Correction Comparing the uncorrected centered spin with the reference :
- Level Shift: Every curve reads above the reference, but by varying amounts: On-Axis is about +1.3 dB high, Listening Window is +1.2 dB, Early Reflections is +0.7 dB, and Sound Power is near 0 dB in the bass, rising to +0.4 dB at high frequencies. The mic is 14 cm closer on-axis than at the sides, and the curves directly reflect that geometry.
- Directivity Index (DI) Error: Because the offsets vary between curves, the Directivity Index reads about 0.85 dB too high across all frequencies. This is the main takeaway: an uncorrected centered spin makes the speaker appear significantly more directional than it actually is.
- Bass Artifacts: The difference map is uniformly blue ("quieter off-axis than reference") across all frequencies, including 100 Hz where a box of this size is omnidirectional. A spinorama that shows directivity at 100 Hz was not rotated about its reference point—a useful sanity check for any measurement rig.
- Summary: Mean level is 0.77 dB high, and the worst curve deviates by 1.13 dB RMS in shape.
Results With Correction Applying the correction:
- Level Error: Reduced from 0.77 dB -> 0.05 dB.
- Shape Error (Worst Curve): Reduced from 1.13 dB -> 0.11 dB RMS.
- DI & Difference Map: The Directivity Index error is eliminated, and the difference map flattens to near-neutral.
Residual Errors The remaining residual is small and comes from two identifiable sources:
- Crossover Dip (~0.2 dB between 1.5–3 kHz): The two drivers are 190 mm apart, and their path difference to the mic differs slightly at 0.86 m vs 1 m. A single geometric reference point cannot resolve driver spacing.
- On-Axis Ripple (+/- 0.3 dB above 3 kHz): Caused by the cabinet's edge-diffraction pattern shifting marginally with distance (~1 mm rim path shift, or 8 degrees of phase shift at 8 kHz).
Importantly, the averaged curves (Listening Window, Early Reflections, Sound Power) smooth these out and remain within about 0.1 dB. Both residuals cut roughly in half if measured at a 2 m distance.
Summary & Takeaway
The bottom line is very encouraging for DIY measurement rigs.
A corrected 1 m spin with the cabinet centered lands within
~0.1 dB of the true CTA-2034 reference on averaged curves. That puts its accuracy right on par with top-tier, multi-thousand-dollar commercial measurement systems, while keeping the physical hardware setup fast, stable, and straightforward.
You don't need awkward counterweights or unstable physical offsets to capture reliable polar data. Simple software geometry does the heavy lifting, bringing high-end Spinorama precision within easy reach of any DIY setup.
