- Adjust measurement parameters to increase coherence (≥95%).
- Log every detail necessary for future you to be able to recreate the setup.
Imagine that you are going to work on a show tomorrow with VRX932 and 918 speakers. You open up a folder of old measurements because you want to be as prepared as possible and you figure it would be helpful to see how those two speakers line up together and what you should expect to see once you get your audio analyzer fired up in the field. You find a 932 you measured on a show in New Orleans and a 918 you measured in a warehouse in Minneapolis.
While it is helpful to see the traces, you still have questions.
- Are they phase compatible? Will you need to add any filters?
- What is their relative phase offset? (so that you can calculate their absolute phase alignment in the field with a laser disto)
- What about power scaling? Will you need a gain offset or a change in system design?
While systems can be very complex and at first it might seem like an impossible task to account for every step in the signal chain, the methodical logging of a few important details will have you on the way to building your own database of high-quality measurements that you can rely on for reliable predictable predictions. 🤷♂️
*I have completed this process several more times since creating this video. As far as I can tell, there’s no way to stop the wind from screwing up the VHF (very high frequency) data. It seems to help to leave auto delay tracking on for better average coherence. Measuring outside in still preferable for better LF data, unless you have a giant airplane hanger where you will be far away from all boundaries.
Software settings 🤖
- Sample rate: 96kHz recommended if hardware compatible, otherwise 48kHz. Keep it consistent so you won’t have a bunch of mismatched data. See more on time records here.
- Coherence: Squared Coh enabled so that 0dB SNR corresponds to 50% coherence
- Blanking Threshold: 95% for visual feedback on SNR
- Averaging Type: Complex (Vector)
- Averages: Inf(inity) for best SNR and stability
- Smoothing: 48dB/oct for accurate visual feedback
- Live IR window visible (cmd+i) so that it will be stored with the other traces
Although many different options are encountered as signal passes through DSP, amp, and loudspeaker, the goal is that any other audio professional (or your future self) will be able to recreate the same setup following your notes. To that end, I recommend using the manufacturer’s defaults at every step.
If you can’t find the manufacturer recommended settings, contact them. (try it, it works!) If that doesn’t work, ask colleagues. If all else fails, record as many variations as you have the patience for while also including a native or naked setting with no processing active so that you can potentially model any adjustments on top of it later.
Mic Placement: Start at 3x the baffle diagonal. To calculate: google “calculate diagonal of a triangle” or drop this in: √(width^2 + height^2)*3.
√(22.47^2 + 11.15^2) * 3 = 25.08 * 3 = 75.24 in = 6.27 ft = 1.91m
We accept the fact that the frequency response may not be benchmark in exchange for a higher quantity of data (until I can build my own 2k sq ft anechoic chamber and invite you over).
Ground plane: the measurement microphone capsule is as close to the ground as possible. Line source elements are in landscape mode, perpendicular to the floor.
(optional: if you don’t own a mic calibrator, the phase data is still useful)
- Calibrate the MIC channel with the microphone calibrator set to 114dB (or your calibrator’s highest setting) and the microphone preamp set to -12dBFS as seen on the audio analyzer’s input meters.
- Adjust the REF loop input so that when the signal generator plays a 1kHz sine wave at -12dBFS, the audio analyzer input meter also reads -12dBFS.
Note: I chose -12dBFS because it is easy to see on the meter where green changes to yellow. If you need more headroom for some reason, you are free to use any other value. As long as MIC and REF match, we can make a linear comparison.
- Start your measurement.
- Confirm that coherence is ≥95% and ripple is ≤6dB. If you have your coherence blanking threshold set to 95% as indicated above, it should be fairly easy to see how much data is being blanked out. To improve, reduce background noise, increase signal generator*, reduce measurement distance, or increase the distance from boundaries (go outside).
- Wait until measurement data stops moving (stability!).
- Store trace.
- Log measurement conditions.
*Remember to use the sig gen controls in Smaart© to maintain matching outputs. Do not adjust the output at the amp or DSP.
You may want to copy the form I created here.
Have you tried building your own personal database of near-anechoic measurements. What are your tips?