Multi-Sub Optimizer FAQ

Introduction

So far, this FAQ only contains one question, but it's by far the most commonly asked one, and the number of possible answers is quite large. Unfortunately, the problem of mismatch between MSO's calculated results and the final measured results is one that's very difficult to solve in a forum Q&A setting. To accurately and efficiently track the problem down would require being physically present during the whole process.

Why Doesn't My Final Measured Data Agree With MSO's Predictions?

This problem is usually measurement-related. If you're not getting reliable measurements, MSO's calculated results will not correspond to final measured results either. If this is indeed the case, then the problem of getting unreliable REW measurements needs to be solved.

See this post for the kind of accuracy (calculated vs. measured) that can be achieved.

Verify Your Timing Reference Before Taking Your Final Measurements

To do this, perform the Two-Sub Timing Reference Test. If you are running Windows on your measurement machine, follow the REW Acoustic Timing Reference Considerations instructions before running the timing reference test.

Important Considerations for MSO's Prediction Accuracy Below 25 Hz

When using REW to perform your measurements for MSO, the default REW right-side window width of 500 msec causes inaccuracies in REW's reported frequency response below about 25 Hz, which affect the accuracy of MSO's predictions. To prevent this, you must specify an increased window width prior to performing the measurements for importing into MSO. For detailed instructions, see Important Considerations for Accurate Data Below 25 Hz on the Performing Measurements page of the User Guide.

Thanks to Greg Rogers for pointing this out.

A Suggested Technique for Verifying MSO Predictions

Before delving into the details of the potential causes of verification errors, it's worth mentioning one technique for measurement and verification that has worked well for many users:

• When you do your measurements for importing into MSO, measure the MLP last. Afterwards, keep the microphone in place so that when you go back to verify the combined sub response after optimization, the microphone is in the same location you used for the original MLP measurements.
• Using this technique, the MLP will show the best correspondence possible between MSO's predictions and the final measurements. If the MLP has problems, other listening positions will as well. Other listening positions will have additional errors caused by the difficulty of putting the microphone back into the exact same position as the original measurements.
• After performing your verification measurement at the MLP using REW, you can use the techniques described on the Verifying MSO Predictions page to add your measured data to the same graph that displays the predicted data. This allows for direct comparison of predicted and measured values.

A List of Possible Causes of Prediction/Measurement Mismatch

Below is a list of some causes of the mismatch between MSO predictions and final measurement that have been identified by MSO users.

• Not using an acoustic timing reference in the measurements. See this post for detailed instructions on how to set it up.
• Not using the same acoustic timing reference for all measurements.
• Using an acoustic timing reference whose path to the microphone may be partially obstructed in some listening positions.
• Attempting to use a subwoofer as an acoustic timing reference. The acoustic timing reference must be a full-range speaker.
• Attempting to use a loopback timing reference with a USB microphone.
• For errors that only occur below about 25 Hz, using an REW right-side measurement window that's not long enough, such as the default REW 500 msec window. See the explanation above for more detail.
• Performing some individual sub measurements under conditions different from the other subs. This could include:
  • Changing the applied signal level within REW from one individual sub measurement to the next.
  • Changing the AVR volume setting from one individual sub measurement to the next.
  • Changing the timing reference speaker from one individual sub measurement to the next.
• When exporting biquads, you might get a warning message that there are not enough available biquads in the device to accommodate all the MSO filters used. If you ignore this warning and instead try to load the biquad files anyway, your measured and predicted data will not match. You must figure out and fix the cause of the warning before using any exported biquad files.
• When using miniDSP devices, after loading the biquad filters into the device, make sure you keep the Advanced Mode active in all channels. If you temporarily switch to Basic Mode in any channel, make sure you switch back to Advanced Mode when you're done. This applies to the crossover sections as well. These devices maintain separate sets of filter parameters in Advanced Mode and Basic Mode, so if you switch to Basic Mode after loading the biquad filters computed by MSO, you'll be using the wrong filters in any channel that's in Basic Mode. The miniDSP documentation says this about the two modes:

  Switching between basic and advanced mode does not convert between Frequency/Gain/Q and biquad parameters. The basic parameters and advanced biquad coefficients for each filter are stored independently and the selected set is used for processing.

• Using third-party software to change the miniDSP configuration before running the miniDSP configuration software. This can result in the miniDSP software displaying incorrect device state information and an incorrect device state being saved. See The miniDSP "Write-Only" Problem in Tech Topics for more information and the solution.
• Using a different set of AVR outputs for initial measurements and validation measurements. For instance, if you use the Trick for Eliminating All Low-Pass Filters from the Subs' Signal Path when doing the initial measurements, use this technique for the validation measurements as well. If you do not, the final measurements will appear rolled off at high frequencies compared to the prediction.
• Possible changes in the acoustic environment between when the measurements for import into MSO were originally taken, and the final verification measurements after running MSO.
  • For instance, if the room is well sealed with a door, and the original measurements were taken with the door open, while the verification measurements were taken with the door closed (or vice versa), this difference in measurement conditions could cause discrepancies in the predicted vs. measured response at the lowest frequencies.
• Using measurement smoothing, and/or accidentally enabling smoothing during REW measurement export. Exported measurements must not have any smoothing.
• The miniDSP 2x4 non-HD devices have multiple software plugins available, and not all are compatible with multi-sub setups.
  • Do not use the 2-way advanced or 2-way advanced 2.1 plugins.
  • Instead, use the 4-way advanced or 2x4 advanced plugins.
  • The 4-way advanced plugin is preferred, as it allows for 6 biquads per channel and a maximum of 7.5 msec delay.
  • The 2x4 advanced plugin only allows 5 biquads per channel and 7.2 msec delay.
• If you're using a Mac and REW for measurements, use one with a native HDMI port when measuring HDMI devices. The use of USB-to-HDMI adapters has been shown to introduce timing errors in certain setups. See this thread for more details.
• Not using a proper tripod microphone stand for measurements. Here is an example of the correct type of stand to use.
• Not using vertical orientation of the microphone. The microphone vendor's 90 degree calibration file must be used.
• Using unconventional measurement sequences. The correct sequence is:
  • Place the microphone stand at the first listening position to be measured.
  • Mute all subs except for the one to be measured.
  • Measure the non-muted sub.
  • Repeat for all subs at that listening position.
  • Move the microphone stand to the next listening position.
• Specifying an incorrect miniDSP sample rate in Tools, Application Options, Hardware. The miniDSP 2x4 HD uses a 96 kHz sample rate, while the non-HD miniDSP 2x4 devices use a 48 kHz sample rate. These MSO settings are stored in the registry, not the project, so if you switch to using another computer, double-check these settings on the second computer.
• Make sure the sample rates chosen in the Windows Sound Control Panel for the input and output sound devices are equal to one another and correct, and that the REW measurement sample rate agrees with what's been chosen in the Windows Sound Control Panel. For example:
  • The miniDSP UMIK-1 microphone uses a fixed 48 kHz sample rate. This means you must do the following:
    • Set the UMIK-1 sample rate in the Windows Sound Control Panel to 48 kHz.
    • Set the HDMI output device sample rate to 48 kHz in the Windows Sound Control Panel.
    • Set the measurement sample rate to 48 kHz in REW.
  • The sample rate of the miniDSP UMIK-2 microphone is user-adjustable. When using this microphone, do the following:
    • Set the UMIK-2 sample rate in the Windows Sound Control Panel to the desired value, say f₀. A safe value for f₀ is 48 kHz.
    • Set the HDMI output device sample rate to the chosen f₀ in the Windows Sound Control Panel.
    • Set the measurement sample rate to the chosen f₀ in REW.
• Failing to manually enter the delay and attenuation values into each miniDSP channel after running MSO. Some new miniDSP users mistakenly believe that loading the exported biquad coefficients is all that's needed. Manual attenuation and delay entry into each miniDSP channel is also required.
• Using delay or gain values from the MSO filters themselves, rather than the filter report. Delay blocks in MSO can contain negative delays. When you run the filter report, the final delays are shown at the end, adjusted so that no negative delays are present. A similar thing is done with gains. The filter report adjusts gain values in dB to be negative, representing an attenuation. This is to accommodate certain miniDSP models that don't allow positive gains.
• Attempting to adjust individual sub levels after optimization using sub amp analog volume controls instead of DSP hardware. After optimization, sub levels should only be adjusted in DSP hardware such as the miniDSP 2x4 HD. If your amp has internal DSP hardware as some Behringer amps do, and that's your only DSP, then that DSP's level control can and should be used.
• Some miniDSP devices (like the original 2x4 devices, the 4x10 and others) use fixed-point DSP processing. These fixed-point processors have degraded accuracy at very low frequencies (see images showing frequency response). For this reason, fixed-point processors are not recommended for use with MSO.
  • The miniDSP 2x4 HD uses floating-point DSP, which does not have this problem.
• Mix-ups between MSO sub channel and miniDSP sub channel causing biquad text files to be loaded into the wrong miniDSP channel. The miniDSP channel naming feature can help with this.
• Some AVR or amplifier DSP settings failing to "take" until backing out of the screen in which they are set.
• Accidentally taking measurements without first clearing out the miniDSP filters, gains, delays and polarity inversions.
• Inadvertently taking measurements with Audyssey Dynamic EQ or Dynamic Volume on. All room correction should be disabled before measuring (except under certain very obscure circumstances).
• Unintentionally using one of your AVR's upmix or other unusual processing modes during the measurements. This might include modes such as "party mode", "all-channel stereo" or some other mode that combines channels rather than allowing them to be energized individually. Using such modes could cause some speakers other than the desired ones to be outputting sound and interfering with the desired measurements. Instead, use straight 7.1 mode or whatever mode avoids undesired signal processing. Some users of Pioneer and Yamaha AVRs report needing to use the so-called "pure direct" mode to avoid unwanted upmixing or other processing.
• If you get a warning about signal level for the acoustic timing reference in REW when doing a measurement, adjust signal levels until the warning goes away before performing any measurement.
• Failing to clear out the default low-pass and high-pass filters from the miniDSP crossover section before measuring.
• Setting "LPF for LFE" too low if measuring using HDMI channel 4, or crossover too low if measuring with a Left, Right or Center channel stimulus. These should temporarily be set to their maximum allowable values (usually 200 Hz or 250 Hz) prior to measuring for a sub-only optimization.
• Measuring at too high an SPL, such that significant subwoofer distortion or room object vibrations are present. MSO does its calculations using the Superposition Principle (like a generalization of REW's "A+B" measurement processing function). The Superposition Principle assumes a linear system (see link), which means that distortion of the subs should be negligible. Try to keep typical subwoofer measurement levels at around 85 dB SPL to minimize distortion.
• The results of including bass shakers and so-called "BOSS" arrangements in the measurements and optimization are likely to be invalid due to nonlinear behavior of the devices.
• One user had been getting very good correspondence between MSO predictions and final results, but then made some changes to his system. The system developed a slight hum as a result of the changes, and then he began getting poor correspondence between MSO predictions and final results. The hum was interfering with the measurements, especially around 60 Hz and 120 Hz.
• At the high end of the subwoofer measurement band, microphone placement repeatability can affect correspondence between MSO predictions and final measurements. Measuring the MLP last can help with getting MSO predictions and final measurements to match.

This FAQ will be updated from time to time as more possibilities come to mind and more user experience is incorporated.