Will this measurement process work?

[mhdelano]

So I have ordered the dayton omnimic system so I can do some measurements instead of modeling everything based on sometimes suspect infinite baffle measurements. However, I'm not sure if what I want to do will work.

1.) Measure response of woofer in box on axis with a gated measurement.

2.) Measure tweeter response on axis in box.

3.) Measure both drivers in box on tweeter's axis.

4.) Import woofer's frd file into response modeler and splice with a box sim at the lowest frequency that the gated measurement was accurate.

5.) Extract minimum phase data from spliced woofer frd file, and tweeters measurement.

6.) Insert frd and zma files into PCD. change woofer z distance so that summed response without a crossover matches the measured summed response.

Will this method work with minimum phase extraction? or do I have to use measured phase? Will this capture the full effect of the baffle step?

[daryl]

I would not assume that a simulated box alignment is at all what you really get.

You will want to measure woofer response full range to see what you really achieved and see if you want to make adjustments.

I don't care for trying to splice windowed measurements to nearfield measurements.

The best technique is the groundplane method (not so good in the winter).

Place the mic on a hard surface (parking lot) and locate the speaker one to three meters away facing the mic (tilted down) so the mic is on the intended listening axis (larger speakers will need to be three meters so the drivers and all diffractions sum properly, small speakers will sum fine at one meter).

The mic being on the ground causes the ground reflection to be coinsident with the direct signal so the ground reflection only adds 6db to spl without any other unwanted contributions.

The mic actually is slightly above the ground because of the diameter of the mic capsule and will cause some high frequency rolloff, so you will want to also measure the tweeter with both the mic and speaker elevated, windowing out the ground reflection and splice this with the tweeters groundplane measurement.

Splicing the tweeters groundplane measurement with it's windowed measurement does not have the problematic nature which you have when trying to splice nearfield and windowed bass measurements because only the highest frequencies are in question with the groundplane measurement making it easy to window out nearby surfaces and still have plenty of valid data.

Also both measurements include the diffraction effects of the cabinet so that the two measurements to be spliced overlap nicely (nearfield measurements don't fully include contributions from baffle diffraction).

With groundplane measurements if you set the speaker on the ground the baffle is mirrored by the ground and becomes twice as big in that dimension.

A tower speaker with all of the drivers at the top is not a problem.

You can stand it up and lean it forward to point it at the mic.

The bottom of the baffle is doubled, but there is little contribution from the bottom of the baffle for a tower with the drivers at the top.

A small two way will be effected more because the sound normally is diffracted around the bottom of the enclosure and setting it on the ground makes it twice as tall and sound will not be diffracted around the bottom of the enclosure.

You can belt a small speaker to a stand and lean it forward like in the tower example.

You would never want to put a skinny speaker on it's side and measure it because that would make the baffle twice as wide.

You are going to remove the excess time from your measurements so you don't have crazy phase wrap in the tweeter measurement (the same value for all drivers, omnimic measurements wont have excess time).

Now you have valid fullrange measurements for your drivers with valid relative phase data which will make it a snap to design you're crossovers in PCD.

[daryl]

I see in a post from Jeff B. in another thread going on right now that the omnimic does not measure phase relative to the outgoing signal but instead starts the clock at the measured impulse which will skew relative measured phase.

Still I would not discard measured phase and calculate minimum-phase.

Instead I would take an additional measurement with the woofer and tweeter connected in parallel.

You can use you're software to compare the parallel connected woofer/tweeter meausurement to the individual woofer tweeter measurements summed in software by adding offset time to woofer measurement until the software summed individual measurements agree with the parallel connected measurement.

also.....reverse polarity comparisons as bwaslo points out below may be better at creating nulls near the intended crossover frequency which are more sensitive to timing errors than if the the drivers were to sum in phase.

[mhdelano]

Thanks for the info, I'll do some more research into ground plane measurements. not sure if i'll be able to do it since i live in an apartment.

[mzisserson]

Thanks for the info, I'll do some more research into ground plane measurements. not sure if i'll be able to do it since i live in an apartment.

For all intensive purposes, your method will work just fine. Bare in mind it is a simulation, not an absolute. Extracting phase is a good thing. Daryl is right, you will not get 100% accurate phase measurements from the Omnimic.

I use a similar measurement technique. I sim below 200Hz, since end-of-day response is HIGHLY room dependent. Advanced measurements like a Ground Plane are extremely useful in accurately tuning bass, but this can be done after the primary c-over is squared away.

I recommend building the cabinets and measuring the drivers on axis at 1M this will capture diffraction and step effects. If designing for a specific room you can go so far as to measure in their position. Then take an accurate in-room response measurement from several locations and hone in the summed response of the two speakers as well.

[daryl]

You'll want an inverter so you can power you're measurement rig from you're car.

A closed store parking lot or parking lot or a lot at an industrial complex at night is what you want.

[bwaslo]

You'll get accurate phase measurements with OmniMic. What you possibly won't get are accurate delay measurements (because of the reference plane being tied to the impulse arrival). For speaker measurement you don't really need to know the absolute delay, only the relative delay between drivers you are crossing over between.

Make individual measurements, with phase of the drivers. Then play both drivers together and measure their combination. If the sensitivities of the two are much different, pad the sensitive one down a little (put a series resistor on it, maybe a parallel one across it too) till they are close at the measuring point. Look for a suckout where the drivers combine. If there isn't one, invert the phase on the tweeter to find where you get the sharpest dip. Then go into PCD, and set it to match the network (series and/or shunt resistors, tweeter polarity) that you have and the vertical listening position relative to the drivers, and distance from the tweeter.

Finally, adjust the distance to the woofer in the software to achieve a simulated response that best matches what you you measured. All you are trying to find is this distance of woofer plane relative to tweeter. After you have it, leave it fixed, remove the resistors and set the tweeter polarity back in the software and in reality (those were just for determining delay difference). You are then ready to begin designing and optimizing the software.

Doing three-way or 4-way crossovers can be done along the same lines, though more difficult of course.

[Jeff B.]

You'll get accurate phase measurements with OmniMic. What you possibly won't get are accurate delay measurements (because of the reference plane being tied to the impulse arrival). For speaker measurement you don't really need to know the absolute delay, only the relative delay between drivers you are crossing over between.

Make individual measurements, with phase of the drivers. Then play both drivers together and measure their combination. If the sensitivities of the two are much different, pad the sensitive one down a little (put a series resistor on it, maybe a parallel one across it too) till they are close at the measuring point. Look for a suckout where the drivers combine. If there isn't one, invert the phase on the tweeter to find where you get the sharpest dip. Then go into PCD, and set it to match the network (series and/or shunt resistors, tweeter polarity) that you have and the vertical listening position relative to the drivers, and distance from the tweeter.

Finally, adjust the distance to the woofer in the software to achieve a simulated response that best matches what you you measured. All you are trying to find is this distance of woofer plane relative to tweeter. After you have it, leave it fixed, remove the resistors and set the tweeter polarity back in the software and in reality (those were just for determining delay difference). You are then ready to begin designing and optimizing the software.

Doing three-way or 4-way crossovers can be done along the same lines, though more difficult of course.

Maybe this will help (downloads a PDF file on the topic):
http://www.box.com/s/ouxjjsx0m8bs00cil5iq

Jeff

[mzisserson]

Maybe this will help (downloads a PDF file on the topic):
http://www.box.com/s/ouxjjsx0m8bs00cil5iq

Jeff

Jeff,
Would this work in an MTM setup? I have read the 3-step method several times, first time I understood it!

[Jeff B.]

Jeff,
Would this work in an MTM setup? I have read the 3-step method several times, first time I understood it!

If the two woofers are driven in parallel for a combined response then it should work fine.

[mhdelano]

Wow thanks for the responses everyone! I've never gotten so much help before from a forum haha.

I noticed in the end of your pdf, you recommend finding the acoustic offset with measured phase data, then do your post processing, and then extract the minimum phase data. How does this "minimum" phase data differ from the omnimic measured phase data?

Another lingering issue is the baffle step transition period. If I were to create the crossover so that it is a flat response from 250 Hz on up, it would actually only be about 4dB of baffle step compensation. Which might be fine given room boundary reinforcement. I guess the simplest thing to do if I wanted to do a full BSC design, would be to design the crossover so that it is flat 1 or 2 dB down from the 250 Hz level.

Once again thanks for fully explaining this to me.

[daryl]

I don't know that you will get reliably down to 250hz indoors without reflections in the window.

Thus the reason for the groundplane suggestion.

An enclosure does have an associated diffraction loss.

However I think the prevalent perspective of 'baffle step compensation' and that you will compensate a certain number of decibels is the wrong perpective from which to view speaker design.

If you start with full range anechoic measurements and design you're crossover filters for flat response.

This does not account for the inherent bass augmentation you have with speakers indoors.

It wouldn't be bad in reasonable rooms because the bass boost is progressive as the frequency gets lower and the lower frequency gets the more benign bass boost is.

For larger speakers even though it doesn't sound bad you are wasting precious cone excursion which is usually limited.

For small speakers with modest bass extension you actually will need some bass boost to restore bass heft to the sound, compensating for the subdued lower frquencies.

To design you're speakers so that they are flat inroom (or so that they have an intensional amount of bass emphasis in the case of small speakers of limited bass extension) you will perform additional measurements inroom.

These will be conducted with the speaker in it's intended place and the mic in the best listening position (or average several if you prefer).

You will use a long measurement window so to include the accumulated effects of the room.

Now you can use the anechoic measurements for designing the crossover and checking for issues (resonances etc.).

And you will also apply you're modelled crossover filters to you're inroom measurements with smoothing applied to verify inroom bass levels.

You will in fact use you're inroom measurements to set you're target spl for you're crossover optimizer.

The most important thing you can take from this is that the concept of 'BSC' and how much do you need to apply, never even came up in the entire design process and is completely unecessary.

The concept of BSC can only hinder you in you're effort.

[mzisserson]

If the two woofers are driven in parallel for a combined response then it should work fine.

Thanks Jeff!

[mhdelano]

Thanks Daryl, so by doing an ungated, smoothed response measurement in the typical listening application I'll know what level to set my flat response at. Makes sense to use the equipment instead of just guessing on a BSC amount.