Woofer-midrang crossover null question

[gainphile]

I am using 4th order LR xo (active) for Woofer-Mid separation. The xo frequency is 160Hz.

Why am I seeing -6db nulls? I thought LR would give flat freq. response? When I invert the connection of the woofer the null is much deeper which seems to be the right way.

I have checked and compensated the phase alignment actively. I checked the phase (2nd pic) and the phase matches. I have moved the mic location, height, etc.

So what's wrong here? btw. ignore the woofer rise as it is due to dipole compensation circuit.

I do not experience the same problem with M-T xo at about 1.8kHz. It is flat there.

[jerome69]

Originally posted by gainphile
...
So what's wrong here?
...

Below 300HZ, it's very difficult to make good measurements in a normal room (not anechoic). The cause is reflection. The best I've seen indoor is 160Hz, the roof was high. Try outdoor if you can and put damping material on the floor.
I see here a post of "john k...", he explains to do low frequency measurement in small space.

[Ferrit37]

Hey Gainphile,
Can you take a measurement of just the LF and then just the MF?
Don't change any delay times and look at the phase responses through the x-over overlap zone.

[Ferrit37]

Hey Guy's seems like it's 1 am in Melbourne so here's the quick answer "time alignment"

I would guess driver acoustic center offset

Here's a couple of screenshots that should illustrate the point.



Here we have a LR 24dB 160 Hz High pass and Low pass filter with the summation in Orange (moved up +6dB for clarity)



In this second shot we have the same thing but 2ms of delay has been added to the High pass and we see the dip.

hope this helps

[BlueWizard]

First, I don't claim to be an expert, but I thought I saw something like this in the 'Loudspeaker Design Cookbook' when one of the speakers needed to be have reverse polarity.

Frequently in a woofer/tweeter configuration, the tweeter needs to be wired backwards - reverse polarity. In the case of a 3-way it is usually the midrange that needs to be reversed.

Just from memory, it seemed a dip at the crossover was a sign that the polarities were not correct.

I'm not saying this IS your problem, only that it is one more thing to consider.

Personally, I think you are seeing a problem with your room as suggested by someone else.

For what it's worth.

Steve/bluewizard

[5th element]

Originally posted by Ferrit37
Hey Guy's seems like it's 1 am in Melbourne so here's the quick answer "time alignment"

I'd be surprised at it being an issue with time alignment. A 2ms delay results in a path difference of 0.68 meters, which is huge.

In my open baffle, the Woofers are in a H baffle, which sets them back 15cm from the lower midrange unit. With the first bass driver 30cm below the lower mid.

This represents, in my opinion, a pretty worst case scenario, usually you've got the drivers mounted in alignment on the Z plane. Only the units physical construction separating the acoustic centres.

And at a measurement distance of 2meters, we only see very minimal freq' response issues due to driver offset, (with the conditions mentioned above). In a theoretical filter set at 160hz there's less then 1dB of ripple through the xover region under these conditions.

Here however we don't know the measurement conditions, which I think are pretty important.

In fact what jumps out at me more is the smoothing at 1/2 octave. This smothers everything completely away from the real world situation which probably looks horrific, hence the applied smoothing.

As Jerome69 says

Below 300HZ, it's very difficult to make good measurements in a normal room (not anechoic).

This is the pertinent point and could be the cause of the peculiar results.

While it is true that applying heavy smoothing to an obviously reflection influenced FR, can show you overall trends, it cannot be relied upon to be completely accurate.

This null that you're showing could simply be due to a room reflection of some sort.

Ferrit37 says

Can you take a measurement of just the LF and then just the MF?

Which is a very good thing to recommend, however I'd add to do this as a near field measurement.

I wouldn't be interested in trying to get these measurements to sum (in a simulation program) in any realistic way. However taking a near field response will remove almost all reflections and allow you to accurately see what your filters are doing.

If the midrange highpass is working properly at 160hz. But the low pass on the bass driver is incorrect and you're getting the filter centred at ~200hz, this would show these kind of results.(See attached picture.)

IMO you need to see that you're filters are doing exactly as they are intended to do. As this is active and you're using ARTA, you can feed the output of the active filter into the line in on the sound card and take a measurement. This will allow you to see the filters transfer function and you can easily verify if they are doing as they are intended to do.

If they are, hook up the drivers again, then take the near field measurements to check that the drivers acoustic response and the filters electrical response are combining into the slopes you are expecting. (Could be somewhat tricky due to the compensation circuit for bass loss)

From simulation it shows that unless you've got some wild cabinet arrangement, path differences shouldn't influence the final result by much. Remember we're operating with a wavelength of over 2 meters at 160hz, c-c spacing isn't that critical.

So if you're final acoustic slopes are showing proper 160hz 4th order LW characteristics. You can be safe knowing that the result will be summing correctly and that any nulls you're seeing are most likely a result of reflections.

[Ferrit37]

Hey 5th Element,
your correct - 2ms is an awful lot of delay, either its in his processor, or as you say a boundary reflection.
your suggestion of measuring the outputs is very good.
I'm used to making ground plane measurements to minimise boundary effects and as you say the 1/2 octave smoothing is masking a lot of the clues.
I'm not that familiar with ARTA does it have any Coherence/cross correlation function? ( I'll go look)

[gainphile]

Thank you!

I will summarise the next steps based on your inputs:

1. Make sure the analog filters are correct as intended - measurement of the outputs and their summation.

2. Once #1 is confirmed, measure the acoustic outputs individually and investigate the phase response (near field and farfield).

About the 1/2octave smoothing, I actually looked at unsmoothed, 1/6, 1/3 etc. with the same trends. Also reversing the mid driver gives the same effect (only with the tweeter xo correctly becoming null).

The speakers being investigated look like this:
http://2.bp.blogspot.com/_COvnfAy4R9...h/DSC_9122.jpg

I tried both with 1 and 2 woofers active with same results.

[John_E_Janowitz]

Since this is an active Xover you have only time alignment and phase to work with. Very slight differences in time alignment will cause this. Take an impulse response. You'll see the tweeter and mid arriving about the same time with the woofer arriving later. Correct this and you'll be fine but only at the point where you take the measurement. Move the mic up, down, forward or backwards and you'll see the dip change as the time difference changes. Optimize for your listening point and that is about all you can do. Time difference between mid/tweeter is much less than mid/woofer.



John