[joshsehn]

Based on that new testing data, I'm also going to implement 200Hz notch filter on the subwoofer. That hump is probably effecting output even with the crossover at 100Hz. You can see the output is NOT down 24db at 200hz like it should be.

[matevana]

Based on that new testing data, I'm also going to implement 200Hz notch filter on the subwoofer. That hump is probably effecting output even with the crossover at 100Hz. You can see the output is NOT down 24db at 200hz like it should be.

Very cool. As long as I've already wasted a few hours of your time, I'm really curious about the alignment you selected for the two mid woofers, the dome midrange and the tweeter. It looks to me like the acoustic centers form a perfect arc, but maybe that just worked out as such. Did you calculate their relative distance based on frequencies?  Just curious. When I experimented with offsets in my current configuration, the measurements showed only a small difference off-axis. With my previous config, (not coax) it was larger.

[Rudolf]

Based on that new testing data, I'm also going to implement 200Hz notch filter on the subwoofer. That hump is probably effecting output even with the crossover at 100Hz. You can see the output is NOT down 24db at 200hz like it should be.

Your way of EQing a dipole is not the usual one afaik. You start with a 12 dB/oct shelving filter at 96 Hz. Obviously this is much too weak. Most others would start with the crossover low pass at 30-50 Hz and second or even third order. You first want to get the response flat from the lowest target frequency  up to the dipole peak region. This way all peaks higher up get much more attenuated than with your method. The shelving filter would only be a second thought if you want to cross over lower than at the natural roll off of the H frame into the first dipole dip.

The filters of the DCX do work as good as they should and they do have enough cut/gain. But they have to be used with a good strategy: Start with the more powerful filter and adjust with the less powerful one - not vice versa.

Rudolf

[joshsehn]

Your way of EQing a dipole is not the usual one afaik. You start with a 12 dB/oct shelving filter at 96 Hz. Obviously this is much too weak. Most others would start with the crossover low pass at 30-50 Hz and second or even third order. You first want to get the response flat from the lowest target frequency  up to the dipole peak region. This way all peaks higher up get much more attenuated than with your method. The shelving filter would only be a second thought if you want to cross over lower than at the natural roll off of the H frame into the first dipole dip.

The filters of the DCX do work as good as they should and they do have enough cut/gain. But they have to be used with a good strategy: Start with the more powerful filter and adjust with the less powerful one - not vice versa.

Can you spell out that strategy in more detail for me? I would be happy to improve the setup.   

To clarify, in my understanding with my existing setup I get full output at bottom end, reducing to -15db at 96Hz with 12db/octave HP EQ ... and I apply LR 24db/octave crossover above 110Hz.

[joshsehn]

I'm really curious about the alignment you selected for the two mid woofers, the dome midrange and the tweeter. It looks to me like the acoustic centers form a perfect arc, but maybe that just worked out as such. Did you calculate their relative distance based on frequencies?  Just curious. When I experimented with offsets in my current configuration, the measurements showed only a small difference off-axis. With my previous config, (not coax) it was larger.

Despite any real proof that the effects of baffle diffraction are well understood and can be modeled accurately, I choose the locations of the drivers and size and shape of the baffle based on may hours of extensive baffle diffraction simulations in Excel Spreadsheets to find the configuration that would have the least possible negative effects of the baffle.

Hence the 3" rounding on the edges of the baffle, the angled side and varied driver distances from the edge as you can see.

I doubt that the baffle works exactly like the simulations said it would but you can see the dip at approximately 6000 off axis was predicted. Otherwise the off axis measurements seem pretty good. I think there is one substantial dip on the other side due to the proximity of the 2" to the close edge. If I did it again I might do that one a little differently.... then again I might be tempted to do without baffles entirely. 

I put all the drivers as close as possible together. The are almost all touching.

After mounting everything and getting the system up and running I looked at the time delayed alignment of acoustic centers:
I measured the distance of the cone to a point in space near listening position to confirm that the time alignment settings for each driver were set as accurately as possible. Also delayed ALL the smaller drivers by almost 6ms to match the delayed impulse response time of the big 10" drivers. The time alignment had a much bigger effect on the sound than I ever imagined it might.

The parabolic layout was kind of just a coincidence. But it made sense based on that layout to use the side where the drivers were closer toward the outside wall because it might have the least output. (It definitely sounded better that way but I can't say why for sure) I suppose part of the reason they sound nice _could_ be that nice arc which makes for kind of a natural alignment in the listening position. Coax would obviously provide better of access phase response.

Here are the spreadsheet simulations I ended up with (I won't bother showing the previous 19 examples I didn't use.     )

Baffle configuration #20 on axis


Baffle configuration #20 @ 30degrees on left


Baffle configuration #20 @ 30degrees right

[Rudolf]

Can you spell out that strategy in more detail for me? I would be happy to improve the setup.   

To clarify, in my understanding with my existing setup I get full output at bottom end, reducing to -15db at 96Hz with 12db/octave HP EQ ... and I apply LR 24db/octave crossover above 110Hz.

Above strategy of yours is the green curve - starting with the (light green) 15 dB shelf filter and moving over to the (dark green) cross over filter. The dark green line is including the part of the light green filter above 110 Hz:



I recommend another strategy: Start with a 12 dB/oct lowpass crossover at 30 Hz (or wherever you find appropriate). This would move everything along the lilac line to the level of the dark lilac line - compensating the dipole rolloff very well. After that you could apply the brown shelf filter at 310 Hz to subdue any remaining ondulations above 200 Hz up to another 15 dB.

Rudolf

[matevana]

After mounting everything and getting the system up and running I looked at the time delayed alignment of acoustic centers:
I measured the distance of the cone to a point in space near listening position to confirm that the time alignment settings for each driver were set as accurately as possible. Also delayed ALL the smaller drivers by almost 6ms to match the delayed impulse response time of the big 10" drivers. The time alignment had a much bigger effect on the sound than I ever imagined it might.

Good stuff! I'm using the 2nd set of outputs on my DEQ to drive the high side so I can delay the signal to approximate the arrival time of the low frequency side. In my case I converted MS to feet. This was a bit easier to visualize the distance the rear wave would have to travel to hit the rear wall and bounce back to the listening poisition. After doing so, the delay came out to 23ms and I halved that time as an approximation since the higher frequencies would also essentially follow the same path, but at a faster speed.  I wound up with 12ms and my listening position is about 10 ft in front of the speakers.

[joshsehn]

After doing so, the delay came out to 23ms and I halved that time as an approximation since the higher frequencies would also essentially follow the same path, but at a faster speed.  I wound up with 12ms and my listening position is about 10 ft in front of the speakers.

I think my comments may have been misleading because I think you are not thinking properly about how to use the delay. The delay should not be set to the time required for sound to reach your ears. Also the sound from the tweeter does not travel substantially faster than the sound from the woofer.

Rather, the delay you set on the tweeter should only be the tiny difference between the distance of the high end and low end.

For example: If your tweeter is physically offset 1cm in front of your midrange, you should set a delay of 1cm on the tweeter to make it line up with accoustic center of your midrange. This is how you set the delay regardless of how far away from the speaker you are positioned or how far away from the rear wall that your speaker is located.

To re-iterate/summarize; Set the delay on the closest speaker based on THE DIFFERENCE in the distance of the acoustic centers from your listening position. This should not change substantially unless you are very close to the speaker or have a separately located subwoofer.

Now if you are using a separate subwoofer and it is right beside you, and your tweeter speaker is 10ft away, then you would set the tweeter speaker to 10ft. (1ft = ~ 1ms)

This might help.
http://www.doctorproaudio.com/doctor/calculadores_en.htm

In addition to the above, the only other delay you might incorporate is the delay in the impulse response of a large driver. In my case the woofer takes 5ms to respond to an impulse so I delay the midrange by 5ms and tweeter by 5ms+acoustic center distance as per above.

[joshsehn]

Above strategy of yours is the green curve - starting with the (light green) 15 dB shelf filter and moving over to the (dark green) cross over filter. The dark green line is including the part of the light green filter above 110 Hz:

The top end of my green curve (the 12db/octave -15db cut) ENDS at 96hz instead of 150Hz. So in my estimation, it would start at about 40hz wouldn't it?

I recommend another strategy: Start with a 12 dB/oct lowpass crossover at 30 Hz (or wherever you find appropriate). This would move everything along the lilac line to the level of the dark lilac line - compensating the dipole rolloff very well. After that you could apply the brown shelf filter at 310 Hz to subdue any remaining ondulations above 200 Hz up to another 15 dB.

Thank you for the clarification. The strategy you recommend does seem to make some good sense and I can try that... but I always thought it would be preferable to use 24db/octave at required crossover frequency.

I will have to test this out and report the results here. Thanks for the suggestion.

[matevana]

I think my comments may have been misleading because I think you are not thinking properly about how to use the delay. The delay should not be set to the time required for sound to reach your ears. Also the sound from the tweeter does not travel substantially faster than the sound from the woofer.

Hmm, I've always used some delay to offset the higher frequencies, to prevent them from arriving at the listening position too early. This is done in pro-audio applications all the time, granted the distances traveled are typically greater. It is also one of the advantages of biamping. I find it to be particularly helpful with dipoles, since yhou are also dealing with a rear wave, that tends to compound the distances and make it more of an issue. It's my understand that various frequencies are affected by both time and temperature, and that some delay (usually people cannot detect delay less than 6ms) can be helpful in cleaning things up at the listening position.

[Rudolf]

The top end of my green curve (the 12db/octave -15db cut) ENDS at 96hz instead of 150Hz. So in my estimation, it would start at about 40hz wouldn't it?

I thought that 96 Hz was your input number for the DCX EQ. If it is the end of the shelf, both filters would merge at the right point, and your strategy would follow what SL shows in http://www.linkwitzlab.com/proto.htm#PW1. This would be valid of cause, but it would not allow to move the cross over to higher frequencies. 

... but I always thought it would be preferable to use 24db/octave at required crossover frequency.

From 200-500 Hz both of my (cascaded) filters sum up to 24 dB/oct  . Look at the blue 12dB/oct marks:



At 500 Hz we should be down about 30 dB. I don't care much about the return to 12 dB/oct above the 500 Hz mark.

Rudolf

[joshsehn]

I thought that 96 Hz was your input number for the DCX EQ. If it is the end of the shelf, both filters would merge at the right point, and your strategy would follow what SL shows in http://www.linkwitzlab.com/proto.htm#PW1. This would be valid of cause, but it would not allow to move the cross over to higher frequencies.

Indeed I have tried to follow Linkwitz and I think I'm as close as I could get with the gear I've got. I tried various tests using the filters you proposed and the results are so very nearly the same it is surprising. 

This test shows in green my traditional setup. ( -15db 12db/octave at 96Hz + 24db/octave LR @ 110Hz)

Both red and yellow use -15db 12db/octave shelving filter ending at 223Hz.
Red uses Butterworth 12db @ 50Hz crossover
Yellow uses LR 12db @ 66Hz crossover
 


Group delay, distortion and all the other characteristics were so similar the differences are not notable. Group delay and/or phase improvements are probably the only place that substantial improvements could be found. Cascaded filters and Phase Linear EQ could perhaps offer some improvements.

[joshsehn]

Hmm, I've always used some delay to offset the higher frequencies, to prevent them from arriving at the listening position too early. This is done in pro-audio applications all the time, granted the distances traveled are typically greater

You may be correct to delay the higher frequencies to prevent them from arriving to early... but this would be for reasons other than the temperature of the air as the speed of sound in air is nearly the same for all frequencies and amplitudes.

http://physics.info/sound/

This is done in pro-audio applications all the time, granted the distances traveled are typically greater. .

In pro audio, normally the delay would be applied to compensate for the multiple speakers at alternate distances. You are right the the speed of sound changes with temperature but high and low frequencies would be similarly affected by the temperature. The reason that the temperature is important however is because as you said, depending on the temperature, the time to travel a distance is affected and you may need to use alternate timers depending on the temperature.

To clarify, when I specified a fixed delay of 2212mm for my midrange this correlates to 6.16ms at 22degrees C.  If I change the setting of the room temperature to 25degrees C on my device, the distance offset needs to stay at exactly 2212mm but the time required to affect this offset is updated to 6.12ms.

Bottom line;
Temperature change affects the speed of sound.
All audible frequencies should be traveling at very nearly the same speed for any given temperature.

It is also one of the advantages of biamping

Right, with biamping you can introduce a delay on one channel to compensate for things you would not otherwise be able to... and so, you might be asking, why did I offset the tweeters than to get them to match the sound of the woofers?

There are some legitimate reasons to delay a driver. Crossovers and filters can add delay as can large speaker cone mass. Maybe there are other good reasons as well.

But to delay the tweeter to match the arrival of the rear wave of a dipole speaker I'm pretty sure would not be a good reason to delay your tweeter. Firstly, if you delay the tweeter by 6ms because your speakers are 3ft from the wall (3ft + 3ft = 6ft = ~6ms) the sound from the tweeter would arrive 6ms before the front firing soundwave of the midrange and that definitely can't be good. Secondly, the rear wave would arrive at the listening position at a substantially reduced volume from the original front firing sound so should care less about it than the primary sound anyway.

I hope my explanation was useful. I'll admit I'm don't know everything so perhaps someone will correct me I have have presented any inaccuracies.

But Linkwitz is a pro and has written much much more specific to this topic of the room reflections, the impact of a dipole etc.

Linkwitz confirms your statement to the effect that humans can't differentiate less than 6ms delay from the original sound. This is part of why you should keep the dipole speakers a minimum from the rear wall so the sound off the rear wall can be interpreted as a reflection rather than a smearing or distortion of the original sound.

http://www.linkwitzlab.com/stereo%20reproduction.htm

Hope this helps.

[matevana]

Very logically stated!  If we think about it, the difference between the initial and secondary wave arriving at the listening position is interpreted as a reverberation and is one of the characteristics so pleasing to dipole fans. In a perfect single driver application, I would agree that little should be done. But many common dipole configs include a small, wide range driver and a rather large (15-18") helper woofer with largish Qts. Due to a typically small (loosely controlling) magnet and spider combination, these drivers often react much more slowly than the smaller, wide range drivers. My first dipole experiment used this config and I found that delaying the small driver tidy'd things up quite a bit.

The Peerless XLS and XXLS use a large magnet, have a fairly controlling spider and a corresponding low Qts. I would agree that my experiments with time delay in this case have not been as beneficial, perhaps addressing some lesser phase anomolies, as opposed to a real need to compensate for the slowness of some larger drivers.   

[joshsehn]

But many common dipole configs include a small, wide range driver and a rather large (15-18") helper woofer with largish Qts

Agreed. And I do delay my entire 4way by 15ms to match the ultra delayed response of my big 15" ported sub.

Within my 4 way:
- I delay the 6.5" drivers by 6.16ms to match the delayed response of the 10" peerless XLS drivers
- I delay the 2" driver by an additional 36mm over the 6.5" to match the acoustic centerline
- I delay the 1" driver by an additional 26mm over the 6.5" to match the acoustic centerline

With this insight, you can could perhaps visualize my baffle/drivers and "see";
- my tweeter is flush with the baffle
- the 2" driver cone sticks out about 10mm in front of the baffle
- the 6.5" driver cones are recessed about 26mm behind baffle surface

my experiments with time delay in this case have not been as beneficial, perhaps addressing some lesser phase anomolies, as opposed to a real need to compensate for the slowness of some larger drivers. 

I suspect you may have introduced larger phase anomaly if you were delaying drivers to match the reflected rear sound wave instead of the direct sound wave.

[matevana]

Im just curious and would like to compare notes with you as far as your physical setup goes. Can you describe how the two dipoles are situated in your room and where your typical listening position is located. Without your seperate sub active, can you describe the roll-off as you walk across the room from one side to the other.

[Rudolf]

Within my 4 way: - I delay the 6.5" drivers by 6.16ms to match the delayed response of the 10" peerless XLS drivers - I delay the 2" driver by an additional 36mm over the 6.5" to match the acoustic centerline - I delay the 1" driver by an additional 26mm over the 6.5" to match the acoustic centerline

  From http://www.musicanddesign.com/Dipole-offset.html:  "In conclusion it is apparent that the conventional wisdom of correcting AC offsets using allpass delays or other means of introducing excess phase in the crossover, as applied to box type direct radiator speakers, is flawed when applied to speaker systems employing multiple dipole sources. This is because the offset of source 1 relative to source 2 reverses as we move from the front side of the dipole speaker to the rear side. If the offset is X cm from the front, it is -X cm from the rear. Thus the delay compensation, while the same magnitude, would have to be applied to the 2nd source rather than the 1st, or vice versa, situations which can not exist simultaneously. Application of a delay will alway introduce greater error to the rear response if it corrects the misalignment relative to the front of the speaker. The only correct solution is to physical align the AC of the two dipole sources."

[joshsehn]

Im just curious and would like to compare notes with you as far as your physical setup goes. Can you describe how the two dipoles are situated in your room and where your typical listening position is located.

My room is approximately 13ft X 18ft X 8ft.

The short wall is the front of the room.
The dipoles are situated along this front/short wall approximately 4ft away from it.
The outside edge of the baffles is approximately 2ft away from the long/side walls.
Of course, my listening position is along the center line of the room and approximately 5.5ft from the rear wall.

This leaves the speakers approximately 7ft apart and 8 feet away from my listening position.

Like this (The brown thing is my couch)

Without your seperate sub active, can you describe the roll-off as you walk across the room from one side to the other.

I don't think I could describe the difference... but here are some measurements:

Blue is center seat (Center of couch)
Green is 18" left of center seat (i.e. Left seat on couch)
Red is 36" left of center (i.e. sitting on arm of couch)

[joshsehn]

  From http://www.musicanddesign.com/Dipole-offset.html:  "In conclusion it is apparent that the conventional wisdom of correcting AC offsets using allpass delays or other means of introducing excess phase in the crossover, as applied to box type direct radiator speakers, is flawed when applied to speaker systems employing multiple dipole sources. This is because the offset of source 1 relative to source 2 reverses as we move from the front side of the dipole speaker to the rear side. If the offset is X cm from the front, it is -X cm from the rear. Thus the delay compensation, while the same magnitude, would have to be applied to the 2nd source rather than the 1st, or vice versa, situations which can not exist simultaneously. Application of a delay will alway introduce greater error to the rear response if it corrects the misalignment relative to the front of the speaker. The only correct solution is to physical align the AC of the two dipole sources."

I've studied his work and in the case of my speakers, I submit that his conclusion is not applicable to my application because it fails to consider a few things;

1. The delay I have set on my 6.5", 2" and 1" drivers by 6ms is to make them line up with the delayed response of the 10" subwoofer. This delay in the response of the 10" is the same regardless if listening from the front or the back.

2. The delay of my 2" and 1" are relative to each other and the rear baffle is the same as the front baffle. So, indeed the delay DOES make the acoustic centers line up. These 2 drivers are NOT open back. I have 2 more drivers the same mounted on the rear baffle.

3.  The delay of the 2" and 1" relative to the 6.5" is still useful because the speaker whether viewed from the front baffle or the rear baffle, the cones of my smaller drivers ARE physically forward of the larger driver.

Here is a side view picture that better illustrates the physical offset of these drivers.







I would agree with him 100% that his conclusions are valid with regard to multiple open cone speakers on an open baffle. (i.e. 8", 6" and 4" open back cones) in which case one should obviously physically align the acoustic centers rather than use time delay.

In conclusion, I am not delaying the 6.5" relative to the physical offset of the 10" but rather I am delaying the 6.5" to compensate for the delayed impulse response of the 10" driver.  If I had not introduced the 6ms delay on the 6.5" midrange, the offset when viewed from the rear would be 6ft - 8" and would still have over 5ft+4" mismatch.

Technically he is right, my delay on the 6.5 relative to the 10" is not quite as accurate on the rear as it would be on the front but I submit because a 6ms delay is equivalent to almost 6ft and I would rather the front response match at the front and not worry about the remaining 8" physical offset when viewed from the rear.

[Rudolf]

Many thanks for the side view and the in-depth explanation of your delays. Both helped me very well to understand what you are doing, and I (believe that I) can follow you completely.

I was really barking up the wrong tree.

Rudolf
www.dipolplus.de/thema11.php