I'm interested in feedback for my first serious loudspeaker design exercise. This took me a lot longer than expected and I learnt a lot in the process, really opening my eyes to the complexities involved. I have written up the "journey" of the design for 2 reasons, firstly to see if anyone can pick up on any mistakes of logic & fact used here, and secondly I have taken some effort to explain the reasons behind the various design choices so that this post can be used as a reference for others.
Design goals:
1) Integrate with my existing mains, Dunlavy Athena/Aletha. This means that I have to select the same drivers (Vifa P17 mid-woofer and Vifa D27 tweeter) and apply similar design principles that Dunlavy used (MTM configuration, time aligned & phase coherent).
2) I like the minimalist concepts of gainclone and nonos DAC, and want to apply that to speaker design. This should lead me to a single driver center speaker (eg. Tang-Band W3 817s – I have 16 of these waiting for a future project). But due to the need for the speaker to be voiced the same as the mains, I will apply the minimalist concept to the crossover:
- The fewer parts the better
- Keep the design simple
- Keep as many of the components shunting the signal to ground rather than in series with the speaker and the amplifier.
- Use high quality parts
3) Ensure the box fits within the asthetics of the custom cabinet, which limits the speaker shape to be longer than it is taller a maximum of 200mm high. Keep the box size below 50 litres.
4) Keep the costs reasonable
5) The low bass is not important, these speakers will be used with a receiver and a subwoofer and the receiver will be set to centre “small” so that below 100Hz is sent to the subwoofer. The design is only for HT use, for stereo use I only anticipate using the Dunlavy mains.
6) Just for HT use means that near enough is good enough. As long as the speaker sounds integrated with the rest, HT use does not demand critical listening and the typical movie soundtrack in DD or DTS is not an audiophile source. The design approach I will use here is to generalise a number of speaker & crossover design rules that will be applied to the design.
Design Assumptions:
1) Room boost assumed to be +3db from approx 200 hz below
http://members.ozemail.com.au/~joeras/sub_index.htm (section Oh What a Curve, Joe Rasmussen)
2) Baffle step diffraction -6db from the formula f3 = 115 / WB (where WB is the baffle width in meters), so with 300mm high centre speakers (the narrowest front dimension) = 400Hz
http://www.t-linespeakers.org/tech/bafflestep/ (Jason Neal)
3) Taking both the room boost and BSD into account, I will assume that I need +3db baffle step diffraction compensation at 300Hz (the average freq between room boost & BSD f3)
4) Using 2 mid-woofers in parallel will give a 3db electrical boost + a 1.5db acoustical boost for a total of +4.5db over its frequency range.
5) The Athena has a sensitivity of 90db, so the centre speaker should also have a sensitivity of 90db. The relative SPL between the centre channel and the mains is not as important as achieving a flat frequency response, as the HT receiver can compensate for speaker level differences if needed.
Box design:
I used WinISD and punched in the P17 parameters from the Vifa web site. I applied a +3db port assisted boost peaking at 70hz (using a 40 litre box and 60Hz tuning freq) to flatten the woofer response curve to +- 1db from 60Hz upwards, but this is at the expense of lower frequency extension as below 60Hz the response drops like a stone by -10db at 50Hz. However as the speaker will be crossed over to the subwoofer at 100Hz I figure a flat response above 60Hz would be ideal.
The only real compromise made in this design is the MTM arrangement. Due to the 1st order butterworth filter and the horizontal MTM alignment, there will be nulls in the horizontal plane, but I hope to minimize these irregularities by keeping the speakers as close together as possible. I don’t have a lot of choice here, due to the aesthetic consideration of using a wide but not tall speaker and I want to use MTM as the Dunlavys uses it and the SPL boost by using 2 woofers makes the crossover simple. The speaker arrangement will squeeze the tweeter and port between woofers.
Crossover:
For the crossover, I considered taking the althena apart and reverse engineering the crossover. However this was not possible without damaging the speaker and I wanted to go through the design exercise myself. I managed to dig up enough info about the Dunlavy speakers on the net to provide enough clues as to how to assemble the speaker so that it is voiced like the Dunlavy. Also as the Athena is a 3 way and the centre speaker will be a 2 way (due to space restrictions) I figure that the crossover will need to be different anyway.
In keeping with the simplistic concept, no over complicated tweaking of the frequency or phase response will be done. By using a simple 1st order crossover, the signal quality & phase / transient response will be preserved, and odd order crossovers are preferable for MTM configurations. 1st order is also simple to design so I’m less likely to stuff it up, and many published 1st order designs use these Vifa drivers.
1st order butterworth filter will be used for its simplicity, its minimum phase effect and reported clarity of reproduction. It also matches the crossover type of the Dunlavy main speakers, which will help to ensure the centre is voiced similarly to the mains. Dunlavy is reported (but not confirmed) to use a 1.8khz cross over frequency for the Athena, but as a 1st order crossover will not attenuate the signal to the tweeter sufficiently at its resonant frequency of 650 Hz (Fs quite low for a tweeter, so a good choice for a 1st order xover and no need for a notch filter) By looking at the SPL graphs of both speakers 2.5Khz looked like a better place for the cutover (as both woofer and tweeter start dropping off around this frequency). Even though the cutover frequency is different from the Dunlavy mains, it is not that different that it should cause audible differences. I used the crossover calculator in winISD which includes the speaker impedance when calculating values (beware that some sites on the ‘net with crossover designs do not compensate for the speaker impedance).
I want to use a series crossovers as they are reputed to sound better for 1st order crossovers (subjective I know but there seems to be some agreement from the different speaker builders who have posted on the ‘net) and are more forgiving to changing driver characterisitics. Dunlavy is also reputed to use a series crossover. Series crossover values 0.3mH and 12 uF. Links:
http://sound.westhost.com/parallel-series.htm#3.1 (Rod Elliot)
I printed the Vifa SPL graph for both the woofer and the tweeter on the same piece of paper (feeding the paper from the initial printed graph back into the printer) which worked well and allowed me to easily compare the frequency responses of both speakers on the same graph. I also found it easiest to design the crossover and box by studying the graphs and applying the empirical rules documented here.
BSD compensation & zobel. A zobel network for the woofer fits this design well. The role of the zobel is to flatten the woofer impedance at higher frequencies, which would otherwise reduce the effectiveness of the high pass filter on the woofer which is already shallow due to the use of a 1st order crossover. Another advantage of the zobel is that it will provide a form of BSC by reducing the higher frequencies of the woofer. In the assumption section I mentioned that I want to do a +3db bass boost above 300Hz to compensate for BSD (the average of the room bass boost and BSD loss). To do this, when equalizing the SPL outputs between the woofers and the tweeter, the woofer output (due to the use of 2 speakers in parallel) was 4.5db higher (see assumptions), so I applied -3db attenuation above 300Hz, which coincidently lowered the woofers SPL in its audible band to 90db, exactly the SPL I need to match the main speakers.
I’m going to use 8 ohms as my zobel resistor instead of 4 ohms (the parallel resistance of the two woofers) as I only want -3db of high frequency attenuation not -6db for BSC, although this will reduce the effectiveness of the zobel somewhat it should be fine as I am crossing the driver over at a low frequency. This gives me a zobel cap of approx 10uf, however I want the zobel to be effective at a lower frequency around 300 Hz to do BSC as well (which is still away from Fs), so I’ll use a 55uF instead.
There are a couple of good articles on BSC and zobels here:
http://www.trueaudio.com (John Murphy)
To equalize the tweeter, an L-pad variable resistor was added to drop the SPL by 2db to match the woofer and the mains output. It also means that I can adjust the tweeter levels later if needed (eg. If future speaker measurements show that this is needed).
Time & phase alignment. As I’m using a 1st order filter, I can simply offset the tweeter so that the acoustical centres of the speakers are aligned. This also means that the tweeter needs to be rebated back into the box, and follow the Dunlavy practice of covering tweeter surrounds with felt which should sit nicely in the rebate. The 1st order filter will help to ensure phase integrity.
Conclusion:
I know this is not the most scientific approach to speaker design, but the above applies the 80/20 rule where I think the design should be 80% correct and if it does not sound good or has poor integration with the other speakers, then I will perform actual speaker measurements and tweak the port & xover.
If anyone has any feedback on this design or can point out any improvements / errors I’d appreciate it.
Regards,
Dean
Design goals:
1) Integrate with my existing mains, Dunlavy Athena/Aletha. This means that I have to select the same drivers (Vifa P17 mid-woofer and Vifa D27 tweeter) and apply similar design principles that Dunlavy used (MTM configuration, time aligned & phase coherent).
2) I like the minimalist concepts of gainclone and nonos DAC, and want to apply that to speaker design. This should lead me to a single driver center speaker (eg. Tang-Band W3 817s – I have 16 of these waiting for a future project). But due to the need for the speaker to be voiced the same as the mains, I will apply the minimalist concept to the crossover:
- The fewer parts the better
- Keep the design simple
- Keep as many of the components shunting the signal to ground rather than in series with the speaker and the amplifier.
- Use high quality parts
3) Ensure the box fits within the asthetics of the custom cabinet, which limits the speaker shape to be longer than it is taller a maximum of 200mm high. Keep the box size below 50 litres.
4) Keep the costs reasonable
5) The low bass is not important, these speakers will be used with a receiver and a subwoofer and the receiver will be set to centre “small” so that below 100Hz is sent to the subwoofer. The design is only for HT use, for stereo use I only anticipate using the Dunlavy mains.
6) Just for HT use means that near enough is good enough. As long as the speaker sounds integrated with the rest, HT use does not demand critical listening and the typical movie soundtrack in DD or DTS is not an audiophile source. The design approach I will use here is to generalise a number of speaker & crossover design rules that will be applied to the design.
Design Assumptions:
1) Room boost assumed to be +3db from approx 200 hz below
http://members.ozemail.com.au/~joeras/sub_index.htm (section Oh What a Curve, Joe Rasmussen)
2) Baffle step diffraction -6db from the formula f3 = 115 / WB (where WB is the baffle width in meters), so with 300mm high centre speakers (the narrowest front dimension) = 400Hz
http://www.t-linespeakers.org/tech/bafflestep/ (Jason Neal)
3) Taking both the room boost and BSD into account, I will assume that I need +3db baffle step diffraction compensation at 300Hz (the average freq between room boost & BSD f3)
4) Using 2 mid-woofers in parallel will give a 3db electrical boost + a 1.5db acoustical boost for a total of +4.5db over its frequency range.
5) The Athena has a sensitivity of 90db, so the centre speaker should also have a sensitivity of 90db. The relative SPL between the centre channel and the mains is not as important as achieving a flat frequency response, as the HT receiver can compensate for speaker level differences if needed.
Box design:
I used WinISD and punched in the P17 parameters from the Vifa web site. I applied a +3db port assisted boost peaking at 70hz (using a 40 litre box and 60Hz tuning freq) to flatten the woofer response curve to +- 1db from 60Hz upwards, but this is at the expense of lower frequency extension as below 60Hz the response drops like a stone by -10db at 50Hz. However as the speaker will be crossed over to the subwoofer at 100Hz I figure a flat response above 60Hz would be ideal.
The only real compromise made in this design is the MTM arrangement. Due to the 1st order butterworth filter and the horizontal MTM alignment, there will be nulls in the horizontal plane, but I hope to minimize these irregularities by keeping the speakers as close together as possible. I don’t have a lot of choice here, due to the aesthetic consideration of using a wide but not tall speaker and I want to use MTM as the Dunlavys uses it and the SPL boost by using 2 woofers makes the crossover simple. The speaker arrangement will squeeze the tweeter and port between woofers.
Crossover:
For the crossover, I considered taking the althena apart and reverse engineering the crossover. However this was not possible without damaging the speaker and I wanted to go through the design exercise myself. I managed to dig up enough info about the Dunlavy speakers on the net to provide enough clues as to how to assemble the speaker so that it is voiced like the Dunlavy. Also as the Athena is a 3 way and the centre speaker will be a 2 way (due to space restrictions) I figure that the crossover will need to be different anyway.
In keeping with the simplistic concept, no over complicated tweaking of the frequency or phase response will be done. By using a simple 1st order crossover, the signal quality & phase / transient response will be preserved, and odd order crossovers are preferable for MTM configurations. 1st order is also simple to design so I’m less likely to stuff it up, and many published 1st order designs use these Vifa drivers.
1st order butterworth filter will be used for its simplicity, its minimum phase effect and reported clarity of reproduction. It also matches the crossover type of the Dunlavy main speakers, which will help to ensure the centre is voiced similarly to the mains. Dunlavy is reported (but not confirmed) to use a 1.8khz cross over frequency for the Athena, but as a 1st order crossover will not attenuate the signal to the tweeter sufficiently at its resonant frequency of 650 Hz (Fs quite low for a tweeter, so a good choice for a 1st order xover and no need for a notch filter) By looking at the SPL graphs of both speakers 2.5Khz looked like a better place for the cutover (as both woofer and tweeter start dropping off around this frequency). Even though the cutover frequency is different from the Dunlavy mains, it is not that different that it should cause audible differences. I used the crossover calculator in winISD which includes the speaker impedance when calculating values (beware that some sites on the ‘net with crossover designs do not compensate for the speaker impedance).
I want to use a series crossovers as they are reputed to sound better for 1st order crossovers (subjective I know but there seems to be some agreement from the different speaker builders who have posted on the ‘net) and are more forgiving to changing driver characterisitics. Dunlavy is also reputed to use a series crossover. Series crossover values 0.3mH and 12 uF. Links:
http://sound.westhost.com/parallel-series.htm#3.1 (Rod Elliot)
I printed the Vifa SPL graph for both the woofer and the tweeter on the same piece of paper (feeding the paper from the initial printed graph back into the printer) which worked well and allowed me to easily compare the frequency responses of both speakers on the same graph. I also found it easiest to design the crossover and box by studying the graphs and applying the empirical rules documented here.
BSD compensation & zobel. A zobel network for the woofer fits this design well. The role of the zobel is to flatten the woofer impedance at higher frequencies, which would otherwise reduce the effectiveness of the high pass filter on the woofer which is already shallow due to the use of a 1st order crossover. Another advantage of the zobel is that it will provide a form of BSC by reducing the higher frequencies of the woofer. In the assumption section I mentioned that I want to do a +3db bass boost above 300Hz to compensate for BSD (the average of the room bass boost and BSD loss). To do this, when equalizing the SPL outputs between the woofers and the tweeter, the woofer output (due to the use of 2 speakers in parallel) was 4.5db higher (see assumptions), so I applied -3db attenuation above 300Hz, which coincidently lowered the woofers SPL in its audible band to 90db, exactly the SPL I need to match the main speakers.
I’m going to use 8 ohms as my zobel resistor instead of 4 ohms (the parallel resistance of the two woofers) as I only want -3db of high frequency attenuation not -6db for BSC, although this will reduce the effectiveness of the zobel somewhat it should be fine as I am crossing the driver over at a low frequency. This gives me a zobel cap of approx 10uf, however I want the zobel to be effective at a lower frequency around 300 Hz to do BSC as well (which is still away from Fs), so I’ll use a 55uF instead.
There are a couple of good articles on BSC and zobels here:
http://www.trueaudio.com (John Murphy)
To equalize the tweeter, an L-pad variable resistor was added to drop the SPL by 2db to match the woofer and the mains output. It also means that I can adjust the tweeter levels later if needed (eg. If future speaker measurements show that this is needed).
Time & phase alignment. As I’m using a 1st order filter, I can simply offset the tweeter so that the acoustical centres of the speakers are aligned. This also means that the tweeter needs to be rebated back into the box, and follow the Dunlavy practice of covering tweeter surrounds with felt which should sit nicely in the rebate. The 1st order filter will help to ensure phase integrity.
Conclusion:
I know this is not the most scientific approach to speaker design, but the above applies the 80/20 rule where I think the design should be 80% correct and if it does not sound good or has poor integration with the other speakers, then I will perform actual speaker measurements and tweak the port & xover.
If anyone has any feedback on this design or can point out any improvements / errors I’d appreciate it.
Regards,
Dean
Hi,
Just some points :
Whatever (dis?) advantages vertical MTM's have for stereo use
its a very poor arrangement for the centre centre channel unless
you can keep the MTM vertical. its fashionable becuase it looks
"right" to the man on the street but horizontal dispersion is poor.
Acoustical boost is 3dB giving a total of 6dB.
Note that the baffle step of a centre speaker is heavily influenced
by its immediate surroundings, this needs to be taken into account.
All 1st order filters are the same, there are no different 1st order
alignments, you need at least 2nd order to have a Q of the filter.
IMO the choice of 1st order slopes and dual mids with the tweeter
will cause intractable problems with horizontal dispersion.
🙂 sreten.
Just some points :
Whatever (dis?) advantages vertical MTM's have for stereo use
its a very poor arrangement for the centre centre channel unless
you can keep the MTM vertical. its fashionable becuase it looks
"right" to the man on the street but horizontal dispersion is poor.
Acoustical boost is 3dB giving a total of 6dB.
Note that the baffle step of a centre speaker is heavily influenced
by its immediate surroundings, this needs to be taken into account.
All 1st order filters are the same, there are no different 1st order
alignments, you need at least 2nd order to have a Q of the filter.
IMO the choice of 1st order slopes and dual mids with the tweeter
will cause intractable problems with horizontal dispersion.
🙂 sreten.
Hi Sreten, thanks for the feedback.
Do you have any pointers to how the baffle step is affected by immediate surroundings?
Regarding acoustical boost, is +3db best case? I was assuming that in practice you would not see as much boost.
To minimise the horizontal dispersion I was hoping that using the woofers as close as possible together, have the tweeter above the centre line between the woofers and use a 1st order xover would reduce horizontal lobing effects.
I was also thinking of sloping back the front baffle on either side where the mid-woofer is mounted, so that the tweeter & port are on the straight part of the baffle, and the woofers are angled to the sides left & right. Although this would look good, I suspect this would make the lobing worse.
Another alternative is to use a 2.5 arrangement where the second woofer rolls off above 300Hz reducing the lobing in the important mid frequencies. However I'm not sure if this affect the voicing with the mains, and the zobel cutover will need adjusting.
Another question about the zobel, if I'm using a 1st order filter with two woofers in parallel (for 4 ohms total impedance), and the zobel frequency is reduced to a frequency where the driver impedance drops below 8 ohms, will the total impedance at that frequency be so low that it risks damaging the amplifier? Is it wise to use the zobel to compensate for BSD as well?
Any other ideas for a MTM centre? Any feedback from anyone who has successfully built a MTM centre and minimised the horizontal lobing effects?
Regards,
Dean
Do you have any pointers to how the baffle step is affected by immediate surroundings?
Regarding acoustical boost, is +3db best case? I was assuming that in practice you would not see as much boost.
To minimise the horizontal dispersion I was hoping that using the woofers as close as possible together, have the tweeter above the centre line between the woofers and use a 1st order xover would reduce horizontal lobing effects.
I was also thinking of sloping back the front baffle on either side where the mid-woofer is mounted, so that the tweeter & port are on the straight part of the baffle, and the woofers are angled to the sides left & right. Although this would look good, I suspect this would make the lobing worse.
Another alternative is to use a 2.5 arrangement where the second woofer rolls off above 300Hz reducing the lobing in the important mid frequencies. However I'm not sure if this affect the voicing with the mains, and the zobel cutover will need adjusting.
Another question about the zobel, if I'm using a 1st order filter with two woofers in parallel (for 4 ohms total impedance), and the zobel frequency is reduced to a frequency where the driver impedance drops below 8 ohms, will the total impedance at that frequency be so low that it risks damaging the amplifier? Is it wise to use the zobel to compensate for BSD as well?
Any other ideas for a MTM centre? Any feedback from anyone who has successfully built a MTM centre and minimised the horizontal lobing effects?
Regards,
Dean
Hi DD,
I'm no expert on HT design , just plain stereo speakers, but it
seems to me most centre designs are crude to say the least.
The effective baffle step will be affected by a large screen
immediately above the unit, the rest of the housing ?
Using the Zobel for BSD won't work well at all as far as I know,
but I'm no expert on series filters, it certainly wouldn't work for
parrallel filters.
Angling the mid/bass units could help matters as it reduces
the off axis interference between those two units. I doubt
it would make the lobing worse, I'd say probably better.
I'd be tempted to use something like this for the Centre and Rears :
http://home.hetnet.nl/~geenius/USB.html
🙂 sreten.
I'm no expert on HT design , just plain stereo speakers, but it
seems to me most centre designs are crude to say the least.
The effective baffle step will be affected by a large screen
immediately above the unit, the rest of the housing ?
Using the Zobel for BSD won't work well at all as far as I know,
but I'm no expert on series filters, it certainly wouldn't work for
parrallel filters.
Angling the mid/bass units could help matters as it reduces
the off axis interference between those two units. I doubt
it would make the lobing worse, I'd say probably better.
I'd be tempted to use something like this for the Centre and Rears :
http://home.hetnet.nl/~geenius/USB.html
🙂 sreten.
Thanks Sreten.
The zobel topology in parallel with a woofer reduces the rising speaker impedance seen by the amplifier due to the capacitor falling impedance with frequency. I would have thought that this effect also acts to attenuate the output of the woofer at higher frequencys so setting the zobel frequency around 300Hz would act as BSD compensation?
Can anyone recommend a good (freeware) crossover modelling program so that I can simulate this better?
I'll try the mid/bass angling. I think I'm at the stage where if the theoretical design is OK, further tweaking should be done by listening/measuring.
Regards,
Dean
The zobel topology in parallel with a woofer reduces the rising speaker impedance seen by the amplifier due to the capacitor falling impedance with frequency. I would have thought that this effect also acts to attenuate the output of the woofer at higher frequencys so setting the zobel frequency around 300Hz would act as BSD compensation?
Can anyone recommend a good (freeware) crossover modelling program so that I can simulate this better?
I'll try the mid/bass angling. I think I'm at the stage where if the theoretical design is OK, further tweaking should be done by listening/measuring.
Regards,
Dean
why not measure first?
Isnt it a practical approach to build a quick box for 2 midbass units and a vertical mid tweeter arrangement approaching your main ideas, and measure all parametres in final room location?
Then you could be much better informed of what is needed.
Isnt it a practical approach to build a quick box for 2 midbass units and a vertical mid tweeter arrangement approaching your main ideas, and measure all parametres in final room location?
Then you could be much better informed of what is needed.
Salas,
I agree. I wanted to get the basics in place first before tuning in room.
Any ideas about the zobel compensation, using it for BSD compensation with a series 1st order crossover & if the zobel impedance with 2 parallel 8 ohm woofers will drop the total impedance much below 4 ohms?
Regards,
Dean
I agree. I wanted to get the basics in place first before tuning in room.
Any ideas about the zobel compensation, using it for BSD compensation with a series 1st order crossover & if the zobel impedance with 2 parallel 8 ohm woofers will drop the total impedance much below 4 ohms?
Regards,
Dean
deandob said:
A zobel tuned very low will act as a shunt, and this way will load the amp much more at low frequencies (-> the impedance will be lower)
a classical BSC (L//R) will add an impedance at high frequencies, and won't surcharge the amp
You could also do this
-pad down the tweeter
-use a larger L value on the woofer, but the correct value for the tweeter's cap
play with the R and L values with speakerworkshop to tune this
-pad down the tweeter
-use a larger L value on the woofer, but the correct value for the tweeter's cap
play with the R and L values with speakerworkshop to tune this
Re: 8 Ohm is better...
Or 2 16R in parallel
series woofers have a huge impedance peak
salas said:
Or 2 16R in parallel
series woofers have a huge impedance peak
Great replies, thanks all.
I'll play with speakerworkshop (great app BTW) to fine tune the design then I'll build and tune to ear.
Regards,
Dean
I'll play with speakerworkshop (great app BTW) to fine tune the design then I'll build and tune to ear.
Regards,
Dean
yes 2x16 Ohm is even better, but I skipped that due to limited models... PHL comes to mind though...
- Status
- Not open for further replies.