Re: Correcting a few misconceptions...
Not without adding further components to the 1st order.
Adjusting the inductor value will move the c/o frequency
for both units.
Unlike the parallel case.
(The butterworth was shown as an example of 2nd order
series filters, it in no way implied Q=0.707 must be used)
Good point about impedance seen by the driver in the cuttoff band.
Interestingly the opposite applies to second order series networks
and the same applies to 2nd order parallel networks.
🙂 sreten.
Andy Graddon said:
Not without adding further components to the 1st order.
Adjusting the inductor value will move the c/o frequency
for both units.
Unlike the parallel case.
(The butterworth was shown as an example of 2nd order
series filters, it in no way implied Q=0.707 must be used)
Good point about impedance seen by the driver in the cuttoff band.
Interestingly the opposite applies to second order series networks
and the same applies to 2nd order parallel networks.
🙂 sreten.
Svante said:
Hey..... leave us rabbits out of this. Well this rabbitz uses series crossovers.
For the small number of speakers I design and build I have found a can get a more cohesive sound with a series than a parallel crossover and have tested plenty of both against each other.
I choose my drivers carefully as I know I can't correct any problems or nasties, get the impedance of the drivers at the crossover frequency, work out the best place to crossover, put it into Andy's spreadsheet, and this gives me the starting values. With only 3 components testing becomes very easy and if you make a little error it's only going to alter the crossover point and slope but the drivers will still blend together. I just keep going until I get the sound I'm after and these days that happens after about 3 - 5 combinations. After over 30 years in Hi Fi I know what good sound is and it's strange that even with hearing loss due to age, I can pick out the differences in equipment better now than ever before.
Do I totally understand series xo's....no. Do I know how tweak them..... yes. Am I happy with them.... yes. Do they sound better... IMO yes for the speakers I did the testing (over 20 crossovers) - could be due to that they are easier to tweak and more forgiving. sreten could be right that they are good choice for the amateur.
The only problem I have is finding inductors with a low DCR at a reasonable price in Australia. Have started buying mine from Parts Express in USA.
Re: Re: Correcting a few misconceptions...
But.... by adjusting the coil and inductor by the same multiple but opposite direction, you can hold the x-o point, and still form a widish dip or rise in the response that can be used to counteract "minor" problems with a woofer's top end. Since you have been VERY careful choosing your woofer, there will only be minor problems that may need correcting..
If I woofer has major peaks, break-ups etc it shouldn't have been chosen for a 1st order set-up anyway !!!
it should also be noted that if you do form a dip this way , using say zeta =1.2, and then trim the tweeter down a bit as well using L-padding and offset impedances, you can actually form a sort of psuedo- baffle compensation without any extra components.
I have a sort of "elastic string" idea that helps understand the effects of altering different components.... I really must put it into an understandable form some day !!! maybe, perhaps.....
sreten said:
But.... by adjusting the coil and inductor by the same multiple but opposite direction, you can hold the x-o point, and still form a widish dip or rise in the response that can be used to counteract "minor" problems with a woofer's top end. Since you have been VERY careful choosing your woofer, there will only be minor problems that may need correcting..
If I woofer has major peaks, break-ups etc it shouldn't have been chosen for a 1st order set-up anyway !!!
it should also be noted that if you do form a dip this way , using say zeta =1.2, and then trim the tweeter down a bit as well using L-padding and offset impedances, you can actually form a sort of psuedo- baffle compensation without any extra components.
I have a sort of "elastic string" idea that helps understand the effects of altering different components.... I really must put it into an understandable form some day !!! maybe, perhaps.....
OK, folks, I would like really to benefit from the series-XO-specialists here as I made my first steps with series XO and they have been very disappointing.
Here is the description of the project: Line-Source, two-ways, with 18 4" Etons (HEX/300/25/4 see www.intertechnik.com) per side and a R45 Ribbon from newform (http://www.newformresearch.com/).
As you can see I wanted to create here my personal referemce system with all the goodies I learned to appreciate over the last years: FAst and smooth ribbons, woofers which are fast and stiff enough for the ribbon, closed enclosure not to mess up the step-response, 6DB-crossover fully impedance corrected and a very low cross-over-frequence.
So far I only have experiences with parallel cross-overs. Nevertheless having heard all the goodies about serries XO, we started to play with series ones. So far without success. Clearly, dynamics are great and everything is very smooth, but I have as well quite a lack of 3D-information, less space, les imaging etc. Interesting enough Doede Douma reports the same thing with his speaker (www.doede.de) even though his concept is very different. His explanation were too many interferences between the tweeter and the woofer when using 6DB in serial cross-overs. His solution was then to use 12 db to get the focus right.
Well...I don't think that this makes sense for my project. I have clearly speakers chosen which are born to be used with a 6 DB-cross-over. So, what causes the lack of 3D-information ? Or to ask the question the other way round: What would be your recommendation for the values ? The ribbon has 5 Ohms and the woofer as well. Cross-over-point should be 1000Hz.
Thanks & Best Regards
Here is the description of the project: Line-Source, two-ways, with 18 4" Etons (HEX/300/25/4 see www.intertechnik.com) per side and a R45 Ribbon from newform (http://www.newformresearch.com/).
As you can see I wanted to create here my personal referemce system with all the goodies I learned to appreciate over the last years: FAst and smooth ribbons, woofers which are fast and stiff enough for the ribbon, closed enclosure not to mess up the step-response, 6DB-crossover fully impedance corrected and a very low cross-over-frequence.
So far I only have experiences with parallel cross-overs. Nevertheless having heard all the goodies about serries XO, we started to play with series ones. So far without success. Clearly, dynamics are great and everything is very smooth, but I have as well quite a lack of 3D-information, less space, les imaging etc. Interesting enough Doede Douma reports the same thing with his speaker (www.doede.de) even though his concept is very different. His explanation were too many interferences between the tweeter and the woofer when using 6DB in serial cross-overs. His solution was then to use 12 db to get the focus right.
Well...I don't think that this makes sense for my project. I have clearly speakers chosen which are born to be used with a 6 DB-cross-over. So, what causes the lack of 3D-information ? Or to ask the question the other way round: What would be your recommendation for the values ? The ribbon has 5 Ohms and the woofer as well. Cross-over-point should be 1000Hz.
Thanks & Best Regards
Correction: THe woofers have right now 7Ohm, and have more sensitivity than the ribbon, so need another 3R3 in series to come down to the tweeter-level, so have an impedance of 10,3 Ohm.
The values listened to where (Z=1,1/1,2):
- 2100 Hz cross-frequence: 0,82mH and 14uF
- 1000 Hz cross-frequence: 1,8mH and 20uF (clearly better option)
BEst Regards
The values listened to where (Z=1,1/1,2):
- 2100 Hz cross-frequence: 0,82mH and 14uF
- 1000 Hz cross-frequence: 1,8mH and 20uF (clearly better option)
BEst Regards
Hard to say without more information, especially the relative
sensistivities of the two arrays but :
But with 6dB/actave at 1kHz , I'd say the bass units need to be
rolled off early depending on baffle width for baffle step correction.
(BSD is a step due to the drivers changing from 2pi to pi radiation)
See :
http://sound.westhost.com/bafflestep.htm
I'd say you need a similar pre-eq circuit to use a series crossover.
The series values for 5 ohms are 32uF and 0.8mH.
But I'd hazard a guess that 18 bass drivers have too much
sensitivity, the ribbon is 91dB, each driver is 88dB, 16 wired
in series parallel sections will end up at 94dB.
Not sure how you've wired up 18 x 6R to get 5R.
Personally I'd go for active biamping with passive line level c/o
and BSC.
edit :
you can use up 3dB of BSC by paralleling the 3.3R with an
inductor, then add another 3dB with a passive network,
good idea to make this adjustable as shown.
🙂 sreten.
sensistivities of the two arrays but :
But with 6dB/actave at 1kHz , I'd say the bass units need to be
rolled off early depending on baffle width for baffle step correction.
(BSD is a step due to the drivers changing from 2pi to pi radiation)
See :
http://sound.westhost.com/bafflestep.htm
I'd say you need a similar pre-eq circuit to use a series crossover.
The series values for 5 ohms are 32uF and 0.8mH.
But I'd hazard a guess that 18 bass drivers have too much
sensitivity, the ribbon is 91dB, each driver is 88dB, 16 wired
in series parallel sections will end up at 94dB.
Not sure how you've wired up 18 x 6R to get 5R.
Personally I'd go for active biamping with passive line level c/o
and BSC.
edit :
you can use up 3dB of BSC by paralleling the 3.3R with an
inductor, then add another 3dB with a passive network,
good idea to make this adjustable as shown.
🙂 sreten.
These are good hints, I will read through it. And once I understand more of it will com back with my questions.
Regarding the values: Again correction:
- 1000 Hz cross-frequence: 1,8mH and 29uF (clearly better option)
Why is 1,8mH wrong when the woofer has 7 ohm it self + 3,3 Ohm. The 7 Ohm of the woofer came from a combination of series-parallel connections of the woofers. The 94 DB are a good estimate, I would estimate this in that range as well.
From the theory of line sources I calculated the speakers in a way that you should have a near field within the 4,5m from the speaker and than go from the 3db to the 6 db scenario.
Your suggestion on activating it are welcome, I guess that might be the ultimate solution. I heard once the Dali Megaline which is a Line Source as well with acitve cross-over - absolutely stunning. Was able to reproduce small events small and big events really big ( so not each recording sounds big like with an Martin Logan ESL). Unfortunately the Ribbons of the MEgaline are not for sale, they are fantastic.
Regarding the values: Again correction:
- 1000 Hz cross-frequence: 1,8mH and 29uF (clearly better option)
Why is 1,8mH wrong when the woofer has 7 ohm it self + 3,3 Ohm. The 7 Ohm of the woofer came from a combination of series-parallel connections of the woofers. The 94 DB are a good estimate, I would estimate this in that range as well.
From the theory of line sources I calculated the speakers in a way that you should have a near field within the 4,5m from the speaker and than go from the 3db to the 6 db scenario.
Your suggestion on activating it are welcome, I guess that might be the ultimate solution. I heard once the Dali Megaline which is a Line Source as well with acitve cross-over - absolutely stunning. Was able to reproduce small events small and big events really big ( so not each recording sounds big like with an Martin Logan ESL). Unfortunately the Ribbons of the MEgaline are not for sale, they are fantastic.
OK, have read it. As my baffle width is only 15cm, the effect might be not extreme. If I understand the concept of the equalisation right, I would make the woofers less sensitive in the cross-over area with this.
Now, just thinking about today's overall sound, I have the fear that this will add to much bass. I have already a significant warm and full sounding speaker. Surprisingly though as this is not the sound auf a single unit usually. As well it sounds much deeper than the resonance frequence of the drivers would suggest. So, there are already some effects by stappling 18 units above each other that makes your bass much louder and deeper than you would have in non-near-field scenario. Like in PA-Systems where you receive a lower bass-frequence by stappling units onto each other.
Now, just thinking about today's overall sound, I have the fear that this will add to much bass. I have already a significant warm and full sounding speaker. Surprisingly though as this is not the sound auf a single unit usually. As well it sounds much deeper than the resonance frequence of the drivers would suggest. So, there are already some effects by stappling 18 units above each other that makes your bass much louder and deeper than you would have in non-near-field scenario. Like in PA-Systems where you receive a lower bass-frequence by stappling units onto each other.
sreten said:
Errh... How did you end up with 6 dB sensitivity change for 16 drivers? The efficiency will go up by a factor 16 due to the acoustic coupling, and since the impedance is the same in the series-parallel (4x4) configuration, so will the sensitivity. 16 times the power is 12 dB, so sensitivity will be 88+12=100dB.
Or did I miss someting?
Agree that biamping would be good in this case!
Blitz said:
The line source theory applies to vertical dispersion.
By my estimates a 6ft line source is good down to 60Hz or so,
before you get a falloff as it becomes hemispherical, but said
fall off should match room gain quite well.
Baffle step applies to horizontal dimensions and I'm trying to get
my head around whether the step is 3dB or 6dB for a line source.
I suspect it may only be 3dB which means you have the right
combination of drivers whether by design or accident.
For BSC paralleling the 3.3R with an inductor should be enough.
You may need to add a Zobel to the bass units to get this to
work properly and this will help with implementing the crossover.
I'm not really familiar with series crossovers and different
impedance drivers, though I suspect the way you dimension
the components is counter-intuitive compared to parallel.
🙂 sreten.
note that at the designed BSC frequency Zl = 3.3R||7R.
sreten said:
Baffle step must be 6dB for line sources as well. I just realised, though that response must be as bad as in the circular baffle case, since there only is *one* distance between the line source and the edge. Big peak of +9.5dB at width=wavelength and then multiple comb-filter like drops down to 0dB.
Ok, the inductor over the resistor have to be tried, what would be the correct value in your eyes ?
I have already a Zobel in there.
What makes me think are the following quote from Jon Risch:
"BTW, (this is for AWP and Rick) it is not a requirement that each driver be identical in impedance, etc. For instance, if the tweeter were 4 ohm, and the woofer 8 ohm, you figure the inductor for an 8 ohm load, and the cap for a 4 ohm load. I know this seems counterintuitive, but the signal for the woofer really is passing through the inductor, and the signal for the tweeter really is passing through the cap. I have verified this via measurement, so I know it works OK. The proper shape and phase responses are maintained, and Zeta can still be adjusted, with regard for the proper load R.
... as I have stated before, and even presented a paper on (AES [reprint #4784, "Loudspeaker Crossover Lookback Impedance", the fact that what the driver "sees" looking back into the crossover network on a series network is a low impedance outside the passband is a big plus. This helps control driver resonances, as well as prevent any out of band peaking from occuring. This is also why a series 2nd order is a not such a good idea, as the second set of components are now in series right at the dirvers, and "looking back" into the network the driver "sees" a high impedance outside the passband. A more conventional parallel network will retain the advantage of low lookback impedance due to the shunt elements being directly across the drivers."
So, I understood two points:
- I guess the domensions fo the components are right.
- If we want to go with series cross-overs, first order is where you want to be (did that).
In conclusion this tells me that I am doing the stuff already as suggest by those who understand series cross-overs much better than I do, but the result is not good in imaging / 3D / space etc.
One factor which has to be taken into accountas well might be: Clearly the tweeter and the woofer are 15 cm apart from each other horizontly, what does this do to our filter ?
I have already a Zobel in there.
What makes me think are the following quote from Jon Risch:
"BTW, (this is for AWP and Rick) it is not a requirement that each driver be identical in impedance, etc. For instance, if the tweeter were 4 ohm, and the woofer 8 ohm, you figure the inductor for an 8 ohm load, and the cap for a 4 ohm load. I know this seems counterintuitive, but the signal for the woofer really is passing through the inductor, and the signal for the tweeter really is passing through the cap. I have verified this via measurement, so I know it works OK. The proper shape and phase responses are maintained, and Zeta can still be adjusted, with regard for the proper load R.
... as I have stated before, and even presented a paper on (AES [reprint #4784, "Loudspeaker Crossover Lookback Impedance", the fact that what the driver "sees" looking back into the crossover network on a series network is a low impedance outside the passband is a big plus. This helps control driver resonances, as well as prevent any out of band peaking from occuring. This is also why a series 2nd order is a not such a good idea, as the second set of components are now in series right at the dirvers, and "looking back" into the network the driver "sees" a high impedance outside the passband. A more conventional parallel network will retain the advantage of low lookback impedance due to the shunt elements being directly across the drivers."
So, I understood two points:
- I guess the domensions fo the components are right.
- If we want to go with series cross-overs, first order is where you want to be (did that).
In conclusion this tells me that I am doing the stuff already as suggest by those who understand series cross-overs much better than I do, but the result is not good in imaging / 3D / space etc.
One factor which has to be taken into accountas well might be: Clearly the tweeter and the woofer are 15 cm apart from each other horizontly, what does this do to our filter ?
Interesting last comment you made. Is this true as well in my setting: I have a twoer with the woofers which has 15cm width and a separate enclusore of the ribbing standing next to the woofer (When you look at the newform R45 you see that it comes with its own enclosure). So, I have two enclosures standing side by side.
Svante said:
Hmmm..... seem to have got my own logic twisted, what
you get for not writing it down and looking at it, you are
of course correct.
(I did think effiency would far too high for the tweeter
until I convinced myself otherwise)
So how to wire up 18 units (!?) for the right ballpark ?
We know efficiency is +12dB, and we want +6dB to allow
3dB of BSC ? Therefore we need 4x nominal impedance.
Easily done with 16 units, 4's in series/parallel all wired in series.
(And 8's in series and then paralled)
Hmmm.... bi-amping is the way to go with plenty of juice for
the tweeter, and a more normal load for the bass amplifier.
🙂 sreten.
Svante said:
I'm not sure. A normal box goes from 180 to 360 degrees
dispersion in both vertical and horizontal planes, bleeding
around the baffle width for the vertical plane.
A line source only changes in one plane - but -
the solid angle still doubles - so you are correct.
I'm not having a good day.......
so for a passive c/o with BSC we need + 9dB over a single unit,
rather than the +12dB we have.
So we need a nominal impedance that is double for the array.
With 18 drivers 6's in series then the 3 sets in parrallel ?
Should give +9dB ? the BSC resistor = twice nominal.
The BSC inductor Z = nominal for the baffle frequency ?
Alternatively for a parallel crossover you can omit the BSC
resistor, use a BSC inductor with Z = 2xnominal for the baffle
frequency, after the baffle step it will rolloff and then match
the treble c/o to this.
🙂 sreten.
My first question is still wether the impression that imaging is not that good is a BCR problem or not ?
If Yes:
Well, let see how far we could come with a passive solution first. If I want a BCR of 9 DB, the ratio of the series resistor to the impedance of the woofer should be 3:2, right ? So 3 parts of the voltage would be consumed completly over the resistor and 2 parts over by the woofer. Right now I have it the other way round: 3,3Ohm: 7 Ohm, so this needs to be changed, fortunately paralleling more devices gives me as well more sensitivity and a lower DCR as the same time.
With a baffle width of 15cm, f3 would be 766Hz...So what would be the right component values for this ?
If Yes:
Well, let see how far we could come with a passive solution first. If I want a BCR of 9 DB, the ratio of the series resistor to the impedance of the woofer should be 3:2, right ? So 3 parts of the voltage would be consumed completly over the resistor and 2 parts over by the woofer. Right now I have it the other way round: 3,3Ohm: 7 Ohm, so this needs to be changed, fortunately paralleling more devices gives me as well more sensitivity and a lower DCR as the same time.
With a baffle width of 15cm, f3 would be 766Hz...So what would be the right component values for this ?
At that high an f3 I'd definetely go for no BSC resistor,
and a parallel network.
Your bass array sensitivity needs to be 90/91dB in
the bass which means 96/97dB in the midrange,
above the baffle step frequency, say unfiltered @ 2Khz.
If your target is 1kHz c/o frequency with the array set
at the impedance for the right sensitivity (should be 12R
with Zobel correction) then c/o the bass units at 500hz,
= 3.8mH by my calculations.
The tweeter capacitor needs to be adjusted for a smooth
match, I estimate the value will be around 1.5kHz c/o.
🙂 sreten.
and a parallel network.
Your bass array sensitivity needs to be 90/91dB in
the bass which means 96/97dB in the midrange,
above the baffle step frequency, say unfiltered @ 2Khz.
If your target is 1kHz c/o frequency with the array set
at the impedance for the right sensitivity (should be 12R
with Zobel correction) then c/o the bass units at 500hz,
= 3.8mH by my calculations.
The tweeter capacitor needs to be adjusted for a smooth
match, I estimate the value will be around 1.5kHz c/o.
🙂 sreten.
- Status
- Not open for further replies.