Crossover point for Scan-Speak Drivers - Opinions needed

[ShinOBIWAN]

Lo all,

I've just recently bought a pair of R2904-7000 ring radiators along with the 18W4531 Mid/Bass driver for use in an active 2-way design.

I'll be using a Linkwitz Riley 4th order 24dB/Octave network but would like to hear peoples thoughts on potential crossover points.

I was looking at around 2Khz personally but would value any input from others.

Here's the manufacturer response/impedance plots:

R2904:

An externally hosted image should be here but it was not working when we last tested it.

18W4531:

An externally hosted image should be here but it was not working when we last tested it.

 

[ShinOBIWAN]

Anyone?

 

[ScottG]

w/ 4th order 1.5

 

[Bill Fitzpatrick]

2750

 

[SpeakerBob]

Hi Shinobiwan,

Bill's on the money with where the dispersion pattern is the closest for the 2 drivers. Best measure the drivers in the finished box to make sure though.

That mids' response is like my s-speak 8530's plot. Make sure you include a shelving low pass to lose the hump around 1 to 5k, otherwise they'll sound very bright. (to my ears anyway )

I take it you got bored with the Visatons ?

Cheers,

Rob

 

[5th element]

Bill is correct saying an xover of 2750.

The ringr will appreciate a higher xover having limited xmax and the mid/bass is sufficiently well behaved to handle 2750.

The bump in response can be managed with the lowpass.

You will require a shelving network to compensate for bafflestep though, linkwitz website details this and it works very well just using textbook maths.

You will also most probably require a delay circuit to bring the drivers into correct (or close enough) phase alignment to get the xover to sum well at 2750 and give you the notch on reverse polarity. Otherwise this should sound pretty darned good.

Tweeters are nice and easy just the filter network and a level adjuster, understand though that you probably will not need 4th order electrical filters on either the drivers to achieve 4th order acoustic.

Active filters are a little bit easier to design because you dont need to worry about impedance, but you still really need to measure the drivers in box, and use software to get the most out of them, especially with the delay network.

 

[ShinOBIWAN]

Scott:
You think the R2904 would handle that? What would distortion be like at that frequency. I know the 4th order would quickly roll off the response before real problem occured but wouldn't SQ be impaired at all?

Even with an FS of 520hz scan-speak recommend crossing over no lower than 2Khz.

Bill:
I was thinking that I should maybe move the crossover up to about 2.5-3Khz myself. What would be the most transparent point to crossover at?

Is it better reason that the R2904 is among the best HF drivers out there ATM so should be used to cover as much of the frequency range as it possible to do without impairing quality - my original reasoning behind the 2Khz point.

Or...

Move the crossover point out of the mid-range and upto around 3.5-4Khz. For the best midrange transparency.

Or...

Go with a different mid/bass all together! Whats passing off as one of the best mid/bass driver at the moment? What about PHL? The sensitivity looks like a great match for R2904.

As ever, any opinions greatly appreciated.

 

[ShinOBIWAN]

Hi Shinobiwan,

Bill's on the money with where the dispersion pattern is the closest for the 2 drivers. Best measure the drivers in the finished box to make sure though.

That mids' response is like my s-speak 8530's plot. Make sure you include a shelving low pass to lose the hump around 1 to 5k, otherwise they'll sound very bright. (to my ears anyway )

I take it you got bored with the Visatons ?

Cheers,

Rob

Hi Rob,

I will of course take some measurement proper once the drivers are in the cabinets. At the moment I just want to throw as many ideas around as I can to get an idea of the directions I can head.

Regarding the hump, I agree whole heartedly with what you've said , since its going to be an active XO, shelving could be added at some point but I'm also strongly thinking about swapping the mid/bass out for something else.

Yep, the Vistatons went to another good home where I hope they are being enjoyed. That project was a great learning experience as far as cabinet building goes, so now I'm confident in that area I'd like to turn my attention to the electronics side and hence my reason for going with full active 2-way's with onboard XO's & bi-amping with dual mono power amps(based on something like the lateral MOSFET design by Rod Elliot).

 

[ShinOBIWAN]

@ Fifth Element

Wow!

Clearly I thought I was beginning to really understand loudspeaker design and then a few new topics get thrown in for good measure, cheers FE

If you don't mind I've got a couple of questions because my brain is hurting after reading that:

Bill is correct saying an xover of 2750.

The ringr will appreciate a higher xover having limited xmax and the mid/bass is sufficiently well behaved to handle 2750.

The bump in response can be managed with the lowpass.

You will require a shelving network to compensate for bafflestep though, linkwitz website details this and it works very well just using textbook maths.

Cheers pretty much what I originally thought until madness set in after looking at driver frequency plots for too long.

The R2904 has a rising response anyway so what do you think bafflestep will add to this? Better still isn't there something on the linkwitz site to calculate it?

You will also most probably require a delay circuit to bring the drivers into correct (or close enough) phase alignment to get the xover to sum well at 2750 and give you the notch on reverse polarity. Otherwise this should sound pretty darned good.

This is the one I'm having trouble understanding. Help please!

Delay circuit? I was planning on using the cabinet design to time(phase) align the drivers. Am I being dumb in assuming this is enough or is their more to be done that I simply don't understand quite yet. I know that out of phase drivers usually results in drops/gains in the response but what would a delay offer that an offset tweeter wouldn't or is it an easier way to achieve the same thing?
What do you mean when you say 'give you the notch on reverse polarity'?
I've clearly still got lots to learn so any advice would certainly help me understand a little more.

Tweeters are nice and easy just the filter network and a level adjuster, understand though that you probably will not need 4th order electrical filters on either the drivers to achieve 4th order acoustic.

Yep I've allowed for variable cermet trimmers in the XO output buffer stage on both the HP and LP. What do you think the final acoustic order of rolloff will be, taking into account the HF drivers rolled off response?

Active filters are a little bit easier to design because you dont need to worry about impedance, but you still really need to measure the drivers in box, and use software to get the most out of them, especially with the delay network.

Will do once the cabinets have reach a suitable stage. I'm using a Behringer ECM8000 measurement mic along with TrueRTA with 1/24th octave measurements. Is this setup OK for measurements or do I need more accuracy?

 

[ShinOBIWAN]

Fifth Element:

Wilmslow Audio cocked my order up and sent me a 15W8530 instead of the 18W4531. So this is the perfect opportunity to swap to something else, maybe something better or something that doesn't have the nasty hump at 1-5Khz? All the Scanspeaks seem to do that strangely Maybe its the SD-1 motor tech?

What do you think about swapping the mid/bass for something like a Volt 2500.4 mid/bass driver? Effeciency is a little better but I'm concerned about using a 10" driver up at around 2.5Khz.

Hmmm... maybe not such a good idea. Always liked the Volt bass drivers though and for me they are unmatched for pure SQ, was wondering what there mid/bass's were like, anyone tried them?

Or what about something like a PHL or Seas Excel.

 

[ScottG]

Scott:
You think the R2904 would handle that? What would distortion be like at that frequency. I know the 4th order would quickly roll off the response before real problem occured but wouldn't SQ be impaired at all?

Even with an FS of 520hz scan-speak recommend crossing over no lower than 2Khz.

Yes it would handle it, but this depends on what your power handling requirements are, (if extreme spl's is your concern then look elsewhere)..

If your not interested in extreme CONTINUOUS spl's then read on:

Look at how Linkwitz himself uses the 24db crossover in the Orion. Consider that there are some design elements at work here that may not be imeadiatly obvious. In particular what will be the off-axis response of the mid/woofer at say 60 degrees, or 90 degrees, (or more)? (i.e. if you think 2750 is the magic number think again - frankly even 1.5 isn't..) Furthermore, how smooth is the response provided by the mid/woof (I gurantee that the manufacturer graph provides a "rosey" picture). What are the distortion levels of the mid/woof at the top of its range? (both linear and non-linear). Finally, what type of driver diameter usually gives better imaging - large diameter or small diameter?

Now note the differences between the 2 tweeters (the Seas in the Orion and your tweeter). The Seas has 1mm of linear excursion and the Scan only has .2mm (but a "max" of 1.6, and I don't think they are specifying x-damage - i.e. look at the size of the surrounds on this driver.) Now look at the differences in eff.. The Scan effectivly is almost 9db more eff. (i.e. the Scan will be seeing a lot less power at normal listening levels than the Seas). Now look at the Impeadance as it approaches fs between the two drivers. Both tweeters have very similar impeadances suggesting similar linear distortion. Sure, the decay won't be as clean at 1.5 as at 2khz or higher BUT what will energy storage (decay) be like for a small diaphram tweeter as opposed to a mid/woofer? (just look at the decay on the seas - its down almost 15dbs before .5 mill sec.s at 1.5kHz.) Finally, look at the recomended crossover slopes and freq.s for both - hmmm they are the same..

 

[Hans L]

The SS ringrad is an utter waste of money if you crossover it that low. It's distortion is much higher than the Seas Excel 002 used by LR, at lower frequencies. They don't even compare below 2.0KHz. Ringradiators have notoriously high 2nd order non linear distortion compared to dome tweeters.

I heard this tweeter in a few two way design and was thoroughly unimpressed. In a 3-way it is good, my guess is crossed not significantly lower than 3000Hz at a minimum of 2nd order. Try a perfect Bessel 2nd order at 3000Hz or a Butterworth 3rd order (these are all acoustic slopes, not electrical) on both slopes and let us know how you like it. You needn't worry about the exact freq. The tweeter won't be different xo'ing at 2800 or 2900, that kind of difference is irrelevant. It's the xo order and how the slopes of the tweet and the mid add up that really count.

The 'nasty' hump of the scanspeak midbasses isn't a problem at all, in fact it may help with the xo design, but you certainly have to deal with it or the speaker will sound shouty and forward.

 

[5th element]

The SS ringrad is an utter waste of money if you crossover it that low. It's distortion is much higher than the Seas Excel 002 used by LR, at lower frequencies. They don't even compare below 2.0KHz. Ringradiators have notoriously high 2nd order non linear distortion compared to dome tweeters.

I heard this tweeter in a few two way design and was thoroughly unimpressed. In a 3-way it is good, my guess is crossed not significantly lower than 3000Hz at a minimum of 2nd order. Try a perfect Bessel 2nd order at 3000Hz or a Butterworth 3rd order (these are all acoustic slopes, not electrical) on both slopes and let us know how you like it. You needn't worry about the exact freq. The tweeter won't be different xo'ing at 2800 or 2900, that kind of difference is irrelevant. It's the xo order and how the slopes of the tweet and the mid add up that really count.

The 'nasty' hump of the scanspeak midbasses isn't a problem at all, in fact it may help with the xo design, but you certainly have to deal with it or the speaker will sound shouty and forward.

Exactly.

WRT the delay circuit.

Both the driver positions and the crossovers used alter the phase of the signal reaching the listener. Time aligning basically means putting the acoustic centres of the drivers, @ the xover frequency in line with each other. Note that the acoustic centre of a driver alters with frequency so the best you can hope for is aligning them at the xover frequency.

Having done this both drivers are going to be radiating the sound in the same "z" axis when a signal is applied, but this does not mean your xover is going to be summing correctly or be phase aligned.

When you design a crossover for a standard passive two way speaker, with both drivers mounted on a flat baffle you have to play around with both frequency, slope and order until you get good summation and notch when the polarity of one driver is reversed.

If you use a delay network you can practically use any slope, order and frequency you wish, then just tweak the delay network to yield a good notch on reverse polarity.

You can do the same thing by moving the tweeter around in the z plane, if you needed to add in 100us of delay for the notch @ reverse polarity then moving the tweeter back by 3.4cm would achieve the same thing.

In the above instance you would have a phase aligned crossover but not time aligned because you have moved the tweeter to achieve the reverse null.

Now to be phase and time aligned you need to place the tweeter in the correct z plane (both the tweeter and woofers acoustic centres matched for the xover frequency) and then use a delay network to bring the phase of the drivers into alignment.

Is there anypoint in time aligning the drivers tho... maybe not, you also add in a negative factor and the time aligning is not perfect. Remember that the time alignment will only be correct for one vertical angle in the listening axis, when you go above or below this you will be adding in additional delay in favour or against one driver. Also the actual physical process of moving the tweeter back can cause deffraction effects to occur that you might not wish to include. Mounting both drivers flush on a flat baffle is the best option for deffraction effects.

Also someone else mentioned somewhere that we cannot actually here the difference in these tiny differences in wavefronts arriving at our ears.

It would certainly be interesting to hear two loudspeakers that were exactly the same as one another, same drivers, same xover and both were phase aligned but one was time aligned, could you hear the difference.

 

[ShinOBIWAN]

Re: Crossover below 2Khz

Thanks Scott & Hans for the opinions.

Personally I'm with Hans on this, ring-radiators are less suited to such a low crossover as 1.5Khz. Even Linkwitz admits that using the Excel HF at this point could could provide problems with trumpets for example, at high SPL's.

Everywhere I've looked everyone has crossed over at 2.5Khz+. LDSG emphatically emphasize that distortion will be audible at a low crossover frequency and suggest 3Khz+

I'm beginning to agree with 3Khz now, even looking at the response on the filter this looks perfect.

So for the first trails with the crossover this will be my reference when experimenting.

Re: Delay Circuit

Thanks for all the info 5th

I'll have a good look through what you've written, research and then take on board what you have written.

For anyone stepping into the world of active XO's the Linkwitz site is an invaluable resource.


Also what do people think about swapping the mid/bass for another brand?

The Audio Tech Flex units look interesting and seem highly regarded, perhaps this is the direction I should be heading?

 

[ScottG]

Re: Crossover below 2Khz

Thanks Scott & Hans for the opinions.

Personally I'm with Hans on this, ring-radiators are less suited to such a low crossover as 1.5Khz. Even Linkwitz admits that using the Excel HF at this point could could provide problems with trumpets at high SPL's.

Everywhere I've looked everyone has crossed over at 2.5Khz+. LDSG emphatically emphasize that distortion will be audible at a low crossover frequency and suggest 3Khz+

I'm beginning to agree with 3Khz now, even looking at the response on the filter this looks perfect.

So for the first trails with the crossover this will be my reference when experimenting.

Re: Delay Circuit

Thanks for all the info 5th

I'll have a good look through what you've written, research and then take on board what you have written.

For anyone stepping into the world of active XO's the Linkwitz site is an invaluable resource.


Also what do people think about swapping the mid/bass for another brand?

The Audio Tech Flex units look interesting and seem highly regarded, perhaps this is the direction I should be heading?

Thats cool, but I think your getting a lot of mis-information from others including the LDSG - Its almost as if one person says something and it spreads like a virus on the web, without understanding the operating conditions that the originator makes his/her basis from..

Case in point the LDSG which states:

"Also, as seen clearly in the FR graph, this unit should never be crossed over too low."
This statement needs to viewed in context. When you refer to the FREQ. RESPONSE of a driver you are NOT talking about the distortion levels but instead the FLAT (freq. response) summation of the driver with a midrange, which becomes increasingly important with lower order crossovers. This has NOTHING to do with distortion levels..

Now lets look at HansL's assesment - does he know what the distortion levels are of this driver? (I don't.)

What about the "notorius" nature of ring radiators? Frankly the only thing notorius about ring radiators is a general lack of off-axis dispersion because the effective diaphram diameter is wide.

Additionally, what two-ways with this driver were less than satisfactory? What was the pairing driver and what was the crossover slope? (i.e. there are a LOT of potentially misleading variables.)

Now there are several measurements available on the XT25 that suggest that distortion is higher around 2kHz (non-linear), BUT the Scan Speak version is NOT the Vifa version - and what is Scan Speak known for? LOW DISTORTION! (i.e. why you would pay 9 times the price for this driver over the Vifa alternative..)

link to XT25:

http://206.13.113.199/ncdiyaudio/mark/Testing_page.htm

and guess what - the XT25's spectral decay is down MORE than 15dbs at 1.5kHz (.5 mill. sec.s).

Finally, once again we have replies placing almost exclusive emphasis on the tweeter without looking at the distortion of the mid/woof (and other attributes)..

As for other mid/woof's:

The AT driver's are far more expensive but the expense allows for a degree of customization. (i.e. if you desire a quality customized driver then contact AT.)

 

[5th element]

Due to the price of the scan ringradiator not many people have used it so actual measurements of this device are few and far between. However there are LOTS of similarities between the scan unit and the vifa unit, most importantly though is they have an xmax of 0.15 and 0.2mm for the vifa and the scan respectively & they have a radiating area of 5.4 and 5.6cm2.

The XT25 has been measured for distorsion on several occasions and it does not do well near 2khz, it would not be hard to asssume that the scan driver also follows suit with this.

Now if you consider the scanspeak d2905/95 tweeter known for its ability to handle low xovers well, it has an xmax of 0.4mm and a radiating area of 8.5cm2.

Now if we compare the amount of air each of these drivers can shift

first the scan rr - 112 cubic mm.
now the 95 - 340 cubic mm.

The 95 has three times the ability to move air over the ringradiator, so if the lowest people like to cross the 95 is around 1800hz with steep slopes it makes sense that around 2800 is the lowest you should take a ringradiator.

Yes ShinOBIWAN you could consider using other drivers if you want to. If you are going to keep the other scan units wilmslow sent you, then I would make a speaker with them anyway and have a listen, you may really like them. If you decide to change the woofer and box at some point then it does not matter, the active xover will ultimately remain roughly the same and only require a quick change a couple of resistors when you alter the woofer. If you changed to a SEAS excel then you would require the addition of a notch filter too but you could build the active xover with provisions for adding one anyway.

The scan speak slitted paper, seas excel and audio technology drivers are all considered to be top of the range. The 6.5" excel mag cones dont really suit themselves well to the ringr with its penchant for high xovers. Seas are releasing their new range of nextel coated paper, you may wish to make a speaker with the 15w wilmslow sent you by mistake and the ringradiator, then try the W18NX when its available. Although it looks like breakup occurs at about 5khz again with this driver, I dont think its in the same league as the mag cones with respect to the massive increase in distortion roughly an octave lower.

Whatever you decide keep us posted on what you think of the scan RR.

Scott, neither me or hans are saying that the scan cant handle low xovers as fact. We are mearly looking at the data supplied by the manufactures and coming to a reasonable conclusion about what it says. Yes the scan unit is probably better in distorsion then the vifa, but everthing else about this driver says that it will not particularly like low xovers. Obiwan is asking for advice on how to best put these speakers together, going on all teh facts and figures presented by D-S-T and other information regarding other tweeters ability to handle low xovers it makes sense that the RR will sound better crossed over at 2.8khz then 2khz.

 

[ScottG]

The XT25 has been measured for distorsion on several occasions and it does not do well near 2khz, it would not be hard to asssume that the scan driver also follows suit with this.

But I think it would be "hard" to assume so..

For the XT25 linear distortion is quite good, non-linear distortion is neither bad nor good (perhaps sub-par).

Non-linear distortion is F A R easier to reduce than linear distortion. Again, I could be wrong - but where do you think the RD of Scan Speak's version went? And again, why would anyone pay 9 times the price?

(And once more you are fixated on the tweeter - what is the level of distortion on the woofer at 2kHz?)

Sure - we are all premising what would aprox. offer the best result. But a statment like: "The SS ringrad is an utter waste of money if you crossover it that low.", is rather definitive. Moreover, the support to this conclusion was "illusive" at best.

 

[5th element]

But I think it would be "hard" to assume so..

For the XT25 linear distortion is quite good, non-linear distortion is neither bad nor good (perhaps sub-par).

Non-linear distortion is F A R easier to reduce than linear distortion. Again, I could be wrong - but where do you think the RD of Scan Speak's version went? And again, why would anyone pay 9 times the price?

(And once more you are fixated on the tweeter - what is the level of distortion on the woofer at 2kHz?)

Sure - we are all premising what would aprox. offer the best result. But a statment like: "The SS ringrad is an utter waste of money if you crossover it that low.", is rather definitive. Moreover, the support to this conclusion was "illusive" at best.

Oh ofcourse the 6.5" would most probably have better distorsion lower down then higher for sure, and I think it would be stupid not to try the ringradiator at about 2khz just to see what its like.

Nine times the price - aluminium face plate over plastic, scanspeak name and patented technology the ability to charge that amount. Lets face it this whole thing follows diminishing returns, yeah the scan will sound better then the vifa, but not 9 times worth it, just as a SEAS 27TDFC will not sound as good as a d2905/97, but as far as value for money goes the seas wins hands down. This tweeter is supposed to represent one of the ultimate tweeters therefore scan can charge what they like. This tweeter is meant to be better then the 99/97 so naturally they are going to charge more. Obviously the basic construction methods are not hard nor expensive because most of them are shared with the XT device and most of the core research and design will have been done on the XT and that doesnt cost very much, so I dont think we can pin the huge price tag on the development of this tweeter. They charge that price because they want to, or they only want to see it in the best of the commercial speakers out there so they make it expensive to stop the mass market throwing it in £1000 bookshelf speakers.

 

[Hans L]

Thats cool, but I think your getting a lot of mis-information from others including the LDSG - Its almost as if one person says something and it spreads like a virus on the web, without understanding the operating conditions that the originator makes his/her basis from..

I do not base strong opinions on the words of others. LDSG is a very helpful resource, nothing wrong with it. It's the first advice I give to newbies entering speakerdesign. But like all subjective sources, it's not the say all be all of speakerdesign.

Now lets look at HansL's assesment - does he know what the distortion levels are of this driver? (I don't.)

I do... That's why I permit myself to say 'they don't even compare below 2.0KHz.' Or do you think I'm making this up while I'm typing?...

What about the "notorius" nature of ring radiators? Frankly the only thing notorius about ring radiators is a general lack of off-axis dispersion because the effective diaphram diameter is wide.

K2 for a ringrad is always relatively high compared to higher order non linear distiortion products. This is a well known fact and relates to the construction of a ringrad, which is essentially a surround without the dome.

Additionally, what two-ways with this driver were less than satisfactory? What was the pairing driver and what was the crossover slope? (i.e. there are a LOT of potentially misleading variables.)

I have heared two two-way designs, a 3-way design and have had the opportunity to tweak the xo in a scan speak Ultimo via LspCad pro's emulator. I think I know what the tweeter can or shouldn't be doing.

Sure - we are all premising what would aprox. offer the best result. But a statment like: "The SS ringrad is an utter waste of money if you crossover it that low.", is rather definitive. Moreover, the support to this conclusion was "illusive" at best.

So far it seems you are basing you opinion on Scan speak's specsheets and you seem to be more interested in my credentials than the Scan Speak ringrad's potential. You may now kiss my





bottom.

 

[Pan]

Hans,

do you know if the 2nd harmonic (and overall dist.) is comparable between the Scan and Vifa ringrad.? Or do the Scan improve measurable over the Vifa?

Think I saw something about -30dB (or was it -40, yea it must have been) 2nd order harmonic of the Vifa at what I assume to be 90dB or so.

/Peter