Go Back   Home > Forums > >
Home Forums Rules Articles diyAudio Store Blogs Gallery Wiki Register Donations FAQ Calendar Search Today's Posts Mark Forums Read

Multi-Way Conventional loudspeakers with crossovers

Curious about ZAPH's designs. Seas L18 and P18RNXP
Curious about ZAPH's designs. Seas L18 and P18RNXP
Please consider donating to help us continue to serve you.

Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving
Thread Tools Search this Thread
Old 6th August 2007, 06:29 PM   #111
sreten is offline sreten  United Kingdom
diyAudio Member RIP
Join Date: Nov 2003
Location: Brighton UK

I think the original post stated that CSD's could be misunderstood
and were not particularly needed with comparable drivers if you
knew what to look for in the frequency response.

Those that cannot read the frequency response are also the most
likely to misinterpret the CSD's, especially say when comparing
very similar overall performance "soft" (plastic) and "hard" (metal)
cone drivers.

But when I look at the SCD graph,
the P18RNXP looks like a clear winner.
What am I not understanding or overlooking?
I think the original statement that CSDs are next to useless has
to be taken in the correct context - of publishing comparative
test data - and the fact they are not much use to an amateur
designer in terms of the design process - which is frequency
response based - but they may be misinterpreted.

I do not think CSD's were being dismissed out of hand,
just they have to be interpreted correctly in context.

  Reply With Quote
Old 6th August 2007, 07:31 PM   #112
dlr is offline dlr  United States
diyAudio Member
Join Date: Mar 2005
Location: Canton, MA
Default I had planned to drop this, but...

Originally posted by Danny
Dave, Mark, Pete, and myself can easily see this, but not everyone.

However, we can't see how much stored energy, how long, or how bad until we look at the CSD.
Before anyone brings up first order approximations, please keep in mind that this hijacked thread so-to-speak is at the point of focusing on the FR and CSD for a given drive level. This perforce means that they are generated from the same impulse response, so any comments about drive level nonlinearities are a separate issue. Valid, but separate.

The CSD is as subject to error as is the FR, maybe more so due to the successively shorter windows used to generate the decay. In any case, the "amount" is a nebulous term that you absolutely refuse to define. Not surprising, as there IS no absolute measure of reference for it, especially since it subject to the same or more vaguaries as is the FR. But all of it is reflected in the FR. Knowledge of some nebulous "how much stored energy" is useless. Useless Useless. Did I say useless? That would mean that one can do all that is required to design without the CSD. Yes, useless.

Why? Because there is no use in a CSD, the energy storage is clearly evident in the FR and all one need do is either compensate for it in the crossover or move on to the next driver. There is nothing in the CSD that gives additional information as to whether or not a compensation network can mitigate. Nothing.

Again, the CSD is useless. All it is good for is showing the neophyte that, yes, indeed, the raw driver has a set of various resonances of some magnitude and Q. Period. It does nothing else. Emphasis on the CSD is misdirected.

Further more, to suggest to a bunch of hobbyist that CSD info is of no value is ridiculous and misleading.
No, it's the absolute truth because whether experienced or not, the CSD provides NO additional USEFUL information that one needs to design. Nothing. Any and all non-linearities that are the result of the resonances are in the FR and include step and diffraction for that measurement position. One need ONLY the FR, coupled with distortion data if available and the impedance to determine and design.

Now keep in mind that this woofer was played to nearly the same level as the other woofer. I think someone said that if two woofers are playing the same output level they will have the same stored energy. Well that is not true at all.
Please direct us to this quote. I'm not aware of anyone making such a blanket statement. This is a specious argument put this way and I believe untrue. Use the direct quote for clarity if there is such. I don't think that you'll find it.

To play at this level this woofer had so much stored energy that the 5ms time window isn't even near long enough to catch it all.
This is so clearly evident in the FR that it would be wasted time inspecting the CSD. Yes, again, useless to look at the CSD.

Looking at its impulse response is like looking at one long curvy wave that just keeps on going. You couldn't even tell where the first room reflections were because the stored energy went right past it.

This woofer was also a paper cone woofer but is only used as a sub woofer. To get it to play to the same output level as the Alpha 8 with only 1/2 of a watt on it, this one needed the gain on the amp cranked up to a 100 watt level.

Yep, it needed 100 watts of input to give it the same output.

Differences in stored energy even below 1khz was tremendous.
If this is the same setup as the others it's useless to try to determine the driver's response below 1KHz, or anywhere for that matter, with contaminated data. You've got the same 2.88ms start and a 7.85ms stop. I'm sorry, this is just lame.

All of which is evidence that if the user cannot properly set up a measurement to exclude reflections, it's the error of the person making the measurement. Contaminated data proves nothing. It's a simple task in such a case to find the proper point for the marker. One would assume that the measurement conditions are under control. This is a strawman argument if ever there was one. This is the same baseless kind of argument that you used in the beginning, using purposely contaminated data. It simply undermines your position.

And for those who will want to claim that the CSD is not accurate below 1kHz can take a reality check here. It may not be completely accurate but who cares. +/-1/4 or 1/2 of a db is nothing. Differences are in such magnitude that we easily see them.
The errors can be and often are much larger. They could easily swing +/- 5db, but in any case it's a limit of the window used. You can't seriously be arguing that a 3ms window of an impulse is valid? Ridges can appear or disappear depending on the users ability and even with small changes in the window length. In any case, there is not enough data below that point to be considered accurate given the short time samples of the examples.

We've all commented, only it's been to point out the serious shortcoming of your argument.

Here is a direct quote from your first post taking us to task:

A crossover will also not filter out problem areas within the drivers pass band.
This is an absolute statement, no equivocation. Do you stand by this? Let's have no equivocating now, yes or no, do you stand by this statement that you said previously or not?

Another quote from that first post:

[QUOTE[QUOTE]Most proper crossovers for metal cone woofers do have the the breakup filtered out, and thus ringing in the CSD is not a problem.

Again, not true at all. Just because a resonance or break-up area can be pushed down below the 20db window shown in a spectral decay does not mean it is gone.
That part is true, but anyone who relies on a 20db limit of design is not doing a proper job of design IMO and it has nothing to do with whether or not that shows up in the FR, which it will. It's up to the designer to take all aspects into account and decide on the tradeoffs that he's willing to accept.

First, you said that you use a 25db window, why did you loosen that upto 20db in that comment anyway? When I did pay attention to the CSD, I prefered a 30db window. I usually didn't go much longer because one has to be careful that it's not the noise floor starting to contaminate the result. You've made no comments about this critical aspect all the while saying that those less experienced should pay the most attention to the CSD, yet one of the common problems is either a disregard of the noise floor or being unaware as to its impact. The CSD, especially at the low end, is susceptible to this.

My claim was and is that everything I need is in the FR and that the CSD is useless. I stand by that and can demonstrate that it is not required to design. The CSD provides no additional information towards design and can in fact be misleading when, as you've tried to do more than once, contaminated data is used to try to make a point.

Since you're so sure of yourself let ME ask a question. If I show you a measurement of a driver with a properly designed crossover with all measurement conditions defined, no contaminated data to mislead, would you claim that you could pick out any and all of the driver's RAW measurement anomolies that are pushed down, say to 20db or maybe even 30db? You said that "a crossover will also not filter out problem areas within the drivers pass band", therefore it must show up in the FR and the CSD since both are derived from the same impulse and window. Certainly you're so concerned about things 20db or more down that you'd still be able to pick out any problems that are 20db or higher in the early stop-band, but let's push it a little and say 30db since you're obviously concerned about that region.

Are you absolutely sure that you could point them out? I will provide impulse, FR and CSD for the test, probably this weekend since I'm still only using the digital filter in SoundEasy for my current auditioning and would have to construct the physical crossover to remove doubts. This might be very enlightening to all of us. You could demonstrate your prowess by pointing out how and where a crossover does not provide for correction only by looking at the final result. We'll then go back and look at the raw measurement to see how close you came in having a CSD. I'll show it for as low as I can get before noise floor contamination sets in. This will all be for 1KHz up since the CSD won't be valid below 1K as an absolute.

The one thing that we won't be able to determine since I'm not yet into making distortion measurements is what is the impact of the area below the 20 or 30db limit is with the crossover. The only thing that actually matters is what impact remains. That impact is measureable and shows up in distortion measurements made with the crossover in place. Various drive levels help, of course, to take drive signal nonlinearities into account, but that's still a separate issue vis-a-vis FR vs. CSD. In the end it's immaterial as to the measurement raw, because in the end no one listens to a system with no crossover in place. Or do you disagree with this as well?

  Reply With Quote
Old 6th August 2007, 07:55 PM   #113
dlr is offline dlr  United States
diyAudio Member
Join Date: Mar 2005
Location: Canton, MA
Default The more I view the CSD...

the closer I am to dismissing them out of hand due to not only the limitations of most impulse responses used that only allow for too-short a window due to reflections, but also for the misunderstandings as to what it shows with regard to what mitigation is possible.

I had questioned SL over at Madisound on them at one time as he didn't like them for the limits of graphical display, but this was a limitation of MLSSA at the time (and Clio apparently) and possibly may still be so. He was unresponsive when I pointed out that LAUD has the scale and time axes fully at the users control. At this point, I do not view the CSD for any design purpose at all. It's not needed.

  Reply With Quote
Old 6th August 2007, 08:41 PM   #114
Danny is offline Danny  United States
diyAudio Member
Join Date: Nov 2002
Location: Texas
Send a message via AIM to Danny
Dave, your long winded ramble wasn't worth the time it took to read it much less respond to any of it.

Same old same old, blah, blah, blah... Say whatever you like...

...there is no use in a CSD, the energy storage is clearly evident in the FR...
Bottom line is that you had plenty of opportunity to demonstrate this but did not. You were too afraid to comment on the posted measurements even when the two drivers were completely different to the largest degree.

The only thing you have demonstrated the propensity of your own opinion.
  Reply With Quote
Old 6th August 2007, 11:50 PM   #115
dlr is offline dlr  United States
diyAudio Member
Join Date: Mar 2005
Location: Canton, MA
Default Ah yes, the classic dodge

Originally posted by Danny
Dave, your long winded ramble wasn't worth the time it took to read it much less respond to any of it.

Same old same old, blah, blah, blah... Say whatever you like...

Bottom line is that you had plenty of opportunity to demonstrate this but did not. You were too afraid to comment on the posted measurements even when the two drivers were completely different to the largest degree.

The only thing you have demonstrated the propensity of your own opinion.
"Wasn't worth reading...". Whenever you can't or won't respond, this is your classic response. It's evident that you can't because you know that you can't back up your previous unequivocal statements. We've seen it before.

I can and still will provide evidence as I said I would. It should make for an interesting page at my site. There will be no question as to the facts.

  Reply With Quote
Old 6th August 2007, 11:53 PM   #116
pinkmouse is offline pinkmouse  Europe
diyAudio Moderator
pinkmouse's Avatar
Join Date: Apr 2002
Location: Rotherham, England
Curious about ZAPH's designs. Seas L18 and P18RNXP

Guys, stick to facts, don't bring personalities into the argument.
Rick: Oh Cliff / Sometimes it must be difficult not to feel as if / You really are a cliff / when fascists keep trying to push you over it! / Are they the lemmings / Or are you, Cliff? / Or are you Cliff?
  Reply With Quote
Old 7th August 2007, 12:42 AM   #117
dlr is offline dlr  United States
diyAudio Member
Join Date: Mar 2005
Location: Canton, MA
Default I did

Originally posted by pinkmouse

Guys, stick to facts, don't bring personalities into the argument.
I did. I also quoted some of Danny's own words. He cannot support those unequivocal statements of his. Also consider his lame attempts at "gotcha's", then trying to make good afterward. His graphs and data are simply not trustworthy given his history in the posts. It's by design a lost cause.

I never use nor do I have any use for the CSD to design, none whatsoever and that has been my point from the beginning. Anyone who knows their stuff also knows this. The FR has what is needed outside of actual distortion measurements. CAD software operates strictly on the FR for all but the most advanced versions that have additional capability, but the basics are the same. The CSD has no place nor any requirement in actual design. I dare say that you will never find any software that uses it for anything other than a flashy graph.

It is also subject to misinterpretation and is also inaccurate and incomplete when short windows are used for the FFT, such as the examples here. The resolution on the frequency axis is the primary problem. The magnitude also suffers at the low end, but it's primarily the extremely limited frequency resolution that makes them suspect, mostly below about 1KHz. That's a fact of the MLS and the FFT from which the CSD is determined. Protests to the contrary and whimsical and indeterminant measures of "most stored energy" are not based in fact. Danny wants everything on his terms, but won't even define them when pressed, as I did more than once.

  Reply With Quote
Old 7th August 2007, 02:24 PM   #118
PB2 is offline PB2  United States
diyAudio Member
PB2's Avatar
Join Date: Sep 2004
Location: North East
Default My point, and Nice Work Zaph

Originally posted by PB2

Let me put this in simple terms, let's take the very low frequency example where most drivers will be easily driven to high levels of distortion. Let's say we wanted to know the frequency response of a closed box woofer from 5 Hz to 100 Hz using pure tones, a nearfield mic and an RMS meter. Many systems driven at 5 Hz will produce over 100 % THD even at moderate levels, yes there is more level in the harmonics than the fundamental and if we measure the output with a meter we will obviously read the fundamental + THD and get a false reading. The solution is to do a filtered measurement, or one using a spectrum analyzer and obviously just take the level of the fundamental. This example was given when I studied audio engineering as an undergraduate.

Measurements in the presence of high levels of distortion are complex, more complex than with well behaved systems. Distortion will introduce an error into a frequency response measurement that assumes a linear system and does not provide filtering. Isn't this obvious?

I'm getting busy with some real work, so I'll bow out of this one also.

Pete B.
I want to clarify something here, some asked if FR tests were influenced by non-linear aspects of the driver or system under test and I provided the above extreme example to show that indeed they are. People seem to be jumping to the conclusion that this was the previous point I was trying to make which is not the case. My previous point was that even if you EQ all 7 peaks in the L18 frequency response, you need to then consider differences in the distortion performance when comparing two drivers and that the linear EQ could not possibly make the distortion of the two drivers the same.

I want to make it clear that this was my point all along.

I also want to mention that I respect Zaph's work, and the P18 was not available when he did that design so it is easy to see why he might have chosen it. Today, I'd choose the P18 for this low cost application. The much more expensive SEAS EXcel metal cone drivers provide much better distortion and FR than the L18 and I would consider these also in a higher cost system. Still, speaker design and psychoacoustic considerations are complex and I can see why Zaph might have been curious to try the L18 to determine their performance in a real world application.

Pete B.
  Reply With Quote
Old 7th August 2007, 09:12 PM   #119
theSuede is offline theSuede  Sweden
diyAudio Member
Join Date: Aug 2006
If the statement that all of the CSD's data is present in the FR (which is quite obviously missing one of the vectors in the linear system [Frequency][Amplitude][Time], namely... [Time] ! - a vector of no small importance), then the following would be true:

This is of course concerning ONE driver - in an almost perfect infinite baffle with no non-linear external influences like reflections or external non-linear hard resonances - someone earlier in this thread stated that all drivers in themselves are Minimum Phase, (mostly) Linear Systems... So lets see

1. A well thought out and implemented Impulse Measurement is made, or the Impulse is synthesized with one of the many available models (MLS, swept sine with inverse convolution or any of the others)

2. A well balanced CSD is plotted from the Impulse in (1.). And it is saved for later purposes. No conclusions will be drawn from this plot now.

3. The mathemathical conversion from f(Amplitude)of(Time) - the Impulse - to f(Amplitude)of(Frequency) - the FR plot - is made (with FFT, windowed or 0-padded). This is then a [Time]-independant plot. Obviously.

4. The mathematical conversion from f(Amplitude)of(Frequency) to f(Amplitude)of(Time) is made (step no. 3. is reversed). This is what some people here claim they can do - in their heads.

5. The Impulse generated by step 4 is used to make another CSD with exactly the same window(s) and resolution as the one made from the original impulse in step 2.

6. Compare the two impulses - the original and the one with [Time] removed from the linear system (the one in step 4. This what we actually have done by converting from the original impulse to a static FR)
They should be the same if the argument that CSD is entirely FR dependant is true.

7. Compare the two CSD's. They should be the same if the argument that CSD is entirely FR dependant is true.


And now we return to reality.

This is what WILL happen:

Steps (6.) and (7.) will show significant differences between the two versions.
These differences will be larger or smaller *(or lets just say - different) depending on the non-linearities and the energy-storage properties of the chosen driver.

These differences will NOT scale in a linear fashion if the amplitude of the impulse is changed (more or less power fed to the driver in step (1.)). The differences will NOT be the same. More non-linearities.

This said - I have to agree that MOST OF THE TIME, most of the energy-storage in a driver CAN be roughly estimated by looking at the FR. But this will be an educated guess, not a truth. Most of the time, this will be enough, but....

This has been done and documented. By more proficient organizations and persons than me. I will not make my own measurements regarding this experiment all over again just to satisfy someone that disagrees with this statement.
I will follow a well functioning and universaly accepted scientific procedure - my statement is true unless someone PROVES it to be untrue.
So - please.....
Make my statement fail by following my 7 steps and publish your results here
This will not happen.

(But btw - I usually only use FR's when designing loudspeaker systems... I cannot afford the luxury of allways buying lots of different drivers to compare their non-linear resonant behaviours. But I would like to... And I HAVE done exactly that a few times - for systems that I myself will use AND there are in my personal opinion audible differences between drivers that i chose from, even when equalized (with X-over) to the same FR! Even if the THD's are very similar. This was a by-comment)

I do hope that haven't made any personal enemies by writing this. That was not my intent.

Greetings from sweden... /Joakim
  Reply With Quote
Old 8th August 2007, 12:07 AM   #120
ShinOBIWAN is offline ShinOBIWAN  United Kingdom
diyAudio Member
ShinOBIWAN's Avatar
Join Date: Feb 2004
Location: UK
I've been making my way through this thread for last few days and I have to say I've learnt a thing or two.

Without getting drawn into the debate and instead just giving some acknowledgement for an interesting thread. I think John sums up much of what has been said in this thread with illustrative graphs and concise, easy to understand text. This can be found on his blog page here:


Hopefully he leaves that info up for awhile so others can take a look.

With regard to CSD proponents, I see where they're coming from. I thought the same way up until I read this thread but John, Mark and David make a lot of sense once I'd bashed that info into my head hard enough.

Cheers for the discussion guys. I'm a little wiser.
  Reply With Quote


Curious about ZAPH's designs. Seas L18 and P18RNXPHide this!Advertise here!
Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump

Similar Threads
Thread Thread Starter Forum Replies Last Post
Help with first Build (Seas CNO MkII or Zaph ZRT) whimsical Multi-Way 2 6th July 2009 05:40 PM
Where to get Seas P18RNXP TurboFC3S Multi-Way 6 22nd October 2006 10:05 PM
RS180 Modula MT vs Zaph Seas Metal 2 way AutoAudio Multi-Way 5 22nd March 2006 12:42 AM
Opinions on Zaph's Seas Kit demon2091tb Multi-Way 6 3rd October 2005 04:19 PM

New To Site? Need Help?

All times are GMT. The time now is 01:48 PM.

Search Engine Optimisation provided by DragonByte SEO (Pro) - vBulletin Mods & Addons Copyright © 2017 DragonByte Technologies Ltd.
Resources saved on this page: MySQL 15.79%
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2017 DragonByte Technologies Ltd.
Copyright ©1999-2017 diyAudio