Curious about ZAPH's designs. Seas L18 and P18RNXP

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Having NO knowledge in speaker design, I'm doing alot of reading and trying to get a better understanding of what makes a good driver (among other things). I was looking at Zaph's L18 design and looking at his 6.5" measurements and comparison... And can't help but wonder why did he choose the L18 over the P18RNXP?

If I look at the responce curve and the harmonic distortion, They are very comparable (to me). But when I look at the SCD graph, the P18RNXP looks like a clear winner. What am I not understanding or overlooking?

Peter
 
Well, for starters, the P18 didn't exist at the time I designed the L18 system. It came out about 1.5 years after that. But, even if I had to choose today, I'd probably still pick the L18 over the P18 by a teeny margin based on a few characteristics.

You're making a very common mistake about cumulative spectrum decays. But don't feel bad, it's a mistake that even industry professionals make. The CSD is the most over rated and useless form of measurement there is. There is absolutely nothing in there that can't be seen in a response curve. They both are generated from the same impulse and are simply different ways of looking at the same linear distortion information.

One might ask why I provide them on my web site. I often ask myself the same thing. The sad fact is that I think I've been brainwashed by people who think they need to see them. If I didn't provide them, there would be hoards of people saying "yeah, but he didn't provide any CSD plots". I'm guilty of posting CSD plots to avoid having to explain linear distortion to people who aren't likely to understand it. There are people who think a CSD contains info that a response curve doesn't because one has time slices and the other doesn't. Consider this question: If both the P18 and L18 were filtered with a proper crossover to shape the response curves into perfect LR4 rolloffs, and a CSD was then taken of the result, what would the differences in the CSD's of each be? The answer is that they'd be exactly the same. This is hardest thing to grasp about CSD plots.

A response curve and a CSD are just different ways of looking at linear distortion. linear distortion can and should be fixed in the crossover when possible. Most proper crossovers for metal cone woofers do have the the breakup filtered out, and thus ringing in the CSD is not a problem.

Examples with tweeters: when I see this response curve with the sharp peak/dip combo right in the middle of it's bandwidth, even before looking at the CSD I know that it will show a ridge at the same spot like this. What's notable about this is that the ringing anomaly is not easy to fix in the crossover. If you want to view a CSD, this is the kind of thing to look for.

On the other hand, let's look at a tweeter with a high Qts bump in the low end like this. Without even looking at the CSD, we know the low end will have some ringing like this. It's ok though, because the low end can and will be shaped into the required rolloff by the filter, making the low end ringing a non-issue.

So, the bottom line is that nobody needs to see CSD plots, but if you're going to look at them, at least know what to look for. You may simply ignore the CSD plots and look at the response curves for sharp/steep anomolies not easily fixed in the crossover. Also be aware that the same gating issues that reduce the number of data points in the low end response curves of drivers, will also affect the low end of CSD plots. Most of the CSD plots I post are only usefull down to maybe 800hz, after which the mush of gating obscures usefull info. On the other hand, some of the response plots I post (not the 6.5" woofer tests though) have nearfield low end responses merged in. This alone makes the response curves much more usefull than the CSD plots.

I hope this clears things up for you and others.
 
So Zaph,
you are waying that a CSD plot doesn't provide anything more at all than a correct FR graph ?

i might ask here, what is the point of the CSD plot at all then ?

or ..are you saying that a CSD doesn't present much usefull information in what we are using to determine crossovers and such when desinging a particular system??

I do not understand fully here,
is the FR curve a sum up of all ( what we see in the CSD)
waves for a certain period of time summing up together?
 
Wow, the Zaph-man himslef, my question wasn't as stupid as I thought afterall :)

"If both the P18 and L18 were filtered with a proper crossover to shape the response... ...what would the differences in the CSD's of each be? The answer is that they'd be exactly the same. This is hardest thing to grasp about CSD plots"

Good to know! Thanks for the explanation.
 
Great write-up Zaph, very informative, I really like the use of examples. :)

This lead me to do some playing on your site to test my new found knowledge and yes, you can see a correlation between the CSD and the freq. response. I think I may have found one example in which it doesn't apply. The Ascendant Audio Poly 6.5 has a raise/hump between 1.5-4 kHz in the freq. response and I except to see some similar information in the CSD, but the CSD looks very clean in that range. Am I missing something? Is this an odd-ball example? Is the hump in the freq. response due to something else that doesn't show up in the CSD?

Again, thanks for the lessons.
 
Assuming that the frequency curve in your tests on your site includes a certain amount of time in the measurement, is it the same time that is sampled in the CSD graphs ?


I am not sure i completly understand CSD though,
aren't the graph supposed to show a linear decay on time of producing sound vs rising frequency ?
i assume that a "perfect" driver would measure in such a way that the sound wave timing would get smaller with the frequency and thus the graph would show more like a 3d 3-sided pyramid towards high freqs?
 
Zaph is exactly right.

The CSD provides no additional information than is present in a well done FR curve. Neither do burst/linear distortion graphs. They are all different ways of showing the same phenomena. You have to think about what you can easily fix in the xover and what you can't. That's all.



Although I don't spend much time on CSD here--

http://www.markk.claub.net/linear_distortion.htm

This plus the three references should make for a very good set of readings.
 
Phase information

I'm 99% in agreement with what is above, but feel it may be worth adding a little.

As Zaph explained there is some information hidden in the frequency response that shows in the CSD.

Generally anything where there are multiple time delays involved (reflections, echoes, cone breakup) introduce the possibility that the resulting peaks or dips cannot be corrected, and that means the driver is hard to use in that frequency range. (Again as explained by Zaph.)

Although not very useful for quantitative information (mainly because it is recorded on one axis and the time evolution often depends strongly on the measurement position), the CSD does hint at whether there is time delay involved.

Another way to provide the same hint would be to record the phase in a (very) near-field measurement (although multipath problems or cavity effects will make this challenging approaching 20 kHz).

In the general electronics and control systems world, it would be normal to plot the amplitude and phase of the response (i.e. a Bode plot or equivalent) to help reveal this information.

(I've deliberately not mentioned minimum phase, and zeros in the postive real half plane, as those who know about that probably won't need the rest of this post.)

Ken

ps. I'm still trying to get a feel for what is useful to post, so please forgive if this is not the kind of thing people want to read.
 
CSD

Someone pointed this thread out to me because it was funny. Actually it is a bit sad.

I hate seeing people being mislead by false information.

Some of this I just had to respond to.

The CSD is the most over rated and useless form of measurement there is.

Actually it's one of the most important measurements one can take. Of all the measurements one could take, this one will tell you more about how a driver will sound and what type of network it might require than any other measurement.

QUOTE]There is absolutely nothing in there that can't be seen in a response curve.[/QUOTE]

Actually the truth is quite the opposite. You can't see the stored energy and inertia (CSD) in a frequency response measurement, but you can see a frequency response measurement in a CSD.

They both are generated from the same impulse and are simply different ways of looking at the same linear distortion information.

One looks only at amplitude (frequency response), the other looks at how fast a driver will bleed off stored energy and inertia. They are very different.

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.

If you play a driver to a 90db reference and the break up is down 20db it is still ringing at a 70db level and easily heard.

A crossover will also not filter out problem areas within the drivers pass band.

Don't be fooled. Look very closely at the spectral decay.
 
Re: CSD

Danny said:
Someone pointed this thread out to me because it was funny. Actually it is a bit sad.

I hate seeing people being mislead by false information.

Some of this I just had to respond to.



Actually it's one of the most important measurements one can take. Of all the measurements one could take, this one will tell you more about how a driver will sound and what type of network it might require than any other measurement.

There is absolutely nothing in there that can't be seen in a response curve.

Actually the truth is quite the opposite. You can't see the stored energy and inertia (CSD) in a frequency response measurement, but you can see a frequency response measurement in a CSD.



One looks only at amplitude (frequency response), the other looks at how fast a driver will bleed off stored energy and inertia. They are very different.


I hate false information too:whazzat:

Mathematically speaking, there is nothing that isn't in an FR curve, that isn't in an impulse. And the CSD is derived purely from an imulse. So, from an information theory standpoint, the CSD offers no additional information that isn't already in the impulse, and thus, the FR.

What you're saying is it's your preference to look at a CSD. Well, OK. The up side of a CSD is that, by slicing up the time domain response into time delayed chunks, yes, you can see how much energy is left after some arbitrary time more easily than by looking at an FR curve where all time is superimposed. However, the down side is that by successively truncating the impulse you introduct significant artifacts.

Again, there is no new information in a CSD that is not present in the impulse, and, hence, in the FR. This is a mathematical fact of no dispute.

Again, I would recommend the links from my previous reference to john k's and SL's comments on energy storage.

Also, a brief easy read on csd's can be found over at liberty instruments-

http://www.libinst.com/wattlar.htm
 
Mathematically speaking, there is nothing that isn't in an FR curve, that isn't in an impulse. And the CSD is derived purely from an imulse. So, from an information theory standpoint,

This part is true. Both are derived from the information found in the impulse response.

But if the impulse response contains all that information why look beyond the impulse response?

Obviously the impulse response does not display amplitude in any meaningful way. The frequency response does and displays amplitude in relation to frequency.

the CSD offers no additional information that isn't already in the impulse, and thus, the FR.

Not true. You can see a response pulse in the impulse response in relation to time, but not in relation to frequency and time.

Again, there is no new information in a CSD that is not present in the impulse, and, hence, in the FR.

Not true. While the single measurement captured the information to display the data for both measurements, you clearly cannot see a decay rate from the frequency response. You can only see amplitude.
 
Hmmm, I think Zaph's response may have been a little misleading, though I wouldn't discount it entirely.

Re: CSD plots, I quite like them if they are done well. What's the point of fussing over accurate data collection when half the time people can't even make the axes legible?

Ucla88, I'm not too worried about artifacts from CSD plots. Windowing functions are usually designed to get rid of artifacts from the classic 'start' and 'stop' clicks, albeit at the cost of losing low frequency information whose wavelength approaches the length of the window.

(edit: on second thoughts, that last bit ^^^ is not quite true, but it'll do :D )

IMHO, it's usually enough to have a close and careful look at the speaker's impulse or step response (assuming that the manufacturer/tester is kind enough to oblige us with that information). And if there's substantial ringing after the initial "bang" then it's often possible to just measure the dominant resonant frequency/ies manually. I guess CSD plots add an element of mathematical rigour to the method and it may be possible to tune them to provide a visual analogy for the way that humans hear things. I.e.: there's necessarily a compromise between frequency accuracy and timing accuracy, and ringing may be masked below a certain point because it is of such a short duration.

Cheers,
 
Csd

Since decay over time is perfectly predicted with FR data, there is no additional information in CSD data.

Not true at all, and I can illustrate it with Zaphs own measurements. Let's look at a few 6.5" woofers.

Here are two drivers from Seas. One is $63 and one is $70.

L18-FR.gif


P18RNXP-FR.gif


Lets look at the top one. First there is some obvious break up going on in the top end. The frequency response does not hide that. How bad is it? We can't really tell. Zaph claimed though that with a properly designed crossover that it won't be an issue. That's not true either. It will still be an issue. Aside from an infinite slope design that break-up will still be there just down in amplitude.

Here is the real problem though. How do they look in the range you want to use them? How do they look from 4kHz and down? Are they both the same? They both look pretty fair from 4kHz and down don't they? Does the frequency response tell you all that you need to know? Let's see.

Here are the spectral decays.

L18-CSD.gif


P18RNXP-CSD.gif


Besides the massive break-up in the top end. Look at all the additional stored energy in the range below 4kHz. One rings like a bell and continues to do so through the intended range of use. The other looks really clean.

Which do you think will have the cleaner vocal range? Could you see it in the frequency response? Certainly not !

Lets look at a couple more. Both of these are poly cone woofers. They must be the same right?

Extremis-FR.gif


AA6.5-FR.gif


What do you think?

The top one looks like it has a pretty good break up peak at 6kHz, and it has a little roughness in the 1.5kHz to 3kHz range that could be something, but we really don't know.

The bottom one could have some break up going on at 4kHz, 6khz and 9kHz, but how bad is it.

Both look pretty good below 2kHz though right? Let's see.

Extremis-CSD.gif


AA6.5-CSD.gif


Well on the top one, that peak at 6kHz wasn't really much of a resonance at all. It's there all right, but it dies fairly quickly. The smaller ones below it are more of a problem.

However, the biggest problem is in the range below 2kHz. This woofer stored a lot of energy in that range (its useable range).

For what its worth, my experience with this woofer is that it has a horribly smeared mid-range and is not useable beyond 200Hz.

Now the lower woofer is not just clean but real clean. The peaks in the frequency response that looked like might be a problem are no problem at all.

This is information that you CAN NOT get from the frequency response.

Hey let's look at one more. Here is one that is $46 and one that is $45.

F6-FR.gif


M-165X-FR.gif


Okay the top one has some small peaks in the top end. How bad is it? We don't know. It looks flat in the lower range though and all the way up to 2kHz so it must be great right? :whazzat:

The lower one looks pretty flat in the lower ranges too but has an elevated area from 3kHz to 5kHz that could be something. There is some small peaks above that. Maybe that is something too.

Lets see what the real tattle tale says (the CSD).

F6-CSD.gif


M-165X-CSD.gif


The top one looks pretty bad all over. Those peaks in the top end are bad. But look at all the stored energy in the range below 2kHz. It is all lumpy. There are areas that are popping up (ringing) where there wasn't even a peak. This thing is ringing all over the place. The mid-range and upper mid-range will be very colored. I have heard this woofer and it is very colored. You couldn't see it in the frequency response either.

Lets look at the lower one. Hey that amplitude peak in the 3kHz to 5kHz area died immediately. In fact the entire response across the board died immediately. You can't ask for it to be much cleaner.

Danny, you understand Zaph and Mark's point. Stop pretending not to.

Give me a break. I don't think you guys understand MY point. Read the data ! It speaks for itself.
 
Re: Csd

Danny said:
Give me a break. I don't think you guys understand MY point. Read the data ! It speaks for itself.

We do understand your point, but that point is just plain wrong. Zaph is without question absolutely correct. There is absolutely nothing in the CSD that is not in the FR. What is evident is that you have a vested interest in maintaining your position. It's wrong, period. All of your comparison examples simply illustrate the fact that the CSD can either misinterpreted or misunderstood as to the data that is there.

The CSD is close to useless. Resonance ridges are obvious in the FR. All one need do is understand the relationship between the Q and magnitude in the FR to not need the CSD. It should be obvious to anyone with experience.

Trying to read too much from the low frequency end also shows a misunderstanding of the limits of the measurement system. The accuracy of the curve and any "resonances" are problematic due to the cutoff frequency. Those CDS examples stop at 300Hz. That also means that there is question about what's happening between successive sample resolution points. A change from a square window to a half Blackman-Harris will likely show two different stories. Even moving the stop time marker a point or two may alter the low end response more than one might think.

No, the CSD is almost useless if you understand the FR curve.

Dave
 
We do understand your point, but that point is just plain wrong. Zaph is without question absolutely correct.

Not at all.

There is absolutely nothing in the CSD that is not in the FR.

Again not true at all. Re-look at the examples I posted.

The CSD is close to useless.

Nope. It is the most telling of any measurement you can take of a driver.

All one need do is understand the relationship between the Q and magnitude in the FR to not need the CSD. It should be obvious to anyone with experience.

You are incorrect. And just what is your experience level?

As for me I design drivers and work with manufacturers developing new drivers for various clients. This year alone approximately 20,000 to 30,000 drivers will be built for my clients that are of my design and about 8,000 drivers will be built for my own company that are of my design.

Trying to read too much from the low frequency end also shows a misunderstanding of the limits of the measurement system.

Decay rates below 200Hz are not accurate and are omitted from the measurements. No misunderstanding here.

No, the CSD is almost useless if you understand the FR curve.

You have it backwards as well. Stored energy and inertia information is not found on a frequency response graph. However frequency response is seen in the CSD.

My illustration clearly indicates the value of the information found in the CSD that is not clear from a frequency response curve.

Feel free to fool yourself if you like. Zaph and or any others can do the same, but there is no sense in trying to mislead others.
 
You're just plain wrong

People can be involved in design and still not fully understand some aspects of measurement. This is the case here. The more you protest, the more obvious it is.

I also was not speaking of the cut off area below 300, (not 200 as you mentioned). I'm speaking of the low end of those measurements. The resolution does not allow for accurately describing the true response much below about 1K down to the 300 cutoff. If you think otherwise, then that's a clear demonstration of misunderstanding the limits of the measurement.

Do you truly understand the indisputable relationship between the impulse response, frequency response and the CSD? At this point I'm questioning that even more.

Dave
 
People can be involved in design and still not fully understand some aspects of measurement. This is the case here. The more you protest, the more obvious it is.

Or so you think (in this case).

The resolution does not allow for accurately describing the true response much below about 1K down to the 300 cutoff. If you think otherwise, then that's a clear demonstration of misunderstanding the limits of the measurement.

In a CSD it is not that there is a point at which everything below point A is not accurate and everything above point B is. It is a graduated curve. Obviously accuracy gets less and less as frequency decreases. Depending on the measurement set up and gated time window accuracy below a 1kHz range becomes less accurate. Below 200Hz (or even 300Hz) it cannot be observed to have any real value as accuracy below that point is not possible.

If YOU think otherwise, then that's a clear demonstration of misunderstanding the limits of the measurement.

Do you truly understand the indisputable relationship between the impulse response, frequency response and the CSD?

Clearly you and I know that both are derived from the impulse response measurement. That does not diminish the value of looking to a spectral decay to see stored energy problems. To dismiss a CSD as having no value is absurd.

Maybe you would like to rub your crystal ball and tell us what the spectral decay will look like from looking at a frequency response curve? I could post a few for you if you'd like to decipher them for us.

Clearly from the examples that I posted one can NOT in any way accurately predict the stored energy of a driver only from looking at a frequency response measurement. The suggestion of such is ridiculous. Furthermore to suggest that knowing the information shown on a CSD is of no value is even more ridiculous.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.