I'm surprised that a search through the archive doesn't find mention of this, since it is such a crucial part of crossover design, working upwards towards a system from a modelled transducer, not a real, acoustically measured one in the target enclosure.
As I see it, the calcualtion of the free-air efficiency, No and hence the "spl" of a driver arrives at a number. Since I have an LS3/5a and its bass/mid driver to hand, let's say that number is 86dB from the manufacturers data sheet, and my own measurement and T-S calculation by hand.
We know that the LS3/5a cabinet is 188mm wide. It is generally stated that the baffle step - being the change in sound pressure relative to frequency and relative to the cabinet width - results in a '6dB transition', the centre frequency, at which the response is up/down by 3dB depending upon your prespective is, I'm told f3 = 115/WB, (where WB, width of baffle is in metres). Hence, for an LS3/5a this calculates as 611Hz.
My initial question is, if LEAP models the drive unit from calculation of the T-S parameters (of an imported impedance curve or manual data entry) and curves the frequency response in a cabinet of 3/5a dimensions, at what frequency would we find the sensitivity of 86dB? Would we find that at the point of highest output - and if so, how can LEAP reliably predict that frequency considering that the driver may well not be behaving as a piston towards the top end of it's operating band, somewhere around 1kHz where that drive unit naturally peaks-out.
What I am curious about is how LEAP would calculate the SPL output in a region where we can expect the driver to be pistonic - say, 100-250Hz, plus make an reliable estimation of the baffle step situation.
I have a couple of follow-on questions acording to your answer. Thanks Chris - hope you are well.
Alan Shaw
Let's start by describing clearly what we are talking about in terms of transducer parameter model sets.
(1) EncShop suports 3 transducer models: STD, TSL, LTD. What you call Thiel-Small parameters is what I call the old STD (standard) model. It is the most limited, and has no ability to repesent the mid band SPL or impedance correctly at all. In fact, for sub woofers with very long voice coils the error of this model can be 6dB even at 100Hz.
(2) EncShop (and even LEAP-4) does not produce simple highpass filter simulations of the SPL response from the transducers, rather it uses an electroacoustical network to model the real current flow through the voice coil and generates the SPL based on that. Therefore, the impedance function is critical. The impedance of a speaker determines the current flow through the voice coil, and therefore what SPL is produced. Remember that all dynamic speakers produce SPL as a function of BL*i. It is the current flowing in the voice coil that matters.
(3) The ability to create an accurate impedance model is therefore crucial. The STD model is horrible at this, the TSL is much better, and the LTD model is by far the best. Just look at the differences between these real drivers and the three models on this page:
http://www.linearx.com/products/soft...p/EncShp05.htm
(4) The Reference Eff No is simply a theoretical number based on the old STD model of what the speaker would produce for SPL on a half-space (infinite baffle) condition based on the assumption that there is NO motor impedance, just Re. Well that is simply untrue, all dynamic speakers have a motor impedance which reduces the current in the voice coil, and therefore the real SPL is nearly always lower than what No and SPLo would suggest. However at the same time the Mmd value is decreasing in the midband because the outter edges of the cone are no longer moving. The apparent mass is lower.
(5) The "baffle step" is not some simple calculation of a single frequency where a 6dB transition occurs. How and where this 6dB increase occurs is entirely dependent on the box shape and where the transducers are mounted on that box. There are peaks, valleys, and ripples produced from the diffraction of the box as the radiation domain changes from 4Pi to 2Pi space due to the box.
(6) For the first time in history, EncShop provides you with the real ability to accurately model the true diffraction of the box and see exactly what the "baffle step" looks like. How can you see this alone? Well, simple. Just run the simulation in free space, and then with your box on an infinite baffle, and divide the curves. That is the diffraction effect of your box.
(7) You can explore these diffraction effects to your heart's content, and answer all of your questions yourself. Just take a box and run multiple simulations while moving the source around to different locations on the baffle board, corner, center, etc. You will see the response shape change due to the changing diffraction.
Thanks for a comprehensive reply Chris.
I have looked at the many examples given of the superior Z modelling capability of LTD and it looks very impressive, and I've previously been convinced about the better Z matching it offers. What I was really looking for was some published examples of my specific need - LEAP's handling of the 'baffle step'.
As far as I can tell, the acqusition of enough data for LTD to characterise a driver is a good deal more complex than that of the TSL, but without any explanation (such as a graph or two) of how much more capable LTD is of modelling the 'baffle step' I simply don't have the time (wish I did) to chase elegant concepts that may/may not give me a result that I can trust. Sorry if this sounds harsh, but those of us still trading in this shrinking industry are all working flat out.
Seems like a missed sales opportunity Chris - a few graphics would really make the point here. Or have I missed something? I bet I have.
I think you are confusing a couple issues. The transducer model (STD, TSL, LTD) has nothing to do with the diffraction analysis, or what you call the "baffle step". All of those models have the same capability to specify a diaphragm profile (flat, dome, cone) and its shape (round, square, triangular, hex, etc). Those transducer parameters are used by the diffraction engine along with the 3D shell design to determine the response around the enclosure. The old common transducer parameters you are thinking of have nothing to do with that.
Maybe I am missing something here, do you have LEAP-5 now? If you do not, then it is going to be difficult to explain much here describing things you have not seen yet. EncShop contains a massive diffraction engine. It calculates the true spatial field around a 3D shell based on exact driver locations. What you call the baffle step, EncShop analyzes as the entire field around the enclosure in full detail. Horz polar, Vert polar, etc.
I certainly want to answer any of your questions I can.
OK Chris, I've had LEAP-5 for a couple of years but not used it for reasons of time pressure.
According to page 72 of Enclosure Shop Apps manual, the tutorial for Transducer Location and diffraction uses a pre-prepared LTD model. I do not have an LTD model of my transducer, and as I noted before (and even in your own text!) acquiring data to make an LTD model is 'no mean feat'. So, it seemed to me that T-L-A-D-A was not applicable to a TSL-dataset user. Hence, dead stop.
Is this wrong?
Yes, that is wrong.
All three of the models STD, TSL, LTD have a common set of extended parameters which enable diffraction analysis. If you look at the transducer parameters dialog, you will see that you still have to choose the shape and profile of the diaphragm even if you are using a STD model. Those fields are ALWAYS used.
They control the modeling of a diaphragm, and all sources must have a radiating surface. How that surface behaves, what it's shape is, etc are independent of the low frequency modeling parameters.
Bottom line, you will get difffraction analysis with any transducer model. You have to, or there would be no SPL curves at all. All of the acoustic radiation is ALWAYS calculated through the diffraction engine. So you do not need a LTD model for that.
Noted Chris.
So, I sort of work my way back to the beginning where I think it's good prectice to predict the result of an experiment before actually conducting the experiment: if I load my B110 SP1228 date into a TSL (not LTD) and create an LS3/5a-like enclosure, would I expect to see the manufacturers free-field sensitivity of 86dB at, say 100Hz of your spl graph or at the very peak of the diffraction prediction - perhaps around 1kHz? This is a very important point.
I am not sure into which space (but I assume half space) the manufacturer's data sheet relates to.
I would add, that you will see from my own notebook:
[urlhttp://www.linearx.com/products/software/LEAP5/EnclosureShop/EncShp05.htm[/url]
(bottom curve) that to make the system response of the LS3/5a flat (with the 1228 bass/mid driver) approx. -9dB of eq is applied at 1kHz, resulting in a system with an spl of approximately 82dB across the pass-band from 100Hz upwards post-crossover equalisation.
Insufficient data. I have no data from you to be able to make such conclusions. I would suggest you run the analysis and find out. That's the best way.
OK Chris, I'll do it your way.
I know, for example, from many years of working with drivers, that the cone material and/or surround material (let alone profile), the method of adhesion of the cone to the surround, cone/surround overlaps etc. all *critically* effect the top end SPL of a drive unit by ± several dB's. I'll be looking for convincing that there is anything like enough data input to the model from the user to accurately model the axial lift. As you say 'insufficient (user) data'?
I don't care how ugly the user interface is, whether its DOS or Windows - I only care about saving my precious design time with acurate predictions of SPL from basic T-S parameters of the driver unit. If you have achieved this with such a small amount of data entry from the user, it's a miracle of programing and I'll be the first one to take my hat off to you.
Will you allow me to follow this up here with the results?
Alan
Well again you are mixing topics here. Now you are talking about cone breakup and high frequency behavior due to the cone etc. I do not try to model that. As you say there are no parameters for that. Rather, what I provide you with is a set of filters to contour the response at the high end to whatever your driver behaves like. So you can recreate those high frequency response details.
But none of that has anything to do with diffraction modeling, which was your original question. Diffraction happens around the box due to its shape, and is determined as well by the shape and location of the source(s). But at high frequencies the baffle appears as infinite and diffraction is insignificant anyway.
Ah ha! Now we are both talking at cross purposes, but with a common objective!
My original entry point was to understand how you would model the 'baffle step', which (admittedly) would exclude any cone break-up etc. and would, presumably be dominated by any on-axis SPL rise due to only the baffle dimesions: your diffraction model.
You say " ... what I provide you with is a set of filters to contour the response at the high end to whatever your driver behaves like. So you can recreate those high frequency response details.". That's fine Chris, but that assumes that I have an actual physical drive unit to hand that I can acoustically measure to determine what effect those break-up conditions have on the SPL.
I don't have a driver to hand. I'm playing with a conceptual driver, the T-S of which I'm modelling in LEAP-5. Hence, where I started - how can I predict what I'm calling 'the baffle step' solely from the T-S model. The answer seems to be ....
1. I can predict something of an SPL rise due to dimensions (but I don't know how, or what this will look like in your model) - you haven't commented to my question where the free-field "86dB" would line up on your curve, and
2. In addidtion, there is un unspecified and non-predicted "factor X" comprising other ±dB's of contribution to the driver's top end which is beyond the LEAP model and can only be 'drawn' by manual adjustment of the LEAP curve - assuming that there is a real driver to hand. A fiddle factor.
If that's correct - and it surely must be since neither you nor LEAP are psychic about cone break-up etc. etc. then the diffraction model is surely completely inadequate to reliably drive crossover design in the absence of a real, measured driver. The errors at the bass/mid drivers predicted response could be deviant with reality by, easily ± 3dB. Disappointing; back to the old way of making the driver (takes days), buring it in (more time), cutting a baffle (time), fitting it and measuring its SPL.
Am I wrong again Chris?
I think we are going in circles here. You have some preconceived ideas of the way you think things work, and in some other cases you are making incorrect assumptions. Again, a lot of your questions would be answered simply by running some designs and looking at the results.
EncShop produces a ton of data. You can change the shape of the box, change the location of the drivers, and see what effect it has on the polar fields all around the box. It is really amazing. You can test many of your assumptions as I have in the past, and the results will sometimes surprise you. I have been surprised on several occasions and learned things about the way things work that were different than how I had expected.
You can try all kinds of "what if" questions and see the real results. For example, What happens to the field response when you locate a port on the front or rear of a box? You can find out exactly what that difference is and answer that question. How much change is there if I put a 1" bevel around the baffle board? You can answer that question exactly. What changes in the response if I mount the driver in the center of the baffle or down at the bottom? Again, the answer is available with EncShop. All of this is due to diffraction analysis.
I enjoy answering these questions, but you must understand that it would take many pages to do so here and I do not have the advantage of drawing you pictures and other visual aids that would help your understanding. The fact is, I have already done all of this in the manuals. It is much easier to read the info there because I have much more space for detailed descriptions and many pictures and drawings as well.
Read Chapter/Section 4.12. It explains how EncShop represents the radiating surface of a transducer for diffraction analysis. Chapter 4 is all about transducer modeling. Here are the sections:
4.1 Introduction .................................................. ............ 43
4.2 References & Standards .................................................. . 44
4.3 Dynamic Analogies .................................................. ........ 45
4.4 Radiation Impedance .................................................. ..... 46
4.5 Acoustic Network Analysis ............................................. 51
4.6 Acoustic Current .................................................. ........... 53
4.7 Electrical Impedance .................................................. ...... 54
4.8 Acoustic Pressure & Directivity ........................................ 55
4.9 Acoustic Power & Efficiency ............................................ 58
4.10 Highpass Filter Approximation ......................................... 61
4.11 Motor Impedance .................................................. ........ 63
4.12 Diaphragm Structure .................................................. .... 75
4.13 Diaphragm Breakup .................................................. ..... 81
4.14 Diaphragm Suspension .................................................. .. 89
4.15 Magnetic Gap .................................................. .............. 95
4.16 Temperature & Power Compression ............................. 103
4.17 Model Performance & Comparison ............................... 105
You should also read chapter/section 5.9 on Diffraction.
I think reading some of these sections will help you far more than anything I can put into words here, without pictures. Then if you have some specific questions about the items covered in the manuals we will have a common base to start a discussion from that will make it easier for both of us.
Thanks again Chris for a comprehensive reply. Yes, I have of course made assumptions (and possibly unreasonable ones) that you/LEAP-5 are the ultimate in loudspeaker modelling. I know that in the entire modelling industry there is no one who has applied more mental effort to getting to the bottom of the science.
My selfish concern is simply one of my precious time. The example graphs you show are indeed interesting, but if they don't predict how a real transducer will acoustically behave from solely the T-S model, then they are of artistic merit and not inherent value to me.
OK here is the deal then: I will give up my sunday (and the fact that my kids have come to visit for the weekend) to running through the entire work list you've given me. I know (and I have verified) the T-S parameters of a selection of drivers. I have confidence in that data (as much as one ever can have). Obviously I can measure the dimensions of the box in which they are mounted. I can measure the 1m SPL response outside, up high, and I'll do that too with my B&K certified microphone. I will then compare against your diffraction model which - honestly - I really want to believe is very close to actual because if the model works it will save me a lot of time in the future.
I'll let you know what I find out.
It becomes difficult to answer your questions when you say you do not have the time to read the answers. You keep confusing diffraction with transducer modeling. They are not the same thing at all. Diffraction is mainly a function of the cabinet shell. The transducer only matters in how it presents the location of the acoustic source, and its shape. Nothing in the 75 year old conventional electroacoustic parameters of the transducers has anything to do with this. Read sections 4.12 and 5.9 those should help the most.
You are correct to the extent that I have mixed together what you have specified as the separate roles of the tranducer and enclosure modelling. This is because I am looking for a global solution to the reason the SPL curve rises (the 'baffle step' +++ whatever else is happening in the transducer or enclosure environment) and an ability to model it, quickly, on a laptop, on the beech with only the T-S driver parameters and LEAP-5.
However, I have made a mistake in my understanding of how LEAP works and it is clearly stated on chapter 4.13, Cone breakup (p.81) "By blending together and adjusting their parameters [of the transducer model] a wide variety of high frequency response shapes can be produced [ie. predicted]. This is largely a trial and error process to find the optimum parameters which fit the transducer". That is perfectly clear. Trial and error.
It means that, setting aside the diffraction issue, it is not possible (as I feared) to ask LEAP to automatically generate a reliable SPL curve solely *from the T-S data of a driver* that could be taken as a reliable model of the measured SPL such that crossover work could be undertaken. As you say yourself, "Finite element methods have been attempted for many years, but these require tremendous amounts of detailed information about the materials and construction of the device".
Hence, the user has to conceive the driver, model it the transducer model, physically build it, measure its *actual* SPL and tweak the model (Fmd, Qmd, Flp, Qlp - page 81) to give a good correlation with actual. I doubt that this will save me time but I clearly expected far too much of LEAP-5 which is not promoted as a finite element system: my over-optimism took hold.
There is no point in gathering the data I previously suggested because I already have the drive unit in existance. That's all I need.
You are not going to find anything to model and predict the mid and high frequency response of a transducer in any practical sense. It's just not practical. Nothing in the electromechanical parameters covers this, and the T-S stuff is just purely for the old fashion simplistic highpass filter approximation. Remember the mid/high freq response shape of a highpass filter is a flat line.
Many do not realize that to even utilize FEA type methods requires 3 independent FEA solutions: one for the magnetic system, one for the mechanical system, and one for the acoustic system. But that's not even half the problem. You need parameters for those models. Young's modulus for the cone, the surround, the spider, glue seams, mag coefs for the steel alloys, etc. You would need a zillion parameters that you have never heard of before.
Yes Chris, you are right. And even a program that we bought from a European supplier to do cone modelling proves to be completely useless for exactly the reasons you state.
I really do wish that you'd made this all crystal clear for dimwits like me. I've rewritten your sales blumb here .... and this is what I'd like to have read ....
------------
"LEAP-5 is capable of remarkable time saving in many areas of speaker design, and has a very reasonable price tag. It does not replace complex Finite Element Analysis as it has neither the data modelling capability nor the user interface to do that, but it is very simple to use. It looks complicated at first, and that may give the impression that it can do more than it really can.
Any speaker system designer knows that when you take a nice flat-measuring drive unit (measured on a big baffle) and put it on the face of a cabinet and re-measure the SPL, it looks horrible. Lumps and bumps appear in the response that are not there on the baffle curves. The system designer doesn't care much how or why these anomalies occurred - they are there whether or not he likes is and he has just one simple choice: to eq them out or not in his crossover design since the marketing guys have probably already fixed the cabinet dimensions, so moving the drivers about on the baffle has limited potential.
However, LEAP-5 very much does care about the reasons for those deviations from flat, and it has to take two quite separate but complimentary approaches to understanding and modelling them. First, the user has to make as T-S transducer model which LEAP-5 can very effectively model up to, say, 200Hz. Then, the user has to manipulate that model by hand until the frequency range above the limit of LEAP-5's capabilities correlates well with the measured response *on a big baffle*.
Then, the user has to define the details of the target enclosure and load those into LEAP-5's Enclosure model and to link it to that tweaked driver. LEAP-5 is now capable of predicting those lumps and bumps, which are a combination of the break-up modes of the driver above, say 200Hz, (not modelled by LEAP-5 at all, but hand-added to the transducer model) plus the cabinet diffraction. Providing that the user has accurately tweaked the transducer model to represent the driver's performance on a big, non-reflective baffle, then the Enclosure model should make a good job of predicting what would be measured as the SPL of the driver mounted on the target enclosure."
--------
I thinks that's about right, and if I'd read that in the blurb, I wouldn't have started this thread! There is no shame in admitting the limitation of the program since *all* software has limitations and the price here for a million lines of code is extremely inexepensive and excellent value for money.
Well, I guess my only comment is that I prefer my listing of features and capabilities to the text you have above. You keep commenting on what you "think" the program does and does not do, but I still have the impression you have not worked with it much. So again you are back to making assumptions. You need to do some designs and then you will start to realize what the capabilities and features really are.
>Again you are back to talking about modeling mid/high frequency transducer response. Like I said there is no practical way to do this, and you state the same yourself when you talk about trying another program and found out exactly what I just said. So that is not a practical capability to expect.
But your thread started off asking about "baffle step", and EncShop is the ONLY program in the world that can model that correctly. That is diffraction. And EncShop is the king of 3D enclosure shell modeling. It is the only software that can do that.
What the LTD model gives you is a much more accurate model of a transducer at different power levels. Same with how it models satuaration and non linearities in the ports. Excellent large signal modeling. EncShop enables you to model any combination of transducers, chambers, and ports however you want to construct them. EncShop gives you field simulations around the box using high order diffraction for different shell designs. You can explore the differences of moving the transducers and ports to locations around the box. Lot's more.
As I see it, the calcualtion of the free-air efficiency, No and hence the "spl" of a driver arrives at a number. Since I have an LS3/5a and its bass/mid driver to hand, let's say that number is 86dB from the manufacturers data sheet, and my own measurement and T-S calculation by hand.
We know that the LS3/5a cabinet is 188mm wide. It is generally stated that the baffle step - being the change in sound pressure relative to frequency and relative to the cabinet width - results in a '6dB transition', the centre frequency, at which the response is up/down by 3dB depending upon your prespective is, I'm told f3 = 115/WB, (where WB, width of baffle is in metres). Hence, for an LS3/5a this calculates as 611Hz.
My initial question is, if LEAP models the drive unit from calculation of the T-S parameters (of an imported impedance curve or manual data entry) and curves the frequency response in a cabinet of 3/5a dimensions, at what frequency would we find the sensitivity of 86dB? Would we find that at the point of highest output - and if so, how can LEAP reliably predict that frequency considering that the driver may well not be behaving as a piston towards the top end of it's operating band, somewhere around 1kHz where that drive unit naturally peaks-out.
What I am curious about is how LEAP would calculate the SPL output in a region where we can expect the driver to be pistonic - say, 100-250Hz, plus make an reliable estimation of the baffle step situation.
I have a couple of follow-on questions acording to your answer. Thanks Chris - hope you are well.
Alan Shaw
Let's start by describing clearly what we are talking about in terms of transducer parameter model sets.
(1) EncShop suports 3 transducer models: STD, TSL, LTD. What you call Thiel-Small parameters is what I call the old STD (standard) model. It is the most limited, and has no ability to repesent the mid band SPL or impedance correctly at all. In fact, for sub woofers with very long voice coils the error of this model can be 6dB even at 100Hz.
(2) EncShop (and even LEAP-4) does not produce simple highpass filter simulations of the SPL response from the transducers, rather it uses an electroacoustical network to model the real current flow through the voice coil and generates the SPL based on that. Therefore, the impedance function is critical. The impedance of a speaker determines the current flow through the voice coil, and therefore what SPL is produced. Remember that all dynamic speakers produce SPL as a function of BL*i. It is the current flowing in the voice coil that matters.
(3) The ability to create an accurate impedance model is therefore crucial. The STD model is horrible at this, the TSL is much better, and the LTD model is by far the best. Just look at the differences between these real drivers and the three models on this page:
http://www.linearx.com/products/soft...p/EncShp05.htm
(4) The Reference Eff No is simply a theoretical number based on the old STD model of what the speaker would produce for SPL on a half-space (infinite baffle) condition based on the assumption that there is NO motor impedance, just Re. Well that is simply untrue, all dynamic speakers have a motor impedance which reduces the current in the voice coil, and therefore the real SPL is nearly always lower than what No and SPLo would suggest. However at the same time the Mmd value is decreasing in the midband because the outter edges of the cone are no longer moving. The apparent mass is lower.
(5) The "baffle step" is not some simple calculation of a single frequency where a 6dB transition occurs. How and where this 6dB increase occurs is entirely dependent on the box shape and where the transducers are mounted on that box. There are peaks, valleys, and ripples produced from the diffraction of the box as the radiation domain changes from 4Pi to 2Pi space due to the box.
(6) For the first time in history, EncShop provides you with the real ability to accurately model the true diffraction of the box and see exactly what the "baffle step" looks like. How can you see this alone? Well, simple. Just run the simulation in free space, and then with your box on an infinite baffle, and divide the curves. That is the diffraction effect of your box.
(7) You can explore these diffraction effects to your heart's content, and answer all of your questions yourself. Just take a box and run multiple simulations while moving the source around to different locations on the baffle board, corner, center, etc. You will see the response shape change due to the changing diffraction.
Thanks for a comprehensive reply Chris.
I have looked at the many examples given of the superior Z modelling capability of LTD and it looks very impressive, and I've previously been convinced about the better Z matching it offers. What I was really looking for was some published examples of my specific need - LEAP's handling of the 'baffle step'.
As far as I can tell, the acqusition of enough data for LTD to characterise a driver is a good deal more complex than that of the TSL, but without any explanation (such as a graph or two) of how much more capable LTD is of modelling the 'baffle step' I simply don't have the time (wish I did) to chase elegant concepts that may/may not give me a result that I can trust. Sorry if this sounds harsh, but those of us still trading in this shrinking industry are all working flat out.
Seems like a missed sales opportunity Chris - a few graphics would really make the point here. Or have I missed something? I bet I have.
I think you are confusing a couple issues. The transducer model (STD, TSL, LTD) has nothing to do with the diffraction analysis, or what you call the "baffle step". All of those models have the same capability to specify a diaphragm profile (flat, dome, cone) and its shape (round, square, triangular, hex, etc). Those transducer parameters are used by the diffraction engine along with the 3D shell design to determine the response around the enclosure. The old common transducer parameters you are thinking of have nothing to do with that.
Maybe I am missing something here, do you have LEAP-5 now? If you do not, then it is going to be difficult to explain much here describing things you have not seen yet. EncShop contains a massive diffraction engine. It calculates the true spatial field around a 3D shell based on exact driver locations. What you call the baffle step, EncShop analyzes as the entire field around the enclosure in full detail. Horz polar, Vert polar, etc.
I certainly want to answer any of your questions I can.
OK Chris, I've had LEAP-5 for a couple of years but not used it for reasons of time pressure.
According to page 72 of Enclosure Shop Apps manual, the tutorial for Transducer Location and diffraction uses a pre-prepared LTD model. I do not have an LTD model of my transducer, and as I noted before (and even in your own text!) acquiring data to make an LTD model is 'no mean feat'. So, it seemed to me that T-L-A-D-A was not applicable to a TSL-dataset user. Hence, dead stop.
Is this wrong?
Yes, that is wrong.
All three of the models STD, TSL, LTD have a common set of extended parameters which enable diffraction analysis. If you look at the transducer parameters dialog, you will see that you still have to choose the shape and profile of the diaphragm even if you are using a STD model. Those fields are ALWAYS used.
They control the modeling of a diaphragm, and all sources must have a radiating surface. How that surface behaves, what it's shape is, etc are independent of the low frequency modeling parameters.
Bottom line, you will get difffraction analysis with any transducer model. You have to, or there would be no SPL curves at all. All of the acoustic radiation is ALWAYS calculated through the diffraction engine. So you do not need a LTD model for that.
Noted Chris.
So, I sort of work my way back to the beginning where I think it's good prectice to predict the result of an experiment before actually conducting the experiment: if I load my B110 SP1228 date into a TSL (not LTD) and create an LS3/5a-like enclosure, would I expect to see the manufacturers free-field sensitivity of 86dB at, say 100Hz of your spl graph or at the very peak of the diffraction prediction - perhaps around 1kHz? This is a very important point.
I am not sure into which space (but I assume half space) the manufacturer's data sheet relates to.
I would add, that you will see from my own notebook:
[urlhttp://www.linearx.com/products/software/LEAP5/EnclosureShop/EncShp05.htm[/url]
(bottom curve) that to make the system response of the LS3/5a flat (with the 1228 bass/mid driver) approx. -9dB of eq is applied at 1kHz, resulting in a system with an spl of approximately 82dB across the pass-band from 100Hz upwards post-crossover equalisation.
Insufficient data. I have no data from you to be able to make such conclusions. I would suggest you run the analysis and find out. That's the best way.
OK Chris, I'll do it your way.
I know, for example, from many years of working with drivers, that the cone material and/or surround material (let alone profile), the method of adhesion of the cone to the surround, cone/surround overlaps etc. all *critically* effect the top end SPL of a drive unit by ± several dB's. I'll be looking for convincing that there is anything like enough data input to the model from the user to accurately model the axial lift. As you say 'insufficient (user) data'?
I don't care how ugly the user interface is, whether its DOS or Windows - I only care about saving my precious design time with acurate predictions of SPL from basic T-S parameters of the driver unit. If you have achieved this with such a small amount of data entry from the user, it's a miracle of programing and I'll be the first one to take my hat off to you.
Will you allow me to follow this up here with the results?
Alan
Well again you are mixing topics here. Now you are talking about cone breakup and high frequency behavior due to the cone etc. I do not try to model that. As you say there are no parameters for that. Rather, what I provide you with is a set of filters to contour the response at the high end to whatever your driver behaves like. So you can recreate those high frequency response details.
But none of that has anything to do with diffraction modeling, which was your original question. Diffraction happens around the box due to its shape, and is determined as well by the shape and location of the source(s). But at high frequencies the baffle appears as infinite and diffraction is insignificant anyway.
Ah ha! Now we are both talking at cross purposes, but with a common objective!
My original entry point was to understand how you would model the 'baffle step', which (admittedly) would exclude any cone break-up etc. and would, presumably be dominated by any on-axis SPL rise due to only the baffle dimesions: your diffraction model.
You say " ... what I provide you with is a set of filters to contour the response at the high end to whatever your driver behaves like. So you can recreate those high frequency response details.". That's fine Chris, but that assumes that I have an actual physical drive unit to hand that I can acoustically measure to determine what effect those break-up conditions have on the SPL.
I don't have a driver to hand. I'm playing with a conceptual driver, the T-S of which I'm modelling in LEAP-5. Hence, where I started - how can I predict what I'm calling 'the baffle step' solely from the T-S model. The answer seems to be ....
1. I can predict something of an SPL rise due to dimensions (but I don't know how, or what this will look like in your model) - you haven't commented to my question where the free-field "86dB" would line up on your curve, and
2. In addidtion, there is un unspecified and non-predicted "factor X" comprising other ±dB's of contribution to the driver's top end which is beyond the LEAP model and can only be 'drawn' by manual adjustment of the LEAP curve - assuming that there is a real driver to hand. A fiddle factor.
If that's correct - and it surely must be since neither you nor LEAP are psychic about cone break-up etc. etc. then the diffraction model is surely completely inadequate to reliably drive crossover design in the absence of a real, measured driver. The errors at the bass/mid drivers predicted response could be deviant with reality by, easily ± 3dB. Disappointing; back to the old way of making the driver (takes days), buring it in (more time), cutting a baffle (time), fitting it and measuring its SPL.
Am I wrong again Chris?
I think we are going in circles here. You have some preconceived ideas of the way you think things work, and in some other cases you are making incorrect assumptions. Again, a lot of your questions would be answered simply by running some designs and looking at the results.
EncShop produces a ton of data. You can change the shape of the box, change the location of the drivers, and see what effect it has on the polar fields all around the box. It is really amazing. You can test many of your assumptions as I have in the past, and the results will sometimes surprise you. I have been surprised on several occasions and learned things about the way things work that were different than how I had expected.
You can try all kinds of "what if" questions and see the real results. For example, What happens to the field response when you locate a port on the front or rear of a box? You can find out exactly what that difference is and answer that question. How much change is there if I put a 1" bevel around the baffle board? You can answer that question exactly. What changes in the response if I mount the driver in the center of the baffle or down at the bottom? Again, the answer is available with EncShop. All of this is due to diffraction analysis.
I enjoy answering these questions, but you must understand that it would take many pages to do so here and I do not have the advantage of drawing you pictures and other visual aids that would help your understanding. The fact is, I have already done all of this in the manuals. It is much easier to read the info there because I have much more space for detailed descriptions and many pictures and drawings as well.
Read Chapter/Section 4.12. It explains how EncShop represents the radiating surface of a transducer for diffraction analysis. Chapter 4 is all about transducer modeling. Here are the sections:
4.1 Introduction .................................................. ............ 43
4.2 References & Standards .................................................. . 44
4.3 Dynamic Analogies .................................................. ........ 45
4.4 Radiation Impedance .................................................. ..... 46
4.5 Acoustic Network Analysis ............................................. 51
4.6 Acoustic Current .................................................. ........... 53
4.7 Electrical Impedance .................................................. ...... 54
4.8 Acoustic Pressure & Directivity ........................................ 55
4.9 Acoustic Power & Efficiency ............................................ 58
4.10 Highpass Filter Approximation ......................................... 61
4.11 Motor Impedance .................................................. ........ 63
4.12 Diaphragm Structure .................................................. .... 75
4.13 Diaphragm Breakup .................................................. ..... 81
4.14 Diaphragm Suspension .................................................. .. 89
4.15 Magnetic Gap .................................................. .............. 95
4.16 Temperature & Power Compression ............................. 103
4.17 Model Performance & Comparison ............................... 105
You should also read chapter/section 5.9 on Diffraction.
I think reading some of these sections will help you far more than anything I can put into words here, without pictures. Then if you have some specific questions about the items covered in the manuals we will have a common base to start a discussion from that will make it easier for both of us.
Thanks again Chris for a comprehensive reply. Yes, I have of course made assumptions (and possibly unreasonable ones) that you/LEAP-5 are the ultimate in loudspeaker modelling. I know that in the entire modelling industry there is no one who has applied more mental effort to getting to the bottom of the science.
My selfish concern is simply one of my precious time. The example graphs you show are indeed interesting, but if they don't predict how a real transducer will acoustically behave from solely the T-S model, then they are of artistic merit and not inherent value to me.
OK here is the deal then: I will give up my sunday (and the fact that my kids have come to visit for the weekend) to running through the entire work list you've given me. I know (and I have verified) the T-S parameters of a selection of drivers. I have confidence in that data (as much as one ever can have). Obviously I can measure the dimensions of the box in which they are mounted. I can measure the 1m SPL response outside, up high, and I'll do that too with my B&K certified microphone. I will then compare against your diffraction model which - honestly - I really want to believe is very close to actual because if the model works it will save me a lot of time in the future.
I'll let you know what I find out.
It becomes difficult to answer your questions when you say you do not have the time to read the answers. You keep confusing diffraction with transducer modeling. They are not the same thing at all. Diffraction is mainly a function of the cabinet shell. The transducer only matters in how it presents the location of the acoustic source, and its shape. Nothing in the 75 year old conventional electroacoustic parameters of the transducers has anything to do with this. Read sections 4.12 and 5.9 those should help the most.
You are correct to the extent that I have mixed together what you have specified as the separate roles of the tranducer and enclosure modelling. This is because I am looking for a global solution to the reason the SPL curve rises (the 'baffle step' +++ whatever else is happening in the transducer or enclosure environment) and an ability to model it, quickly, on a laptop, on the beech with only the T-S driver parameters and LEAP-5.
However, I have made a mistake in my understanding of how LEAP works and it is clearly stated on chapter 4.13, Cone breakup (p.81) "By blending together and adjusting their parameters [of the transducer model] a wide variety of high frequency response shapes can be produced [ie. predicted]. This is largely a trial and error process to find the optimum parameters which fit the transducer". That is perfectly clear. Trial and error.
It means that, setting aside the diffraction issue, it is not possible (as I feared) to ask LEAP to automatically generate a reliable SPL curve solely *from the T-S data of a driver* that could be taken as a reliable model of the measured SPL such that crossover work could be undertaken. As you say yourself, "Finite element methods have been attempted for many years, but these require tremendous amounts of detailed information about the materials and construction of the device".
Hence, the user has to conceive the driver, model it the transducer model, physically build it, measure its *actual* SPL and tweak the model (Fmd, Qmd, Flp, Qlp - page 81) to give a good correlation with actual. I doubt that this will save me time but I clearly expected far too much of LEAP-5 which is not promoted as a finite element system: my over-optimism took hold.
There is no point in gathering the data I previously suggested because I already have the drive unit in existance. That's all I need.
You are not going to find anything to model and predict the mid and high frequency response of a transducer in any practical sense. It's just not practical. Nothing in the electromechanical parameters covers this, and the T-S stuff is just purely for the old fashion simplistic highpass filter approximation. Remember the mid/high freq response shape of a highpass filter is a flat line.
Many do not realize that to even utilize FEA type methods requires 3 independent FEA solutions: one for the magnetic system, one for the mechanical system, and one for the acoustic system. But that's not even half the problem. You need parameters for those models. Young's modulus for the cone, the surround, the spider, glue seams, mag coefs for the steel alloys, etc. You would need a zillion parameters that you have never heard of before.
Yes Chris, you are right. And even a program that we bought from a European supplier to do cone modelling proves to be completely useless for exactly the reasons you state.
I really do wish that you'd made this all crystal clear for dimwits like me. I've rewritten your sales blumb here .... and this is what I'd like to have read ....
------------
"LEAP-5 is capable of remarkable time saving in many areas of speaker design, and has a very reasonable price tag. It does not replace complex Finite Element Analysis as it has neither the data modelling capability nor the user interface to do that, but it is very simple to use. It looks complicated at first, and that may give the impression that it can do more than it really can.
Any speaker system designer knows that when you take a nice flat-measuring drive unit (measured on a big baffle) and put it on the face of a cabinet and re-measure the SPL, it looks horrible. Lumps and bumps appear in the response that are not there on the baffle curves. The system designer doesn't care much how or why these anomalies occurred - they are there whether or not he likes is and he has just one simple choice: to eq them out or not in his crossover design since the marketing guys have probably already fixed the cabinet dimensions, so moving the drivers about on the baffle has limited potential.
However, LEAP-5 very much does care about the reasons for those deviations from flat, and it has to take two quite separate but complimentary approaches to understanding and modelling them. First, the user has to make as T-S transducer model which LEAP-5 can very effectively model up to, say, 200Hz. Then, the user has to manipulate that model by hand until the frequency range above the limit of LEAP-5's capabilities correlates well with the measured response *on a big baffle*.
Then, the user has to define the details of the target enclosure and load those into LEAP-5's Enclosure model and to link it to that tweaked driver. LEAP-5 is now capable of predicting those lumps and bumps, which are a combination of the break-up modes of the driver above, say 200Hz, (not modelled by LEAP-5 at all, but hand-added to the transducer model) plus the cabinet diffraction. Providing that the user has accurately tweaked the transducer model to represent the driver's performance on a big, non-reflective baffle, then the Enclosure model should make a good job of predicting what would be measured as the SPL of the driver mounted on the target enclosure."
--------
I thinks that's about right, and if I'd read that in the blurb, I wouldn't have started this thread! There is no shame in admitting the limitation of the program since *all* software has limitations and the price here for a million lines of code is extremely inexepensive and excellent value for money.
Well, I guess my only comment is that I prefer my listing of features and capabilities to the text you have above. You keep commenting on what you "think" the program does and does not do, but I still have the impression you have not worked with it much. So again you are back to making assumptions. You need to do some designs and then you will start to realize what the capabilities and features really are.
>Again you are back to talking about modeling mid/high frequency transducer response. Like I said there is no practical way to do this, and you state the same yourself when you talk about trying another program and found out exactly what I just said. So that is not a practical capability to expect.
But your thread started off asking about "baffle step", and EncShop is the ONLY program in the world that can model that correctly. That is diffraction. And EncShop is the king of 3D enclosure shell modeling. It is the only software that can do that.
What the LTD model gives you is a much more accurate model of a transducer at different power levels. Same with how it models satuaration and non linearities in the ports. Excellent large signal modeling. EncShop enables you to model any combination of transducers, chambers, and ports however you want to construct them. EncShop gives you field simulations around the box using high order diffraction for different shell designs. You can explore the differences of moving the transducers and ports to locations around the box. Lot's more.
) geht eher in die Richtung, den Ansatz von AkAbak weiterzuführen: Beschreibung der Oberfläche eines Gehäuses durch ein feinmaschiges Netzwerk aus Wellenleitern.
Kommentar