Thiel small parameter testing is a subject always under discussion.
I saw in Voice coil magazine Mike Klasco & Vance Dickason are testing the speakers with LMS LTD method.
I saw one magazine testing with some other method.
I thought to discuss here with others having expertise & knowledge.
Report of a magzine : Testing methods :
Using the very best computer based hardware and software products available:
a. Cliowin 7QC
b. Goldline TEF 25
Thiele & Small Parameters, hereafter referred to as T&S, are parameters derived
from impedance measurements (resistance at varying or fixed frequencies).
To get a full set of T&S’s, two impedance curves are needed: the first is the
loudspeakers Free Air Impedance, the second, is impedance obtained by
either Delta Mass or Delta Compliance method. The first consists of
adding a known mass to the cone of the driver, the latter in loading the
cone with a box of known air volume. We supports the viewpoint; there is no
better substitute for air - then, air! Therefore, we measure our second
impedances using Delta Compliance.
There are four popular methods of impedance measurement: Constant Voltage,
Constant Current, Sinusoidal and MLS. The most accurate is MLS then Sinusoidal
followed by either of the remaining. For this and other reasons, results often vary.
We have decide to implement just Sinusoidal, as it is the most widely used amongst
manufactures, however, we do use MLS to rule out driver defect. .
Lastly we use a more sophisticated calculation method, know as Least Square Error (LSE).
LSE improves the theoretical extrapolations and can reveal defects in drivers measured at
high output levels. In short, LSE calculations provide more accurate T&S parameters
providing the driver isn’t defective.
About LTD testing in LMS:
The LTD model derivation utilizes advanced optimization techniques and highly
refined algorithms to distill sophisticated characteristics solely from impedance
data. The data may be measured at different power levels and temperatures.
The measurement of transducer impedance has traditionally been accomplished
through the use of constant current source CCS techniques
The LTD model uses Constant Voltage Source CVS.
The operating conditions imposed by the CCS method are in reality the exact
opposite of normal loudspeaker operation. Typically a loudspeaker is driven
directly from a power amplifier at virtually zero source impedance. The drive
method is by constant voltage source CVS. In CVS operation the speaker is driven
with a linear voltage vs. frequency characteristic, with virtually zero source
impedance and at a much higher power level.
In order to determine impedance using the CVS method, a current shunt is
employed to measure the current flow through the device. Knowing the current
flow through the device, and the voltage across the device, the true impedance can
then be calculated using basic Ohm’s law as Z=V/I.
I saw in Voice coil magazine Mike Klasco & Vance Dickason are testing the speakers with LMS LTD method.
I saw one magazine testing with some other method.
I thought to discuss here with others having expertise & knowledge.
Report of a magzine : Testing methods :
Using the very best computer based hardware and software products available:
a. Cliowin 7QC
b. Goldline TEF 25
Thiele & Small Parameters, hereafter referred to as T&S, are parameters derived
from impedance measurements (resistance at varying or fixed frequencies).
To get a full set of T&S’s, two impedance curves are needed: the first is the
loudspeakers Free Air Impedance, the second, is impedance obtained by
either Delta Mass or Delta Compliance method. The first consists of
adding a known mass to the cone of the driver, the latter in loading the
cone with a box of known air volume. We supports the viewpoint; there is no
better substitute for air - then, air! Therefore, we measure our second
impedances using Delta Compliance.
There are four popular methods of impedance measurement: Constant Voltage,
Constant Current, Sinusoidal and MLS. The most accurate is MLS then Sinusoidal
followed by either of the remaining. For this and other reasons, results often vary.
We have decide to implement just Sinusoidal, as it is the most widely used amongst
manufactures, however, we do use MLS to rule out driver defect. .
Lastly we use a more sophisticated calculation method, know as Least Square Error (LSE).
LSE improves the theoretical extrapolations and can reveal defects in drivers measured at
high output levels. In short, LSE calculations provide more accurate T&S parameters
providing the driver isn’t defective.
About LTD testing in LMS:
The LTD model derivation utilizes advanced optimization techniques and highly
refined algorithms to distill sophisticated characteristics solely from impedance
data. The data may be measured at different power levels and temperatures.
The measurement of transducer impedance has traditionally been accomplished
through the use of constant current source CCS techniques
The LTD model uses Constant Voltage Source CVS.
The operating conditions imposed by the CCS method are in reality the exact
opposite of normal loudspeaker operation. Typically a loudspeaker is driven
directly from a power amplifier at virtually zero source impedance. The drive
method is by constant voltage source CVS. In CVS operation the speaker is driven
with a linear voltage vs. frequency characteristic, with virtually zero source
impedance and at a much higher power level.
In order to determine impedance using the CVS method, a current shunt is
employed to measure the current flow through the device. Knowing the current
flow through the device, and the voltage across the device, the true impedance can
then be calculated using basic Ohm’s law as Z=V/I.
There are problems with all of these methods since they all require data from two different runs and the assumption is made that nothing changes between these two runs except what we intended to change. This latter assumption is not very accurate.
Some years ago a study was done at Ford, when I was there, on all these methods and they were all found lacking. The parameters changed with the amount of mass added as well as the amount of compliance added. The actual amounts of these changes (perturbations) should all yield the same answers but they don't.
In the early 90's I patented a method which uses three measurements taken simultaneously; they are the current spectrum, the voltage spectrum and the sound output spectrum (which can be inside a closed box for best accuracy. Since there is no "perturbation" of the system this technique is very accurate and stable.
Using LSE (or any fitting technique) the parameters of the voltage, current and SPL are calculated. From these curve parameters one can calculate all of the TS parameters and with one change in procedure - but no purturbation - one can actually calculate the radiating area - which is always assumed to be know, but never actually is.
Some years ago a study was done at Ford, when I was there, on all these methods and they were all found lacking. The parameters changed with the amount of mass added as well as the amount of compliance added. The actual amounts of these changes (perturbations) should all yield the same answers but they don't.
In the early 90's I patented a method which uses three measurements taken simultaneously; they are the current spectrum, the voltage spectrum and the sound output spectrum (which can be inside a closed box for best accuracy. Since there is no "perturbation" of the system this technique is very accurate and stable.
Using LSE (or any fitting technique) the parameters of the voltage, current and SPL are calculated. From these curve parameters one can calculate all of the TS parameters and with one change in procedure - but no purturbation - one can actually calculate the radiating area - which is always assumed to be know, but never actually is.
jaya000 said:
Go to www.USPTO.gov and look up my name Earl Geddes and find the patent. Thats the best source for information. No one has put this into a system yet.
Dr. Gedee, have you ever considered working with one of the many software companies to sell your approach for the betterment of the measurement methods. Looking at your patent, I think that I could probably replicate your proposed method in a very ham fisted way using a tone generator, impedance meter, and amplifier, along with a relatively simple spreadsheet application. None the less, its not something I would want to do, especially when it could be integrated into an analysis suite for not much money. If I had my druthers, I would probably look to Liberty or ATB, since then I might stand a chance of actually owning the software as well.
Can anyone tell me if the method of measurement itself could throw off the results much, with regard to driver mounting, and test leads, etc. I know with measuring inductance, you have to be very very careful, yet I find most of these products use magnetic elements on or around the leads. Does it matter? Then you have the mounting method of the driver. I noticed in the Linear X "white paper" on their preferred method, the drivers were clamped, by the magnet, with a large steel engine stand. My concern here would be the attachment of the steel, which I would think would impact coil inductance, BL, and maybe a few other parameters I can't think of at the moment. If thats the case, how should we mount drivers for testing, again, non-ferrous methods?
Can anyone tell me if the method of measurement itself could throw off the results much, with regard to driver mounting, and test leads, etc. I know with measuring inductance, you have to be very very careful, yet I find most of these products use magnetic elements on or around the leads. Does it matter? Then you have the mounting method of the driver. I noticed in the Linear X "white paper" on their preferred method, the drivers were clamped, by the magnet, with a large steel engine stand. My concern here would be the attachment of the steel, which I would think would impact coil inductance, BL, and maybe a few other parameters I can't think of at the moment. If thats the case, how should we mount drivers for testing, again, non-ferrous methods?
pjpoes said:
The easiest way is just to use a PC and sound card. The three channels are best done at one time, but it could be done in two passes. Two sound cards could also work. Fitting the curves takes some computer power, but its not difficult.
I have talked with a couple of measurement companies Nobody is interested in using someone elses inventions. So in the meantime users just have to use what the developers choose to program.
I would expect your concerns to not be too significant. But the fact is that the TS parameters don't really tell us much anyways. Heck, I don't use them at all, so I guess that I don't actually care how accurate they are
Oh wait, I see that you have software available, which appears to use an alternate method of inductance modeling, but it looks like we would have to build the actual test apparatus our selves to use your method. None the less, after reading the patent more thoroughly, I imagine that, again, it could be accomplished relatively simply. I guess I need to add to what I wrote earlier, I would want either a clamp on multimeater or at least an inductive loop to measure the current flow, as well as a handful of power resistors. Looks like no impedance meter is needed, we would be creating that essentially.
I have to say, looking at your patents, and what they were meant for, I have a whole new respect for manufacturers efforts in car audio. I honestly thought that they just found some random convenient places to stick holes for the drivers, had the cheapest possibly drivers created, and connected them in the cheapest possibly fashion. While some of that may be true, it appears that strong science and engineering went in to much of the design, at least as far as Ford is concerned.
I have to say, looking at your patents, and what they were meant for, I have a whole new respect for manufacturers efforts in car audio. I honestly thought that they just found some random convenient places to stick holes for the drivers, had the cheapest possibly drivers created, and connected them in the cheapest possibly fashion. While some of that may be true, it appears that strong science and engineering went in to much of the design, at least as far as Ford is concerned.
One would simple measure the current with a small series resistor. Use say .1 ohm in series with the driver, and the voltage across it will be 1/10 of the current. Sample the voltage across the driver and you have all the electrical quantities. Put the driver on a small closed box and measure the pressure inside. All of this can be done with a PC.
I have to chuckle about the "Ford" comment. Ford became totally disinterested in audio about ten years ago. Basically got rid of all those "scientists" - had no use for them in a commodity marketplace. The last seven of my patents, including the one being discussed here, I did on my own. Ford had no interest in them.
Ford sold off the Audio Group as Visteon and Visteon "cash cowed" it into oblivion. There is nothing left of the audio group that I worked with. Visteon is on the sales block with virtually no interested buyers - they have nothing left to buy.
"I honestly thought that they just found some random convenient places to stick holes for the drivers, had the cheapest possibly drivers created, and connected them in the cheapest possibly fashion. "
This is all basically true. Some "premium" systems are a little better, but ALL standard ones are just as you imagine.
I have to chuckle about the "Ford" comment. Ford became totally disinterested in audio about ten years ago. Basically got rid of all those "scientists" - had no use for them in a commodity marketplace. The last seven of my patents, including the one being discussed here, I did on my own. Ford had no interest in them.
Ford sold off the Audio Group as Visteon and Visteon "cash cowed" it into oblivion. There is nothing left of the audio group that I worked with. Visteon is on the sales block with virtually no interested buyers - they have nothing left to buy.
"I honestly thought that they just found some random convenient places to stick holes for the drivers, had the cheapest possibly drivers created, and connected them in the cheapest possibly fashion. "
This is all basically true. Some "premium" systems are a little better, but ALL standard ones are just as you imagine.
jaya000 said:
What difference does that make? For a closed box?
My subs are such that the TS parameters don't matter there either as the box parameters dominate the problem.
I'm quite serious, I haven't a clue what the TS parameters are or what the final Q's of any of my systems are. (I know all the drivers have very low Q's, thats all I need to know.) The room dominates the LF situation in the end and once you fit the mains and subs into the room (properly setup of course), the LF parameters of the drivers just end up not being a factor.
And in the big picture of sound 20 Hz -> 20 kHz how important could a set of a few numbers that don't mean a thing above about 50 Hz be?
I have never understood the audio worlds obsession with a single HP filter that operates in an environment where there are virtually hundreds of high Q complex modes.
I must say that these comments reinforce some suspicions I have myself about the usefulness of T/S parameters. I read over and over again about how bad the Thiel Small method is, how we need to use these other forms which can take into account thermal compression, behavior at higher levels, etc. In the end though, my experience has been that the T/S gives me a very rough ball park starting point, and the more boxes I design, the better a feel I have for, well, winging it.
As some may have noted, I began designing some crossovers around these drivers I picked up (Which I finally fixed by dismantling the tweeters and reassembling them fresh). I've taken a lot of grief both on forums and here at the University for not relying heavily enough on my t/s parameters and impedance curves. I finally got measurements of the woofers in a test box, and what do you know, my best judgments ended up giving a better response (1.8cf) than those of some so called desk jockey experts (1cf). Additionally, the crossover, which I haven't really mocked up yet, I am relying on simulations for the moment, relies very heavily on the impedance curve. Here again I was pushed to get to a lab and measure on the Linear X LEAP system so that I could model the crossovers more appropriately. I did so over the weekend, and found that, after all that work, my crossover design and baffles were close enough. There was never more than a 1db difference between the system modeled using the LEAP impedance curve, and that of the WT3. I'm far from an expert (I'm in my second year of my PhD program in Child Development, a soft science pretty far separated from acoustic engineering) but I sometimes feel that we pay too much attention to minor differences in the numbers.
As some may have noted, I began designing some crossovers around these drivers I picked up (Which I finally fixed by dismantling the tweeters and reassembling them fresh). I've taken a lot of grief both on forums and here at the University for not relying heavily enough on my t/s parameters and impedance curves. I finally got measurements of the woofers in a test box, and what do you know, my best judgments ended up giving a better response (1.8cf) than those of some so called desk jockey experts (1cf). Additionally, the crossover, which I haven't really mocked up yet, I am relying on simulations for the moment, relies very heavily on the impedance curve. Here again I was pushed to get to a lab and measure on the Linear X LEAP system so that I could model the crossovers more appropriately. I did so over the weekend, and found that, after all that work, my crossover design and baffles were close enough. There was never more than a 1db difference between the system modeled using the LEAP impedance curve, and that of the WT3. I'm far from an expert (I'm in my second year of my PhD program in Child Development, a soft science pretty far separated from acoustic engineering) but I sometimes feel that we pay too much attention to minor differences in the numbers.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.