This is an outgrowth of this other thread
https://www.diyaudio.com/community/...s-really-matter-for-sound.409770/post-7624912
… which turned into a wide ranging discussion about new forms of distortion, new ways of measuring and presenting distortion, and several other sub-topics. While that thread is informative and entertaining, it glosses over the practical application of measurements.
I would like to focus on those practical measurement techniques that are available today to a speaker designer. I want to discuss
(1) those measurements which should be used to select the best drivers for a project.
(2) The measurements needed for simulation and to implement the design.
(3) The various techniques and equipment that people use to perform the measurements.
I would also like for us to distinguish between those measurements which are crucial to the design process, and those which are nice to have because they make the design process faster. Which measurements become important when aiming for top performance?
So I will start.
Impedance sweep: A driver impedance sweep both in free air, AND as installed in the cabinet, is one of the most crucial measurements. It reveals the driver Fs, the Le, the Re. Impedance wobbles show the existence and frequency of potential resonances. The impedance data is used in simulation to model passive crossovers. I use DATSv3 for this. It automatically calculates the Fs, Re, Le, and for sealed box designs, the Qt of the driver+cabinet.
j.
https://www.diyaudio.com/community/...s-really-matter-for-sound.409770/post-7624912
… which turned into a wide ranging discussion about new forms of distortion, new ways of measuring and presenting distortion, and several other sub-topics. While that thread is informative and entertaining, it glosses over the practical application of measurements.
I would like to focus on those practical measurement techniques that are available today to a speaker designer. I want to discuss
(1) those measurements which should be used to select the best drivers for a project.
(2) The measurements needed for simulation and to implement the design.
(3) The various techniques and equipment that people use to perform the measurements.
I would also like for us to distinguish between those measurements which are crucial to the design process, and those which are nice to have because they make the design process faster. Which measurements become important when aiming for top performance?
So I will start.
Impedance sweep: A driver impedance sweep both in free air, AND as installed in the cabinet, is one of the most crucial measurements. It reveals the driver Fs, the Le, the Re. Impedance wobbles show the existence and frequency of potential resonances. The impedance data is used in simulation to model passive crossovers. I use DATSv3 for this. It automatically calculates the Fs, Re, Le, and for sealed box designs, the Qt of the driver+cabinet.
j.
Driver suspension breakin is advised when measuring TSPs and its a good idea to measure both before and after results. Averaging the results and applying them to the cabinet / xover design will yield more consistent results, plus it gives you an indication later on when something isn't right. I save my impedance measurements before and after, including the finished product to have a reference for further development (xover tweaks) or troubleshooting..
I think measuring TSP, and then designing around them, is most important in a passive vented box speaker where no DSP augmentation is planned. If I was going to build a vented box speaker, I go beyond my DATS tool, and measure the parameters at both low voltage and high voltage... 0.1 V (using DATS), and then something like 2 V, using the well known methods.
Since I mostly design closed box speakers, I never bother to measure Vas. I just measure Fs and Qts using DATS at a low signal level, and compare it to factory specs to make sure I am reasonably close. Sealed box designs are so forgiving in this way.
If anyone has an unconventional or clever way of measuring impedance and TSP, that would be a very interesting post.
j.
Since I mostly design closed box speakers, I never bother to measure Vas. I just measure Fs and Qts using DATS at a low signal level, and compare it to factory specs to make sure I am reasonably close. Sealed box designs are so forgiving in this way.
If anyone has an unconventional or clever way of measuring impedance and TSP, that would be a very interesting post.
j.
TSP has never been intended to be anything else than small signal parameters, thus not necessary representative for higher drive levels. There seems to be an obsession among diy builder to be extremely focused on optimized-to-the last-digit vented systems, whereas i.m.o. we are not so sensitive to optimum tuning at all. Next thing is the room itself and next item after that placement in the room.
I use REW for TSP measuring with added mass method. Very easy and fast when using some coin what mass is known, most important is to tape the coin with tape (I use electrical insulation tape) properly to the cone and later take it off not damaging the cone.
To start with TSP evaluation at first "TspCheck" was installed :
https://hifi-selbstbau.de/index.php?option=com_content&view=article&id=199&Itemid=82
The parameters from the manufacturers data sheet were filled in. If this dataset is internally consistent it may be valid. If not it is a set of lies!
To measure the most expensive to manufacture and most lied on parameter: BL (!!) the following was performed:
The cheap continuously variable DC power supply 1.5-15V / 1.5A was pulled from the stash.
The best (for current measurement) multi meter was used.
Some plastic bags were filled with rice. 102 grams mass are 1 Newton.
The transducer was placed with its magnet on a safe true horizontal surface. Cone directing ceiling.
A small piece of graph paper was creased and taped to the underside of the cone. To make up a vertical scale. From wire a pointer was made to read the position (excursion) of the cone. A desk lamp was used to light the graph paper and the wire pointer. A sharp shadow may help.
The best current reading multi meter was connected in series with the transducer and the DC-supply.
The rest position of the cone/voice coil was meticulously established with the piece of graph paper and the wire pointer.
Then the prepared rice bag(s) (labeled in Newton) were carefully put on the cone. The Newtons/(rice grains) were increased until the coil started to leave the magnetic gap (more than negative X-max). Then the load was lifted with the current through the voice coil, back to rest position. The current was read off the multi meter.
Please, please rehearse this meticulously: Lift to rest position, read current, quick, immediately turn DC down to zero ...again and again. The DC will toast the voice coil.
BL is...
Next:
Best regards
Bernd
https://hifi-selbstbau.de/index.php?option=com_content&view=article&id=199&Itemid=82
The parameters from the manufacturers data sheet were filled in. If this dataset is internally consistent it may be valid. If not it is a set of lies!
To measure the most expensive to manufacture and most lied on parameter: BL (!!) the following was performed:
The cheap continuously variable DC power supply 1.5-15V / 1.5A was pulled from the stash.
The best (for current measurement) multi meter was used.
Some plastic bags were filled with rice. 102 grams mass are 1 Newton.
The transducer was placed with its magnet on a safe true horizontal surface. Cone directing ceiling.
A small piece of graph paper was creased and taped to the underside of the cone. To make up a vertical scale. From wire a pointer was made to read the position (excursion) of the cone. A desk lamp was used to light the graph paper and the wire pointer. A sharp shadow may help.
The best current reading multi meter was connected in series with the transducer and the DC-supply.
The rest position of the cone/voice coil was meticulously established with the piece of graph paper and the wire pointer.
Then the prepared rice bag(s) (labeled in Newton) were carefully put on the cone. The Newtons/(rice grains) were increased until the coil started to leave the magnetic gap (more than negative X-max). Then the load was lifted with the current through the voice coil, back to rest position. The current was read off the multi meter.
Please, please rehearse this meticulously: Lift to rest position, read current, quick, immediately turn DC down to zero ...again and again. The DC will toast the voice coil.
BL is...
Next:
- coil Rdc
- Fs
- Qes
Best regards
Bernd
Last edited:
Thanks Hörnli for pointing this out. Da Aboriginal elders Thiele & Small were academics and da TSPs tend to obscure more important information on units.
The important parameters are Bl, Rdc, moving mass & effective cone area. In production we would measure these directly; tearing down a sample of units to cut out the moving parts and weigh them directly. You can measure the effective cone area with a ruler from the top of the surround to the other side.
You might like to try this out with one of the simulation programmes by keeping these 4 parameters constant while allowing Fs, compliance, etc to vary by 10% or more from an optimum B4 vented box or a B2 closed box. You'll see the changes are hardly worth noting.
So use any method to dream up TSPs but measure the above 4 parameters directly and perhaps put them back into your dreamt up TSPs to refine them.
PS You don't have to use such high currents for Bl but you might want a travelling microscope to see if the cone has returned to the 'rest' position.
The important parameters are Bl, Rdc, moving mass & effective cone area. In production we would measure these directly; tearing down a sample of units to cut out the moving parts and weigh them directly. You can measure the effective cone area with a ruler from the top of the surround to the other side.
You might like to try this out with one of the simulation programmes by keeping these 4 parameters constant while allowing Fs, compliance, etc to vary by 10% or more from an optimum B4 vented box or a B2 closed box. You'll see the changes are hardly worth noting.
So use any method to dream up TSPs but measure the above 4 parameters directly and perhaps put them back into your dreamt up TSPs to refine them.
PS You don't have to use such high currents for Bl but you might want a travelling microscope to see if the cone has returned to the 'rest' position.
Break in, no such thing.
Warm up, shure that is a thing. Flex and warm up the driver each time that you measure. A cold driver does not measure consistently or reliably.
TSP no such thing as "small" either other than the guys name. Measure parameters at 2 or so volts, something near the voltage where the drivers are used most of the time. Not to the point of compression due to voice coil heat.
Thanks DT
I often drill a 1/4" pluss a little in the enclosure and measure SPL inside the box. The measurement shows a lot about resonance and standing waves inside the enclosure. I have a couple of GRAS 1/4" microphones made for this and similar high SPL applications.
Last edited:
Small signal parameters is just a way of isolating the considered behaviour from power concerns, nonlinearity and such.
@DualTriode Break in is definitely a thing with any dynamic driver possessing an elastic suspension. The suspension has small fracture points which permanently give after the first higher input operation. The resonant frequency Fs will decrease, therefore changing other TSP related perameters as well. This shouldn't be confused with heating from internal suspension losses, temporarily lowering Fs, then returning back to higher Fs after cooling back down to ambient temp.
The amount of permanent Fs change depends on suspension materials and their age, suspension adhesive type, total time and state of long term cure after driver assembly. An average lowering of Fs can be from a few percent to over several percent.
Its practical to use warmed up suspension TSPs to design your enclosure. The problem with measuring TSPs consistently is using a standardized drive level, as dampening Q and Fs measure lower with higher drive amplitude levels.
The amount of permanent Fs change depends on suspension materials and their age, suspension adhesive type, total time and state of long term cure after driver assembly. An average lowering of Fs can be from a few percent to over several percent.
Its practical to use warmed up suspension TSPs to design your enclosure. The problem with measuring TSPs consistently is using a standardized drive level, as dampening Q and Fs measure lower with higher drive amplitude levels.
Actually there is one more important step, which we basically can call poormans Klippel LSI.
Which is taken a look in how much the Fs shifts at higher voltages, as well as how much the height of this impedance peak changes.
The first one gives us an indication how well the Cms (Kms) behaves, the other one how well the BL behaves.
Especially the BL will have a huge impact not just on distortion, but quite literally system response as well.
Something that is basically being overlooked by everyone btw.
Unfortunately we don't have any information as function of cone excursion, but the more stable these two parameters behave, the better.
I find a DATS impedance WAY limited for any impedance testing in general.
It will just work for simple passive filtering, but that's about it.
I have measured and tested hundreds of speakers and TS at this point.
The story most people tell that it needs to be super accurate is just not true at all, just an absolute myth
(from people who haven't tested much themselves?)
Basic parameters like Fs, Qms, Qes, Qts and Re can always be determined easily.
Even just added mass method will work just fine.
The biggest problem is that manufactures make up their own way of testing, incl random testing signals.
Often not equal to the AES standard of 0.1V !!
I have shown this multiple times in other topics, so people can look that up.
Or maybe I will show that here more once again.
Breakin depends much on the driver. In the distant past for sure, today with purifi drivers the diff is small. I use sealed enclosures so little need for all the tsp stuff. I did measure drivers after some 15 years use and the measurements were very close. So a few hours breakin is enough to get a stable situation. I also record the room temperature, as that can have some impact.
More of interest to me is woofer resonance behaviour at 1 to 4 watt. Also to assure the damping material (i use a kind of sheep wool) does not introduce variations. Basically Fs and Qts, and measured with Arta Limp.
For the recent imp meas with the purifi drivers i use a low value Rsense, about 0.135 ohms. And the parameters measured are quite close to what purifi publishes.
Last edited:
I always got very good results using T/S parameters for box design.
I used both the added mass method, and known enclosure size. I had a couple baffles for different sized drivers that would fit on two known volumes. You had to know the volume of air from the front of the driver basket to the cone surface (measured).
For added mass, I used carefully weighted putty which was stuck around the dust cap joint to the cone. Again, different masses to match the size of the woofer.
These were always accurate and repeatable. The two methods agreed with each other closely enough and I ended up using the added mass method most often. The directions I followed were outlined by Douglas Weems in his first book.
For enclosed midranges and tweeters, you could only get "system values" if you wanted non-destructive testing. To be honest, that's all you needed to know.
I don't think I would ever run direct current through a voice coil to see how much correction is needed to restore the centre position, certainly not with enough mass to deflect the coil out of the gap or even anywhere near x max. I think that just deflecting a 1 cm would be enough - and probably give more accurate results. IF you don't test a parameter the same way exactly that the published values were obtained, your results are very probably not valid.
I used both the added mass method, and known enclosure size. I had a couple baffles for different sized drivers that would fit on two known volumes. You had to know the volume of air from the front of the driver basket to the cone surface (measured).
For added mass, I used carefully weighted putty which was stuck around the dust cap joint to the cone. Again, different masses to match the size of the woofer.
These were always accurate and repeatable. The two methods agreed with each other closely enough and I ended up using the added mass method most often. The directions I followed were outlined by Douglas Weems in his first book.
For enclosed midranges and tweeters, you could only get "system values" if you wanted non-destructive testing. To be honest, that's all you needed to know.
I don't think I would ever run direct current through a voice coil to see how much correction is needed to restore the centre position, certainly not with enough mass to deflect the coil out of the gap or even anywhere near x max. I think that just deflecting a 1 cm would be enough - and probably give more accurate results. IF you don't test a parameter the same way exactly that the published values were obtained, your results are very probably not valid.
whether you use the manufacturer provided (Neville)Thiele/(Richard)Small parameters, Advanced Parameter Model, or other models eg. (Wolfgang) Klippel’s suggested Large + Cold models, you can always check your actual tuning frequency with whatever power level you want, using a good old multimeter and a few resistors:
To determine the resonance frequency of the system, set the multimeter to current measurement, then hook it up in series with the subwoofer to your amplifier, then use (a tone) frequency generator to drive the (sub)woofer. At the resonance frequency, the meter will read the lowest result (sealed enclosure) or the highest result (ported and bandpass enclosures)…
… A more accurate version of the above method involves using a resistor in series with the (sub)woofer system, and connecting the multimeter (set to voltage mode) across the resistor. In this case, the meter will measure the lowest voltage at the resonance frequency of a sealed system, and the highest voltage at the resonance frequencies of the ported and bandpass systems.
Reference:
http://www.diysubwoofers.org/faq.html
Section 1.3 1.03 - Tuning - checking the resonance frequency
Courtesy
@Brian Steele
So you get to have your cake and eat it too- model with whatever you have (or nothing at all) and check the tuning frequency with whatever drive level you want (small (voltage) signal, Large signal, driver warmed up, or broken in or not).
So go ahead and drive your (sub)woofer with 0.1V, 1V, 2V, 2.83V, heck 28.3V or 120V (if you dare😂) and check the tuning frequency. Was it close (enough to your simulation?
As with a lot of DIY things; we can all get carried away (we’re not getting paid by the hour, so why stop?).
But having had done this experiment some 20+ years ago, I can attest that the tuning frequency doesn’t sway as much as you might have been led/read to believe.
Yes it changes a bit, but In the grand scheme of things, there are other things to be mindful of, like the woofer excursion going through the roof / Power handling falls off a cliff below the tuning. There’s a real risk of permanently damaging your (sub)woofer if you don’t have a subsonic filter “but I have was 40W amplifier!” but that’s a topic for another day….
Remember that simulations, are, well, sims.
In many fields including sciences and engineering, theoretical models sometimes fail to predict real-world behaviour. The gap that can exist between a clean, idealized model and the messiness of reality is a reminder that practical experience is invaluable and that theories must be tested.
So just a proceed, and build and have fun!
To determine the resonance frequency of the system, set the multimeter to current measurement, then hook it up in series with the subwoofer to your amplifier, then use (a tone) frequency generator to drive the (sub)woofer. At the resonance frequency, the meter will read the lowest result (sealed enclosure) or the highest result (ported and bandpass enclosures)…
… A more accurate version of the above method involves using a resistor in series with the (sub)woofer system, and connecting the multimeter (set to voltage mode) across the resistor. In this case, the meter will measure the lowest voltage at the resonance frequency of a sealed system, and the highest voltage at the resonance frequencies of the ported and bandpass systems.
Reference:
http://www.diysubwoofers.org/faq.html
Section 1.3 1.03 - Tuning - checking the resonance frequency
Courtesy
@Brian Steele
So you get to have your cake and eat it too- model with whatever you have (or nothing at all) and check the tuning frequency with whatever drive level you want (small (voltage) signal, Large signal, driver warmed up, or broken in or not).
So go ahead and drive your (sub)woofer with 0.1V, 1V, 2V, 2.83V, heck 28.3V or 120V (if you dare😂) and check the tuning frequency. Was it close (enough to your simulation?
As with a lot of DIY things; we can all get carried away (we’re not getting paid by the hour, so why stop?).
But having had done this experiment some 20+ years ago, I can attest that the tuning frequency doesn’t sway as much as you might have been led/read to believe.
Yes it changes a bit, but In the grand scheme of things, there are other things to be mindful of, like the woofer excursion going through the roof / Power handling falls off a cliff below the tuning. There’s a real risk of permanently damaging your (sub)woofer if you don’t have a subsonic filter “but I have was 40W amplifier!” but that’s a topic for another day….
Remember that simulations, are, well, sims.
In many fields including sciences and engineering, theoretical models sometimes fail to predict real-world behaviour. The gap that can exist between a clean, idealized model and the messiness of reality is a reminder that practical experience is invaluable and that theories must be tested.
So just a proceed, and build and have fun!
Last edited:
Why going through all this trouble beat me?
Same for this.
At this day an age, there are multiple solutions to very quickly measure this on any computer with just a very simple and cheap audio interface for less than 25 bucks?