• Welcome to ASR. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

The Heresy of the T/S Parameter Bl

smowry

Senior Member
Joined
Nov 20, 2024
Messages
457
Likes
397
Location
Phuket
Although this claim could be considered controversial, unless you are using a current source amplifier with high output impedance to drive your loudspeaker, my claim is Bl with unit (Tm) is Heresy!

Kindly let me show you why. I will start with the equation that defines a moving coil transducer's efficiency,


1768433208333.jpeg


The expression ρ/2πc is a constant and can be replaced by the value 5.445×10−4 (m^2)s/kg for dry air.

Where ρ is the density of air and c is the speed of sound in SI units.

B is the DC magnetic flux density with unit (T).

l is the length of the voice coil winding with unit (m).

S
d is the effective area of the diaphragm with unit (m^2).

M
ms is the moving mass including air load with unit (kg).

Re is the DC resistance of the voice coil with unit (Ω).

Using the transducer's efficiency equation above, the SPL at 1.0 (m) in dB can be calculated.


1768433353163.jpeg

From the efficiency equation above a term β or the motor efficiency factor can be defined.

β = (Bl)(Bl)/R
e with unit (N^2 /W)

Where Re = l /(Aσ)

σ is the electrical conductivity of the voice coil wire with unit (1/(Ωm)).

A is the cross-sectional area of the wire's conductor with unit (m^2).

Then β = (Bl)(Bl)(Aσ)/l

β
=
(B)(Bl)(Aσ)

But lA is the volume of the conductor, V with unit (m^3).

Then β = (B^2)()

Looking back at the equation for transducer efficiency, η
0 goes as β, Sd^2, and 1/Mms^2.

This shows that we don't care about the length of the voice coil wire, l.

What we care about is the volume of the conductor, V.

Now we can see that the SPL also depends on, β and not on Bl alone.

For example:

Motor A has a Bl of 10 (N/A).

Motor B has a Bl of 8 (N/A).

Which motor do you prefer? (Trick question!)

Motor A has an R
e of 6.4 (Ω).

Motor B has an R
e of 3.2 (Ω).

Then Motor A has a β = 15.6.

Motor B has a β = 20.0.

Motor B actually has 1.1 (dB) higher sensitivity with 20% less Bl!

The lesson here is clear. We don't care about Bl; what we care about is β.

With a large signal input, we can use Bl to evaluate motor linearity, Bl(x).

However for small signal evaluation Bl(0) = Bl, the force factor if considered alone is misleading at best.

Whereas β(x) allows for evaluation of both linearity and motor efficiency regardless of R
e.

β is the true figure of merit for transducer motor evaluation, while Bl alone is effectively meaningless.

Final note: 1/β is also responsible for most of the damping, where QesQts.


1768439881888.jpeg
 
Last edited:
Before I joined the Transducer Research Group (TRES) at Bose in 1995, I had studied Dr. Richard Small's classic AES papers. However, after I joined Bose as an R&D Engineer 3, I was quickly told in no uncertain terms, not to refer to the term Bl or I would be ridiculed or shunned. I was instructed by colleagues to use the term Beta when referring to transducer motor assemblies. Contained above is the derivation of the reasoning for that instruction.
 
Last edited:
Last edited:
Intriguing.

One question though - how much does ρV generally contribute to Mms? Clearly there have to be some deviations once its effect is no longer negligible, or else highest sensitivity would just mean slapping on an absolute chonker of a voice coil and calling it a day.

And then there's the slight complications of fitting all that conductor inside the air gap (give or take whatever tolerances you consider acceptable, plus the separate issue of potentially permitting air movement for cooling), introducing an interaction between B and V that's not immediately obvious. Given the B²V factor, you'd think B would have a greater effect, although V seems a lot easier to modify. And then there's the considerations of overhung vs. underhung voice coils in terms of linearity and whatnot.

Having V in there does make a good bit of sense, given that it's known that you can make transducers of very similar properties with wildly different nominal impedances (think headphones - 32, 250, 600 ohms). The thought process is explained here:
Now the part that slightly puzzles me is that higher impedance tends to have a slight edge in bandwidth in practice, and I'm not entirely sure why that is. When generally L ~ n² in inductors, you'd rather be expecting the contrary.
 
Intriguing.

One question though - how much does ρV generally contribute to Mms? Clearly there have to be some deviations once its effect is no longer negligible, or else highest sensitivity would just mean slapping on an absolute chonker of a voice coil and calling it a day.

And then there's the slight complications of fitting all that conductor inside the air gap (give or take whatever tolerances you consider acceptable, plus the separate issue of potentially permitting air movement for cooling), introducing an interaction between B and V that's not immediately obvious. Given the B²V factor, you'd think B would have a greater effect, although V seems a lot easier to modify. And then there's the considerations of overhung vs. underhung voice coils in terms of linearity and whatnot.

Having V in there does make a good bit of sense, given that it's known that you can make transducers of very similar properties with wildly different nominal impedances (think headphones - 32, 250, 600 ohms). The thought process is explained here:
Now the part that slightly puzzles me is that higher impedance tends to have a slight edge in bandwidth in practice, and I'm not entirely sure why that is. When generally L ~ n² in inductors, you'd rather be expecting the contrary.

Mass is simply density times volume and again Bl is no help. Then with regards to bandwidth, this tends to the first order low pass break frequency (Hz).

1768442214144.png

Moving coil transducer inductance, L
e is a much more complex calculation than just N^2, where N is the number of turns. Le depends on magnetic permeability which is typically nonlinear. Le is typically controlled by using shorting rings and/or magnetic saturation and needs to be simulated with AC FEA, restarted from the DC solution. https://pearl-hifi.com/06_Lit_Archive/14_Books_Tech_Papers/Mowry_Steve/A_Closer_Look _at_L_etc.pdf

The exotic approach for controlling inductance is a Steallus or an almost air core motor assembly.
These approaches require about ten times (10x) the magnet volume or more compared to typical moving coil transducers!

A few Transducer Engineers like to use anti-coils to reduce inductance, analogous to an active shorting ring. Dan Wiggins claims that he can achieve negative inductance. https://patentimages.storage.googleapis.com/84/72/ee/7d675945e54587/US7873180.pdf

However, for me the most intriguing approach is EMI or transformer drive. With EMI it's almost the case of driving the shorting ring with a stationary voice coil. The induced voltage in the secondary moving coil goes as frequency, f . This is why induction heaters operate at very high frequencies. Where N is the number of stationary coil turns and Φ is the magnetic flux. Where is that ultra low moving mass EMI Beryllium Tweeter with stationary lead dress guys?

1768464696747.png


 
Last edited:
Let's consider the power amplifier. Then β = (Bl)(Bl)/(Re + Ro). Where Ro is the output resistance of the amplifier, which is in series with Re.

Most high performance amplifiers available today are voltage sources and Ro is very small. The above derivations assume Ro = 0. Where Re/Ro is the "Damping Factor". Typical Dampical Factor is greater than 100 and can be as high as 1,000. Low frequency transducers, especially subwoofers have high β, which results in "Tight Bass". The can be represented by Qes, the dominant factor in the voice coil damping.


1768515361017.jpeg


For a typical high performance subwoofer Qes < 0.3.

To better understand, Ohm's Law is considered, u = (Re + Ro)i. Where u is the voltage and i is the current.

Below is Dr. Klippel's transducer model with the addition of Ro.


1768515361041.jpeg


So if Ro is small then the transducer's impedance, Re + Zl, dominates; however if Ro is large, we don't care about Re + Zl, Ro dominates. The amplifier become a current source and the effective β goes to zero and Qes blows up..

1768515415773.jpeg


The point here is i becomes essentially constant but the electrical damping is gone and the back EMF becomes ineffective with the mechanical damping Q
ms dominating.


1768515529730.jpeg


Where Rms is a measurement of the losses, or damping, in a driver's suspension and moving system. An acoustic resistive enclosure or a U-frame terminated with an acoustic resister (e.g. fiberglass or rockwool) can be designed to increase effective Rms

So what if, a conventional voltage source power amp is used to drive the low frequency transducer(s), while a current source amplifier is used to drive the high frequency transducer(s) in an active loudspeaker system? Some audiophiles claim that by placing a 1k Ω resistor between the amplifier and the transducer they can measure significantly reduced distortion but obviously sensitivity is also significantly reduced. Potentially, using voltage drive for low frequencies and current drive for high frequencies could raise the level of the state of the art. Reductions in distortion are potentially greater than 10 dB with current drive along with bandwidth limitation related to Le is minimized. Ironically, then with current drive β becomes the Heresy and the force, Bl(x)i (N) is what we care most about.

KEF is presently investigating current source drive with a closed loop control system they call VECO to control velocity and compensate for the high effective Q
es. In this way, they can use a current source to drive a low frequency transducer. There is feedback control. https://www.audiosciencereview.com/forum/index.php?threads/kefs-breakthrough-veco-speaker-technology-to-lower-distortion–explained-by-its-inventors.64033/

Folks are enthusiastically waiting for a LS50 Meta/VECO! :)
 
Last edited:
https://ptt.purifi-audio.com/shop/ptt10-0x04-nab-01-ptt10-0x04-nab-01-2043/document/583

Bl = 12.8 Tm
R
e = 3.9 Ω
β = 42.0 N^2/W
SPL (1W@1m) = 85.7 dB



Bl = 15.7 Tm
R
e = 6.4 Ω
β = 38.5 N^2/W
SPL (1W@1m) = 83.9 dB

This is typical. The 8 Ω nominal voice coil has more turns, N and greater l but less volume of conductor in the gap due to more insulation and bond coat,
We don't care about Bl we care about β. Some poor soul will think, the 8 Ω version has more Bl and make a system design decision based on Bl. Note that a 2 Ω version would have even more β (and less Bl) without any increase in manufacturer's cost. Bl is misleading at best when the amplifier is a voltage source.

At Bose the 901 4.5 inch transducer had Re = 0.7 Ω x 9 in series, with single layer rectangular anodized aluminum magnet wire. Packing factor was very high. Note that TAD also uses single layer rectangular anodized aluminum magnet wire in their high frequency pro transducers. https://www.technicalaudiodevices.com/pro-hf-units/
The ultimate magnet wire is retangular anodized aluminum clad copper wire; however, it's not available. It was invented in 1956. Nobody want's to pay for it.


So when @Lars Risbo stops by, he will claim that β is not the industry standard. And that's correct; however, the industry in general is lame! https://www.audiosciencereview.com/...the-commutated-voice-coil.61642/#post-2260401 Folks call transducers, drivers. They call loudspeakers, speakers. PhD's are doing this! When I worked in Malaysia and Thailand, I had a driver. He had a car and he would take me where I asked him. I have been a speaker at conventions in China, Hong Kong, Malaysia, Singapore and Thailand. Check the dictionary. Pardon me but I am a champion for the advancement of the Loudspeaker Industry, yet I am retired and have not practiced transducer engineering in more than 10 years. R&D is not free and needs to be funded. An example is Parts Express and Dayton Audio. Purifi is a small company and they are clearly on the high side of the Industry; however, folks have asked me when they will update their outdated spiders. Purifi developed a great surround; however, they patented it and now that's it for 20 years. Could Purifi defend that Patent? Maybe :rolleyes:
 
Last edited:
@AnalogSteph

I think I misunderstood your question. "One question though - how much does ρV generally contribute to Mms?" Where ρ was defined as the density of air with respect to the efficiency equation.

1768568765168.png


I am still not sure what you are asking but I will try again.

As a general low frequency transducer design guideline, the voice coil mass should be approximately 50% of the total moving mass, Mms.
 
Last edited:
@Lars Risbo I am curious about your thoughts on this.
i use beta internally. it’s a useful parameter for several things. In some litteraturen it’s called electrical damping - R_D or something like that and if you add the the mechanical Rms you get the total damping. in current drive, electrical damping disappears and we only have Rms.

One interesting aspect is that fs/Qts is equal to the ratio of beta and Mms and this is again a proxy for the F3 obtainable in a box alignment. so to lower F3 we need to decrease beta and or increase mass. The sensitivity scales with sqrt(beta)/Mms. This means beta and mass are the principal parameters to determine the Hofmann trade off between sensitivity, volume and F3

cheers

Lars
 
Potentially, using voltage drive for low frequencies and current drive for high frequencies could raise the level of the state of the art
FWIW, that's what I'm doing for more than a decade, and I have designed a few commercial studio monitors using the principle. Mixed impedance drive on the midwoofers and pure current drive on AMT tweeters. Privately, I've even successfully used an "augmented voltage drive" at LF so to say, a negative source resistance blending into current drive (conceptual info here), therefore even more local velocity feedback in the driver.

Personally, I've always used (BL)^2/Re as a figure of merit for motor strength, many years without knowing that it is called beta.
 
Hi @Lars Risbo

Thanks for stopping by. Please don't make fun of my concept sketch below. There is no simulation; I was just playing around. However, I can use it to illustrate some talking points for discussion purpose.


1768587125901.png


This midrange transducer concept is intended for current drive without feedback control.

1. The mid-engine is my Almost Aircore topology from 2009 with 4 x NdFeB magnets and the gap plate is non-ferritic SS with an aluminum alloy 2-piece basket. https://pearl-hifi.com/06_Lit_Archi.../Mowry_Steve/Air_Core_Tweeter_Magnet_Assy.pdf
2. There are 2 x auxiliary supplemental voice coils that are in the fringing magnetic fields (return paths) and are intended to complement and linearize Bl(x). These aluminum wire coils could be used to in a sense tune Bl(x) and help to reduce magnet volume.
3. The transducer topology is complementary symmetrical (push-pull) with wide wheel base suspension and no spider. Note that the assembly creates isobarik loading between the backsides of the diaphragms. The pressure gradients are on the outsides of the diaphragms.
4. Acoustic resistive material can be placed between the diaphragms to increase R
ms. I have placed acoustic resistors between 2 x transducers in a complementary isobarik array. I was able to get almost any reasonable Qms by varying the thickness of the acoustic fiberglass, 48 kg/m^3. I used fiberglass because that is what I had on hand to test the concept. However, there are numerous alternative materials such as BASF Basotect.
5. This concept could be mounted in an acoustic resistance enclosure which could help to control directivity and again increase Rms.

6. Note that the connectors are copper tubes that will accept copper z-type banana plugs.
7. The low frequency transducers could then be Purifi 10 inch woofers driven by a voltage source amplifier (Purifi) within a hybrid active loudspeaker system.
8. The tweeter could then be that EMI Beryllium Dome that I am waiting for.

All serious comments, questions, and concerns are welcome and encouraged.
 
Last edited:
4. Acoustic resistive material can be placed between the diaphragms to increase Rms. I have placed acoustic resistors between 2 x transducers in a complementary isobarik array. I was able to get almost any reasonable Qms by varying the thickness of the acoustic fiberglass, 48 kg/m^3.
I would think this is a most critical part, isn't it? Both in terms of manufacturing tolerances/repeatability, and in linearity.
Gut feeling suggests that when the pressure gradient starts to build up, the fibers at first will start moving/flexing with the air a bit until they hit sort of an end stop condition where actual flow resistance comes fully into effect. Therefore, at very low levels, the damping could be less than nominal, also leading to some additional distortion.
 
@Lars Risbo in your post you stated, "One interesting aspect is that fs/Qts is equal to the ratio of beta and Mms"

PTT10.0X04-NAB-01

1768600298453.jpeg

fs (1/s) / Qts
24 (Hz) / 0.33 = 72

β (N^2/W) / M
ms (kg)
42 (N^2/W) / 0.098 (kg) = 429

429 / 72 = 6 Ouch! Something is wrong Josée. Did Carsten tell you that?

However: Q
tsQes

1768602216848.jpeg

Then dividing both sides by 2πfs and inverting the resultant ratios, Bingo!

ωs / Qes = 431 Where ωs = 2πfs
β / Mms
= 429 (Round off errors in manufacture's specifications)

What is the point of these ratios? They are simply a representation of 1/Q
es.
 
Last edited:
@Lars Risbo said, "In current drive, electrical damping disappears and we only have Rms."

Now as my derivation including the nonlinear models contained above indicate that I firmly agree with Lars' claim. However, Dr Klippel has indicated that R
ms is inherently a nonlinear parameter essentially because it is a dynamic parameter. Where v = ωx = 2πfx is the velocity (m/s). Whereas any added acoustic resistance is static. So any claim that adding static acoustic resistance to increase mechanic losses will increase nonlinearity of the transducer is controversial and must be proven by math model and/or experiment. Then my claim is that static acoustic resistance is essentially linear, where Rms(0,0) is static and v and f are zero. This further indicates that the additional Rms(0,0) is linear, where Rms(0,0) = Rms and will in effect mitigate the dynamic Rms(f,v) nonlinearity that results from the moving assembly. Furthermore, the inventor of the Isobarik Loudspeaker, Ivor S. Tiefenbrun from Linn, claims that the "acoustic curtain" between two diaphragms reduces distortion.
Technical rhetoric alone tends to arguments of convenience rather than fact and should be avoided unless some proof can be shown. Personally, if I cannot prove my claim, then I will not make such a claim. Having said that, frivolous or what I call "cowboy claims" are related to engineering ethics. Suppose MD's acted in that manner?
 
Last edited:
@Lars Risbo I hate to do this but I disagree with you again. You made the following claim, "The sensitivity scales with sqrt(beta)/Mms."

I did a derivation of 1.0 W @ 1.0 m Sensitivity in my opening post and your claim contradicted my derivation. So I will present my derivation again. I will start with the equation that defines a moving coil transducer's efficiency,


1768433208333.jpeg


It then follows that SPL at 1.0 W @ 1.0 m can be determined using the equation contained below.

1768629254341.png


Then I claim that 1.0 W @ 1.0 m Sensitivity goes as [β S
d^2] / Mms^2. Am I incorrect? If I am then can someone please help me with a derivation.

I think it's correct, 2 x power goes to +3 dB. Then if I double β, 10log(2) = 3 dB.
 
Last edited:
@Lars Risbo in your post you stated, "One interesting aspect is that fs/Qts is equal to the ratio of beta and Mms"

PTT10.0X04-NAB-01

View attachment 504742
fs (1/s) / Qts
24 (Hz) / 0.33 = 72

β (N^2/W) / M
ms (kg)
42 (N^2/W) / 0.098 (kg) = 429

429 / 72 = 6 Ouch! Something is wrong Josée. Did Carsten tell you that?

However: Q
tsQes

View attachment 504752
Then dividing both sides by 2πfs and inverting the resultant ratios, Bingo!

ωs / Qes = 431 Where ωs = 2πfs
β / Mms
= 429 (Round off errors in manufacture's specifications)

What is the point of these ratios? They are simply a representation of 1/Q
es.
yes, i should have said proportional. there is, as you show, a factor 2pi missing.
 
Hi @Lars Risbo

Yes, off by a factor of ; however, the math still will not work with Q
ts. It only works with Qes. I derived that for you. Qts considers both β and Rms.

1768652118570.png


Where 1/Q
es = β / [Mms(2πfs)]

Multiply both sides by fs

Then fs / Qes = β / Mms Bingo!

However, what's the point in rearranging the equation for Qes? Is it a system thing? From the transducer side, sometimes I would increase both Mms and β (i.e. bigger voice coil) and equalize the response to reduce the effective f3. Purifi has a 500 W amplifier. Remembering that the larger β is, the smaller I can make the box (Carver) but I must equalize to reduce the effective f3. Note it is not often that I give credit to 30% Bob.

1768654977896.png


fs goes as 1/ Mms^0.5 regardless of β. Whereas it is Qes that goes as Mms / β.


So how about your other claim? "The sensitivity scales with sqrt(beta)/Mms." I cannot derive that.

So I will present my derivation again. I will start with the equation that defines a moving coil transducer's efficiency,


1768433208333.jpeg


It then follows that SPL at 1.0 W @ 1.0 m can be determined using the equation contained below.

1768629254341.png


Then I claim that 1.0 W @ 1.0 m Sensitivity goes as [β Sd^2] / Mms^2 or β / Mms^2. Am I incorrect? If I am, then can you please help me with the derivation.

However, I think it's correct, 2 x power goes to +3 dB. Then if I double β (N^2/W), 10log(2) = 3 dB. Where did you get the square root from?

Lars, young engineers look up to you. You need to resolve this matter for their sake and not mine. This has nothing to do with what folks like. You stopped by and made two claims but I found issue with both. What was I to do, ignore this? Frankly, I also look up to you. That is why I took the time to study your post.

Thank you for your attention to this matter.

Steve
 

Attachments

  • 1768655017807.png
    1768655017807.png
    13 KB · Views: 19
Last edited:
Then fs / Qes = β / Mms Bingo!
yes that is correct. I did not include the mechanical damping. In this case we have:

fs/Qts = (beta +Rms) /Mms/2/pi

this frequency fs/Qts is where the damping force eclipses the drive force. I have seen it called f_D in an old paper. Similar to EBP=fs/Qes. F_D is what to look at to assess the obtainable F3 of a box alignment. Eg a sealed box with butterworth tuning achieves F3=fc=f_D•0.71. Note that the is independent of the suspension compliance.
Then I claim that 1.0 W @ 1.0 m Sensitivity goes as [β Sd^2] / Mms^2 or β / Mms^2. Am I incorrect? If I am, then can you please help me with the derivation.

That is correct but note that your expression is formulated as the power out to power in ratio. if express it as sound pressure per voltage (linear) then we take the sqrt.

i hope that helps.

cheers

Lars
 
However, what's the point in rearranging the equation for Qes? Is it a system thing? From the transducer side, sometimes I would increase both Mms and β (i.e. bigger voice coil) and equalize the response to reduce the effective f3. Purifi has a 500 W amplifier. Remembering that the larger β is, the smaller I can make the box (Carver) but I must equalize to reduce the effective f3. Note it is not often that I give credit to 30% Bob.
i find the two scaling rules (F3 scales with beta/Mms and sensitivity scales with beta^0.5/Mms) very useful when designing to meet certain box size and F3 targets.

Eg if you double both beta and Mms then F3 is unchanged but sensitivity is lowered. Such driver gives the same F3 but in smaller box (Vas unchanged by Qts reduced).

this of course only for passive boxes without active EQ.

For the small active box then high beta is king since this reduces the power needed to compress the air in the small box.
 
Back
Top Bottom