Thiele-Small reloaded with LTspiceIV

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Here is a LTspiceIV toolkit aiming at simulating a loudspeaker driver in a closed box.

See attached files.

Steph
 

Attachments

  • Thiele-Small reloaded - closed box.jpg
    Thiele-Small reloaded - closed box.jpg
    129.9 KB · Views: 480
  • Thiele-Small reloaded - closed box (Bode plot).jpg
    Thiele-Small reloaded - closed box (Bode plot).jpg
    97.6 KB · Views: 440
  • Thiele-Small reloaded with LTspiceIV.zip
    162.3 KB · Views: 108
Here are some MONACOR loudspeaker drivers, simulated in free air.
See attached .zip file.
You need the toolkit in the above test for running the simulations.
 

Attachments

  • some Monacor loudspeaker drivers.zip
    18.1 KB · Views: 99
  • Monacor SPH-210 - free air.jpg
    Monacor SPH-210 - free air.jpg
    140.2 KB · Views: 402
  • Monacor SPH-210 - free air (Impedance & Bode plot).jpg
    Monacor SPH-210 - free air (Impedance & Bode plot).jpg
    91.1 KB · Views: 424
Hi Steph
thanks for posting your model.
I have run some simulations for a 10 woofer looks good so far.
what are some formulas for
1.) C_compl values relating closed box volume in liters
2) " " R losses relating to air leaks > any useful empirical values
assume Lmass is moving mass of driver
4) acoustical stuffing affects all mech terms RLC, but mainly C? changes by some empirical percentages?
Finally why choose a power amp with gain of 1+(33/1.5) ? Does this normalize the acoustical output somehow?

Note > I suppose I could sort all this out using Unibox for awhile.
 
Last edited:
What are some formulas for
1) C_compl values relating closed box volume in liters
2) " " R losses relating to air leaks > any useful empirical values
assume Lmass is moving mass of driver
3) acoustical stuffing affects all mech terms RLC, but mainly C? changes by some empirical percentages
4)Why choose a power amp with gain of 1+(33/1.5) ? Does this normalize the acoustical output somehow?
Hello, I'm glad you asked this.

1) There are so many people wanting to think in terms of "litres" when dealing with suspension compliance. My opinion is that one should call a cat a cat. The idea of suspension compliance being expressed in "litres" is a nasty complication. It makes everything fuzzy in my poor, limited mind. Using "litres", I need to remember that this kind of quantification also involves the speaker surface ! No no, here we shall never again use "litres" when dealing with the suspension compliance. Whe shall call a cat a cat. Compliance is the mechanical membrane displacement expressed in meters when your finger (or anything else) applies a force of 1 Newton (= 102 gram) on the membrane. This is so easy to remember, easy to figure !

2) So you seem to say that R_losses is quite an empirical value, quite difficut to objectivate ? Must say I had the same feeling. You are introducing the concept of air leaks. What's this ?

3) Point 3 is unclear for me. Can you detail it ?

4) Excellent question ! The power amplifier gain is the gain of the Kuroda 1982 amp. From there comes the feedback network 33k / 1k5. A few days later, I realized that when specifying an .AC simulation using 1 Volt at the input, that the voltage applied on the speaker model would be more than 30 Volt, much more than the standard "1 Watt" test. I thus lowered the .AC simulation voltage to 100 mV. A few days later, after having defined the internals of the "*s" (differentiator) LTspiceIV component, I needed to define the internals of the "Piston Diameter" LTspiceIV component. Over there I took the decision to tweak the scale factor in such a way that using the .AC 100mV at the input, we get a SPL value equal to the "SPL per watt at 1 meter distance" you can read in any datasheet. This way, looking at the LTspiceIV Bode plot, you get matched with the usual datsheets. This being said, I think I need to slightly reverse engineer this now, for applying a more scientifically precise .AC voltage on the speaker, and accordingly trimming the .AC voltage source, and the scale factor in the "Piston Diameter" block.

Actually, what is the voltage amplitude to be applied on the speaker if one wants to mimic the "1 watt at 1 meter test" leading to the SPL sensitivity figure that gets documented on any speaker datasheet ?

P.S.
I'm here anticipating the remark you will do about putting the driver in a closed box enclosure, effectively decreasing the overall compliance. In this thread I will make a dedicated post explaining in detail the "PV = constant" equation, and the effect of this law when putting a driver in a closed box. It will then become clear why "box volume" may get associated to "compliance", hence "compliance" being sometimes quoted as "equivalent volume" for a particular driver. It will become clear why the closed box is showing as a C_Box capacitor fitted in series with the driver C_Compl. Actually, C_Box needs to be an "intelligent" capacitor, with the capacitor value being computed using a) the box volume and b) the speaker diameter.

Cheers,
Steph
 
Last edited:
Once the C_Box automatic size calculation gets done, there will be another LTspiceIV .asc template, called "closed box", showing C_Box in series with C_Compl.

I intend using a dedicated symbol for C_Box, for not confusing it with the suspension compliance, and also for allowing the user to enter the two parameters (closed box volume in cubic litres, driver membrane diameter in meters).

I'll chose cubic liters for the volume, for staying in the normalized MKS physical system.

Later on, we'll experiment other enclosures like 4th order bass reflex. While doing this, I intend having the same "bare hands" approach when dealing with the vents.

In a bass-reflex enclosure, a vent is a piston (vent diameter) having a mass (air mass) that gets mobilized by the force existing between the enclosure inside, and the enclosure outside.

Those 32 words, properly understood, should enable us to determine how and where connect the vent in the LTspiceIV equivalent schematic, and how and where collect the associated acoustic emission, with the right polarity.
 
Last edited:
Hi Steph
Can you suggest to the Mods that this thread s/b moved to 'Multi' forum. It should generate much more input and excitement from members there, even though not too many use spice.

Firstly I use LTspice to model passive xovers. In the past I ignored the Low Frequency (LF) resonance assuming the filters I modeled where at least a couple of octaves higher. HF effects were simply to use Re and Le as a 1st order approx. Taking the pseudo acoustic output across Re. My crude modeling also assumes the HF output of interest is mainly off axis, ie no acoustical cone effects. This was helpful only using wide band drivers well in their pass band.
Now with your model I would like to investigate the LF effects accurately. From your responses above I think you are still early on your work.

I think we should use a simple voltage source with a constant amplitude of 2.83 V RMS and not watts/m and setting the source impedance equal to the amplifier damping factor plus cable/connector losses. The output sensitivity of your model should match other well known LF T/S modeling programs like "Unibox". Also consider the loss effects of passive filters, or zobels, etc inserted on the drivers terminals. RE> losses/ air leaks and stuffing> If you play around with Unibox and re-read R Smalls paper on closed box modeling all your other questions will be answered.
 
Air Leaks intentional or not, are important to model.

Intentional air leaks are used quite often for high Qts boxes as sort of dynamic reduction in Q. see aperiodic alignments ala restrictive flow vents "Vario vents".
Others maybe less so intentional air leaks are built-in some high power sub drivers that use porous dust caps for VC cooling. Other drivers may use pole pieces instead of dust caps and can be open here as well.
Other air leaks are caused by cabinets not completely sealed, usually much lower in comparison to the first two examples.
 
Now with your model I would like to investigate the LF effects accurately. From your responses above I think you are still early on your work. I think we should use a simple voltage source with a constant amplitude of 2.83 V RMS and not watts/m and setting the source impedance equal to the amplifier damping factor plus cable/connector losses. The output sensitivity of your model should match other well known LF T/S modeling programs like "Unibox".
I have updated the toolkit. Please replace the old files with the newly supplied here, attached. Now the toolkit is using the "1 Watt on 8 Ohm" standard using a 2.82 V RMS .AC stimulus. The "Piston" block has been updated accordingly. Please have a look to the "Monacor SP-100-8 variations on a theme" file. There you see how powerful and easy LTspiceIV is. If you want to apply motional feedback using an acceleration sensor, there is already a piezo sensor in the toolkit, as symbol and schematic.
 

Attachments

  • Thiele-Small reloaded with LTspiceIV.zip
    25.7 KB · Views: 68
  • Monacor SP-100-8 variations on a theme.zip
    10.6 KB · Views: 66
  • Monacor Monacor SP-100-8 variation on a theme.jpg
    Monacor Monacor SP-100-8 variation on a theme.jpg
    129.1 KB · Views: 342
I went to Elak (an electronic shop in Brussels) grabbing some info about loudspeaker drivers one may eventually purchase at decent prices here in Europe. They gave me the 2009 MONACOR catalog, no less than 710 pages. I did an arbitrary selection of drivers :

Celestion Truvox 1225 ... 500 Watt thay say, but excursion is only +/- 2.4 mm. A joke ? Not suited to deep bass where excursion matters ?
Monacor SPH-100/8 ..... inexpensive, wideband, +/- 3.5 mm excursion
Monacor SPH-100C ...... looks hi-tech but beware the nasty resonance at 9.0 kHz
Monacor SPH-135C ...... looks hi-tech but beware the nasty resonance at 6.5 kHz
Monacor SPH-165 ......... inexpensive, wideband, +/- 3.5 mm excursion
Monacor SPH-165CP ..... looks high-tech but beware the nasty resonance at 7.0 kHz
Monacor SPH-170 ........ inexpensive, wideband, +/- 3.0 mm excursion
Monacor SPH-170C ...... looks hi-tech, wideband, manageable controlled resonance at 4.5 kHz, +/- 4.0 mm excursion
Monacor SPH-176 ........ wideband, +/- 5.5 mm excursion
Monacor SPH-210 ........ nice traditional driver, no pronouced high frequency resonance, +/- 3.5 mm excursion
Monacor SPH-225C ...... looks high-tech, wideband, manageable controlled resonance at 3.0 kHz, +/- 6.0 mm excursion !
Monacor SPH-255 ....... traditional driver, not to be used above 2 kHz, +/- 4.5 mm excursion
Monacor SPH-275C ...... looks high-tech, wideband, small manageable controlled resonances at 3 kHz, +/- 6.0 mm excursion !
Monacor SPH-315 ....... traditional driver, not to be used above 2 kHz, +/- 4.5 mm excursion

The MONACOR catalog provides the impedance curve along with most electrical, mechanical and electromechanical data in tabular form. Looks nice at first glance.

I was thus able to enter the necessary data in the LTspiceIV Thiel-Small toolbox.

I must however report a few facts.

1.
The equivalent loss resistor is never quoted by Monacor. However, it is easy to retrieve it when looking at the impedance plot. When all the other parameters have been duly entered in the LTspiveIV toolbox, an impedance plot (voltage / current) tells you if you need to increase or reduce the equivalent loss resistor. The maximum impedance at resonance is governed by the value of the equivalent loss resistor. I have thus set the equivalent loss resistor for obtaining the same max value, under LTspiceIV toolbox, than showed by the impedance plot published by Monacor.

2.
The coil inductance specified by Monacor seems to be systematically exagerated. The coil inductance specified in the Monacor tables doesn't fit with the value showing on the Monacor impedance plot at 20 kHz. So, as expected, if one enters the tabular value of the coil inductance in the LTspiceIV toolbox, the impedance displayed by LTspiceIV at 20 kHz is ways above the impedance showed in the Monacor impedance plot at 20 kHz. In order to get a impedance at 20 Khz equal to the value displayed by the Monacor impedance plot, I have adjusted (lowered) the value of the coil inductance in the LTspiceIV models. This needs to be investigated. The value of the coil inductance is of importance in the Servo-Sound applications (back EMF voltage extraction).

3.
After having entered all the data in all the models, we can't say that there is always a perfect match between the free-air resonance frequency, comparing the Monacor impedance plot with the Monacor tabular data, and with the LTspicsIV impedance plot.
Let us take an example : the SPH-225C (quite my favourite driver, from the supplied data) has the resonance peak at 46 Hz on the graph that is supplied by Monacor. But when you read the Monacor tabular data, they say that the resonance is at 39 Hz. And when you ask LTspiceIV, the impedance peak occurs at 39.9 Hz.
We have the same case with the SPH-210 (a traditional regular driver). It has the resonance peak at 34 Hz on the graph that is supplied by Monacor. But when you read the Monacor tabular data, they say that the resonance is at 28 Hz. And when you ask LTspiceIV, the impedance peak occurs at 27.8 Hz.
So what ?

Would be nice if somebody takes some time explaining me why the inductance values (tabular data) specified by Monacor are always incompatible with the Monacor impedance plots, and why the Monacor resonance frequency value (tabular data) are incompatible with the Monacor impedance plots.

This being said, I have a doubt about the Design Data Base for Unibox that's available here : UniBox - Unified Box Model for Loudspeaker Design
 
Let us take an example : the SPH-225C (quite my favourite driver, from the supplied data) has the resonance peak at 46 Hz on the graph that is supplied by Monacor. But when you read the Monacor tabular data, they say that the resonance is at 39 Hz. And when you ask LTspiceIV, the impedance peak occurs at 39.9 Hz.
Hi steph, yes they are rather nice aren't they :D I have four of them here ready to go into a BMTMB arrangement along with Yamahas beryllium midranges & tweeter unit. Deep bass will be supplied by 2 x 15" per channel.

Not sure if you happen to have one or more there but if you need some measurements of the drivers let me know & i'll see what i can do ;)

Interesting thread by the way, i'm not sure i fully understand it but i have dropped in previously. Had to comment when you brought up the SPH-225C though...

SPH-225C.jpg

Bests, Mark.
 
Last edited:
Would be nice if somebody takes some time explaining me why the inductance values (tabular data) specified by Monacor are always incompatible with the Monacor impedance plots, and why the Monacor resonance frequency value (tabular data) are incompatible with the Monacor impedance plots.

This being said, I have a doubt about the Design Data Base for Unibox that's available here : UniBox - Unified Box Model for Loudspeaker Design


I never take any model handed from the data base. FYI even many OEM data sheets are conflicting with some of their own documents.
Please Do Not evaluate a program based on 1 driver in a DB.


Note> usually the advanced OEM measures Le at 2 frequencies. This of coarse is only a small signal value. For example Le depending on VC position is dynamic, so no accurate static value can be determined esp for higher power overhung VC. So if your feedback is to work at large signals, then you are on your own. Maybe measure an average value at the max level.

Resonance can be measured in different ways and has accordingly many errors and really should be measured after the driver has had much use or exercise. Sometimes the impedance plots is a different test set up. So the devils in the details as always.

You think that is fun.
When you get your drivers samples fully understood in your lab and equipment. And finally get something working in a prototype. Then the OEM will discontinue that product or change the process/ and materials. QC is virtually non existent. or Murphy's Law kicks in. LOL
 
Last edited:
If you want to apply motional feedback using an acceleration sensor, there is already a piezo sensor in the toolkit, as symbol and schematic.

Hi Steph
Piezo? It looks like current sampling to me. I remember motional feedback needs 2 feedback sensors with trans-conductance drive?
You want to do everything at once now. Seems it makes more sense to me to characterize each of the transducers/sensors independently before crunching it all together. What are the goals here anyway?
Get the small signal LF speaker model correct first. With losses and interaction of the electrical input? I don't think you've done that yet.
 
Piezo? It looks like current sampling to me.
I wrote that the piezo was in the toolkit. I didn't wrote that the piezo was in the schematic. You need to see what's in the .zip of the previous post. There you will find a piezo sensor as component. You wrote you like the Monacor SPH-225C. Attached you will find new schematics showing how one can combine acceleration feedback (piezo sensor) and velocity feedback (back EMF coil extraction) on a Monacor SPH-225C.

At this stage, I don't know if the velocity feedback is really a velocity feedback because for avoiding a nasty resonance at 2 kHz, I needed to add a 2k2 resistor in parallel with the x100 coil resistance + x 100 coil inductance.

How can I determine we still benefit from velocity feedback between 120 Hz and 2 kHz, or if it reverts back to a kind of voltage equalization ? In order to know this I need to put my finger on the membrane, and see if it looks harder than the overall compliance (presence of velocity feedback). Can you help me designing such finger, using LTspiceIV ?

See attached files.
You need the rest of the toolkit to run these.

Steph
 

Attachments

  • Monacor SPH-225C - variations on a theme.zip
    487 KB · Views: 62
  • Monacor SPH-225C - R_AFB varying.jpg
    Monacor SPH-225C - R_AFB varying.jpg
    169.2 KB · Views: 167
  • Monacor SPH-225C - R_AFB varying (Bode plot).jpg
    Monacor SPH-225C - R_AFB varying (Bode plot).jpg
    109.7 KB · Views: 171
  • Monacor SPH-225C - C_VFB varying.jpg
    Monacor SPH-225C - C_VFB varying.jpg
    172.9 KB · Views: 165
  • Monacor SPH-225C - C_VFB varying (Bode plot).jpg
    Monacor SPH-225C - C_VFB varying (Bode plot).jpg
    197.2 KB · Views: 163
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.