TS modelling is a good guide, I'm not knocking it. But you have to bear in mind that it is a small signal measurement, and has limitations.
It takes no account of non linear behavour from cone or surround, frequency response curves of drivers, voice coil heating, etc. all of which play a major part in the summed response curve of a driver in a box once you start driving it with anything over a few volts...
It takes no account of non linear behavour from cone or surround, frequency response curves of drivers, voice coil heating, etc. all of which play a major part in the summed response curve of a driver in a box once you start driving it with anything over a few volts...
I talk about this.
When you done measurements right, get exact T/S parameters, with high Qts you must go with smaller box for best results. Other way you got tiny bass. With small Qts, must go with bigger box then simulated.
I eksperiment last years. Not only that, but if you look at some projects, it is like that.
Example.
I worked with RTO GF-650. Measured.
Fs=46 Hz
Mms=15 gr
Sd=129 cm2
Qms=3,86
Qes=0,525
Qts=0,462
Vas=17,7 L
BL=7,52 N/A
Re=6,7 Ohm
With simulations you got 30-32 L, with Fb=40 Hz.
I made box 30 L netto, and I got tiny bass and much higher mid. Whatever I do.
In the start I doubt this bass would work realy good in that big box. Because of that I leave bottom movable. So I try different box volumes, and find 18 L and Fb=47 Hz is the best.
Example 2.
I worked with Monacor SPH-165. Measured.
Fs=41,8 Hz
Mms= 16 gr
Sd=137 cm2
Qms=2,83
Qes=0,268
Qts=0,245
Vas=23,8 L
BL=9,7 N/A
Re=6,0 Ohm
Best results with enclosure of about 11,5-12 L netto.
Before 6-7 years I start to think and work about corections of calculation with T/S parameters.
Today, I got mathematic model for this corections. With empirical corection coefficients. And I think its working good at all Qts range (0,2 - 0,6). It includes most of T/S parameters, not only Vas, Fs and Qts.
I neeed more authoritative measurenaments and compare with simulated results. For that, first I must purchase better equipment. When you live in Belgrade with 200 € at month ( average month pay here) it is not that easy.
For now, give here T/S parameters of some bass unit, best if there is project for that unit. We can compare real project in which gives the best results, with simulation for his T/S parameters.
Remark.
About Qts=0,32 is value I call "ideal" which means what you simulate, in real is about the best.(not constant "ideal". larger Sd, "ideal" value is smaller, and opposite. 0,32 is for 8" bass) Smaller Qts bigger box than simulated. More far from 0,32 more differecies from simulated. Same for high Qts, but opposite, smaller box.
When you done measurements right, get exact T/S parameters, with high Qts you must go with smaller box for best results. Other way you got tiny bass. With small Qts, must go with bigger box then simulated.
I eksperiment last years. Not only that, but if you look at some projects, it is like that.
Example.
I worked with RTO GF-650. Measured.
Fs=46 Hz
Mms=15 gr
Sd=129 cm2
Qms=3,86
Qes=0,525
Qts=0,462
Vas=17,7 L
BL=7,52 N/A
Re=6,7 Ohm
With simulations you got 30-32 L, with Fb=40 Hz.
I made box 30 L netto, and I got tiny bass and much higher mid. Whatever I do.
In the start I doubt this bass would work realy good in that big box. Because of that I leave bottom movable. So I try different box volumes, and find 18 L and Fb=47 Hz is the best.
Example 2.
I worked with Monacor SPH-165. Measured.
Fs=41,8 Hz
Mms= 16 gr
Sd=137 cm2
Qms=2,83
Qes=0,268
Qts=0,245
Vas=23,8 L
BL=9,7 N/A
Re=6,0 Ohm
Best results with enclosure of about 11,5-12 L netto.
Before 6-7 years I start to think and work about corections of calculation with T/S parameters.
Today, I got mathematic model for this corections. With empirical corection coefficients. And I think its working good at all Qts range (0,2 - 0,6). It includes most of T/S parameters, not only Vas, Fs and Qts.
I neeed more authoritative measurenaments and compare with simulated results. For that, first I must purchase better equipment. When you live in Belgrade with 200 € at month ( average month pay here) it is not that easy.
For now, give here T/S parameters of some bass unit, best if there is project for that unit. We can compare real project in which gives the best results, with simulation for his T/S parameters.
Remark.
About Qts=0,32 is value I call "ideal" which means what you simulate, in real is about the best.(not constant "ideal". larger Sd, "ideal" value is smaller, and opposite. 0,32 is for 8" bass) Smaller Qts bigger box than simulated. More far from 0,32 more differecies from simulated. Same for high Qts, but opposite, smaller box.
pinkmouse said:TS modelling is a good guide, I'm not knocking it. But you have to bear in mind that it is a small signal measurement, and has limitations.
It takes no account of non linear behavour from cone or surround, frequency response curves of drivers, voice coil heating, etc. all of which play a major part in the summed response curve of a driver in a box once you start driving it with anything over a few volts...
That is right. Because of that, with high Qts, we must find way to calculate, not exact, but near Vb. If we calculate 2-3 times bigger by using simulation, we can throw a way that box.
Feyz said:
Don't knock T/S modelling that much
http://users.ece.gatech.edu/~mleach/papers/spice_electro.pdf
It is apparent to me that Vifa is recommending such box sizes because they know their drivers and with a box that gives the flattest and deepest extension, it will distort and likely get destroyed. Think about this way: do you use a tweeter at 800Hz, just because it can play down to 800Hz flatly? Same thing here. T/S modelling is not that bad to give twice the box volume of what in reality would be. The small 14lt box WILL not play as low as the box 30lt, as the T/S modelling shows correctly. But it will play music better, because will not distort as much as 30lt box. IMHO this is about it. Don't blame it all on T/S modelling, they work. But T/S models are not only there to produce FR curves, they also produce Excursion, group delay, maximum SPL, maximum power curves as well. If you just look at FR curves of T/S models, and then your woofer box sounds lousy, who is to blame?
BTW, I find it very hard to make a good and accurate measurement of ported boxes with near mic techniques. It is not easy, there are lots of pitfalls, from the mic distance to cone differences to estimating how to adjust port SPL to woofer SPL etc.
Good explanation. Of course, there is much more. Its not so easy to write all about this.
Notax said:
and temperature of the voice coil!!! maybe my world has been turned upside down
nothing like the occasional shake up!! could be that I was lucky in that my speaker was the ideal, gives me something to think about when choosing drivers in future.... Have you tried MJK's mathcad worsheets for bass reflex?? something I've been meaning to look at for a while, not sure if he has "extra" stuff in them.pinkmouse said:
yeah I must say I was a little dissolusioned when I read the bit in LSDC about how they are only valid at very low voltages..... and even more dissolutioned when I read about the effect of voice coil heating on crossovers freqs and the undesirable effects that can have... but I think that the sims at least give us something better than just taking a wild guess
Tony.
edit: wow quite a few posts since I started writing this reply, interesting stuff Notax!
pinkmouse said:TS modelling is a good guide, I'm not knocking it. But you have to bear in mind that it is a small signal measurement, and has limitations.
It takes no account of non linear behavour from cone or surround, frequency response curves of drivers, voice coil heating, etc. all of which play a major part in the summed response curve of a driver in a box once you start driving it with anything over a few volts...
All good what you are saying, but on the context of the thread the question was why the T/S model is showing 30lt, and Vifa is recommending 14lts? You can't explain such discrepancy with nonlinearity of the driver and the box. I had looked into changes in T/S parameters with drive level, you can see the results here:
http://members.fortunecity.com/pirimoglu/Design/DriverComparison/WooferComparisonImpedance.html
For drivers to be used in domestic enviroments Voice coil resistance heating is overrated IMHO. You can see from the curves in that link that Re is not changing much with drive level, if it had, the minima region after resonance peak should change its level. In any case Re changes with temperature can be accounted for by running the driver for a while to reach an internal temperature and then take the T/S measurements. The biggest changes appear in Fs, Vas and Qms because of nonlinearity of the suspension, and Qe changes because nonlinearity of BL. Le also changes but that is related to high frequency region, not LF which T/S modelling is used for. For subwoofers with very high Le this may also be a problem with LF production though. One other thing, drivers with good linearity also go through little change of T/S parameters with drive level, example Scan-Speak 21W/8554 in there. The ratio of Fs/Qts go through very little change, and IIRC this means on a closed box the alignment will go through almost no change. So good quality driver means T/S modelling will work better. One big problem is what kind of drive level measurments T/S parameters to use. I go with a drive level that makes to cone move but doesn't stress it in free air (stress can be seen from the impedance curve starting to have ripples around the lower parts of the resonance peak). I use these parameters to model T/S than go from there. I know this is against the "low-signal" model, but I don't listen to low-signal.
The FR of the driver is not important for T/S modelling because it is used for LF region, and for LF almost all drivers are pistonic.
What I am trying to say is, T/S modelling can be very useful to start. There is no other good way that I am aware of. I haven't looked at MJK's sheets, but I would guess they are also using T/S parameters like Vas, Qms, Qe etc, which are nonlinear in reality. A full nonlinear model, which would be very complex, would give FR for different drive levels, but it still won't tell you how it will sound.
Re: I talk about this.
How much baffle step correction are you applying with the crossover? Just speculation on my part, but you may be compensating for some of the baffle step by putting some peaking on the low-mid low frequencies with the different box tuning. If that's the case, there is nothing wrong with it, how it sounds matters in the end.
Notax said:
How much baffle step correction are you applying with the crossover? Just speculation on my part, but you may be compensating for some of the baffle step by putting some peaking on the low-mid low frequencies with the different box tuning. If that's the case, there is nothing wrong with it, how it sounds matters in the end.
Feyz said:
Great post and very usefull link.
When I modeling system of calculations, I notice very good behavior of most Scan-Speak units. Because of that I take TS parameters of SS 7" mid-basses as some kind of reference or could say as boundary values, to form corrections. We can see that most of "standard" 7" mid-basses with 120 mm magnet and SD-1 drive system has very similar TS parameters. Scan-Speak establish a system and they hold simular values.
I imagine basic values of 7" series as:
1. size of 7" as "ideal" for 2-way systems
2. SD-1 motor system
3. all the same 120 mm magnet
4. all the same 42 mm voice-coile with Re=5,5 Ohm
5. all the same lenght of winding and air gap hights, which resulat with Xmax=+/-6.5 mm and max. p-p +/- 10 mm
6. and many other "all the same" in the series.
It is not coincident, because they went to some (for now) great solutions, and use it in several different models of units with little variations. Most the materials of the cone and little different Cms.
Re: Re: I talk about this.
None.
I dont like to complicate when I dont must.
I like as simple as it could be, or 24/24.
Most the first version.
Its better to go with two units in parallel, one with low 6 or 12 dB/oct. and the other with higher 6 or 12 dB/oct. then made complicated crossover.
I dont think most important is baffle step. OK that produce lower SPL at low frequencies, but when unit has high Qts, have lower Mms, lower BL, lower Xmax and its paradox bigger box (said simulation which is made trought normalized transfer funcions) for "tiny" capability for low Fb and long excursion. That produce compression and big distorsion. That is problem.
Tiny unit have higher Qts, and all worse parameters for long excusion, but by tranfer funcion must go into big enclosure with low Fb. That unit can handle with it, and because of that better sounds in a smaller enclosure with higher Fb. And that compensate lower SPL at low frequencies.
Feyz said:
None.
I dont like to complicate when I dont must.
I like as simple as it could be, or 24/24.
Most the first version.
Its better to go with two units in parallel, one with low 6 or 12 dB/oct. and the other with higher 6 or 12 dB/oct. then made complicated crossover.
I dont think most important is baffle step. OK that produce lower SPL at low frequencies, but when unit has high Qts, have lower Mms, lower BL, lower Xmax and its paradox bigger box (said simulation which is made trought normalized transfer funcions) for "tiny" capability for low Fb and long excursion. That produce compression and big distorsion. That is problem.
Tiny unit have higher Qts, and all worse parameters for long excusion, but by tranfer funcion must go into big enclosure with low Fb. That unit can handle with it, and because of that better sounds in a smaller enclosure with higher Fb. And that compensate lower SPL at low frequencies.
OK I got curious and did some modeling of the different volumes and a few tuning freqs. results are here ---> http://home.exetel.com.au/wintermute/XG18/http://home.swiftdsl.com.au/~tonywww/XG18/
personally I don't like the look of the 14L 60Hz one, but I guess the topic of this thread is that the modelling is inaccurate so it would be interesting to actually build the boxes and see what the "real world" results were.... not something I'm going to be able to do though.....
Tony.
personally I don't like the look of the 14L 60Hz one, but I guess the topic of this thread is that the modelling is inaccurate so it would be interesting to actually build the boxes and see what the "real world" results were.... not something I'm going to be able to do though.....
Tony.
wintermute said:
A note, with most music, there is not much energy below ~40Hz. At a few CD tracks I had looked from different genres, the peak spectrum at low frequencies also follow average spectrum shape, meaning with most music the peak excursion requirements also start to decrease below ~40Hz. So with most music, the real critical region in terms of excursion is somewhere above 40Hz. I think this is another aspect that is overlooked. There are of course exceptions with CD tracks that contain real canon shots etc, but they are exceptions
. Just my 2 cents.
good point, I haven't actually done a spectrum analysis of any cd's.
One that I have which I suspect does have sub 40Hz is Enya Watermark.... it has some seriously deep bass in it, but without checking it I could of course be wrong
If you are doing HT then that's another story completely but I guess most people would have a dedicated sub for that, pretty much eliminating the problem....
Tony.
One that I have which I suspect does have sub 40Hz is Enya Watermark.... it has some seriously deep bass in it, but without checking it I could of course be wrong
If you are doing HT then that's another story completely
but I guess most people would have a dedicated sub for that, pretty much eliminating the problem.... Tony.
wintermute said:
Very good. Its good direction for thinking about this problem.
Which program you use for this simulations of Xmax etc. ??
OK. Here I go with corrected calculation for this Vifa XG18.
Voice coil resistance [ohm] 5.7
Sensitivity [dB] 86
Free air resonance [Hz] 34
Voice coil height [mm] 14
Air gap height [mm] 4
Voice coil inductance [mH] 0.53
Eff. diaphragm Area [cm2] 129
Moving mass [g] 11.8
Force factor [Bl] 5.6
VAS [l] 45
Qms 1.77
Qes 0.46
Qts 0.36
1. Xc (Xmax for calculation) = k1*Sd*sqr(Re)/BL = 0,934*0,0129*sqr5,7/5,6=5,14 mm
k1 = 1/(Qts+0,68)*0,68+0,28
2. Fb=h*Fs (for Qts=0,36 - h=0,382/Qts+0,14*Qts=1,1115)
Fb=37,8 Hz (not finnaly Fb)
3. Rt (half-diametar of tunel)= 0,501*sqr(Xc*Sd*Fb)=25,08 mm
4. Lt (tunel lenght) = 1,272*(sqr(sqr(BL*BL/Re)))*k2/Fb=133,6 mm (I got table for k2 depend of Sd)
5. Vb=(Rt*Rt*C*C)/(4*Pi*Fb*Fb*(1,7*Rt+Lt))= 23,47 L
6. Qt calculated from Alpha as Vas/Vb - Qt=0,343
7. new Fb, say Fb1 from Qt and Fs - Fb1=39,5 Hz
8. Fb2=(Fb+Fb1)/2=38,65 Hz
9. correction of Fb2. Vb*Ro (Ro is specfic density of the air) = Ma=27,77 gr. Mms=11,8 gr - (Ma/Mms)at (1/7)=1,13
1,13*Fb2=Fb3=43,676 Hz
Vb=23,5 L
Fb=43,7 Hz
Rt=0,501*sqr(Xmax*Fb3*Sd)=0,501*sqr(0,005*43,7*0,0129)=26,6 mm (this about minimum. if it is less, at hihger volumes air velocity would be more than 1/10 of C, more than 35 m/s. that means turbulent flow of the air in BR tube. not good.)
Lt=103 mm (formula at point 5.)
And this is if the unit works good at higher volumes, which is not like that in the most of cases. It must be measured, at different volumes (V) of the signal and that"s it.
I dont like recommendations (from manufacturer), too. Only measurements and calculations.
Hi Notax,
I used unibox ----> http://home20.inet.tele.dk/kou/ubmodel.html free for non-comercial use, but you need microsoft excell... It has more graphs but I just posted the cone excursion and spl comparison ones. It also has a more advanced modelling system which uses two additional parameters, le2 and re2 which are inductance and resistance at higher freq's I think...... I've never used these though because I haven't had that data.
I put the xmax in at 5mm (14mm voice coil height - 4mm gap height)/2
I've tried to follow your calculation, but I think I need to go to sleep, I just realised it is nearly 1:30AM!
Tony.
I used unibox ----> http://home20.inet.tele.dk/kou/ubmodel.html free for non-comercial use, but you need microsoft excell... It has more graphs but I just posted the cone excursion and spl comparison ones. It also has a more advanced modelling system which uses two additional parameters, le2 and re2 which are inductance and resistance at higher freq's I think...... I've never used these though because I haven't had that data.
I put the xmax in at 5mm (14mm voice coil height - 4mm gap height)/2
I've tried to follow your calculation, but I think I need to go to sleep, I just realised it is nearly 1:30AM!
Tony.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.