Vented alignments

[Mikeng43]

Let's talk alignments. Are they necessary in todays computer world? I know many pioneers on the subject created these alignments (tables) and many thanks to their enduring work. But, are they necessary with todays computer calculator design programs (i.e. SE, LspCAD etc.). It seems these tables are not linear and the question is... is it best to use the alignment diagrams (alignment magic) per Qt (with Qb) or is it better to just use the software to find the best curve per application? I have been reading the work done by Thiele, Keele, etc., and it's great but am I constrained to these tables? Keele's work showed sensitivity to parameter variations that make a noticable difference (practice vs. theory) that has me scratching my head. I think Ron E said (it sticks in my head) that the alignments are just points on a continuum and alignments were just a reference for getting from A to B in the old days. Is this true? Should I use accurate software now vs. the alignment tables? Is the software as accurate as the tables.

Thanks,

Mike

[richie00boy]

Good software is as accurate as the tables. The tables can be useful for a starting point, but they are just a bunch of useful examples, there's many many more other good ones out there that you can just make up with looking at the modelled responses to suit your needs.

Also taking room influences into account renders the validity of the tables into question for some cases.

[Svante]

IMO, designing loudspeakers using simple formulas ("alignments") is a leftover from the past. These formulas typically generate a Butterworth-like solution in an anechoic chamber but completely neglect the effects of the room, the baffle step, voice coil inductance etc.

I would say that they can be starting points for a design, but they should by no means be regarded as final solutions.

In my software Basta! one can apply such formulas to get a starting point. From there one will usually tweak the parameters of the design to acheive whatever design goals there are.

Below is a simulation using three common "alignments" with the room gain of an average room, the baffle step of a 20 cm wide baffle and also considering the voice coil inductance.



Interestingly, the same holds for crossover filters; table filters are rarely optimal when the response of the entire system is considered. Using a simulator that simulates everything, or even better using measured data and a simulation of the electrical parts gives a much better opportunity to end up where you want.

For example, the wide hump around 500-1000 Hz in the example above could be brought down by tweaking the values of the filter components, or adding some more components.

... so my bottom line is that today's simulators are much better at helping you reach a reasonable design than the simple alignment formulas from the past are.

[Ron E]

Points on a continuum is one way of looking at alignments. I am not a proponent of designing to alignments, but I do think they can serve a reasonable starting points to design. In the simplest case, an alignment is just a chosen box size and tuning frequency. Some of these choices work better than others. Alignments represent an ordered, mathematical way of looking at the relationship of Vas/Vb and Fb/Fs to frequency response.

The strength of a chain is defined by its weakest link, and the weakest link in the T/S theory is the simplifying assumptions behind it. In most T/S models, linear halfspace behavior is assumed, this means a box behind an infinite baffle, and no room or furniture in front of it. Obviously being on a finite baffle in a room changes things somewhat, as Svante's sim shows admirably.

Trying to account for uncontrolled variables (room size, construction, furnishings, driver nonlinearity, etc..) is not a simple thing to do. How would you make a program that a layman could use that would allow entry of the complex situation at hand? You have to stop somewhere and T/S theory helps design to the simpler driver variables quite well. Once you include damping terms and other variables like voice coil inductance you can't even represent the system with the traditional alignments anyway.

Parameter variations are important to understand, the most important of which is compliance variation over the life of the speaker. The compliance will increase and Fs will decrease in turn because the mass of the speaker won't change unless you are in a REALLY dusty environment Luckily, response changes due to compliance variation are pretty minor. As long as you measure your speaker before designing a box for it, you shouldn't have to retune due to aging.

For Svante, I am curious what defines the Ohman curve. Is it another curve fit like the others? BTW, I wouldn't call the Keele/Small/Margolis design methods alignments, although they are two ways of curve fitting the maximally flat QB3 and Equiripple C4 alignments to get a simplified approach.

[Svante]

Originally posted by Ron E
Parameter variations are important to understand, the most important of which is compliance variation over the life of the speaker. The compliance will increase and Fs will decrease in turn because the mass of the speaker won't change unless you are in a REALLY dusty environment Luckily, response changes due to compliance variation are pretty minor. As long as you measure your speaker before designing a box for it, you shouldn't have to retune due to aging.

Indeed understanding is important, and that is where simplifications can have its place. A while ago, when I implemented those alignments I tried to derive my own version of the "design suggestions" as I call them in Basta!. It was based on the fact that changes in compliance make very small changes on the response. The math gave me a formula which was rather similar to the three more well known equations, so I actually left it out. But the journey through the derivation of the equation was certainly worthwile anyway. This is why I encourage people to try to dig beyond the formulas and try to understand when they are valid and how they are derived. They are good starting points, but almost all formulas/models have some sort of limitations.

Originally posted by Ron E
For Svante, I am curious what defines the Ohman curve. Is it another curve fit like the others? BTW, I wouldn't call the Keele/Small/Margolis design methods alignments, although they are two ways of curve fitting the maximally flat QB3 and Equiripple C4 alignments to get a simplified approach.

Ah, yes, maybe I am wrong in calling them "alignments". They will result in near butterworth alingments if Qts is near 0.38.

For the Öhman equations, they are visible on the "design suggestions" tab in Basta!. The .../(1+Viso) is added by me to account for the virtual box volume increase due to isothermalisation from the stuffing.



FYI: Ingvar Öhman is sort of a Hifi Guru in Sweden who has devoted most of his life to loudspeaker design. He made up those formulas some 10-20 years ago, I think, but he actually isn't very fond of them himself (he calls them "toy formulas"). Stereophile recently made a review of one of his designs:

http://blog.stereophile.com/he2007/051607sjofn/

[oshifis]

Group delay is frequently neglected in speaker design. Here is another "toy" that is able to display group delay:

http://tube.fw.hu/Speaker_impedance.xls

Room interaction is not accounted for in my model, though.

[Svante]

Originally posted by Svante

The .../(1+Viso) is added by me to account for the virtual box volume increase due to isothermalisation from the stuffing.

Sorry, it should be .../(1+0.4*Viso), obviously.

[sreten]

Hi,

the alignments of yesteryear are based on an arbitrary performance
criteria, e.g. maximally flat for butterworth, equiripple for Chebyshev
etc, but as described above they do suit real room acoustics.

So you change your abitrary criteria. B&W for example use near 4th
order Bessel alignments which match well with rooms, and can be
combined with an active high pass 1<Q<2 filter depending on the
room to give a more extended 6th order alignment.

As stated above classic say 4th order and 2nd order L/R alignments
are alive and well as acoustic alignments but the practise of
using acoustic alignments has not transferred to the bass end.

Presumably due to the unknowns, room size, shape and speaker position.

My favourite vented alignments have an accelerating slope the
lower you go and you try to suppress an obvious "kneepoint".

No doubt the mathematically inclined could find the "criteria" that
would cause this and define the "family" of filters it is part of.
Detuning a standard Butterworth by 0.7 seems to work well :

/sreten.

[Svante]

Yes, room compensation is tricky. As a DIYer one could in principle optimise the speakers for a particular placement in a particular room. Then again one will probably want to use the speaker in some other room later on. Commercial speakers must optimise towards an average room.

Typical rooms are resonant in the bass region, but resonances can hardly be compensated for unless the exact room and placement is known. OTOH all reasonably sized rooms exhibit some sort of bass lift, and this can be taken into account. It will not remove resonances, but to some extent compensate the overall bass lift that occurs in all rooms.

The bass lift I showed in my simulation above is almost a straight line from 100 Hz to 20 Hz, lifting 20 Hz by ~10 dB. It is not a simulation of an actual room, but a lift that is reasonable to compensate against.

Given this compensation, new equations could be derived, but they will not be scalable (frequency-wise) as the old ones were. The room compensation should not move if the driver fs does. So, while this is possible, I think time has run away from the "design equations" since almost everyone can afford a computer and simulation software today.

[sreten]

Originally posted by Svante


The bass lift I showed in my simulation above is almost a straight line from 100 Hz to 20 Hz, lifting 20 Hz by ~10 dB. It is not a simulation of an actual room, but a lift that is reasonable to compensate against.

Hmmm.......

The smoothest bass lift you are going to get is with the speakers
1/3 along each room diagonal - not a typical placement. so failing
that the bass unit needs to be distanced away from the room
boundaries by dissimilar distances, the most important criteria
to me is the length of the shortest distance.

Who Stole The Bass? / No One Stole The Bass
Anthony H. Cordesman & Martin Colloms, April, 1987
http://www.stereophile.com/features/44/index.html

Is where that model comes from - and it does not apply to
placement near boundaries or speakers with bass drivers
near the floor, see page 4 of the article for the context.
The whole article is quite entertaining ..... relatively .....

/sreten.