Website updates

[John Warren]

Some updates to the website-

I put a page together that describes how to derive the impedance response for a driver (K33E in free air) from T/S parameters using a very user friendly version of the SPICE III simulator that you can download and use.

http://home.comcast.net/~wooferboy/Page_9.html

Hopefully this will generate some real interest in modeling!

[boom3]

Thanks John!

[m8o]

Thank you John. A long time audio hobbiest and speaker builder myself but about 15 years out of the hobby, I have a new-found interest in audio engineering spurred on by first, purchase of La Scalas and Heresys, then getting into the headphone hobby with purchase of several pure tube headamps and hi-end headphones. One of those headamps, the Little Dot III+ can actually drive the La Scalas decently; a DH-SET OTL head amp that comes with speaker terminals capable of 2.7watts in triode mode and 3.5watts in pentode mode.

I have a desire to experiment with constant current amplification of the La Scalas -- my end goal is a tube-based crossover integrated with a special purpose constant current triamp sections driving each horn independently. But for now I'll stick with a single amp with the passive crossovers in the picture (yes, I know that constant current drive is best used to a single driver w/o passive crossover)

I found an intriguing MOSFET Single Ended Constant Current Class-A Power Follower circuit I could build to be used as a buffer and placed between the Little Dot III+ amp and La Scalas to really get things [air] moving --- enter the spice modelling! I'm experimenting with changes to the circuit using a speaker subcircuit, and that's what spurred me to post here. (btw, after trying many software packages, I too find myself ending up on LTSPICE.)

I'm hoping to create a -complete- spice model of the 3-way horn-loaded drivers plus crossover network, to use as the load on the circuit. Btw as an aside, coming from one who has built quite a few custom crossovers with great success, your crossovers look exceptional! Even given that my goal is to replace the passive crossovers with active x-over and constant current amplification, I'm tempted to request if you're interested in making a set and selling them to me until I get done ... which I should be honest and admit will end up being a -long- time from now given I'm prone to procrastination [:$]

But I digress....

I read your article on Modeling Loudspeaker Performance using LTSPICE with great interest. It was great to come upon this thread at this time. If I make progress on my own I'll post back here.

-steve

[John Warren]

Steve-<?xml:namespace prefix = o ns = "urn:schemas-microsoft-com:office:office" />

I'm glad you found the site of some interest.

Keele provides an equivilent circuit for a bass horn (see my website under Technical Articles and look for "Low Frequency Horn Design Using T/S Driver Parameters"). The paper does not provide any experimental results to validate the model predictions (for example, predicted impedance vs. actual). Also, the model considers a straight horn so complexities of folding are ignored. The paper was written when Keele was working for Klipsch. If he extended the model to consider folding, I'm not familar with it. Leech has written extensively on using SPICE to model loudspeaker performance. His website (at Georgia Tech) might be a place to look for other models.

To your point regarding loading the crossover circuit. Attached is a simple model of the Klipsch K28e woofer in the Heresy cabinet. The Heresy is a seal box so the model is simply the loudspeaker eq. circuit in parallel with an inductor that represents the effect of the air compliance in the box (= BL^2 * Vb / rho*c^2 *Sd^2) where BL is the motor energy product (T*m), Vb is the box internal volume (m^3), rho is the density of air (kg/m^3), c is the speed of sound (m/s) and Sd is the radiation area of the cone (m^2). I think the Heresy uses a 2.2 mH inductor in series with the woofer (that's what I modelled anyway).

I also show the response of the inductor with a 4 Ohm resistor in series. This is typically how most "canned" crossover packages would model the behavior i.e. by treating the woofer as a purely resistive load.

jw

[m8o]

Thanx for the references. Found one and will look for the other shortly.

I realize I actually should go the reverse engineering route with manual interpolation I think. I took Terrasonde's offer to upgrade my AudioToolbox to the latest one (haven't actually used it since though!). I can very accurately measure the impedance of the whole speaker system across the audio frequency spectrum, or each driver individually, each network node, whatever; the system's what I'm looking for right now I guess. I can then create a complex LCR circuit to match the impedance I measured through manually interpolation. I did this with my Martin Logan Sequel IIs actually as just a test to see if I could a few years back. The impedance of my circuit was darn close to what I'd measured. ...call it the "brute force" method I guess. []

EDIT: BTW, I see you like the JBL 2404H a lot. I was thinking about using them on my La Scala then discovered they were discontinued. What do you recommend be used now? (and why on earth would JBL do that?)

[John Warren]

The brute force method is certainly one way. The Keele model will predict both the peak at Fc and the increase in radiation resistance due to horn loading. It will (naturally) not predict the various resonance peaks associated with reflections back to the throat due to folding. Thus, if you are so inclined, plot the woofer free air impedance and then the same in the LaScala. This will give you some idea of the upper limit of the radiation resistance provided by the horn (and I'd like to see that plot if you get time to do it) and Fc. Then compare it to the Keele prediction using the T/S parameters for the K33e woofer and the physical parameters of the LaScala (Vb, St). In the Klipschorn for example, the impedance of the K33e mounted in the horn is, for all intents and purposes, the same as free air above 300Hz. The nets used on the Klipschorn attenuate at ~400Hz so what comes out of it above 300Hz may be less than desirable. I suspect that one of the design goals of the Jubillee was to improve horn loading above where the Klipschorn is today and thus reduce distortion.

The 2404H was a nice unit, now NLA. Not sure why, never asked. They might have been on life support for a few years (they were never in stock at JBL when I needed them). Other suppliers, Beyma, Fostex, BMS line arrays. One could also go to a JBL, Radian or McCauley compression driver with horn assy and crossover at 4kHz on up but that does make for a pricey "tweeter".

[hwatkins]

Very nicely done John (and helpful). I am starting to collect data to help determine the end result of new speakers to be built for a redesigned living area and you are, as usual, delivering quality info.

[John Warren]

Why thank you Henry!

What are you going to (or leaning towards) building for you new system? I'm all ears!

jw

[m8o]

The brute force method is certainly one way. The Keele model will predict both the peak at Fc and the increase in radiation resistance due to horn loading. It will (naturally) not predict the various resonance peaks associated with reflections back to the throat due to folding. Thus, if you are so inclined, plot the woofer free air impedance and then the same in the LaScala. This will give you some idea of the upper limit of the radiation resistance provided by the horn (and I'd like to see that plot if you get time to do it) and Fc. Then compare it to the Keele prediction using the T/S parameters for the K33e woofer and the physical parameters of the LaScala (Vb, St). In the Klipschorn for example, the impedance of the K33e mounted in the horn is, for all intents and purposes, the same as free air above 300Hz. The nets used on the Klipschorn attenuate at ~400Hz so what comes out of it above 300Hz may be less than desirable. I suspect that one of the design goals of the Jubillee was to improve horn loading above where the Klipschorn is today and thus reduce distortion.

I finally got my Terrasonde out to do impedance measurements (had been so long the battery won't hold a charge anymore). I need to transfer them to the PC before I can display them (unless I take a digital photo of the display []) I did each speaker as a unit and then each driver individually w/o network. I don't think I'm up to the task of the simulation based on what you're describing mixing T/S parameters and horn theory sadly though.

Regarding my measurements, all I can say is I'm floored, shocked and dismayed! The impedance hump in the midrange is frightening in its 'massiv-isity-iness'! Now I see what Al K had been 'soapboxing' about and against in the past. Also surprising is the bass horn got into the low 3 ohms, with the mid driver+horn remaining above 10 or 11. I'm going to have to produce a higher current constant current amp then I had 1st planned for the bass; easy though with the LM4780 'gainclone' chip I settled on.

BTW, very interesting what you're saying about the mid & bass horn response. Based on 'by ear' tone & sweep tests (yes, will do those advanced time domain tests I purchased for the Terrasone that can measure frequency response and basically removed a room's interaction from the measurements) directly driving the driver (no network), I see how I want to x-over the mid horn at 450 or 500. I thought the bass horn did fairly well above 400 hz myself (again, sans network). I don't know about the bass horn's polar response though.

BTW, do you happen to know the K-55 mid horn driver's electrical power handling? I've come to discover that's somewhat of an enigma based on a search of this board.

-steve

[DrWho]

Do these models work for higher frequencies too? I remember reading somewhere that T/S breaks down when you start dealing with frequencies above 200Hz. [^o)]