Delta
Gattiweb

Loudspeaker Design


It is reasonable to have perfection in our eye that we may always advance toward it,
though we know it can never be reached - Samuel Johnson

Things should be made as simple as possible - but no simpler - Albert Einstein

This is my latest loudspeaker project which began in January 2004.
It encompasses some of my own theories on enclosure, filter and aesthetic design .
They are expensive, but extravagance has been avoided, and when seen as a long term investment,
the cost is reasonable. They are also 80-90% cheaper than an equivalent commerical product,
and the drivers make up 60-70% of the cost.
Exotic, fashionable or unproven technology is avoided - instead the emphasis is on using mature,
high quality technology and methods, and applying it the best way I can.
Since these speaker will be a permanent fixture in my living room, a sensible size and
pleasing appearance are essential.

Note: This design detail is for the benefit of DIY speaker builders.
It is not to be used for commercial purposes without my permission.

Features

- 3.5 way, 2 box design
- compact footprint and appearance
- linear phase crossover design
- numerous diffraction control measures (pyramid shape, curved baffle)
- heavily braced, constrained layer, cabinet construction
- high sensitivity 90dB/2.83V/1m
- high power handling, thermally and dynamically
- very high quality components
- external crossover & removable back panels for future upgrades and tweaks
- very good low frequency extension suitable for locating close to the front wall
- excellent off-axis performance

Delta Goodrem

The following manufactures have provided some of the inspiration and influences for this design, in particular, Thiel, Duntech & Vandersteen (click image for website).

Sonus Faber Duntech VAF
Thiel
Vandersteen Wilson Audio Usher

Design decisions

Filter topology. In my opinion, non linear-phase speakers cannot be considered truly accurate, as they discard musical information, distort the waveform shape and can exhibit ringing. This design uses mainly 6dB filters with response shaping and a time-aligned baffle to meet the linear phase target. I believe this is a "perfection" worth striving for in any true high fidelity system. See Rod Elliot's excellent site for lots of info on linear phase speakers. Other benefits of low-order crossovers are more uniform power response & less components in the signal path.

Passive or active? I considered a full active system, but rejected it on the grounds of complexity and inefficiency of resources. There is some potential for bi-amplifying to allow low frequency equalisation - an option that may be investigated at some stage in the future.

Radiation pattern - Monopole. Dipole and bipole designs were ruled out due to their requirements for steep crossover filters, large cone area and baffles, complex filters, multiple amps, unsuitability for home theatre, and listening space requirements. I also have little experience with them, so consider it a high risk option.

Depth. Enclosure depth is determined by volume requirements given the specified height & width of the enclosure.

Height. For a realistic image height, a speaker needs to be of an appropriate physical height. I have designed for a listening axis in the centre of the midrange driver, 83cm from the floor, which places the tweeter about 1 metre off the floor.

Width. To minimise visual impact on living space, the baffle width will be as narrow as possible. This also helps baffle rigidity.

Weight. For practicality, the speaker needs to be liftable by one man. To meet this requirement, a 2 box (satellite/woofer) system is used, where the heavier woofer enclosure will weigh about 40-50kg (100lbs), and the satellite about 20-30kg.

Sensitivity. I aim for a sensitivity of about 90dB/2.83V, which suits my preference for solid-state amplifiers of 100W into 8ohms. This specification will influence the choice of drivers and the need for a 3.5 way design.

Max SPL and low frequency response. A peak linear SPL of 110dB at 1 metre will be adequate (musical programme), with a -3dB point of about 40Hz, system Q of .5 to 0.7 (12dB/octave). Experience has shown this to be the optimum for my living room and tastes.

Drivers. The 6dB/octave electrical filters require drivers of wide, well-behaved frequency response. This dictates the use of self-damping diaphragm materials like polpropylene or paper. In my opinion, in an overall system context, a dedicated light poly or paper cone is still the best material for a midrange - a combination that only requires a simple crossover, with minimal energy storage, signal loss and/or ringing in the filter/driver combination. This is also the lowest risk approach for the DIYer. Furthermore, as I listen to a lot of music from the 1950s to 1970s, where recording quality is not always perfect, I've found that poly & paper cones present a more "forgiving" & "pleasing" sound. The final choice of drivers are among the finest currently available today (Jan 2004).

Finish. Floorstanding speakers are large items of furniture, and their visual impact can be softened with a wood-grain veneer. These are long-term items, so a conservative look is preferred.

Midrange - Audiotechnology CQuenze 15H520613SDK

Midrange contenders included the Scan 12M/4631 and Seas Excel M15CH001, but these were ruled out due to low impedance, poor sensitivity, or limited excursion capability.
I've been a long time fan and user of Dynaudio drivers, with their simple crossover requirements, wide dynamic range and clean, articulate sound when used appropriately. Dynaudio drivers come from the same designer (Skaaning) as Audiotechnology. Variations of the Audiotechnology midrange are used by the very highly regarded Peak Consulting range, Verity Parsifal, Sonus Faber Stradivari, and Rockport Antares.
The Cquenze 15H52016 was ordered with the kapton former which produces a cleaner midrange due to the avoidance of eddy currents.
It's low moving-mass (8.9g), small effective diameter (11cm), and excellent bandwidth makes it perfect for this type of system. It's shallow profile reduces the necessary baffle slope for time-alignement & minimises it's diffraction on the tweeters output.

Tweeter - Morel Supreme 110 (pdf)

I considered the popular Seas Millenium, but it's sensitivity appears to be too low for this application.
The Supreme 110 easily meets the 90dB sensitivity requirements, and with it's underhung voicecoil, it is suitable for 6dB/octave filters. It does not use ferrofluid which should improve low-level detail at the expense of some power handling, but I don't this will be an issue. It also has excellent off-axis performance, approaching that of 18mm tweeters. Mark K has done some testing on his site.

Midbass - Scanspeak 18W8531G00

Having heard the 18W8531 in the Sonus Faber Cremona, I was very impressed with it's bass performance and midrange neutrality.
It is perfectly suitable for the midbass area and it's low distortion bass (for a 7" driver) is vital in this design, as it will not have a high pass filter, instead relying on enclosure damping. This driver is used to provide baffle-step compensation and a 6dB/octave quasi transient-perfect transition to the CQuenze midrange. Without it, voltage sensitivity would be 3-4dB lower, there would be less LF extension, and the main woofer would be called upon to work an octave higher. It's directivity and moving mass suggest a crossover point below 1kHz is preferable, a condition easily met here.

Woofer - Scanspeak 26W8861T00

I was looking for a specialist low frequency driver rather than a compromised midbass driver. I prefer sealed enclosures because of their lower cone excursion, superior group delay (tighter bass) and better match to the typical low frequency gain of most listening rooms.
12" woofers were ruled out due to their large baffle and volume requirements, so a long throw 10" woofer was sought after and found in this new, state of the art 10" woofer from Scanspeak . It's has an Xmax of 9mm, titanium former, thick and rigid carbon/paper cone, and it will suit my target enclosure volume of below 50 litres.

The combination of the 7" and 10" woofers effectively produces a wideband low frequency driver with a narrow baffle/small volume requirement, and output capability approaching a 12" driver. It also reduces floor cancellation effects and assists in time-alignment with the midrange.

Enclosure plans
The design listening axis is the centre of the midrange driver, 83cm from the floor.
2-enclosure design makes the weight manageable and provides better vibration control. The sloped baffle time-aligns the drivers. Curved edges and the pyramid shape breaks up internal standing waves, reduces edge diffraction and enhances dispersion. The slim, narrow pyramid-style baffle also improves aesthetics and makes the overall size appear smaller. All drivers are flush-mounted except for the 10" woofer which is surface mounted to assist time-alignment and provide the strongest possible mounting.
The construction will be from mdf and particleboard, with a barium-loaded-PVC/MDF constrained layer for very good wideband vibration absorption. There will be shelf-bracing for rigidity and seam bracing to mechanically couple all sides, thereby providing the optimum sink for any vibrations from the front baffle. The base will be optional and not fixed, to allow optimisation of the vertical listening axis.

At a height of 83cm and distance of 300cm, the listener's ear is at the centre of an arc that intersects the acoustic centre of all drivers. Thus, the output from all drivers arrives at the listener simultaneoously.

Filter design concept, using LspCAD

Note that the following graphs are preliminary simulations based on manufacturer data. Actual measurements will follow later. See crossover section. Blue is the Scanspeak 26W8861, red the Scanspeak 18W8351, green the Cquenze 15h5206, yellow the Morel Supreme 110, and black the total.

Note the theoretically flat phase response (+/-15 deg), hence "transient perfect" result. The low frequency response is into free space - in a normal room, boundary reinforcement will raise the response below 75hz.
Low frequency response will be subjectively fine-tuned with internal damping materials and/or lowpass inductor series resistance.

The woofers have different resonant frequencies, and thus their parallel combination reduces their impedance phase angle, easing the amplifier load. 
Impedance minimum will be about 3 - 3.5 ohms at 200 Hz, depending on the final crossover and the lowpass inductors I use. 
Delta will require a good quality solid-state amplifier, around 100W or more.

Below is the simulated response into half-space, which is a good indicator of the in-room low frequency performance. -3dB occurs at abuut 40Hz.

Low frequency power handling.

At Xmax :

Cone excursion versus frequency is shown below (using Unibox). 
This graph shows how the 18W8531 reaches it's Xmax at 12.75V (at 50Hz and below), and below that is the 26W8861's performance at the same voltage.

18W8531 in 10 litres at 12.75V (20W into 8ohms).
26W8861 in 40 litres at 12.75V (20W into 8ohms).
At Xsus :

Below 50Hz, at 21.54Vrms (58W into 8ohms) the 18w8531 reaches it's 11mm peak suspension limit (see below).
At the same 21.54Vrms, the 26W8861 has almost, but not quite, met it's suspension limit (14mm peak).

From the above individual drive limits, a maximum linear SPL can be determined for the 2 woofers combined. This is the 18W8351s maximum SPL + 9dB (relative response of the 26w8861). Given that the low frequency content of music falls rapidly below 100Hz, the system overload point can be estimated as 10dB higher than the 50Hz overload point (Ref: Martin Collums High Performance Loudspeakers). Thus it's safe to say Delta can produce 115dB (with music), which equates to 300 Watts of power handling, With a typical 100W amplifier, 110dB per channel at 1 metre can be comfortably reached.

Off axis responses

Below is the horizontal off axis LspCAD simulation, showing a smooth overall power response, another benefit of "soft" cones and 6dB filters.

Below is the vertical off-axis simulation..
This illustrates the relatively narrow vertical listening window of this design. Above the listening axis, severe dips occur due to cancellations between midbasss & mid, and between mid and tweeter. Listeners should be between 0 and 5 degrees below the midrange axis, which at a distance of 3 metres (10 feet), corresponds to a 40cm high listening window. Given the listening axis is 83cm from the floor, then the listeners ears should ideally be from 45 to 85cm from the floor. Fortunately, when listening seriously, most of us sit at a defined height (our favourite sofa), so this condition is easily met. Delta is not designed for near-field monitoring, standing rooms or floor sitters. Realistically, I've never found this to be an issue.
 
home                     construction