A/V
Speakers
This was my first attempt at speaker building. It was to be a learning
excercise in both design and woodworking.
Design Goals.
- Small, 2 way, ported a/v type speaker.
- Passive crossover
- Easy to drive impeadance.
- Cheap !
Driver search
I spent a long time looking at various drivers, their parameters and
plotting them using "WinIsd".
It didn't take me long to realise that the cone material played a very
important part in how the driver was going to perform. Even with
advances in modern materials, some compromise is still neccesary.
Ideally the cone acts like a piston, with all parts moving at the same
speed and direction. As the cone is driven to higher frequencies,
resonances can and do occur causing parts of the cone to "break-up"
from the ideal piston movement. This is more prevalent with larger
diameter cones. It is far easier to maintain rigidity with smaller
cones. Break-up causes the cone to become a directional radiator
at higher frequencies as indicated by the variation in the "off axis"
frequency response. As can be seen in the graph below, this driver is
quite good below 3KHz but gets very directional above this. For
this reason, I chose a crossover frequency of 3KHz.
One interesting point to add here. I took several published designs and
plotted the specified driver with the actual enclosure used. This
was quite revealing, try it sometime. Most of them just dont
work. One favourite trick seems to be to tune the port a little
higher than it should be, or make the box too small. This gives
an upward bump in the bass response before things start rolling
off. This seems to be a very popular technique to give the
(false) impression of "good bass response".
Another oddity I noted whilst plotting the response of drivers in
WinIsd. Some drivers are just impossible, needing absolutely
massive enclosure volumes in order to work. (anything from 200-400
litres) Fortunatley, most are designed to work in "real" boxes of
modest size.
Bass
To produce bass, the driver has to move a large quantity of air.
To move large quantities of air needs a large cone area.
Large cones are more prone to break-up.
Two solutions
1. Use a driver with a small cone and accept a poorer bass response.
2. Use a driver with a large cone and limit its frequency range to that
where break-up does not occur.
For the purpose of this design, as it was intended to be a small
speaker, I chose solution 1 and the Audax AP100Z0 100mm driver.
The main reason for this being that, in this two-way design, the driver
has to cope with a relatively wide frequency range.
Because of the small diameter of this driver, it has a low maximum
sound pressure level before the maximum cone travel (Xmax) is
exceeded. I realised that this was not going to be adequate
but I still wanted to use this driver. The solution to this was to use
two drivers which gave me an extra 3dB max. sound pressure level.
The design
The AP100Z0 drivers were plotted using "WinIsd". Because I am
now using two bass/mid drivers, this had the effect of doubling the
required enclosure volume to 17.9 litres but that was the penalty I had
to pay.
From the graph, it can be seen that the response is -3dB at 47 Hz,
remarkable for a 100mm dia. driver.
From the manufactuers data, the minimum impeadance was 6.1 ohms at
200-300 Hz. If I connected the two drivers in parallel, this
would give me a minimum impeadance of about 3 ohms. As one of my
design goals was to make this speaker easy to drive with any amplifier,
I thought that this figure was a bit low. I therefor made the decision
to connect the two drivers in series giving me a minimum impeadance of
12.2 ohms. This has the added advantage in that any amplifier driving
this will have limited output power capability into 12 ohms and the
chances of blowing this speaker up would be reduced.
"Running in" speakers
I thought I would mention this bit here because of what I have seen
written in the Hi Fi press. I have often seen comments along the lines
of "they will need running in for a couple of weeks.....". This sounded
to me like another excuse from the HI Fi industry. There is, however a
sound scientific basis for this and it has to do with the mechanical
characteristics of the speaker drive unit. When I recieved the AP100Z0
drive units, I did a few quick measurements on them to check that the
published data was correct. I found that the self resonant frequency,
which was quoted as 64Hz, was actually around 80Hz. To see if the
mechanical suspension characteristics would change, I left one driven
with about 10w of power at 20Hz for 24 hours. I again measured its self
resonant frequency. It was 68Hz, quite a large change, indicating that
the suspension characteristics had indeed changed. After a further 24
hours of "running in", the frequency was close to what it should be. I
repeated the procedure on the other three drivers, which all behaved
similarly. My speakers were "run in" before I had even assembled them.
I have also seen this comment applied to electronic equipment. One such
pre-amp actually had a large label stuck on the front indicating that
"optimum performance would only be acheived after a period of running
in." Oh dear! no mechanical changes here. As an engineer involved
with electronics for a long time, I'm afraid I don't accept this. If
the characteristics of a piece of electronic equipment change
significantly over time, then the design is faulty. Specifications
should be maintained throughout its lifetime. OK, so you could
say that it is the capacitors that change. In my experience, I have yet
to "hear" a capacitor, unless it is faulty of course.
<rant>So this is again the Hi Fi retail sector taking half a
scientific fact and applying it to something un related to perpetuate
their perceived superiority over joe public. </rant>
The tweeter
For this design, I chose the Audax TW025M0. It wasn't the best
tweeter in the world but for its price, had received good reviews. This
has a self-resonant frequency of 900Hz which is well away from the
crossover frequency. Its minimum impeadance is quoted at 6.5
ohms. One of the most important characteristics of a tweeter is
its off-axis frequency response. For this reason, the cone usually
takes the form of a hemispherical dome. Diffraction from cabinet edges
can also degrade the off-axis response so the tweeter should be
flush-mounted on the baffle and the enclosure corners rounded to
minimise diffraction.
Crossovers
This is not intended to be a constructional article, merely a record of
my learning process and design technique. Therefor, only basic details
are given. All crossover components were calculated using the tools
from Rod Elliot's
site which
makes essential reading.
The optimum crossover frequency was determined by looking at the
off-axis response of the bass driver in conjunction with the self
resonant frequency of the tweeter. It is best to keep the crossover two
octaves higher than the tweeter self resonance otherwise there will be
a peak in the response at that frequency.
Crossover units only work properly into a constant impeadance. If you
see a crossover design that takes no account of this, then treat it as
suspect. As can be seen
from the above graph, the impeadance of the AP100Z0 is anything but
constant and varies between 6.5 and 100 ohms. The tweeter behaves
in a similar manner with a hugh peak in impeadance at its self resonant
frequency.
Both drivers exibit a rising impeadance with frequency. This is due to
the inductance of the voice coil. Fortunately this is very easy to
compensate for simply connecting a series C/R network(Zobel network) in
parallel, with the driver effectively cancells this.
The large impeadance peak at resonance is more difficult to compensate
for. This can be acheived using a series resonant L/C/R circuit in
parallel with the driver. The resistance component is calculated so
that the Q of the network, matches the driver.
This can readily be acheived for the tweeter, but becasue its self
resonance is so much lower, component values become impractical to
correct this on the bass driver. We'll just have to live with this I'm
afraid.
So far, we have two drivers in series for the bass/mid with thier Zobel
network and one tweeter with its resonance compensation.
Note that the polarity of the tweeter is reversed. This is due to the
phase characteristics of the proposed second order, 12dB/Octave
filtering in the
crossover sections.
The next step was to take into account, the variation in sensitivity of
the two drivers. The sensitivity of the tweeter is quoted as 92 dB/w
wheras the bass/mids are quotes as 85 dB/w. A difference of 6dB.
An extra 3dB also needs to be added to this due to the fact that there
are two drivers in series for the bass/mid. At total of
9dB. Firstly, I made the impeadance of the tweeter equal to that
of the two
bass units in series. This was acheived using a series resistance
of 8.2R. The loss of this was calculated and almost gave me the
required attenuation, the further attenation required was acheived by a
simple L-pad. In this case, I didn't use a zobel network for the
tweeter as when I measured its impeadance, there was only a slight rise
which I didn't think was worth correcting. The impeadances used in the
crossover calculations were those actually measured with the impeadance
correction circuits connected. Measurements of impeadances were simply
calculated by doing measurements of voltage drop accross a
series resistance with a manual frequency sweep from an audio generator.
The HF section of the crossover network.
The LF section of the crossover network
Note that in the two crossover circuits, the values of L and C are the
same in each. This is consequence of making the tweeter and the
combined bass speakers, have the same impeadance. The 1734uH inductors
were made from off the shelf 1800uH models with a few turns taken off
whilst measuring with an inductance meter. The 1.62uF capacitors were
made from 1.5uF in parrallel with 0.1uF, giving a total of 1.6uF.
Not perfect but probably close enough.
The ported enclosure
I have mentioned WinISD before and this really proved invaluable in
this design. Using the box designer, I played around with the
dimentions whilst keeping the volume constant. Eventually, I came up
with the box that I wanted. External dimensions worked out as 150W,
780H, 247D. (mm)
Port design was taken care of with WinIsd, indicating a tuning
frequency of 52.46 Hz. I had bought a pair of 68mm dia. flared port
tubes for this project. (Maplin Electronics-UK) and again WinIsd
calculated the length required to be 167mm to which I dutifully cut
them.
I actually made the box slightly higher than that calculated, the
reason being that I wanted to add an internal brace to the enclosure
and to compensate for the volume taken by the drive units,crossover
conponents and port tube. When completed, all internal dimentions were
carefully
measured and the actual volume calculated. If this is too large,
it is only neccessary to add a couple of blocks of wood to
compensate. If its too small, well you were warned!
The inside panels were damped by means of "Dedshete panels" (Wilmslow
Audio-UK). These are self-adhesive butuminous pads laminated with 10mm
thick foam. These really do stick and do a good job of damping the
enclosures.
The crossovers were constructed on PCB material with wide tracks and
mounted on the rear panel behind the bass/mid drivers. I covered the
crossovers with sculptured foam insulation to help prevent internal
reflections.
Construction
Rough drawings were produced and the panels were cut from sheets of
18mm MDF. The cutouts for the drivers in the baffle were made using a
router with 6mm round bit and a home-made circle jig. The edges of all
cutouts were slightly rounded using the router with a "rounding over"
cutter. All drivers were recessed flush with the baffle front, this is
especially important for the tweeter but makes the whole thing look
more proffessional. The tweeters were offset from centre, this was
another attempt at minimising edge diffraction effects.
Drivers were mounted using 4mm bushes glued into
place from the rear of the baffle. No grills were considered in this
design.
For final finishing, I applied several coats of black paint using a
gloss roller. I must admit that I am not happy with the finish
and intend to sand them back to bare wood and start again. Next time,
they will be better...
The finished article
Performance
No matter what goes into a loudspeaker design, how it actually sounds
is the only thing that matters. To judge this it is neccessary to
compare them against some kind of reference. My reference, in this case
was my Spendor BC1's. These are not perfect by any means but were the
best I had at this time. Against the BC1's I was immediately
dissapointed with the bass response. I had thought that -3dB down
at 47Hz would be adequate but that is definately not the case. Midrange
and treble were however very acceptable, the bass I could do nothing
about but consider a subwoofer,
which is exactly what I did. After constructing the subwoofer, I was
very pleased with the overall sound. I did notice a slight directivity
in the vertical plane. This I attribute to the dual driver design and
the speaker centre being below ear height. I found that I could make a
worthwhile improvement by tilting the cabinets backwards. I simply used
a DVD case under the front of each cabinet. This produced a worthwhile
improvement in stereo imaging.
Another test I carried out involved placing the cabinets on stands,
even though I designed them for floor mounting. This was a disaster as
I lost the bass enhancement due to close coupling to the floor.
By this time, I had built up a simple kit to measure the frequency
response of my design. This consists of am ECM8000 measurement
microphone and balanced line pre amp with phantom power supply. I setup
the kit with the speakers in my "anechoic chamber" (ouside in the
garden) and performed a manual sweep, measuring the preamp output at
each point. I plotted the response using Open Office spreadsheet.
The above chart shows the measured response of the speaker design. The
undulation in response, I mainly atribute to reflections from the house
and fences. Also bear in mind that the vertical scale is 2.5dB/div.
Commercial response charts are usually 5dB/div making them look
flatter. Allowing for measurement inaccuracies, the response is
probably OK.
Conclusion
Well, I certainly learned a lot during this excercise. I have lived
with these with the subwoofer and on the whole, they produce a nice
"noise". There are no noticeable resonances and the response sounds
very un-coloured. You might even say they sound boring but should a
loudspeaker add anything to the original sound ? I think not.
<-- Back