This is a project prompted by this
thread on diyAudio.com alerting people to the end-of-line deals
available on Audax drive units - in particular the AP100Z0, a 4 inch
mid-bass drive unit. These normally cost £15.00, but were down
to just £4.99 each - I just couldn't resist buying a load;
six in fact!
It's been a while since I seriously played with loudspeaker design, so this was a cheap way to refresh and improve my knowledge. If you're learning about speaker building, you might find some useful bits of information in this section...
The Audax AP100Z0
This driver has a HDA (high definition aerogel) cone, rubber surround and a shielded magnet assembly. The compact dimensions and shielded magnet gave me two ideas - a pair of ultra-compact monitors for PC use, or perhaps in the kitchen or a bedroom, and a slim centre speaker that could fit under the TV. With this in mind, I decided to buy 3 tweeters - choosing the Audax TM025F1, which has a diameter of only 70mm, and looks "right" with this woofer. A "normal" tweeter is nearly as big as the AP100Z0 - it's hard to get a sense of scale from the pictures, but these things are seriously cute!
I ordered the drive units and kept them in the office at work. They didn't go home because I was busy getting my house on the market, and I didn't need any distractions!
Coincidentally, I'd rescued an old pair of Musical Fidelity Reference 2's from the skip a while ago, and have been wondering what to do with them ever since. The Reference 2's were quite good in their day - I bought a pair for my Dad back in 1989 - but both bass drivers had failed in this pair, and there was no chance of finding replacements. So I decided to rescue the crossovers, tweeters and acoustic foam, and junk the chipboard boxes. But as I removed the crossover, I decided to use it to quickly test the TMO25F1 tweeters... One thing led to another, and before the end of the lunchbreak, I'd built a speaker!
The "enclosure"
is made from a plastic storage bin, also rescued from the skip! The
thick cardboard is folded to fit inside, while being free to move in
and out - this way I could determine the optimum enclosure volume!
The inside was lined with the recycled acoustic foam, and the crossovers were modified with additional resistors to bring down the level of the tweeters. Despite everything, it sounded surprisingly good! I just had to build a second, and found that despite a few rattles at low frequencies, and a rather prominent mid-bass because of the resonating cabinet walls, they produced a convincing stereo image. And. most importantly, they amused everyone at work!
Speaker Design
My cardboard games persuaded me to think more seriously about doing something with these little drivers. The first stage is some calculations to see what sort of bass response we might be able to get. In the old days, I used to do this by hand. Any text book will talk you through the procedure - I have an old copy of Designing, Building and Testing Your Own Speaker System With Projects by David B Weems, which is a good introduction to speaker building.
These days, there are a large number of software packages that can speed up the process. I used to play with an old copy of Perfbox, which runs under DOS, and is still a useful tool. But for ease of use, you can't beat WinISD, which is freeware. It comes supplied with a huge driver database, and if your driver is already detailed in there, it's literally a 30 second job to model the behaviour of the AP100Z0 in sealed and ported boxes. It also allows you to experiment with parameters to see how the response changes. It's highly recommended.
Using the Audax-supplied TS parameters, here's a prediction of the bass response:
The yellow trace is the sealed enclosure, falling away at around 12dB per octave as you would expect from a second-order system. The -3dB point is 90Hz, and there is significant output at 50Hz. This is using a 4.4 litre enclosure...
The green trace is achieved with a 9 litre ported enclosure tuned to 52.5Hz. This is a big box for a 4" driver, but then the F3 of 47Hz is pretty impressive. The response falls away much faster, indeed note that below 30Hz the sealed enclosure has a higher output. But this is pretty academic because at these frequencies, the excursion of the drivers will exceed the Xmax specification at relatively low levels.
Measuring the drivers
The above results look promising, but we're assuming that the specifications supplied by Audax are accurate. Reading through the thread on diyAudio.com, I found that most people have found the specs to be wildly different to the supplied data. Modelling the response based on other peoples specs gives different performance and requires different enclosures. So before going further, I needed to make my own measurements.
There are a number of different ways to do this. A popular method is to use software packages that work in conjunction with a PC sound ard and home-made test jig - these can automate the measurement process, but rely on the performance of the sound card and many other variables. I prefer to use the "old fashioned" method, which requires a signal generator, oscilloscope and frequency counter. You also need a power amplifier and power resistor of known value.
The calculations are easy enough, but to make the process faster, you can download a spreadsheet from The Subwoofer DIY page - and while you're there, check the rest of the site for much more information about speaker design.
| Audax | #1 | #2 | Units | |
|---|---|---|---|---|
| Fs | 64.0 | 66.9 | 65.6 | Hz |
| Qms | 2.16 | 1.65 | 1.78 | |
| Qes | 0.63 | 0.67 | 0.74 | |
| Qts | 0.49 | 0.47 | 0.52 | |
| Vas | 4.72 | 3.46 | 3.56 | Litres |
These are the results from the two drivers that I used in the MDF prototypes below. As you can see, they compare reasonably well with the supplied values; Vas and Qms values are a bit off, but the remaining spec's are pretty close.
Running in
These measurements were taken with no "official" run-in, just a little light use with music. Many people report that the drivers require significant run-in time, so after making these initial tests, I wanted to see if there was any evidence of this.
As the system is basically a moving mass on a spring, the only thing that can change during "run-in" is the compliance of the suspension. The mass of the cone can't change, and I can't think of any electrical or magnetic variables - voicecoils burn out, not burn-in! The most direct measure of suspension compliance is the resonant frequency and conveniently, this is the easiest parameter to measure. Pretty much all other parameters are dependent on suspension compliance, so there's no real point in going through the all of the tests until the resonant frequency has stabilised.
| #1 | #2 | #3 | #4 | #5 | #6 | ||
|---|---|---|---|---|---|---|---|
| New | 66.9 | 65.6 | 84.0 | 78.3 | 77.9 | 80.4 | Hz |
| 1 Hour | 63.5 | 62.1 | 69.0 | 65.5 | 63.7 | 68.2 | Hz |
| 4 Hours | 63.2 | 61.6 | 68.6 | 64.7 | 64.6 | 67.3 | Hz |
| 9 Hours | 63.1 | 61.5 | 68.1 | 64.4 | 64.1 | 66.9 | Hz |
| 20 Hours | 63.0 | 58.4 | 64.6 | 60.0 | 60.2 | 59.7 | Hz |
| 30 Hours | 62.9 | 61.8 | 64.4 | 58.9 | 59.6 | 58.0 | Hz |
The drivers were run in using a 15Hz sine wave driving them to a peak-peak excursion of 4 to 5mm - close to the stated Xmax, but nowhere near the mechanical limits. They were left for around half an hour before measuring Fs, eliminating any thermally induced errors. As stated above, drivers #1 and #2 had already been used, both in the initial tests, and during the week with my MDF prototypes (see below). Also, drivers #3 and #4 had briefly been used with the cardboard prototypes at the top of the page. Drivers #5 and #6 were genuinely brand-new at the start of these tests.
It's also worth remembering that we are at the mercy of the tolerances of a mechanical system. The compliance of these drivers seem to be affected by temperature, humidity and just about everything else! The running in happened during the day, and the Fs measurements were made at the end of each day. Before commencing the run-in procedure the following morning, I quickly checked Fs of all the drivers. This table shows how much Fs rises the morning after each run-in period:
| #1 | #2 | #3 | #4 | #5 | #6 | Average rise | |
|---|---|---|---|---|---|---|---|
| After 4 hours | 5.8 | 4 | 5.2 | 2.9 | 2 | 3.7 | 3.9 Hz |
| After 9 hours | 3.5 | 4.7 | 5.6 | 3.9 | 4.5 | 2.1 | 4.1 Hz |
| After 20 hour | 3.6 | 4.3 | 4.3 | 3.1 | 2.4 | 3.9 | 3.6 Hz |
| After 30 hours | 1.3 | 0.7 | 2.0 | 1.4 | 1.4 | 0.5 | 1.2 Hz |
The final results show less rise than before, but the workshop was noticeably warmer that morning, indicating that temperature is definitely a factor. Also, it's worth saying that you can easily get Fs back down to its previous value by running the driver at 15Hz for about 30 seconds.
This is something that I've noticed before - the Rogers dB101's that I use in the workshop sound noticeably "thin" first thing in the morning, especially in the winter months. And as you can read on that page, I found that dB101's require a period of running in, which is not something I'd noticed to that extent with other speakers - it appears that these drivers are particularly sensitive.
Here are the full results after run-in. As you can see, they are close to the Audax specifications - Qms is rather different, bringing down Qts. Fs and Vas is broadly lower. Driver #3 stands out as being slightly away from the average, and if you discount it, the average Vas is 4.46 litres, which gets it a bit closer to the Audax specification...
| Audax | #1 | #2 | #3 | #4 | #5 | #6 | Average | |
|---|---|---|---|---|---|---|---|---|
| Fs | 64 | 62.9 | 61.8 | 64.4 | 58.9 | 59.6 | 58.0 | 60.9 Hz |
| Qms | 2.16 | 1.64 | 1.65 | 1.54 | 1.63 | 1.61 | 1.55 | 1.60 |
| Qes | 0.63 | 0.67 | 0.66 | 0.62 | 0.58 | 0.62 | 0.60 | 0.63 |
| Qts | 0.49 | 0.48 | 0.47 | 0.44 | 0.42 | 0.45 | 0.43 | 0.45 |
| Vas | 4.72 | 4.17 | 4.20 | 3.85 | 4.66 | 4.49 | 4.78 | 4.36 Litres |
How does these results affect the enclosure design? Using the average value of the drivers, I redid the plot to see how the response and box size changes:
Results look much the same at a glance. But happily, the box volumes required are quite different. Taking the sealed case first, the -3dB point is up from 90Hz to 95.8Hz, but the enclosure volume is down from 4.4 litres to 3 litres.
For the ported version, the -3dB point is up from 47Hz to 50.1Hz, and the enclosure volume is down from 9.9 litres to 6.3 litres. The tuning frequency is 54.5Hz. Unfortunately, I can't work out how to build a suitable port that will actually fit in the box. For example, a 2 inch diameter port would need to be 27cm long - probably longer than the enclosure would be tall! Reducing the port diameter to 1 inch brings the length down to 6cm, which is much more practical but the air speed in the port becomes too high, and you would hear "chuffing" noises at certain bass frequencies - not good! A 9cm 30mm port is more practical.
Although I'm generally a big fan of clean sealed bass, the extra bass of the ported version is appealing. But the 6.3 litre enclosure is not, so can you reduce the size of the box while not loosing bass output?
The green plot is as above, but the red shows the result of a 3.9 litre box tuned to 50Hz. The -3dB point is 66.2Hz, which is pretty respectable. I fixed the box volume and experimented with the tuning frequency to get the best response. WinISD has a number of options for this, including a click-and-drag method with live updating.
Again, the problem comes when you try to design the port. For a 2 inch port, the length is 58cm. Reducing the diameter to an inch requires a 13cm port length, which might be practical but again airspeed is a problem. Increasing the diameter to 30mm again reduces the airspeed to a safer value, but the length is 19cm!
Trying a slot doesn't seem to help either. If you fix the slot at the internal width of the box (say, 10cm on the assumption that we're trying to make the box as small as possible), then a slot height of 1cm requires a length of 26cm. And this will increase the size of the box because of the volume occupied by the slot. This also applies to the normal port, but at least the wall thickness of a pipe is much less. A 6mm high slot is the smallest you can get away with before air speed becomes a problem again - this is 14cm long port. One major problem with using a slot occurs when you try to tune the box. Invariably, once constructed you'll find the port frequency will be slightly away from the intended frequency and you'll need to adjust the length of the port. This is clearly a bit tricky with a slot that is part of the enclosure!
So, a ported enclosure for the mini-monitors is looking somewhat impractical, but it might be an option for a larger centre speaker. I'd want to prototype it and audition it carefully, however, because small ported enclosures with a high tuning frequency and a high Qts driver can suffer from the one-note boomy bass problems that you find with a lot of budget commercial speakers - they might sound impressive initially, but quickly become tiring.
MDF Prototypes
To measure VAS, you need an enclosure of known volume, and it has to be a rigid, sealed box. I didn't have anything suitable to hand, so I decided to knock something up using scraps of MDF. Whenever I make anything from MDF, it's a disaster unless I've had the wood cut for me on a table saw at work. But, I decided to try using a straightedge to guide the sole-plate of the jigsaw, and was amazed at the results! Obviously, straightedges are used all the time with circular saws and routers, but I'd never heard of anyone using this method with a jigsaw before. Probably because a jigsaw is really the wrong tool for this job! I used a power-plane to "sweeten" the edges slightly, and the final result was surprisingly good.
For the sides, top and bottom, I used 18mm MDF. Clearly this is much thicker than necessary for such a small box, but it was a case of using whatever I had to hand in the workshop. For the front and back, I used 6mm MDF. This is possibly a bit too thin, but I wanted to screw the panels to the box and trim them flush using the router - unfortunately, my flush-cutting bit can only deal with 13mm stock.
The enclosure had a volume of 3.28 litres, which is a reasonable size for a sealed enclosure using this driver. I had a quick listen to the woofer in the box, and was quite impressed with the quality and extension of the bass. So I couldn't resist building another box!
I'd left the tweeters at
work, so I had to wait for Monday morning to complete the
"system". The crossover is mounted externally as it needs
some optimisation (remember, it's designed for a completely
different system). Two 10 ohm resistors were added to bring the
tweeter level down.
And even at this early stage, these sound very good. The bass is tight and clean, and reasonably well extended given their small size. Remember, these are 3.3 litres - an LS3/5A is 5.5 litres. They measure 22.5 by 15.5 by 16 centimeters, but could be made even smaller if WinISD and my driver measurements are to be believed.
Crossover Details
While not strictly relevant to this project, I've included details of the Musical Fidelity crossovers which might be of interest to anyone who still owns a pair. They are assembled on 6mm MDF, and held to the enclosure using M4 machine bolts and inserts, as you can see here.
The circuit is assembled using point-to-point wiring and copious amounts of hot-melt glue. Despite the cheap components and poor assembly, there is some evidence of good design practice here - for example, all of the electrolytic capacitors are bypassed by 1uF non-electrolytic capacitors.
As you can see from the schematic, the woofer has a second-order low-pass section and the tweeter is third-order. I don't know what the crossover frequency would have been in the system, but measured with an 8 ohm resistive load the bass section gives a -3dB point of 1.2KHz and the treble filter has a -3db point of around 4.7KHz. The original drive units would have had a different impedance compared to the 8 ohm resistor, so the crossover points would have been more similar in reality. This is probably especially the case for the woofer.
While I'm not planning to use these crossover modules for this project, they might be useful to have around when experimenting with speaker design.