Tweeter: Scan Speak D2905/990000
After the disillusion during the dreaming stage, it was a rather painful experience (in terms of time spent!) waking up to reality. All the past 2 months effort seems to have gone down the drain. It was disheartening and the general reaction was 'the hell with it, just go out and pay for a commercial unit'. Fortunately, deep in me is a will that refuses to give me. In fact the more a situation beats me down, the more drive I get to see things through. Guess I've a sadistic liking for self-inflicted pain!
One thing an engineer does well is in systematic redesign. Just think through the processes and align them in a systematic way in order to approach the goal. Simple.... so the 1st thing I needed to do was find out exactly how those simulation software tick. What are the inputs and outputs... after which it is a matter of coming out with a step by step procedure to execute, to cut the story short, the steps I came out with is;
Theorectical
1. Drivers selection
2. Software simulation
i - SPLTrace to obtain the FRD and ZMA files from the manufacturer charts.#1
ii - Unibox simulation - export FRD, ZMA files and note the T/S values such as Vb and etc.
iii - Baffle diffraction simulation.#2
iv - Save the individual response profile from the above (i.e. FRD files).
v - Use FRC to combine all the FRD files and remember to regenerate the phase information.#3
3. Speaker Workshop simulation.#4
i - Import all the combined FRD and ZMA files.
ii - Build the crossover network and begin the optimizing process.
Actual
1. Get your parts as determined in the Theorectical process.
2. Build the xover network.
3. Build the enclosure.
4. Perform critical listening tests.
5. Alternatively, perform measuring via SPL meter in a controlled environment.
6. Fine tune the enclosure or redesign the network via the Theorectical process and back to step 2.
Unfortunately, being the real world, nothing is finite until the actual test. Thus whatever the theorectical values may throw out, the final judgement is still the actual listening test. The complexity of room acoustics and it's influences can never be simulated (using current technology). As such, there would be some amount of reinteration of the above steps until we approach an acceptable level.
Putting the $ where my mouth is, here are the actual results from the above process.
This is probably the easiest part since I've already some ideas on what I wanted. The SF Cremona being the inspiration, the ScanSpeak drivers are the one to beat. I took a look at some other high-end drivers such as Accuton, SEAS and Peerless. In the end, based on the desired response curves and price confirms my choice to be the right one (for me at least!). Here are the drivers:
Tweeter: Scan Speak D2905/990000
Mid-Bass: ScanSpeak 18W/8531G00
This is one marvelous software! You can input the easily (for SS drivers) available T/S parameters and then just play around with the enclosure parameters to obtain your desired frequency response. Note that physic rules here, you can only go so far with any particular driver. Thus, if you are unhappy with the result, you may want to change your mind and go back to driver reselection.
In my case, things turn out pretty well.. Thanks in great parts to the lesson learned in my dreaming stage! The final design proves to be a vented box though I set out to build a closed box. The final charts and results are as show below;
Next up is the ....
Baffle Diffraction Simulation
This is probably one of the things in life we can do without. Alas, baffle diffraction is something of a physics phenonmena which is inevitable#5. After the rather exciting Unibox simulation above, it's time to pay the price of mother nature.
At this point, the important decision on how to layout the drivers on the baffle need to be determined. After much research, I do not think the tweeter/midbass offset will give that significant or audible difference. Considering the fact that there are countless commercial designs without the offset and yet sounded extremely good. As a guide, I adopted the Cremona symetric layout of tweeter, midbass, midbass.
Drv2 - tweeter; Drv3&4 - midbass
as can be seen from the chart, the BDiff effect will rob you of -6dB at the lower register!! Take heart though, in a real room, the back wall reinforcement will restore back about 2-3dB. Thus, for my design I am assuming in advance this 'benefit' and thus, scaled down the effect by a conservative 2dB. This is using the next FRD manipulating tool - the FRC.
Frequency Response Combiner
I will not elaborate in great detail on how to use this rather complex excel xls#6. In short, this is where all the FRD and ZMA files get combined into a single FRD and ZMA complete with phase information. Generic steps are;
1. Import all the FRD files from Unibox, BDiff and SPLTrace.
2. Normalize - I use a high amount of point (>1000) for better accuracy.
3. Combine the response.
4. Generate the phase information.
5. Export result to a FRD file.
6. Repeat steps 1-5 for each driver.
For the ZMA file;
1. To assign individual ZMA to the 2 parallel drivers, I resimulate the Unibox to obtain ZMA for a single driver.#7
2. I then splice the Unibox file with the SPLTrace file.
3. Normalize - use the same data point and freq range as its driver FRD files.
Speaker Workshop
Again, the following is not meant to describe how to use this extremely useful software. I am not making the most of it due to consideration that I will not be able to create the right environment for proper speaker measurement. Probably when I get more ambitious I'll get to it.
For now, just theorectical simulation will do. Continuing from above, the FRD and ZMA are imported in. A sample FRD and ZMA will give the following graph.
I proceed with the creation of all the driver resources and begin the network xover design. Initially I tried to standard text book design of HP/LP filter but it did not yield the result that I was looking for. In the end, I just do trial and error. The end network is;
and the corresponding Freq response and Imp response:
Everything looks fantastic on paper - now to go out and prove it is so...
Next is to get the parts ready and do a mock-up.
#1 This is assuming that we are taking using the manufacturer's values and not perform our own measurements.
#2 By this time you should already have an idea on how you would layout the drivers on the baffle.
#3 Also generate the phase information for the ZMA files.
#4 Download and read the SW manual on how to use this software.
#5 Surprising thing is that I cannot really detect this from the open baffle mock-up. I have my doubts as to what great extent this play in real life.
#6 Somehow the 2PI setting for the ideal BDiff does not work for my system. As an alternative, I created a FRD file with constant 6dB.
#7 I am not sure this is the right way about doing it, but this is what makes sense to me. It may not be 100% accurate, but something is better than nothing!
#8