0 Members and 1 Guest are viewing this topic. Read 133652 times.
Thanks, fellas. If I hear you right, it depends on the execution but there is a potential for this. Supposed to get a listen to Duke's six piece system pretty soon now. And from a cat who sold his big Sound Labs. I'm thinkin Duke's system has wider dynamic range than the SL's. It's the what else I hope to experience, also. Get to bring some of my own records. Trust me, I'm excited.
Thanks for your reply, Dan, and invitation. I'm tryin to get my wife to let me venture solo (I gotta a free ticket but only one) to SF, my old stompin grounds. That happens, I will. Gotta get down to Big Sur, too. I'll be hearing the AK system right down the road from me, actually. Duke knows that cat. Dude has a top of the line SOTA table. Think I'll send him a reminder. And, sure, I'll report back. Cheers
I have a little bit of experience, Dan. I built an OB speaker using ToneTubby 12" drivers and the Beyma CP380M on a narrow-coverage conical horn.And as you say, they're quite a bit of work and devolves into a balancing act. Also, too often I think DIYers pay too much homage to simplicity -- this stuff is complex. Here is my crossover. For me, measurments got me very close then the last bit of tweaking was done by ear -- say, comparing 3.7, 4.0, and 4.3 ohm series resistor in the LPad/ CD compensation. Also, I spent much more time worrying about the power response than any one or several curves -- I use SpeakerWorkshop to simulate, and I'd export all the freq response curves, and then average them using Excel to get an approximation of the shape of the power response. In your curves one thing I'd think needs addressing is the widening in response at ~2.2khz -- you definitely have a S-shaped bump in power response there.Also, what kind of amp are you using with these? A high-dampening SS amp will lead to a freq. response very closely approximated by your curves below. A 2- to 3-ohm output impedance no-feedback DHT amp will tend to not match your simulated response depending on the impedance of the speaker, unless you account for that. Any idea what the impedance curve for your speaker project looks like? Could you post more data then just the final curves?JohnBTW, here's what the reponse for my speakers look like (0- to 35-deg off-axis in 5-deg increments):
About doing xo/eq for constant directivity speakers....I am thinking out loud here. However, what I've thought might work, but haven't tried is doing the -30º to +30º measurements in 7.5º steps as Geddes does. Then take a weighted average of the resulting curves--weights depend on power response, i.e the 0º curve gets the least weight. This average curve then becomes your raw response curve that you use to design your crossover around. Therefore, you are not correcting one curve but the amalgamation of them all.
Hey Dan,Unfortunately, I was unable to get decent measurements further off-axis. Not such a big deal for me, as the compression driver is on a nominally 60deg conical horns.I did something similar to what Josh suggested. I exported a text file for each simulated freq. response -- eight in all, 0- to 35-deg off-axis -- imported them into Excel and did a simple average, without any weighting. My rationale on the weighting is that each one is equally important.I suspect that that is NOT how Earl does it; he's smart enough that he probably has a script in his software that calculates the actual power response from the individual curves. Simply averaging the curves would only give you an idea of how the power changes as you move along the x-axis (freq. response), which I think is good enough at this point in the design process. I mean, by now, you would have fixed big-ticket issues, like relative levels between mid and treble and you're looking to develop refinement in the design. At that point, I think the focus is less on what the power response is specifically and more on how it changes relatively; i.e., no sudden changes or 'bumps'. So, for me, the simple averaging works here.With SS amps, the resulting impedance is mostly a non-issue. It's important for me as I'm pretty tied to using amps with 2-3ohms of output impedance AND would like to option of feeling comfortable with the resulting freq. response if someone where to ever bring over a very low output impedance amp.I use SpeakerWorkshop for simulating crossovers. I think it'll do a couple axis simultaneously, but it won't do more than that. I ended up creating eight drivers each for both TT12 and CP380m -- one for every axial response -- and coming up with eight networks (obviously, all identical), simulating all eight networks, and then combining all the data onto one graph. It's a pain -- as you make changes, you have to make the changes in eight places and simulate eight different networks -- but it works.I use the ARTA suite for measurment, both freq. response and impedance. I built a couple of simplified jigs instead of just one a la Wallin or the one described on the ARTA website.Definitely get them into a cabinet. If your posted measurments are sans a box, you'll want to re-run them once the drivers are boxed up so as to capture any diffraction effects.I hope that helps,John
I don't know exactly what Geddes does, but I suspect it is something like what I suggested. He has eluded to some of what he does that is why I know he weights the on axis less. The weights, the way I think of it, are pretty simple, but not a simple average. Let's assume you make 8 measurements from 0º to 35º. Think of it as slices of a pie where your measurement axis is down the middle of the slice, not the edge. That is, take for example the 5º measurement (0 to 35 in 8 measurements is 5º increments). The 5º measurement is the center of the slice of pie from 2.5º to 7.5º (easy to see if you draw it). The 0º slice goes from -2.5º to 2.5º. The 5º slice has a mirrored slice at -5º, 10º has a mirror at -10º, etc. The only slice that doesn't have a mirror image is the 0º slice. Let's assume all the mirror image responses are the same as their counterparts (reasonable thing to assume since the speaker is symmetric. So you don't have to measure -30º to -5º. Now when you average, you need to give 0º half the weight of the other lines if you are basing on the power response.
