Site Links
Howdy, Stranger!
It looks like you're new here. If you want to get involved, click one of these buttons!
Quick Links
Categories
In this Discussion
Tell us a story
I would like to know the various and or the most common methods used to complete a speaker build. I built speakers back in the day and used a calculator to come up with crossover component value. I then took a long break for 20+ years. Now I have an Omnimic and a DATs but I still consider myself a new guy. I have only played around a bit with Xsim. I do my crossovers with my Omnimic at 30 inches or so from the finished speaker on tweeter axis. Not sure if this applies to 3 ways. I have a prototyping board that I made that has a polarity reversing switch before the tweeter. I then start with a 1st order determine that it sounds like crap, go to a second order, look for a reverse null. Listen for hearing fatigue then try 3rd or forth or a combination. I do not know how to voice a speaker. I assume it means talking nice to them. I do not know how to do quick A B comparisons on two crossovers. There has to be a better way. Please share.
Sincerely,
New Guy
Comments
I could tell you how I do it. But I am sure it is wrong and causes me all sorts of headache. It has been a struggle to just get the sims to even remotely resemble real life measurements.
I started with pcd and moved to winpcd. Not ready for xsim yet, though I have used it to make schematics.
So far I have found that accurate initial measurements are everything. The tighter controll you have of those, the closer the sim will be to real life. Make sure the mic is on the same plane as the tweeter. As in measure distance from the floor to both and make that measurement the same (hopefully the floor is level). Be sure the speaker is plumb as discrepancy will throw off the relative acoustic center estimation. I have heard to keep the mic distance 2-3x farther than the baffle width so that as much of the baffle interactions can be captured in the measurement. I then use the combined measurement method in winpcd to estimate the relative acoustic centers which affects the phase timing and can really make or break your sim.
I don't worry too much about reverse null. I only check it at the end just in case. If your drivers are summing close to 6db at the crossing point then you are pretty much there.
I started off just doing quick flying lead solder job to test crossovers but now I use clip leads. That is still a bit of an issue for me due to how much space large crossovers take up.. but I'm dealing with it as best I can.
Voicing is more of an experience driven thing. Comparing what sound you like best to the measured response. Do you tend to like the "bbc dip"? A slightly dipped response centered around 3-4khz. Or maybe a tilted response that drops around 3db from the bass down to the top octave? Whether those responses are possible depends on the drivers, baffle, and your skills massaging the crossover to get that shape. After that is just listen to the result to see if you like it. Another wrench to throw in is the off axis response can affect it too. But Im not quite ready for that as I have yet to build a speaker turntable for accurate off-axis measurements. Just working based on best guess atm.
Skip XSim, and go straight to VituixCAD. Consider it a spiritual successor, but with regular updates and way more modern features. All of those steps like baffle simulation and data preprocessing for which Jeff wrote his Excel suite are baked into Vituix. It's not easy to do, but still way easier than SoundEasy. LOL
Best thing I ever did was ditch USB mics and move to a 2ch audio interface with a normal XLR mic. Details of why I've tried to capture in the single vs dual ch article linked below. For getting started, a special jig is not required, just a patch cable from output to input on the other channel can get you measuring with accurate timing today.
Best resource for getting started:
https://kimmosaunisto.net/Software/VituixCAD/VituixCAD_help_20.html#How_to_start_with_VituixCAD
Some supplementary documentation I've written, specifically with VituixCAD in mind.
https://www.htguide.com/forum/articles/do-it-yourself-diy/927364-single-vs-dual-channel-measurements-or-why-friends-don’t-let-friend-use-usb-mics
https://www.htguide.com/forum/articles/do-it-yourself-diy/927384-dual-channel-measurement-jig
https://www.htguide.com/forum/articles/do-it-yourself-diy/927407-how-to-measuring-impedance-and-t-s-with-arta-limp
https://www.htguide.com/forum/articles/do-it-yourself-diy/927389-how-to-filter-simulation-with-vituixcad-and-eq-apo
Other excellent reference information:
https://drive.google.com/drive/folders/1SdiWAVgyRHFoXtQBL1M4eKlcezk6h4gJ?usp=sharing
There's some good bedtime reading stories there I'm sure, night night.
Agree on the avoiding the usb mic which I used for a while and just recently switched to emm6 and motu2 interface (which actually is nice sounding dac-outputstage as well). Those measurements from new setup seem to be working much better in the simulations. Also been using the merger in vituixcad recently as learned the hard way guess-timating the woofer mid has one meandering all over the place, all the domino effects.
Come and listen to my story bout a man named Jed. Poor mountaineer barely kept his family fed...
Come on Craig, don't be all top secret. Spill it. You'll feel better
Certainly not easier but definitely better
I do see one immediate problem however. How do I get this new measurement microphone approved by purchasing?
New thread. Tell your wife a story, and it works.
I will be doing a study this weekend and/or next comparing 3 measurement mics and 3 pre-amp interfaces and will post it at MAC.
Mics are $35, $70, $600.
Preamps are $50, $170, $200.
My hypothesis - for crossover design, with a good calibration file, the two low-cost ones are just fine. We'll see.
I would argue the opposite.
Not necessarily better. (I think it likely is better, but there are people smarter than me who disagree.) I'm quite confident that Jeff Bagby could design a speaker with a USB mic and his programs as good as anyone else could do with any other equipment. But definitely better for a noob like me.
But, yes, easier - once you know how to do it. Because with the USB mic you have to do a LOT of extra steps, each of which introduces the chance for error/bad inputs. Garbage in, garbage out. With the actual phase and all of the merger tools built into VituixCAD, it really is "close" to automatic as it probably can get. Extracting minimum phase, converting the port output correctly, merge LF and HF - pretty much a few clicks in VituixCAD. And, as a bonus, you get all of the directivity-related info so you can focus on more than on-axis SPL.
OK Beans spilling (all over the floor)...
I do one channel OmniMic measurements a lot as well as some two channel measurements with Arta. They are both useful and both have their place. I don't think many folks would say that Jeff Bagby's designs sounded like junk. In fact his designs are generally regarded as spot on and I agree. As far as I know he wasn't a two channel designer. Don't think Vituix was even around before his passing. That said I have heard wonderful DIY speakers designed and voiced with one channel measurements and some that were just eh! I have heard the same from two channel measured designs. No clear winner there IMHO.
I think VituixCad "can and does" save some designers time in the voicing/tweeking/measuring process if they don't have great ears or just don't have the time or enjoyment of doing that process. But at the end of the day it's all about what sounds great to you and how you get there based on what process you enjoy.
Probably pretty soon AI will be designing speakers way better than any of us can
I guess when you edit a post it gets kicked to the bottom of the thread...
No, it doesn't from my experience - did you actually edit and repost the edit or did you forget like I do a lot . . .
I use Omnimic (original clear cased V1) and am not ashamed to say so. As to Jeff, he even voiced with a single speaker most of the time. I have to have a pair, or my tweeters end up hot.
I STILL use the old S&L WT2, as I prefer it to the problem plagued WT3 and original DATS1. I know Jon Marsh still uses his too.
I use whatever software is required, with most of my previous time spent with PCD, Xsim, and RM for modelling. I have Vituix, and am trying to learn it.
InDIYana Event Website
Wolfy's Photobucket
You have good ears and xo design intuition...
Microphones are for sissies, just pick out a few parts that look good, wire them up and bingo, you have the greatest speaker ever constructed by a human being.
Thank you, Craig! Likely the nicest thing you've written or said to me. I will say, it has been a long fun road to learn and get to where i am in this hobby. Help from others, shared knowledge has been a blessing that only this hobby entails so deeply. I try to openly and accurately share speaker info as it has been shared with me.
Voicing by ear is a skill in itself, and xover design by ear is even harder, but can be done. Sometimes, Nick's "throw nice looking parts at it" can work rather nicely, though occurrences are fewer.
In the last couple years getting more acquainted with some newer builders, I have found there are questionable spectral tilt applications thereof that I've not really questioned as starkly myself as they have as to what is more accurate. For example, BSC for me is just a "flatten the curve" to yield proper 6dB, and these fellows stress more about that linkage than I ever put stock in the decision. If I felt it was lean, increase the coil. Too full, decrease coil. This of course with proper box tuning in tow.
I guess I have learned when something should be more important to the better endgame to me, and I just don't stress some things other builders do because I expect it to fall into place. It's a wonderful hobby, both perplexing and relaxing, and the work is almost always worth the results. I've heard a lot of good, great, bad, and awful and seen ugly, interesting, pretty, and exceptional workpieces; and have found what I like and don't. That is a large part of the fun!
InDIYana Event Website
Wolfy's Photobucket
I've been an OmniMic user, and have been getting good results since I followed the instructions on how to get the relative phase offsets. That said, I've got a Focusrite sitting here, and I really want to explore the other capabilities of REW. The REW community is sharp, and always coming up with crazy new ways to use the tool. I used it years ago to equalize a few sub setups (with the old Behringer Feedback Destroyer!), but I'd like to have better command over what it can do now.
It should make a decent general purpose 0-40khz spectrum analyzer for the bench, too.
Thanks guys, I love it. Since I am not much of a sit down listener, I tend to believe that power response is very important and I need to spend much more time in that area. I have previously placed a speaker on a stand on top of a lazy susan and sat behind it with my computer and laptop, tweaking components and switching tweeter polarity while rotating the speaker with my foot. I did not compile any off axis data, but I did observe that odd order networks seem to have a larger main lobe.
Tell us a story about your main lobe
That's getting personal
Like others have said with Omni mic and DATs you can design a great sounding speaker. Pick you favorite sim program. I always recommend PCD to start with as you will learn filter types.
I think the key is learning how to make repeatable measurements. Does your raw speaker measurements plus sim match the completed speaker? If so then you are on your way to designing a great speaker.
Listen and tweak with Omni mic to teach you ears what frequencies you are hearing. My $.02
It will be a bit time consuming, but I think you should take a complete set set of off axis measurements, both horizontal and vertical. The vertical can be limited to +/- 16 or 20 degrees. Then do your listening with the lazy susan set up, while looking at the measurements and you'll see a correlation between them, and establish how important the lobing and power response at various angles is to you.
"I built a speaker with USB mic and it doesn't suck" is missing the point of the 2ch method (I did too). It's about enabling the most accurate information with as few steps and potential for error as possible. Who doesn't want that? Whether you can design a great speaker with the information available is beside the point, and dependent on a lot of other factors than the method used to generate the design data. Anyway, I'm done preaching about it, life's too short, so just enjoy the hobby in any way you decide is best.
So trying to make progress on this. First step is creating a calibration file for the Behringer ECM8000 microphone. It is $35 but does not have a calibration file. Going back to my hypothesis, I expected the Behringer to be pretty weak but with a good calibration file it should be just fine for speaker measurements for crossover design. Off course, that requires someone to create a calibration file, but with all the DIY help that is avaialble I'd think that would not be an obstacle.
But I'm a little shocked. Red line below is $600 Earthworks M23R. Blue line is the raw (uncalibrated) Behringer (SPL adjsuted). These are just the raw measurements I was going to use to create a calibration file. I was so surprised that I thought I'd post this result here. More to come in a separate thread when I'm done.
(Hmmm...what else could I have done with that $600?)
Nearfield
I think the Earthquake can take much louder measurements with lower distortion, however.
Yes, and that will be part of my study. It is a more capable microphone, but for just crossover design my hypothesis is that we don't need a lot of that capability. Also, I did buy it to be able to take better distortion measurements.
Sorry, but I think that was $600 well spent. There's a fairly large deviation between 5 & 10k. You can hear a deviation of less than .5db over that wide a frequency range.
These sweeps don't have calibration file applied, so which one is actually more accurate?