Frequency/Phase Measurement in Praxis

[Amphiprion]

Hey guys,

Finally on on topic post! I'm working up that little L15/27TBFCG I started a while back. I've taken some Praxis measurements and imported them into LspCAD and done a crossover design. Both slopes are hitting LR4 targets quite accurately, but my phase and reverse nulls are terrible. I've got essentially the same xo for the woofer as Zaph did for his L15/27TDFC just with less baffle step compensation (smaller series L) and a slightly different tweeter crossover topology but similar freq response. Problem is, I have to set LspCAD to have a huge -28mm offset for one of the drivers (the tweeter I think, I can't remember, but that makes sense from an acoustic center POV). I'm assuming I'm screwing something up in Praxis with windowing and that propogation time setting thus affecting the phase measurement. It shouldn't be this far off from my past experience. I used to have a guy that was very experienced with Laud do all my measurements after which I worked up the xo on my own, but this is my first time doing everything from build-measure-xo design and whatnot.

Can anyone give me some tips on how they do it in Praxis?

Also, a secondary question - everyone seems to do measurements on tweeter axis. For some reason I think it makes sense to do it on an axis perpendicular to the line and midway between the vertical locations of the drivers' acoustic centers, so that the delay times between the drivers are more closely matched at any distance from the speaker. Or in simpler terms, halfway between the center of the woofer and the tweeter. It probably has minor effect on the overall phase relationship especially as you get farther away from the speaker, but I was wondering what peoples thoughts on these matters are. I own measuring loudspeakers by D'Appolito, but I've also been on a book binge with Art of Electronics, that Bob Pease book on Troubleshooting Analog Ckts, The Master Handbook of Acoustics, and the Walter Jung book on opamps, and with my little time I haven't really had a lot of chances to get into anything other than AoE of MHoA.

So, any help or comments are appreciated. If this design turns out well I'll amke it publicly available. If not, it will go into my blooper reel

[TacoD]

I do a measurement of all drivers paralleld and with the mic on tweeter axis. This gets the summed FR of the drivers without a x-over. Then in lspCAD I set the measured result of that as background in the plotting form. I also make measurements of every driver on axis. These measurements I asign to the corresponding drivers in lspCAD.

Now starts the tricky part, fill in the x and y offset (you can measure these ), and alter the z-axis as long as that the measured curve and the simulated curve match as good as possible. Important is to disconnect any filter in your simulator.

Doing this compesates for the offset differences of the drivers on your baffle. In Praxis windowing will cost you resolution, but as far as I know the phase is quite accurate. You have to fill in the time of flight, but if you use the scripts in Praxis these steps are explained.

Oh and do post the result Ohters can learn of that.

[Mark K]

Hi Mark,

Taco is right.

You should not need to put in 28mm of offset. This is way too much. Remember, you're measuring the exact measured phase. The offset is built into the measurement. Usually, the AC offset is close to zero, with only a small offset, really to compensate for small errors.

If you use the measured phase, that has the AC offset built into it.
If you use minimum phase, then you'd have to add more realistic "AC offset"

Hope that makes sense.

[Amphiprion]

I'm kind of lost with Mark's post, but will try to do some more tweeter measurements this week some evening and post back with the results. Thanks!

[Dennis H]

There are two ways you can go. Either measure all drivers from the same point or measure each driver on its own axis.

I think JonMarsh does each on its own axis and subtracts the time of flight (distance from the baffle to the mic) from the measurement. The advantage of doing it that way, for the simulation in LspCAD, is you can then position the drivers anywhere on the baffle (Z=0 for a flat baffle, X and Y reflect their actual positions relative to the imaginary mic) and place the imaginary mic anywhere (like 4m away at ear height.)

If everything is measured from the same point, you set X,Y,Z to 0,0,0 for all the drivers and the sim is only valid at that one imaginary mic point unless you fudge the Z distance -- and that's a royal pain.

[Mark K]

I'm kind of lost with Mark's post, but will try to do some more tweeter measurements this week some evening and post back with the results. Thanks!

Oops, sorry about that.

There are a couple ways of doing this. The easiest is the method above.

Think of it this way.

You measure both drivers at a single point, hypothetically 1m, halfway between woofer and tweeter.

Now STOP and think about what you just did. Forget AC, baffle location, etc.
What you've measured in praxis is the actual measured FR and phase for those two drivers at that point. You can now use lspCAD to use that data to design an xover that will give the correct response exactly at that point. All information about baffle and AC are built into the actual, real world measurements.

This method is the most straightforward and least prone to error. But, it requires you to choose an appropriate axis. Still, if you do this at two meters and on your listening axis, this works very well.

The other method is to measure each driver at 1m on its own axis and reference the phase to the baffle. See JohnK's SE tutorial. It allows for a lot of modeling flexibility, but a bit more error can creep in. It's just very hard to do multiple on axis 1m measurements on different drivers and be exact. However, similutations can then include off axis behavior, etc.