Working on crossover, wondered about something...

[BAM]

I've graduated to Audua's excellent Speaker Workshop, for speaker modeling. I have used the FRD Consortium's SPL Trace software to import the frequency response of my drivers that I am using in my developmentof my BHF (Buszka High Fidelity) MTM-5, which is the next generation redesign of my Project MTM, with the same drivers but a new enclosure and computer-optimized crossover. I have been seeing something strange in my use of Speaker Workshop, and I want to know if this is something I should be worried about.



Here, you see the impedance (red line) peaks at about 20 ohms on this plot.



Here's the plot of the predicted response of my new crossover which is a second-order Bessel, with 5dB of padding on the tweeter. You'll notice that near 50 Hz, the frequency response dips unacceptably leaving the speaker with an electrical F3 of about 100 Hz. Does this have anything to do with the huge impedance peak that I see in the woofer's impedance plot? And what can I do about it?

Also, I am willing to entertain the idea that perhaps I entered the impedance incorrectly. Here is the Clio-measured impedance plot supplied by Parts Express.



Any help you can provide is appreciated very much. In addition, if there's anything anyone can tell me about how to get rid of that peak at 3100 Hz, I'd be glad.

[sreten]

I don't know what has caused such a deep dip in your modelling
but its a dip of far higher Q than the impedance peak of the bass
driver due to driver Q.

If anything the dip appears to be perfect phase cancellation at
that point, how your modelling has arrived at that I'm perplexed.

sreten.

[5th element]

Peak at 3100 notch filter or if the xover freq is near this point clever fiddleing with the crossover.

[BAM]

Crossover is at 3000 Hz. I want to put a notch filter on just the woofers at that point, though. Where can I learn how to calculate one of those?

Could I have read in the impedance plot incorrectly? I will have to try using the raw Clio data there instead of the graph.

[5th element]

The peak is in the tweater response ooooooh. I thought it was in the woofs. Seems strange to have a peak like that in a tweeters response. What is this tweeter and does it contain the peak in the natural response, ie no xover?

Notch filter (this is a cap and a resistor and an inductor in series and then in parallel with the driver).

C = 0.1592/(Re*Qes*Fs)

L =0.1592(Qes*Re)/fs

R = Re+ Qes*Re/Qms

Where fs is the desired frequency. These are design formulas to get rid of a resonance peak in impedance when a tweeter doesnt have ferrofluid. They should give you good starting figures for a notch.

Dont worry about using a notch on the tweeters if its necessary. You can use cheapo iron core inductors and non polarised eletrolytic caps.

[okk]

If the dip or peak appeared at cross frequency, always by the 180 degrees phase between low and high part. You can simulate it with alternative orders or Q of the crossover.

[MarkMcK]

I posted a spreadsheet formula for determining notch filter values in the following thread:

http://www.diyaudio.com/forums/showt...023#post251023

Since, however, the cone resonant peak is only 1/3 to 1/2 octave spread at the base, a passive, speaker level notch filter will not get rid of the peak. It will reduce the magnitude, but at the cost of increasing the depth of the "anit-resonant" dips either side of the peak (not shown in the graphs you included). Resonant notch filters are wonderful tools for canceling resonant structures of drivers, but they are not a panacea. Very high Q resonances of low magnitude cannot be successfully canceled. A notch filter is unlikely to work with this driver (and is unlikely to work with the W3-871S full-range).

Now, as to your simulations. Did you specify an enclosure for the woofer? If not, what does Speaker Workshop default to? If it defaults to open air, then this explains the reduction in output at woofer Fs. Anyway, unless you have specified an enclosure, ignore this low frequency stuff in your simulation. The saying is "garbage in, garbage out."

Next, doesn't Speaker Workshop allow simulation of notch filters? If not, then this is a limitation of the application. If it does not allow simulation of notch filters, it may also not handle the cone resonant peak. You have this peak in the shared stop band regions of each driver. This is a problem. This resonant breakup will dramatically change the acoustic center and the complex impedance of the driver. Both will interfere with the summing of the crossover (as well as interaction with the components in the passive speaker level crossover network) in both magnitude and time (transient).

Lastly, I looked up your original post on this speaker project. Neither driver is still available from parts express. They still have their frequency response test for the driver posted and it is a bit worse than the frequency response in imported into the simulation application. This will also reduce the value of your simulations.

Mark

[BAM]

I used Excel to graph the Clio data for the driver, and then I used FRD Consortium SPL Trace software to trace the graph. That is the same data. How should I get rid of that peak in the woofer's response?



This graph is just the graph of the response of the woofer after crossover.

[BAM]

Ok, I discovered an error in a couple component values. Here is the newly-calculated response with the correct Bessel crossover. Midbass suckout is still there for some reason. Perhaps now the peak can be cut out since it is bigger??



This is the new response. I discovered I had specified farads instead of microfarads. Also, I tweaked the L-Pad to find the smoothest response.

[yeshu26]

I had similar problem, I tried a third order electrical slope crossover and have been smiling ever since!