2nd order LR accoustic slopes will only give a shallow and wide null. Nothing sharp like what JR got with what I assume is closer to LR4. I've never done an odd order crossover like 1st or 3rd order Butterworth so I don't have first hand experience of what their reverse nulls look like. I remember Vance's Loudspeaker Design Cookbook described all of the filter types in depth but I loaned my copy out and it was never returned
I am in agreement with this. Usually the shallower orders' reverse nulls do not dip as deep either. For example, an LR2 reverse null could have a range of 4 octaves, and only dip to -10dB below the fundamental. LR4 tend to be less than 2 octaves wide and -25dB or more deep.
Jeff Bagby made a comment at his seminar in Kokomo that LR4 was probably the best crossover choice for most DIY folks. I can't remember why he said that though.
I went LR4 on these because that is how things clicked. I tinkered with BW3 but couldn't get things to gel correctly. I will try again on the upcoming TMWW. Constant power topology has some real benefits.
I end up needing to relearn each time, but I just play around with values in sims until I get what I need. The resistor only shrinks the amplitude of the dip in the response/impedance. So in sims you can start with a resistor value of 0 to make it the most obvious what the L and C are doing, then add resistance towards the end to tweak.. I start to get the hang of what changing each component value will do pretty quickly (changing frequency and width of the notch). Just remember the series notch is in parallel with the driver so you do want some resistance in there, otherwise impedance dips to the inductor DCR in that spot. Unless you are suppressing something far enough outside of the passband that the crossover circuit prevents that dip getting back to the amp.
The way I understand JRs tweeter circuit is that the parellel shunt resistor dampens the Fs peak a bit and allows for some smaller component values on the notch maybe. I use my DATS to give me tweeter FS Qe and Re, then I use a hand held calculator and formulas from cook book . But for a response peak I assume that I might need to calculate Qe from the 3Db down points or is there a less cumbersome method?
I will echo what Andrew said above about tinkering within the modeling software re. notches.
I wish there was a better answer, but I just insert the components with any set value and move L and C around until I hit my target. The notch on the woofer is to take care of the typical bump left behind just below Fc and the notch on the tweeter was to tame a broad hump. Nothing revolutionary, but the notch on the woofer is not a common technique.
Looking at the above picture you can clearly see the baffle step loss. I targeted my notch as shown below:
This results in the following driver response:
With a traditional topology, a peak remains just below the crossover point - and using the notch removes that peak. That allows a smaller coil to roll off the woofer and eliminates the broad depression below that peak often left behind when using a larger coil.
Here is a snapshot of what a large coil and no notch yields:
THese will be at Iowa next year, and possibly DakotaDIY next June (still in early planning stages for that one). IMHO the results speak for themselves on the value of this approach. I have done this several times in the past, but usually only when the quality of the drivers warrants it. Otherwise I live with the irregular response since it still usually falls in the +/-3db paradigm.
I have also been thinking about the reverse null on odd order networks. It seems to me that it is simply the result of either an upward or downward tilt of the main lobe where both drivers are operating, caused by a time arrival difference
@Eggguy said:
I have also been thinking about the reverse null on odd order networks. It seems to me that it is simply the result of either an upward or downward tilt of the main lobe where both drivers are operating, caused by a time arrival difference
IIRC 3rd order Butterworth are 135 degrees out of phase at the crossover point. That's why you don't get a deep reverse null.
Comments
LR are only even order.
In my experience the higher the slope, the narrower the reverse null is.
2nd order LR accoustic slopes will only give a shallow and wide null. Nothing sharp like what JR got with what I assume is closer to LR4. I've never done an odd order crossover like 1st or 3rd order Butterworth so I don't have first hand experience of what their reverse nulls look like. I remember Vance's Loudspeaker Design Cookbook described all of the filter types in depth but I loaned my copy out and it was never returned
I am in agreement with this. Usually the shallower orders' reverse nulls do not dip as deep either. For example, an LR2 reverse null could have a range of 4 octaves, and only dip to -10dB below the fundamental. LR4 tend to be less than 2 octaves wide and -25dB or more deep.
InDIYana Event Website
1st and 3rd order Butterworth filters have no reverse null. They measure the same either polarity. In a perfect world.
No null for 3rd order slopes when flipping polarity, just 90 degrees out of phase.
Jeff Bagby made a comment at his seminar in Kokomo that LR4 was probably the best crossover choice for most DIY folks. I can't remember why he said that though.
Easy to confirm via measurements.
I went LR4 on these because that is how things clicked. I tinkered with BW3 but couldn't get things to gel correctly. I will try again on the upcoming TMWW. Constant power topology has some real benefits.
How do you calculate your series notches?
I model them using measurements.
FNG need more info
I guess I am unsure what you are asking - calculate as in what frequency to cover?
I end up needing to relearn each time, but I just play around with values in sims until I get what I need. The resistor only shrinks the amplitude of the dip in the response/impedance. So in sims you can start with a resistor value of 0 to make it the most obvious what the L and C are doing, then add resistance towards the end to tweak.. I start to get the hang of what changing each component value will do pretty quickly (changing frequency and width of the notch). Just remember the series notch is in parallel with the driver so you do want some resistance in there, otherwise impedance dips to the inductor DCR in that spot. Unless you are suppressing something far enough outside of the passband that the crossover circuit prevents that dip getting back to the amp.
makes sense
The way I understand JRs tweeter circuit is that the parellel shunt resistor dampens the Fs peak a bit and allows for some smaller component values on the notch maybe. I use my DATS to give me tweeter FS Qe and Re, then I use a hand held calculator and formulas from cook book . But for a response peak I assume that I might need to calculate Qe from the 3Db down points or is there a less cumbersome method?
I am referring to the woofer circuit using my omnimic
Let me give some thought on how to articulate this. Will be from a real PC, phones are difficult for me.
PCD has notch calculators. I use that to get starting values then adjust in XSim. I use the Re at the mid point where I want the notch.
I will check it out, thanks
I will echo what Andrew said above about tinkering within the modeling software re. notches.
I wish there was a better answer, but I just insert the components with any set value and move L and C around until I hit my target. The notch on the woofer is to take care of the typical bump left behind just below Fc and the notch on the tweeter was to tame a broad hump. Nothing revolutionary, but the notch on the woofer is not a common technique.
Looking at the above picture you can clearly see the baffle step loss. I targeted my notch as shown below:
This results in the following driver response:
With a traditional topology, a peak remains just below the crossover point - and using the notch removes that peak. That allows a smaller coil to roll off the woofer and eliminates the broad depression below that peak often left behind when using a larger coil.
Here is a snapshot of what a large coil and no notch yields:
THese will be at Iowa next year, and possibly DakotaDIY next June (still in early planning stages for that one). IMHO the results speak for themselves on the value of this approach. I have done this several times in the past, but usually only when the quality of the drivers warrants it. Otherwise I live with the irregular response since it still usually falls in the +/-3db paradigm.
Hope this helps.
I like it. Thanks. That last roll off looks very familiar to me. Might I propose that we therefore call it the FNG Roll off?
I have also been thinking about the reverse null on odd order networks. It seems to me that it is simply the result of either an upward or downward tilt of the main lobe where both drivers are operating, caused by a time arrival difference
Not really an FNG curve, it is something most of accept as a compromise.
IIRC 3rd order Butterworth are 135 degrees out of phase at the crossover point. That's why you don't get a deep reverse null.