I'm building a small PA system for when I DJ at house parties. I've made a pair of 12in bass bins and pair of 8in + tweeter mid/top cabs. I was planning on using a traditional 4th-order Linkwitz-Riley active crossover at 300Hz for the crossover between cabinets, but I'm wondering how a 4th-order LR for the high-pass and subtractive for the low-pass would fare. As I never thrash the system I reckon the stopband attenuation would not really be a problem.
Any comments on the sound subtractive vs traditional in this application?
Any comments on the sound subtractive vs traditional in this application?
Thanks bjorno. From your graph I pick out that the combined phase response is flat, this is to be expected from subtractive crossovers. The combined amplitude response I'm not sure is flat though, but maybe the bump at crossover will just give a bit of vocal projection, no bad thing really for a PA system. I've attached my simulated filter responses.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
That bump is not that severe. If phase is taken into account it's flat. The subtractive process occurs in real time, not values merely added together, as in the graph.
What you need to watch is the roll off on the bass channel. Its only about 6db/oct. That could be a problem with edge resonance of the bass driver.
Regards,
Geoff.
What you need to watch is the roll off on the bass channel. Its only about 6db/oct. That could be a problem with edge resonance of the bass driver.
Regards,
Geoff.
Thanks Geoff that has helped. I did worry about the shallow roll-off on the bass as the graph showing the much steeper mid roll-off accentuates the problem.
The circuit is fairly simple so I might have a go and see how it turns out. After all a lot of PA systems run the bass driver fullrange or just with a passive crossover, at much higher frequency. The one I'm using has a foam roll surround so edge resonance might not be too badly damped.
The circuit is fairly simple so I might have a go and see how it turns out. After all a lot of PA systems run the bass driver fullrange or just with a passive crossover, at much higher frequency. The one I'm using has a foam roll surround so edge resonance might not be too badly damped.
Your welcome Richie00boy.
If you ever want to shift the xo on the subtractive filter, you only have one set of components to change. You can even make it switchable.
I prefer simple passive xos if I must. But my favourite 15" with foam surround was a bit edgy about 750Hz on a 6db xo at 500.
Geoff.
If you ever want to shift the xo on the subtractive filter, you only have one set of components to change. You can even make it switchable.
I prefer simple passive xos if I must. But my favourite 15" with foam surround was a bit edgy about 750Hz on a 6db xo at 500.
Geoff.
I decided to check time domain amplitude response of 300Hz sine input vs the two outputs. The HF output is out by 180 degrees, surely this should be in phase at the crossover frequency?
An externally hosted image should be here but it was not working when we last tested it.
Geoff H said:
I thought that with 4th-order phase was correct, only 2nd-order needed the connections reversing?
phase_accurate said:
I suspect that it will be fine with a 2nd-order filter. I did a check as above with a 2nd-order Butterworth filter and the phase I thought didn't look as friendly. See below.
An externally hosted image should be here but it was not working when we last tested it.
You're comparing phase shifting at 300 Hz on a xo at 380. There will be a shift, even on a passive 6db/oct.
You might be right about the no reversal on 4th order, the graph didn't indicate that.
I with Charles. 12db/oct under 380 Hz takes a lot of power of the upper system. If you're concerned about blowing the top driver, try a $15 8" 10 watt ceiling speaker on a 2 ft square baffle on the combo.
Cheers,
Geoff.
You might be right about the no reversal on 4th order, the graph didn't indicate that.
I with Charles. 12db/oct under 380 Hz takes a lot of power of the upper system. If you're concerned about blowing the top driver, try a $15 8" 10 watt ceiling speaker on a 2 ft square baffle on the combo.
Cheers,
Geoff.
The main reason for using this type of constant-voltage crossover is improving the transient accuracy.
So polarity inversion etc in order to get a flat summed response is definitely a no-no.
Taking the drivers' responses (especially phase) into account is even more critical for this type of filter than wirth normal crossovers.
The hump on the lowpass gets larger with increasing order of the highpass filter so I'd go for 3rd order or even lower.
If you use 2nd order response (this is including driver and highpass !) for instance you will not have significant excursion reduction at very low frequencies. So this type of x-over is meant to be used with sturdy drivers.
If transient accuracy isn't a goal of your design you are better off using conventional crossovers.
Regards
Charles
So polarity inversion etc in order to get a flat summed response is definitely a no-no.
Taking the drivers' responses (especially phase) into account is even more critical for this type of filter than wirth normal crossovers.
The hump on the lowpass gets larger with increasing order of the highpass filter so I'd go for 3rd order or even lower.
If you use 2nd order response (this is including driver and highpass !) for instance you will not have significant excursion reduction at very low frequencies. So this type of x-over is meant to be used with sturdy drivers.
If transient accuracy isn't a goal of your design you are better off using conventional crossovers.
Regards
Charles
Geoff H said:
You're right, now I see I did the 4th-order one at the -6dB frequency.
phase_accurate said:
The mid cab is actually a sealed box hifi speaker housing a bass-mid, so it will probably even be fine with a 1st-order filter @ 300Hz. Cabinet resonance is under 70Hz so phase response should be a small issue.
I'm going to try a 2nd-order Butterworth filter and see how it works out. That will reduce the hump but still provide enough cut for the mid cab. To be honest the primary goal is just good sound at decent SPL, and I'm interested in a little experimentation.
Or maybe a Bessel highpass ? This will lead to a smaller hump on the lowpass !
Willing to experiment is definitely a prerequisite for using this type of crossover.
Don't forget that this type of crossover is more susceptible to phase response aberrations of the drivers used. Don't be surprised/disappointed if you have to use some equalising to get a flat response. The other price you will have to pay for the improved transient accuracy is the not so ideal polar response of this type of crossover (hint: always use the tops and subs as stack).
If you are willing to experiment a little more then I might have some proposals for the practical implementation of this type of x-over that is taking the driver's response into account.
Regards
Charles
Thanks Charles. Here is the circuit I will be using as a starting point. I've used a Bessel filter with F3 @ 300Hz.
However, the peak is at 173.78Hz, is this the actual crossover frequency?
Also, what is the polar response like? I will be using the cabs as stacks with the drivers fairly close together.
An externally hosted image should be here but it was not working when we last tested it.
However, the peak is at 173.78Hz, is this the actual crossover frequency?
An externally hosted image should be here but it was not working when we last tested it.
Also, what is the polar response like? I will be using the cabs as stacks with the drivers fairly close together.
The answer to this question is the one that everyone fears most: It depends !
If you have a response nonlinearity that is minimum-phase then you can correct it with the inverse transfer function of the error and you will correct the phase error also !!
The response will be tilted down a little in the vicinity of the crossover frequency - depending on the dimension and the crossover frequency itself.
I usually simulate this things by introducing a delay with the trasmission-line model of SPICE. Although this approach is assuming point sources and thus making the simulated result look more severe than it actually is in practice.
Regards
Charles
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.