Yes, that's what I understand (I think) but what is confusing is the acoustic null in regards to distance to set the drivers but that is essentially a reflection? I found that with a .16ms delay on the mids and around 2ms on the low mids (bandpass 6" woofers in horn) things start sounding right and response flattens out at xover points. What is peculiar is setting any delay no matter how much on the 4x10's or 15's reduces the output and causes issues around the xover points, so these two channels are set to 0ms. Also the 15's work way better inverted phase otherwise another big hole around 50hz exists.
Hi Bill,
So this leads me to a question regarding L-R crossovers as implemented in digital as opposed to analog. The Wikipedia article on L-R filters states 4th order is in phase but 1 cycle delayed on the low pass, so wondering if this is not the case with MiniDSP?
sometimes when you can't find the right delay for the woofer or the mids its because its the tweeter that needs to be delayed...
but that isn't my point. I think its dangerous to set the delay according to how well things sound or how flat it gets around XO - because the delay setting is cyclic and you could be off by cycles, or half cycles if you try flipping polarities. Better to first determine the delay - e.g. by using timing references in your measurements of each driver separately with XO filters in place. Do this after EQing each driver to its textbook target acoustic slope and you will get textbook like performance from your XO.
Hi njones,
if you mean the MiniDSP 2x4 or similar (non Sharc-based), then the behaviour is pretty much the same as analog with a couple exceptions. Those are IIR type DSP devices.
With FIR types (Sharc based like the 2x4HD), the phase can be adjusted independently of the dB magnitude shapes of the crossover filters if you wanted.
For IIR DSP, since there are conversion and processor functions, that causes some constant-with-frequency fixed delays (though if all the drivers go through the DSP then that delay has no effect since it is relative delay between drivers that matters). And up in around the highest octave, there phase and magnitude response shapes can be different than an analog filter -- one way of looking at it is that because this is sampled data, the magnitude curve must be at -infinity dB by the time it reaches the 'Nyquist frequency" --- 1/2 the sample rate.
And pretty much all DSPs, FIR or IIR, allow you to add fixed delays to any of the outputs so you can match up overall fixed delays. That is close to impossible with analog filters, though there are networks called "All-Pass" that can be used to add frequency dependent delays without affecting the magnitude response. You could in theory add a huge number of All-Pass filter stages to reach something that resembles a fixed delay over a given frequency band, but it's really hard to do with passive filters. An inherent limitation with analog all-pass filters is that making a fixed amount of delay takes more and more elements for higher frequencies since there would be more and more cycles that needed to be shifted per second of added delay.
if you mean the MiniDSP 2x4 or similar (non Sharc-based), then the behaviour is pretty much the same as analog with a couple exceptions. Those are IIR type DSP devices.
With FIR types (Sharc based like the 2x4HD), the phase can be adjusted independently of the dB magnitude shapes of the crossover filters if you wanted.
For IIR DSP, since there are conversion and processor functions, that causes some constant-with-frequency fixed delays (though if all the drivers go through the DSP then that delay has no effect since it is relative delay between drivers that matters). And up in around the highest octave, there phase and magnitude response shapes can be different than an analog filter -- one way of looking at it is that because this is sampled data, the magnitude curve must be at -infinity dB by the time it reaches the 'Nyquist frequency" --- 1/2 the sample rate.
And pretty much all DSPs, FIR or IIR, allow you to add fixed delays to any of the outputs so you can match up overall fixed delays. That is close to impossible with analog filters, though there are networks called "All-Pass" that can be used to add frequency dependent delays without affecting the magnitude response. You could in theory add a huge number of All-Pass filter stages to reach something that resembles a fixed delay over a given frequency band, but it's really hard to do with passive filters. An inherent limitation with analog all-pass filters is that making a fixed amount of delay takes more and more elements for higher frequencies since there would be more and more cycles that needed to be shifted per second of added delay.
Thanks for the detailed answer. I'll try the suggested methods to get the crossover producing that [somewhat] square wave.
Thanks,
Nate
Thanks everyone for your suggestions and comments.
I think I've got these dialed in now. The difficulty in the mids not having enough output around 1kz is still there but its really only a few db drop and hardly noticeable, despite all the things I have tried - this seems to be the most common issue.
I'm mulling over an idea for my next speaker: a 2-way full range MEH with Radian 1.4" or 2" compression drivers and 2x15" woofers, using a 2x4hd. This one I would definitely model in Hornresp before making sawdust. A guy named Chris I think who built one from a K402 horn and had great things to say about it.
A two way horn seems appealing if the compression driver can handle a low xover point along with good high frequency dispersion.
I think I've got these dialed in now. The difficulty in the mids not having enough output around 1kz is still there but its really only a few db drop and hardly noticeable, despite all the things I have tried - this seems to be the most common issue.
I'm mulling over an idea for my next speaker: a 2-way full range MEH with Radian 1.4" or 2" compression drivers and 2x15" woofers, using a 2x4hd. This one I would definitely model in Hornresp before making sawdust. A guy named Chris I think who built one from a K402 horn and had great things to say about it.
A two way horn seems appealing if the compression driver can handle a low xover point along with good high frequency dispersion.
They are great fun. I admire you guys (built your own); duffers like me were lucky to buy a trashed Unity pair suitable for tweaking. Bwaslo makes good points about the complexity of passive x-over. Probably an over-simplification, but others have said that much of Danley's success is in the no doubt very complex crossovers. Not only R&D costs, but it must be a large fraction of the cost of a Synergy horn (e.g. Sh50). If instead one can resort to active EQ and measurements, even a hacker can dial in his speaker in just a few measurements that is the equal of what it'd take DSL's team months and millions of dollars to do via passive. AT least that is what my ego tells me 
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.