Larger cap?
...and over the weeks it will take me to get moving on this, I'll be starting out passive then evolving from there. I have some active sections in mind. Good luck with yours...
Attachments
OK, so you're offering help again? Man, you are a champ! I used the standard formula for Zobel networks based I think on the actual measured values from WT2. Let me look at it again and get some files together for you...
Well, if I was nearby, I'd certainly be willing to help out with that. I'm just about done with the cabinets on the other project that you helped me on. I posted a couple of photos up in the projects thread.
Steph tsf said;
“If you target a perfect acoustic wave reconstruction, you need a linear phase.
If you target a wide and smooth directivity, you need zero relative phase shifts in the transition band.
Analog crossovers (passive or active) can't deliver both. With passive crossovers, even if it is possible to get a linear phase, you still get a relative phase shift in the transition band, something like 90 degrees.”
Actually with the right combination of acoustic relationships one can have cake and eat it too.
One can have a high order crossover where the summation has no apparent phase shift and the resulting radiation remains a constant beam width with a high degree of angular “confinement”.
In defense of passive filters I would offer the up side. When a driver is driven by an intended source impedance, one finds that is several ways it is not quite the same as actively driven. For example, in the case where a source is driven directly, one finds all of the effects of power compression and parameter alteration are large for a given temperature rise than when driven by a higher source impedance. At the extreme, when driven by a current source (high source impedance) power compression is non-existent up to the point of letting out the magic smoke. Keep in mind too that us poor designers are often faced with the situation of “say, we need a passive box to do X with”.
The down sides would be that one is usually dealing with loudspeaker drivers and not resistors. Drivers only add like resistors (unilaterally in all directions) when they are acoustically very close.
At a spacing of larger than about 1/3 wavelength, drivers begin to produce an interference pattern instead of adding coherently. While the spacing is say less than ¼ wavelength apart, inverting one of two sources results is a near complete cancelation of the energy, when you invert one of two
sources at a large spacing, one only re-arranges the interference patterns pattern of lobes and nulls to even the idea of “speakers adding” only happens in specific conditions.
Once one has an interference pattern, then one is concerned with where the lobe (s) and nulls are relative to where the people are.
Driver placement and interaction is then an X, Y and Z issue.
As if dealing with how drivers add or cancel spatially, the next problem is the two fold issue that the drivers are not simple resistive loads and in a horn system, each range has it’s own magnitude and phase response before adding filtering and one has to equalize the response in addition to providing a crossover. Here, in addition to the crossover part, one usually has to add 3 parts (R,L,C) to address a single bump.
Clearly, the only way to design crossovers like this is using measurements and a computer program, there is no way to design a proper crossover for real drivers even in a simple cabinet by look up tables and tweaking as was the practice in the old days.
Attached is a measurement for a 3 way full range horn I was working on some years ago. This was an SH-50 synergy horn measured on a tower about 25 feet off the ground with the mic at two meters using a TEF-20 at 1/20th octave smoothing. The crossover frequencies are around 250Hz and 1200Hz. The phase rise in the upper part is due to where the TEF places “time =0” which here is about 5/16 inch late relative to actual, otherwise, it looks like one driver instead of 7. It is a 3 way system and can reproduce a square wave from fair to very good looking on an oscilloscope, from about 260Hz to about 2900Hz and has no nulls or outside lobes in it’s pattern.
For what we do, the concern about having everything add up into one source is because for large spaces one needs a lot of acoustic power, for that one needs a lot of drivers. The down side of “a lot of drivers” is those systems usually radiate a complex interference pattern such that if the wind blows even a small amount or if you move around, there are audible changes. Also the large interference arrays that are so popular now, produce a difference spectrum at every seat and distance. With a constant directivity point source like this, the sound spectrum is the same essentially everywhere in the pattern and only gets quieter with distance.
It is hard to describe how large a difference reducing or eliminating the self interference makes in large scale sound but it really is large. We have been selling a lot of speakers for stadium sound upgrades.
If you have headphones hooked up to your computer, try this video Mike from the shop took at a stadium demo Tuesday. Listen to the sound as he walks around and keep in mind, this is all coming from the three speakers (one aimed right, one left and one forward) are the little black blob under the scoreboard.
Danley Sound Labs, Inc.'s Videos | Facebook
That system is active as the power levels add an entirely new layer of issues to passive crossover design, active is so much easier.
Best,
Tom Danley.
“If you target a perfect acoustic wave reconstruction, you need a linear phase.
If you target a wide and smooth directivity, you need zero relative phase shifts in the transition band.
Analog crossovers (passive or active) can't deliver both. With passive crossovers, even if it is possible to get a linear phase, you still get a relative phase shift in the transition band, something like 90 degrees.”
Actually with the right combination of acoustic relationships one can have cake and eat it too.
One can have a high order crossover where the summation has no apparent phase shift and the resulting radiation remains a constant beam width with a high degree of angular “confinement”.
In defense of passive filters I would offer the up side. When a driver is driven by an intended source impedance, one finds that is several ways it is not quite the same as actively driven. For example, in the case where a source is driven directly, one finds all of the effects of power compression and parameter alteration are large for a given temperature rise than when driven by a higher source impedance. At the extreme, when driven by a current source (high source impedance) power compression is non-existent up to the point of letting out the magic smoke. Keep in mind too that us poor designers are often faced with the situation of “say, we need a passive box to do X with”.
The down sides would be that one is usually dealing with loudspeaker drivers and not resistors. Drivers only add like resistors (unilaterally in all directions) when they are acoustically very close.
At a spacing of larger than about 1/3 wavelength, drivers begin to produce an interference pattern instead of adding coherently. While the spacing is say less than ¼ wavelength apart, inverting one of two sources results is a near complete cancelation of the energy, when you invert one of two
sources at a large spacing, one only re-arranges the interference patterns pattern of lobes and nulls to even the idea of “speakers adding” only happens in specific conditions.
Once one has an interference pattern, then one is concerned with where the lobe (s) and nulls are relative to where the people are.
Driver placement and interaction is then an X, Y and Z issue.
As if dealing with how drivers add or cancel spatially, the next problem is the two fold issue that the drivers are not simple resistive loads and in a horn system, each range has it’s own magnitude and phase response before adding filtering and one has to equalize the response in addition to providing a crossover. Here, in addition to the crossover part, one usually has to add 3 parts (R,L,C) to address a single bump.
Clearly, the only way to design crossovers like this is using measurements and a computer program, there is no way to design a proper crossover for real drivers even in a simple cabinet by look up tables and tweaking as was the practice in the old days.
Attached is a measurement for a 3 way full range horn I was working on some years ago. This was an SH-50 synergy horn measured on a tower about 25 feet off the ground with the mic at two meters using a TEF-20 at 1/20th octave smoothing. The crossover frequencies are around 250Hz and 1200Hz. The phase rise in the upper part is due to where the TEF places “time =0” which here is about 5/16 inch late relative to actual, otherwise, it looks like one driver instead of 7. It is a 3 way system and can reproduce a square wave from fair to very good looking on an oscilloscope, from about 260Hz to about 2900Hz and has no nulls or outside lobes in it’s pattern.
For what we do, the concern about having everything add up into one source is because for large spaces one needs a lot of acoustic power, for that one needs a lot of drivers. The down side of “a lot of drivers” is those systems usually radiate a complex interference pattern such that if the wind blows even a small amount or if you move around, there are audible changes. Also the large interference arrays that are so popular now, produce a difference spectrum at every seat and distance. With a constant directivity point source like this, the sound spectrum is the same essentially everywhere in the pattern and only gets quieter with distance.
It is hard to describe how large a difference reducing or eliminating the self interference makes in large scale sound but it really is large. We have been selling a lot of speakers for stadium sound upgrades.
If you have headphones hooked up to your computer, try this video Mike from the shop took at a stadium demo Tuesday. Listen to the sound as he walks around and keep in mind, this is all coming from the three speakers (one aimed right, one left and one forward) are the little black blob under the scoreboard.
Danley Sound Labs, Inc.'s Videos | Facebook
That system is active as the power levels add an entirely new layer of issues to passive crossover design, active is so much easier.
Best,
Tom Danley.
Attachments
It is more question of system concept.
If you just replace a passive XO with an active one with the same identical speaker design nothing essential is gained, and even some things are lost as not all drivers run best with low impdeance drive at MF/HF (of their passband)
But active means more than using standard drivers and standard amps, merely placing the XO at line level as the only difference vs passive.
Active systems start to excel when you make use of the added degrees of freedom, which are i) arbitrary drive impedance and ii) choice of drivers in general. Simpler implementation of more complex filter structures is just the starting point of where active design really start to shine.
Active design frees the designer of the amps and the drivers (which should be the same person, ideally) from a set of constraints that are usually taken as "set in stone" by most any industry or DIY designer of speakers and amps.
Active, to me, automatically implies DSP-methods, and one thing where DSP procecssing is really useful is when you want linear phase *without* any compromise and drawbacks (like non-zero inter-driver phase).... maybe unless you are as smart as Tom Danley is.
- Klaus
Active, to me, automatically implies DSP-methods, and one thing where DSP procecssing is really useful is when you want linear phase *without* any compromise and drawbacks (like non-zero inter-driver phase).... maybe unless you are as smart as Tom Danley is.
- Klaus
Having been involved with the development of DSP linear phase crossovers I would not say there are no compromises.
The transfer function won't change for the active crossover, but it will for the passive crossover. It's the interaction between the impedance of the driver, and the crossover component, that makes the transfer function in a passive crossover. It's impedance independant in an active XO.
Last edited:
Not to an appreciable extent - the center frequency will remain constant for the passive notch xover since it is wholly determined by the L and C that comprises it and is independent of even the presence or absence of the driver - only the driver's parameters that you hypothesize will significantly change due to thermal shift so that any external xover will thus become misaligned whether passive or active.
I don't see that passive xovers are 'cut only'. In each of my last 3 speaker designs, there are frequency regions where I have obtained up to 15 db of boost, and without dropping the impedance to unrealistic levels.
For my basement blasters, I have implemented a passive LC boost that peaks at about 6db ~25hz - this does drop the impedance at this frequency to about 3 ohms compared to the equivalent Rdc of ~4.5 ohms (3 2226J's in parallel), but I could have had 5db without dropping the Z at all at this frequency. Just saying.
Also for my basement blasters, which are a 2 1/2 way system, I peak the mids from the upper 2226J by up to 4db just below xover (which boosts that driver's midband 97db sensitivity to the system's 101db). All passive and overall impedance still safely above 5 ohms in that octave.
Finally, for my basement blasters, I peak the HF by up to 6db at 22Khz into the 2445J's - gives me sizzling highs with no loading issues since the 2445J impedance has more than doubled compared to below 5K.
With my Iron Lawbreakers, I'm using an air core transformer for both impedance transformation and bass boost - allows me to tune the 2220A woofer BR to 32hz and fills in the response saddle very well between 50 hz and 130hz. Plus, the passive xover equalizes the impedance of the overall speaker to 18 ohms +/- 20% from 40 hz to 250 hz, gives me 100db/w/m sensitivity and useable bass response down to 30 hz in a 100 liter cabinet......Never saw any of that in a digital xover advertising blurb.
Finally, for my HT semi - line arrays - I am getting an outrageous 15 db peaking at 22khz into one of the Aura NS3 knockoffs (the center one) that equalizes the whole nine speaker array close to flat out to that frequency using just 2 passive parts, and without overdriving the speaker or overloading the amplifier since the speaker inductance gives around 100 ohms reactive at that frequency. They haven't stopped laughing yet at the 130WRMS (manufacturer's rating) a channel my Onkyo receiver throws at them.
All real world implementations where there is more voltage at speaker terminals than the amp terminals with all passive techniques.
I must have missed all the active eqs claiming to be able to do any of the above.
For my basement blasters, I have implemented a passive LC boost that peaks at about 6db ~25hz - this does drop the impedance at this frequency to about 3 ohms compared to the equivalent Rdc of ~4.5 ohms (3 2226J's in parallel), but I could have had 5db without dropping the Z at all at this frequency. Just saying.
Also for my basement blasters, which are a 2 1/2 way system, I peak the mids from the upper 2226J by up to 4db just below xover (which boosts that driver's midband 97db sensitivity to the system's 101db). All passive and overall impedance still safely above 5 ohms in that octave.
Finally, for my basement blasters, I peak the HF by up to 6db at 22Khz into the 2445J's - gives me sizzling highs with no loading issues since the 2445J impedance has more than doubled compared to below 5K.
With my Iron Lawbreakers, I'm using an air core transformer for both impedance transformation and bass boost - allows me to tune the 2220A woofer BR to 32hz and fills in the response saddle very well between 50 hz and 130hz. Plus, the passive xover equalizes the impedance of the overall speaker to 18 ohms +/- 20% from 40 hz to 250 hz, gives me 100db/w/m sensitivity and useable bass response down to 30 hz in a 100 liter cabinet......Never saw any of that in a digital xover advertising blurb.
Finally, for my HT semi - line arrays - I am getting an outrageous 15 db peaking at 22khz into one of the Aura NS3 knockoffs (the center one) that equalizes the whole nine speaker array close to flat out to that frequency using just 2 passive parts, and without overdriving the speaker or overloading the amplifier since the speaker inductance gives around 100 ohms reactive at that frequency. They haven't stopped laughing yet at the 130WRMS (manufacturer's rating) a channel my Onkyo receiver throws at them.
All real world implementations where there is more voltage at speaker terminals than the amp terminals with all passive techniques.
I must have missed all the active eqs claiming to be able to do any of the above.
Last edited:
You are right, it isn't any easier, quite the contrary, and there is no free lunch. I would think a promising "no-compromise" way to go is a well balanced mixed-domain approach. DSP (convolution) restricted to the very front end, while the basic XO is done in analog. In fact the XO is not a band splitter from a common input, rather a set of passband shaping amplifiers that are fed from individually preprocessed DAC outputs. The preprocessing only does the fine-print in shaping to the magnitude targets and of course all of the time domain correction to get the desired summing and it adjusts the total response (mag and phase) at the L.P. to the target.
The pain lies in the process to find the "best" correction kernels (so that they are effective and benign, technically and in perception) and of course in the design of the rest of the system. DSP is only the polishing, not the body. But even a finish as simple as a smooth phase rollback on the stereo input signal for a conventional speaker's allpass is giving much more than is taking.
- Klaus
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.