Sloped baffle

Lewinskih01,
I took the time to read the XO White Paper by Dr. Uli. In many places he makes the same points that Roy is making i.e. lower order x-over ckts are better than higher order x-over ckts. He talks about the time delays getting worse with higher order x-over ckts - same point that Roy has made many times.
Dr. Uli talks about using minimum phase filters for the analog x-overs &
using linear phase filters (which are digital FIR filters. there is no equivalent in the analog domain) for his Acourate digital x-over software.
Dr. Uli makes a general statement that low-order minimum phase filters used in analog x-overs have limitations & create time distortions & cannot be used......
BUT he conveniently starts off with a 2nd-order x-over ckt while completely glossing over a 1st-order x-over ckt. Does the 1st-order x-over ckt have the same limitations as the 2nd-order x-over Dr. Uli discussed? Dr. Uli would like you to think so but I don't think so......

I created a simple 1-order network for a tweeter, midrange & woofer. I assumed a 6 Ohms resistance for each of the 3 drivers (totally arbitrary). I arbitrarily chose x-over frequencies of 300Hz & 2KHz. I simulated the frequency, phase & step responses of this 1-order x-over. I've labeled the curves in each of the 3 graphs so you can see which curve belongs to which driver. I've also put markers on various curves so you can see the phase shift at the x-over frequency.

For the frequency response - look at the sum of the frequency responses. There's only a 2dB hump at the x-over points.

For the phase response - look at the sum of all the phase responses/ There's a phase shift of only +/- 8 degrees over the entire audio band of 20Hz - 20KHz.

And, for the step response - you can clearly see that all 3 drivers act in unison to create unified step response (rather than the spikes you see in time-Incoherent speakers where the tweeter acts first, the mid second & the woofer third).
From these simulations, a 1st-order passive x-over looks quite good.
And, I don't have the music signal going thru somebody's DSP algorithm which is doing a great deal of signal processing to massage the music signal thereby imparting its sonic signature to the music signal.
Sure the passive x-over components are also imparting their signature to the music signal but by using top quality components I can minimize this.
In the DSP software, if I don't know what I'm doing, I can botch thing pretty badly because the music signal is so heavily modified by the DSP algorithm.

Here is the link to the simulations, if anyone is interested:
https://picasaweb.google.com/bombaywalla9/FirstOrderXOverFreqPhaseStepResponses?authkey=Gv1sRgCOz6xv6RnMDeUA#

In the XO White Paper, Dr. Uli says that "So the crossover has to be selected so that the good properties of the driver are used ! If the driver does not have a good behaviour we should not use it."
I am assuming the "good properties" of a driver are that it has flat freq response over its passband & rolls off at a frequency beyond the x-over freq chosen in Acourate by the user. BUT............
The degree that Acourate can compensate for any driver depends on how well you can characterize the driver. And, we of course, do not know if the drivers in our existing speakers have these "good properties" or not.....

Bombaywalla,

Not sure what to take away from your post. 1st order passive crossovers are better than any type of digital crossover? That would be in line with Roy's explanation. But I wasn't arguing otherwise...

I'll take as given that Roy's approach is the best one could hope for. My question to him is how close to that would my described approach get me.

Cheers!

hi Lewinskih01,
yes, with some engineering proof, that's what I was trying to say. And, the reason that seemed to make sense to me is that signal processing is happening correctly, real-time thru the passive x-over components without any intervention by a human-being. In a time-coherent loudspeaker with passive x-overs, drivers with "good properties" have already been selected & the x-over designed around them & the whole system would be working to benefit the user.

With digital x-overs the correction is as good as the skill of the user to characterize the drivers & to come up with the appropriate filter response to yield a time-coherent delivery. And, from reading Roy's letter to Six Moons - the link to which he provided earlier on - it's no easy feat to characterize a driver in the room. One cannot use 1 type of test tone, one needs to use many different types. And, one needs to measure the driver response in many ways to get an accurate characterization of the driver. Otherwise, the DEQX or Acourate correction will be (very) limited leading to less than stellar benefits.

I don't think that Roy can tell you how well DEQX or Acourate will solve your problem because the answer lies in how skilled you are in understanding the science behind how the driver response is affected by your room,
how skilled you are in DSP algorithms to come up with a filter that corrects for your room & your particular choice of drivers
how skilled you are in understanding the science behind reflections of drivers off the front baffle,
how skilled you are in understanding what the requirements are for selecting a microphone to do the driver characterization,
how skilled you are in compensating for this mic's own frequency response so that you don't misunderstand the mic's response to be that of your driver's,
etc, etc.

My understanding is that if you room correct like HT Receivers do & plug in the correction into some pre-designed filter in the software, you'll get a correction that's average at best & you might not like the results.
The thing that Roy has been saying all along is that we don't listen to test tones (which is what the room correction tones are) - we listen to music which is a bunch of partial wavelengths of various frequencies.
You use full cycle tones to characterize the driver then do the correction & then play partial wavelengths of various frequencies thru that driver - the correction to the driver, in my understanding, is invalid.
Of course, I could be totally off-base here....
FWIW.

Good questions.

I do agree with what Bombaywalla just posted- knowledge and experience in many different areas is required. I know of no way out of that, to simplify a home-designer's life.

Driver selection is by far the most important factor. If all we care about is making the best sound, instead of spending money on the newest technology (usually inferior, I find), then here are the important questions to ask before selecting any drivers:

- How far away will I be from the speakers?
- What kinds of music will I play most?
- How loud will I play, even if only on occasion?
- How large is my room?
- How low in the bass do I want the speakers to go? Here, it is best to use 'body feel' as your guide. If you want to shake the house and your lower pants legs on electric bass, then the speakers need to have good output to 40Hz, but not any lower.

Listening at ten feet away in a room that is not entirely open into the rest of the home, this amount of low-bass output requires a low-distortion eight-inch woofer with a large-diameter bass port tuned to ~40 Hz, or a sealed-box ten-inch woofer, flat to 40Hz (good luck finding that in today's marketplace), at the minimum. There is no reason to use multiple 8 or 10-inch woofers per cabinet.

Which means this will be a three-way design to be able to use a first-order crossover, since no 8 or 10-inch woofer can meet a tweeter.

On the top end, choose ~1" dome tweeter, not one made of metal nor of 'ring radiator' design. That means ~3kHz crossover point. The eight or ten inch woofer means ~300Hz crossover point, or slightly higher. And that means using a 4 to 5-inch mid driver showing no cone breakup nor the HF resonance of metal-cone drivers.

All these drivers need very flat frequency responses. Avoid drivers with impedance-curve wiggles, as those indicate resonances and cone breakups. Avoid molded plastic cones and metal cones.

Sorry- got carried away. I cannot put out my version of the Loudspeaker Design Cookbook here.

Do know that, by careful manipulation of the Zobel parts in my passive crossovers, I can fine-tune the time-coherence between drivers (their individual phase responses), for a better blend. This cannot be achieved digitally without custom programming and the consequent extra signal processing (assuming the right measurements can be made, which is not likely).

But you can always listen to your adjustments, and for that process, I recommend you listen to only your left speaker, but not in mono. Start with getting that speaker's voice range right, such as on a older Diana Krall recording. And get rid of cabinet reflections with wool felt for at least the tweeter, or you are screwed from the beginning.

For a home designer, the results with a simple passive crossover with Zobels or with a digital first-order crossover/EQ/time delay setup will be satisfying on most music. However, the sound would still 'not be quite right' on enough other music to make you think there's something wrong with your source or room or cables or amplifiers.

That turns out to be the residual phase shift of the speakers, which is what I finally fixed .

I will continue to think about questions Bfwynne and Lewinskih01 posed and get back to you.

Best,
Roy

My goodness, I just glanced through the XO paper by Dr. Brüggemann. With all due respect, he is not right in many ways about how crossovers work!

The technical details are far too lengthy for here, but I will point out that, in Fig. 8 on his page seven, he described 'lining up the peaks' from a woofer, mid and tweeter. Instead, what must be done is to line up WHEN each driver's pulse JUST BEGINS to turn upwards from Zero. That's a point easily judged for the beginning of a tweeter's spike, but not on a woofer's slow rise (hence a measurement problem). Thus I advise not bothering with his paper, sorry.

The diagrams from Bombaywalla on his Picasa page DO get that starting alignment correct, although I see some problems:
- The scale used shows a definite starting point to the woofer's pulse. That point is not well-defined when the horizontal scale is expanded.
- The loudness of the mid driver seems low, but I could be wrong.
- The summation pulse is not close enough to the ideal.

But it is late now, and no one is paying me to analyze what may be wrong there- just wanted to point out some suspicious items.

Best,
Roy