LR2 experiment with delay network for 8945P/810921 *PIC*

[Jeff B.]

Going back to the original topic---sorry, dlr, I couldn't resist for the discussion is already beyond the let-the-past-go mode---, it is now unlikely that my preference of the LR2 crossover w/ Usher 8945P/Peerless 810921 to the LR4 one with the same drivers is due to their different characteristics of phase rotation. Though it could still be a possible reason because their xover points are a bit different (LR2 at 2.2 k vs LR4 at 1.6 kHz) and as a result the frequency at which a maximum phase change occurs should be different from each other, I believe it should be more of a negligible factor.

If we now ignore this factor, the rest are 1) smoother transition of system power response, and 2) smoother transition of system harmonic distortion profile from woofer to tweeter. I believe the first should be in most cases a major factor because it must be much more discernable. Concerning this, we may want to note that the major cause of this different power responses of LR2 vs LR4 with non-coincident drivers is due to their different vertical polar responses, as their horizontal off-axis responses are not very different from each other. This can be seen in Zaph's designs: ZRT (LR4 at 1.7 kHz) vs ZD5 (LR2 at 2.4 kHz).

ZRT 2-way Vertical Polar Response (1000 to 2400Hz, 200Hz step)


ZD5 Vertical (and Horizontal) Polar Response (1400 to 3400Hz, 200Hz step)


As you can see, LR2 has slower transition of null patterns across frequencies than LR4. This causes LR4 to have a more abrupt hump right after a dip on its power response (see a simulation at Roman's web page). This can have an adverse effect on our perception. Acoustical room treatment usually does not help much since the treatment is generally on the sides, back and front while the cause of the effect is the system's vertical response behavior for which ceiling and floor treatment is required.

Another factor that makes the effect even more pronounced is the fact that a usual crossover point of LR4 design with modern high quality tweeters is 1.5 kHz to 2.5 kHz versus LR2's 2 kHz to 4 kHz. This means that the power response hump of LR4 is located right at 3 kHz to 5 kHz to which our ear is most sensitive (see here for the recent results of human loudness perception studies). In contrast, LR2 should have a very mild hump on its power response at 4.5 k to 8 kHz.

I have been a proponent of the use of a low xover point, whenever suitable, for reasons like, to avoid the woofer's beaming effect as much as possible, to avoid breakup-induced harmonic distortions (epecially with metal cone midwoofers), to cover up the general 'cone surround dip' of 5" to 7" drivers, etc. But factors in the above made me rethink. Speaker designing is a process of making compromises. It is entirely possible that this power response aspect may actually have a more impact on our subjective impression of sound quality than some of us have been thinking.

-jAy

Yes, that's exactly what I was referring to in my reply to dlr earlier. Here is the quote:

Given that the on-axis response is flat with either LR2 or LR4, the power response and the off-axis response are directly related. So, whether you are saying you like LR2 better due to the smoother power response or due to the smoother off-axis response, you're really saying the same thing when you get right down to it. The power response is simply the integral of the other, but both are composed of the same data in the end. The bottom line is that with the primary acoustic difference between LR2 and LR4 is what is happening off-axis one octave above the crossover point. And for most of us, this is in a very sensitive range for our hearing.

Jeff

Now, with LR4 a higher crossover point will move the power response dip to a frequency range that we are more senstive to and possibly sound very smooth. The peak will be higher and may not be nearly as objectionable, so it is all about where we are crossing and how we perceive the affect.

By the way, the calculator section of PCD has an all-pass delay calculator in it that will tell you the ultimate delay of the circuit as well as the delay at the resonant frequency. It would be easy to covert the time delay to phase if you wanted. If you want I can send the formulas used, or post them.

And finally, I agree with the others, although you may be tired of hearing it , but I also appreciate the way you have evolved in your communications skills through this forum.

Jeff

[jkim]

Yes, that's exactly what I was referring to in my reply to dlr earlier. Here is the quote:

Right, I knew that you made that post and also other similar statements before. I also believe that many other people are aware of the issue. I just wanted to rephrase and elaborate related aspects.

Now, with LR4 a higher crossover point will move the power response dip to a frequency range that we are more senstive to and possibly sound very smooth. The peak will be higher and may not be nearly as objectionable, so it is all about where we are crossing and how we perceive the affect.

Another possible method with LR4 with a low xover point is deliberately placing a slight dip over at 2.5 k to 4 kHz on the system's on-axis response. This can usually be done by shaping the tweeter's rolloff.

By the way, the calculator section of PCD has an all-pass delay calculator in it that will tell you the ultimate delay of the circuit as well as the delay at the resonant frequency. It would be easy to covert the time delay to phase if you wanted. If you want I can send the formulas used, or post them.

I was aware of your calculator. But I simply couldn't focus on the fact that an all-pass filter's delay is bounded by maximum 180 degrees of phase lag, (and why it should be), which is the reason why a non-time-aligned LR2 design with a delay circuit on the tweeter has total system phase rotation of 360 degrees.

And finally, I agree with the others, although you may be tired of hearing it , but I also appreciate the way you have evolved in your communications skills through this forum.

Yes, I'm tired But I appreciate your appreciation.

-jAy

[dlr]

To expand on my take and to re-clarify, I am not in any way referring to 2-way systems. I don't do 2nd order on a 2-way in most cases. I limited my comments to 3-way systems and in particular those with smallish midranges.

In fact, in thinking about it further, my preference extends more to the woofer/midrange than it does to the midrange/tweeter, though I feel LR2 to be slightly better there, too. I've compared the same 3-way with only a change to the M/T and noted the LR2 to be better, but not dramatically.

Consider this (old) graph I had posted (temporarily) at my site. This was when I was working with the 3way with the 13m/8640 and before I improved the mid with more felt. The same holds for my current 12m without the bump, so I feel it's representative. Let me state up front that this is using the built-in power response function of CALSOD that is more a trend than it is precise with a simple model. It doesn't accurately reflect driver rolloff (obvious from the tweeter top end power).



The woofer/mid is LR2 in both, 290 vs. 340, while the mid/tweeter is 4.5K LR4 and 2.5K LR2. The LR4 version has a dip/peak/dip/peak response. Had the W/M been LR4, the dip/peak would have been narrowed a bit more. Move the M/T down to 2.5K and the response is still similar. There would a broad peak roughly centered on the midrange peak.

Before driver rolloff is added, even the LR2 has a midrange peak, though much less pronounced. Add in tweeter rolloff and it's likely that the broad band power response would be smoother overall. Add in some midrange rolloff and it's smoother yet again. It could even be +/- 2 (maybe better?) above the woofer midband (for this simple model).

But again, in my case, it's not so much the M/T response power response as it is the W/M response. I think that rolloff of each driver adds to the power response smoothness and, this is key, only for a 3-way. A good 3-way LR2 is much more likely to have a broad, smooth power response. A 2-way cannot avoid a dip in the power response, whether LR2 or LR4. I don't often find 2-ways that are very amenable to LR2, however.

So again, I'm sure it's partly power response due to crossover, but not exclusively. I can't say which one has what impact.

This is still outside of any distortion issues that just complicate it more. I resist making any conclusions as to any kind of smooth distortion transition. I don't see that as having any validity at this point. There's no data on which to base this, it can't be any more than conjecture for now. Show me some comparative system distortion measurements and I might be inclined to agree.

dlr

[Jeff B.]

To expand on my take and to re-clarify, I am not in any way referring to 2-way systems. I don't do 2nd order on a 2-way in most cases. I limited my comments to 3-way systems and in particular those with smallish midranges.

In fact, in thinking about it further, my preference extends more to the woofer/midrange than it does to the midrange/tweeter, though I feel LR2 to be slightly better there, too. I've compared the same 3-way with only a change to the M/T and noted the LR2 to be better, but not dramatically.

Consider this (old) graph I had posted (temporarily) at my site. This was when I was working with the 3way with the 13m/8640 and before I improved the mid with more felt. The same holds for my current 12m without the bump, so I feel it's representative. Let me state up front that this is using the built-in power response function of CALSOD that is more a trend than it is precise with a simple model. It doesn't accurately reflect driver rolloff (obvious from the tweeter top end power).



The woofer/mid is LR2 in both, 290 vs. 340, while the mid/tweeter is 4.5K LR4 and 2.5K LR2. The LR4 version has a dip/peak/dip/peak response. Had the W/M been LR4, the dip/peak would have been narrowed a bit more. Move the M/T down to 2.5K and the response is still similar. There would a broad peak roughly centered on the midrange peak.

Before driver rolloff is added, even the LR2 has a midrange peak, though much less pronounced. Add in tweeter rolloff and it's likely that the broad band power response would be smoother overall. Add in some midrange rolloff and it's smoother yet again. It could even be +/- 2 (maybe better?) above the woofer midband (for this simple model).

But again, in my case, it's not so much the M/T response power response as it is the W/M response. I think that rolloff of each driver adds to the power response smoothness and, this is key, only for a 3-way. A good 3-way LR2 is much more likely to have a broad, smooth power response. A 2-way cannot avoid a dip in the power response, whether LR2 or LR4. I don't often find 2-ways that are very amenable to LR2, however.

So again, I'm sure it's partly power response due to crossover, but not exclusively. I can't say which one has what impact.

This is still outside of any distortion issues that just complicate it more. I resist making any conclusions as to any kind of smooth distortion transition. I don't see that as having any validity at this point. There's no data on which to base this, it can't be any more than conjecture for now. Show me some comparative system distortion measurements and I might be inclined to agree.

dlr

I'm not arguing with your points, but your on-axis response is quite a bit different between these two also. The LR4 system has be peak in the midrange that you do not have in the LR2, and the LR2 is generally more depressed as well. These systems will sound very different even is power response differences didn't exist. That's why I was comparing textbook flat summations to evaluate the differences in power response. In my model it is the combination of the noncorrelated power response and the driver directivity, particularly that of the tweeter, that create the the apparent peak in the lower treble.

[dlr]

I'm not arguing with your points, but your on-axis response is quite a bit different between these two also. The LR4 system has be peak in the midrange that you do not have in the LR2, and the LR2 is generally more depressed as well. These systems will sound very different even is power response differences didn't exist. That's why I was comparing textbook flat summations to evaluate the differences in power response. In my model it is the combination of the noncorrelated power response and the driver directivity, particularly that of the tweeter, that create the the apparent peak in the lower treble.

Yes, the on-axis is different, but had both summed flat, the same basic characteristics would remain. And again, with my 12m/4631 that does sum flat, the same occurs, I just don't have a set of comparison graphs to display. I think we are agreeing, I just think that there's not a simple way to determine which characteristic is predominantly responsible for the perceived differences when "all else is equal".

The focus seems to be on 2-ways when the Fc is now often 2K or even lower. I just don't hear the differences in 2-way systems that I do in a 3-way with a small midrange, partly because without an offset woofer, the AC difference doesn't allow for optimal summed response anyway, especially LR2 for a 2-way. The woofer rolloff is bound to alter the power response and be more sensitive to even small Fc changes, being closer to the area of changing directionality for the woofer. Even in a 3-way, the M/T response changes are nearly indistinguishable to me if the Fc is unchanged, though there is a small difference, and I offset the mid so LR2 can be easily achieved. But then I use 3K as the crossover most often in a 3-way, since I prefer small mids. It's the W/M that I find so much better using LR2.

dlr