Wayne Parham's H290C waveguide measurements - Page 7

I generally agree with you.

I know in the case of the SEOS18 (because I have them on hand) the catenary part is at least 90% of the profile, but I think on the smaller devices that becomes less and less due to the size. I was part of the discussion on choosing the right amount, but ultimately I wasn't the one coming up with the CAD models.

I'd say if you ever had interest in checking out the SEOS, the SEOS15 would be the one, but its probably bigger than horn.

Wayne, I didn't mean to suggest the thread should become a SEOS vs. H290C debate ad infinitum. In fact, the additional posts have wrapped things up nicely and clarified things for me . Thanks gentlemen.

I think we're all in agreement, and I value each of your comments. I think we got sideways for a moment, but only because we're all pretty passionate about these designs. They are much more alike than different.

wayne, in really like how you are handling this. respect!

Thanks guys. Here are notes on loudspeaker design, how to implement the H290C waveguide. I was thinking how it probably makes more sense to talk about its use in a loudspeaker system.

• Notes for the DIYer
• High-Fidelity Uniform-Directivity Loudspeakers

I've received a lot of emails and private messages over the last several weeks, with questions about common design issues. I put a lot of that in the "Notes for the DIYer" thread, but there is a lot of information that sort of "built up" to that thread but wasn't included. Things like how to perform certain measurements, how to calculate the volume displaced by a horn, driver or port and why baffle step compensation doesn't make sense for a large speaker designed for constant directivity. So I've made a FAQ page that has a lot more information from the dozen or so years on the Pi Speakers forum:

• FAQ

Some Pi Forum members asked me to create this FAQ. I think most of the Econowave and SEOS enthusiasts are familiar with these things, because I see many of these concepts described on Econowave and SEOS related threads, often quoted from old Pi forum posts or if not quoted, written verbatim. So that's good, I think, we have a dissemination of the information. But instead of having it scattered here and there, it might be helpful to have it all in one FAQ.

Tremendous resource. Bookmarked.

JSS

Definitely! Great work and thanks a ton, Wayne.

I've noticed several waveguide/horns with midband ripple that was equalized using response-shaping tank circuits in the crossover.

In the past, I used to rarely see this practice used on compression horns, except as a peaking circuit for the extreme high-frequencies where compression drivers just had no output. Top-octave compensation was first-order to conjugate mass-rolloff, but sometimes manufacturers added an additional pole around 12kHz - 15kHz to counter the extra rolloff from voice-coil inductance. I don't know of anyone that tried to counter the final pole from the front chamber, which doesn't usually set in until around 18kHz or 20kHz anyway.

As for me, I simply avoided using compression drivers until they had evolved to the point where top-octave performance was good and no peaking circuits were required. My crossovers use a simple first-order slope to counter mass rolloff. It's pure and simple.

In fact, Pi Speakers have never used tank circuits for response shaping. It is nearly a philosophical choice for me, and the reasons are described in various documents. But to be brief, I tend to try and find acoustic solutions to acoustical problems rather than to equalize on-axis response, which often adversely affects off-axis response.

Only if the response is the same in all axes will "voicing" be right in all axes. If a designer uses tank circuits to mitigate ripple caused by directivity changes, then they are using electrical equalization to smooth on-axis response, but the end result is that power response is made worse.

One thing that quadratic waveguides and catenary (OS, PS or EC) waveguides have in common is that they are very nearly conical horns. These tend to be peaky if made too small. This is both because of internal quarter-wave modes and because of directivity shifts. Both are problematic, in my opinion.

If the horn is so short it is suffering from 1/4λ peaking in its midband, then power response is varying as the horn is more efficient at modal frequencies than in between modes. So equalization can fix both on-axis and off-axis SPL, but the problem is the acoustic center is moving back and forth too. The movement of the acoustic center makes the forward lobe oscillate with the movement of the apparent sound source. Naturally, off-axis response suffers.

Alternatively, if the ripple is caused by directivity shifts, then what is really happening is the horn is getting louder on-axis when beamwidth narrows and quieter on-axis when beamwidth widens. Conversely, it is getting quieter off-axis when beamwidth narrows and louder off-axis when it widens. These two are equal and opposite because power response is flat. So if response shaping is used to equalize on-axis response, then power response is modified and off-axis response suffers.

Likewise, I'm not sure I would want to incorporate response shaping for the ripple caused by the baffle edge, for the same reason. It's a directivity shift, so equalizing on-axis flat makes power response non-flat. I'd rather have flat power response in a speaker designed for uniform-directivity. That is the point, after all.

For these reasons, I question the efficacy of response shaping in waveguide crossover circuits, beyond the simple first-order compensation for mass-rolloff. It may seem attractive to smooth response ripple in some devices, but I think if response ripple is excessive enough to require tank circuits, the waveguide itself is probably not worthwhile.

Interesting post. I'm not sure what prompted you to post it, but thank you and I have two question.

I don't have a good example, but perhaps this will sorta work:



That's raw response, a single small series cap response, and just off axis. Ignore the off axis for now. The before and after applying a filter. Is this what you mean by the above quote?

Ok I follow you here. What if the waveguide has uniform off axis response. As in, it's shape matches on an off axis. So then flattening the on axis would flatten the off axis. Shaping the on axis frequency response would not be an issue then, in fact, desireable ?

There are numerous threads about the R1/R2/C1 network I use in my crossovers to conjugate mass-rolloff. Here's one:

• Tweeter circuits for constant directivity horns and waveguides

You'll notice in the post above that the basic curve for a compression driver is flat to around 4kHz, then 6dB/octave rolloff over the next two octaves for a total of 12dB. The R1/R2/C1 network in the Pi crossover provides a conjugate for that. If the driver has excessive breakup, then it might have too much peaking up high. Or if there are 1/4λ modes from the horn/waveguide, then the R1/R2/C1 tweeter circuit cannot compensate for those either. But then again, would you really want to use such devices?

Which brings me to your second question. If a sound source has the same spectral balance off-axis as it has on-axis, then it provides constant directivity. Where there is constant directivity, equalization on-axis has the same affect off-axis. But where you see ripple - and the ripple isn't caused by the driver - that leaves the cabinet and/or horn. And don't forget that even if the waveguide alone provides constant directivity, when paired with a midwoofer, it won't have contant directivity any longer because of the complex interaction between the two sound sources.

Ripple from the cabinet or the horn is almost always caused by some sort of directivity fluxuation or a resonance which moves the acoustic center. In either case, electronic response shaping adversely affects the power response for the reasons I mentioned in my last post, above. Here's a write-up on one such cause of directivity ripple that is sometimes dealt with using response shaping, but that I think is ill-advised in a DI-matched two-way loudspeaker:

• Baffle Step

I mentioned this here because I was getting a lot of emails asking about complex crossovers in DI-matched two-ways. My crossovers have taken on a pretty well-defined form, and I've written a ton of papers about them, but still, I found myself replying to a lot of emails on this subject.

These topics and many more are explored in the Pi Speakers FAQ.