Stepped Baffle Study
Copyright 2010 Troels Gravesen
       

Making 1st order Butterworth or 2nd LR crossovers often requires sloped or stepped front panels in order to provide proper frequency and phase integration, thus this study on stepped baffles to learn the impact on tweeter performance from nearby added baffles. How close can they be to the tweeter, how thick can we make the panels, should they be chamfered or not and can we improve performance from making them round or pointed as can be seen from the drawing below?

The overall aim is to make speakers with reduced time and phase distortion and the results from this study will hopefully be integrated in future designs. Often we see speakers claiming use of 1st order filters and perfect transient response made from 3-4 driver units as the drivers must provide a huge overlap to make a reasonably perfect 1st order Butterworth roll-off well beyond points of crossover and very few drivers have an intrinsic response suitable for this. A minor discussion on these topics can be found in "Siri's Killer Note" article.

Two-way speakers with 1st order filters take serious high-performance drivers as the midbass preferably should have a frequency response up to 10 kHz and the tweeter must tolerate a lot of heat from the shallow sloped crossover and quite often we see a dedicated midrange driver inserted to relieve both drivers from doing what they may only be able to deliver at moderate levels. Going 3-way adds to the overall complexity and fortunately (good) moderns tweeters are more capable than we might think and the use of magnetic oil may come in handy in lowering resonance peak and cooling the voice coil.

Generally tweeters fitted with a waveguide is much less susceptible to nearby boundaries compared to a "naked" tweeter. As the waveguide determines the angle of projection, reflection from the front panel and baffle edge diffraction is reduced considerably. This was observed during experiments with the TW034 + JA8008 (DTQWT) where the TW034 + waveguide was recessed some 30 mm relative to the 8008 driver without noticeable impact on tweeter performance. The 8008 driver was mounted in a rounded panel as seen below where the edge was partly chamfered 45 deg.


Sheets piling up for initial study on straight baffles of 19-30 mm thickness, +/- chamfering. To the right an example of stepped panel design using 1. order XO (
Sequerra).


From left: Straight baffle. Middle: Rounded baffle. Right: Pointed baffle.
All forms will be tried from being cut 90 deg and chamfered 45 deg.


Starting point: Tweeter response on non-stepped baffle


Frequency response of SEAS T29CF002 tweeter on 210 mm wide baffle mounted 65 mm from top to center of dome.
This dome is 110 mm in diameter and for some of the measurements the baffle was covering part of the faceplate. Not quite ideal.


Straight 19 mm baffle


19 mm baffle thickness, cut 90 deg.
Distances from center of tweeter to edge of baffle: Red = 45 mm, blue = 55 mm, green = 65 mm, yellow = 75 mm.


19 mm baffle thickness, cut 45 deg.
Distances from center of tweeter to edge of baffle: Red = 45 mm, blue = 55 mm, green = 65 mm, yellow = 75 mm.


Straight 22 mm baffle


22 mm baffle thickness, cut 90 deg.
Distances from center of tweeter to edge of baffle: Red = 45 mm, blue = 55 mm, green = 65 mm, yellow = 75 mm.


19 mm baffle thickness, cut 45 deg.
Distances from center of tweeter to edge of baffle: Red = 45 mm, blue = 55 mm, green = 65 mm, yellow = 75 mm.


Straight 25 mm baffle


25 mm baffle thickness, cut 90 deg.
Distances from center of tweeter to edge of baffle: Red = 45 mm, blue = 55 mm, green = 65 mm, yellow = 75 mm.


19 mm baffle thickness, cut 45 deg.
Distances from center of tweeter to edge of baffle: Red = 45 mm, blue = 55 mm, green = 65 mm, yellow = 75 mm.


Straight 30 mm baffle


30 mm baffle thickness, cut 90 deg.
Distances from center of tweeter to edge of baffle: Red = 45 mm, blue = 55 mm, green = 65 mm, yellow = 75 mm.


19 mm baffle thickness, cut 45 deg.
Distances from center of tweeter to edge of baffle: Red = 45 mm, blue = 55 mm, green = 65 mm, yellow = 75 mm.


Measurements clustered for same distance to center of dome. Only 45 and 55 mm shown.


All 45 mm from center of dome.
Left: 45 deg chamfered baffle. 19-22-25-30 mm panel thickness (red, blue, green, yellow).
Right: Baffles cut 90 deg. 19-22-25-30 mm panel thickness (red, blue, green, yellow).
45 mm distance to center of dome provides the smoothest response of all regardless of chamfering.


All 55 mm from center of dome.
Left: 45 deg chamfered baffle, 19-22-25-30 mm panel thickness (red, blue, green, yellow).
Right: Baffles cut 90 deg. 19-22-25-30 mm panel thickness (red, blue, green, yellow).
55 mm distance to center of dome regardless of chamfering does not perform as well as 45 mm distance in the 2-4 kHz range.

Conclusion/straight baffles:
Thickness of additional panel for middriver has very little impact on performance.
Distance to center of dome plays a major role in getting a smooth frequency response.
Very little difference between panels cut 90 deg. or chamfered 45 deg.
45 mm to center of dome has the disadvantage of midbass panel covering part of tweeter front panel.

Next:
1. Midbass panels rounded, +/- 45 deg. chamfering.
2. Midbass panels pointed, +/- 45 deg. chamfering.

In the following measurements, only 22 and 30 mm baffles were tested as differences in performance vs. baffle thickness doesn't justify the huge numbers of measurements needed.


Rounded baffles


22 mm rounded baffle


22 mm baffle thickness, round, cut 90 deg.
Distances from center of tweeter to edge of baffle: Red = reference, blue = 45 mm, green = 55 mm, purple = 65 mm.


22 mm baffle thickness, round, cut 90 deg. Only 55 mm distance shown here against reference.
Distances from center of tweeter to edge of baffle: Red = reference, blue = 55 mm. Not bad at all!


22 mm baffle thickness, round, chamfered 45 deg.
Distances from center of tweeter to edge of baffle: Red = reference, blue = 45 mm, green = 55 mm, purple = 65 mm.


22 mm baffle thickness, round, chamfered 45 deg. Only 55 mm distance shown here against reference.
Distances from center of tweeter to edge of baffle: Red = reference, blue = 55 mm. Excellent!


30 mm rounded baffle


30 mm baffle thickness, round, cut 90 deg.
Distances from center of tweeter to edge of baffle: Red = reference, blue = 45 mm, green = 55 mm, purple = 65 mm.


30 mm baffle thickness, round, cut 90 deg. Only 55 mm distance shown here against reference.
Distances from center of tweeter to edge of baffle: Red = reference, blue = 55 mm. Very nice!


30 mm baffle thickness, round, chamfered 45 deg.
Distances from center of tweeter to edge of baffle: Red = reference, blue = 45 mm, green = 55 mm, purple = 65 mm.
45-55 mm distance looks excellent.


30 mm baffle thickness, round, chamfered 45 deg. Only 55 mm distance shown here against reference.
Distances from center of tweeter to edge of baffle: Red = reference, blue = 55 mm. Doesn't get much better than this!



The Pointed baffle


22 mm pointy baffle


22 mm baffle thickness, pointy, cut 90 deg.
Distances from center of tweeter to edge of baffle: Red = reference, blue = 45 mm, green = 55 mm, purple = 65 mm.
Some problems at 4 kHz.


22 mm baffle thickness, pointy, chamfered 45 deg.
Distances from center of tweeter to edge of baffle: Red = reference, blue = 25 mm, green = 35 mm, yellow = 45 mm, purple = 55 mm.
Not really much difference from these experiments. This panel can be placed anywhere.


30 mm pointed baffle


30 mm baffle thickness, pointy, cut 90 deg.
Distances from center of tweeter to edge of baffle: Red = reference, blue = 45 mm, green = 55 mm, purple = 65 mm.
Like 22 mm panel, some problems at 4 kHz.


30 mm baffle thickness, pointy, chamfered 45 deg.
Distances from center of tweeter to edge of baffle: Red = reference, blue = 25 mm, green = 35 mm, yellow = 45 mm, purple = 55 mm.
Not quite as nice as 22 mm panel, but appear useful at all distances.


Conclusion


Examples of panels used in study.

Although the pointed baffle, +/- chamfering, performs well, it will increase distance between tweeter and middriver and worsen crossover lobing, which is one of the major trade-offs from low-order filters, so unless listening distance is large, I wouldn't use this solution.

The rounded baffle, cut 90 deg. or chamfered 45 deg. as seen above, leaves only little footprint on tweeter performance and for a given tweeter we only need to find optimum distance to center of tweeter. What's shown here goes for the SEAS T29CF002 tweeter and other tweeters may perform differently. From own experience, tweeters fitted with a waveguide is much less susceptible to any nearby boundaries compared to the "naked" dome.

Crossover lobing is not symmetrical when we move across driver axes. Moving from a point between drivers towards the bass driver we often have tolerable lobing, where moving towards tweeter often leaves serious dips in frequency response. Thus, turning the speaker upside down, we have a better listening window over a larger vertical line. The old Dynaudio practice :-)

An interesting discussion can be found here: http://www.silcom.com/~aludwig/Time_alignment.html

To determine midbass panel height needed for our 1st order crossover is tricky business! Initial measurements has to be done at exactly e.g. 500 mm distance from tip of microphone to front panel plane (if drivers are flush mounted) and the acoustic distance to the drivers must be recorded from the impulse or step response. For the T29CF002, I measure 509 mm and for an e.g. AT 18H52 driver I measure 530 mm, thus we have a 21 mm difference which must be remembered when we model the crossover from dZ = 0 mm in the LspCAD program*. If we can model a 1st order crossover with proper phase integration from these settings , the midbass panel must be 21 mm thick (or actually 22 mm standard MDF). Plus/minus 1 mm doesn't ruin modelling but there are limits to how much we can offset drivers in the simulation program before it starts looking strange. Part of the developing work will be setting up the actual drivers with proper offset and make test crossovers to see if reality fits modelling. If the speaker also has to be tilted, this will further complicate modelling and fine-tuning of crossover.

*The acoustic difference between driver based on step response is not entirely correct as it is frequency dependent, but based on numerous construction with a point of crossover in the 2-4 kHz region it works fairly well and can easily be corrected during test crossover build up.
If we record the minimum phase of the two drivers we can see these aren't liniear, thus we have to look at minimum phase at the suggested point of crossover and take this into account. In LspCAD 6 we can apply recorded amplitude and minimum phase from individual drivers and drivers connected in parallel and exclude the recorded parallel recording in LspCAD's calculated summed response. We can then find the correct dZ when the recorded summed response coincides with the calculated summed response in the e.g. 1,000-5,000 Hz range.