A quick theory thrash out...
My understanding is that the width of the baffle determines the frequency at which it will additionally begin radiating to the rear - effectively moving from 2pi radiation to 4pi (?). This causes a sudden decrease of 6db in output at that frequency. By introducing a curved baffle edge, the change is less sudden.
In theory, the radius of the curve can be calculated to achieve a particular rate of descent in output. Has anyone done this? I think it might be plausible to model this in orders, the same as with XO's, and it would then be possible to compensate a n order baffle rolloff with a n order filter. Or am I just chattin' shite...
Also, I challenge the consensus that a rounded baffle edge is the best. Does it not seem intuitive that a straight slope would provide a more linear rate at which forward radiating frequencies meet the rear - and thus a smoother transition?
My understanding is that the width of the baffle determines the frequency at which it will additionally begin radiating to the rear - effectively moving from 2pi radiation to 4pi (?). This causes a sudden decrease of 6db in output at that frequency. By introducing a curved baffle edge, the change is less sudden.
In theory, the radius of the curve can be calculated to achieve a particular rate of descent in output. Has anyone done this? I think it might be plausible to model this in orders, the same as with XO's, and it would then be possible to compensate a n order baffle rolloff with a n order filter. Or am I just chattin' shite...
Also, I challenge the consensus that a rounded baffle edge is the best. Does it not seem intuitive that a straight slope would provide a more linear rate at which forward radiating frequencies meet the rear - and thus a smoother transition?
Way to go, Vikash. It would be great to have more on this issue.Vikash said:
My designs have gone from a straight slope (System IV) to a curved baffle (Nonsuch 4). I know that the curved design sounds considerably better than the sloped but there may be other reasons than baffle step for the improvement.
As best I can tell, most 'experts' are still quoting the limited experiments that were performed 50 years ago (was it Olsen?). You always see the same tired old graphs dusted off and taken out. If you've got the time to do the work, give us some new stuff, please.
Vikash
This topic isn't straightforward and I'm not attempting to give an authorative answer, but here's my understanding. Firstly terminology. The unit of solid angle measurement is the steradian and there are 4pi of them in a sphere surrounding a point in space just like there are 2pi radians of angle in a circle surrounding a point. When the baffle step frequency is reached the speaker goes from seeing just the space in front of it (2pi steradians) to seeing the whole room (4pi). I'm not sure having rounded edges helps in baffle step as you still reach a point where the sound reaches the sides and sees the whole room. Rounded edges help with imaging due to reducing diffraction effects at sharp edges. B-step correction usually involves first order low pass filters on the midbass, but is often less than 6db as the designer is thinking about room placement - how near to the wall etc. Some speakers are desgined with little correction and are more suitable for near wall placement, or it may be rejected for purist reasons.
Hope this helps
IJ
This topic isn't straightforward and I'm not attempting to give an authorative answer, but here's my understanding. Firstly terminology. The unit of solid angle measurement is the steradian and there are 4pi of them in a sphere surrounding a point in space just like there are 2pi radians of angle in a circle surrounding a point. When the baffle step frequency is reached the speaker goes from seeing just the space in front of it (2pi steradians) to seeing the whole room (4pi). I'm not sure having rounded edges helps in baffle step as you still reach a point where the sound reaches the sides and sees the whole room. Rounded edges help with imaging due to reducing diffraction effects at sharp edges. B-step correction usually involves first order low pass filters on the midbass, but is often less than 6db as the designer is thinking about room placement - how near to the wall etc. Some speakers are desgined with little correction and are more suitable for near wall placement, or it may be rejected for purist reasons.
Hope this helps
IJ
Since the subject has been opened, again, I've seen two different networks for baffle step. One is R1//L1 and another is R1//L1+R2. My first question is basic, why is the resistor there? Does it provide a stable load or is there another reason? Second, why do some use 2 resistors while others only one? Sorry if this is remedial but most of what I'm learning right now probably is.
There is a combing effect that is associated with the baffle step - peaks and troughs all the way up the spectrum. I believe that the rounded edges ameliorate this, provided the radius is sufficiently large. Again, I'm going on the Olsen graphs and subjective impressions.Ian J said:
I get a partial compensation for the diffraction step by using four drivers. The efficiency of this combination is increased over a single unit but only below a certain frequency and this compensates.
You can find a few articles in the JAES that explain and derive how to calculate it, but it's a lot easier to download a program that calculates it.
BDS is the one you are looking for:
www.pvconsultants.com/audio/frdgroup.htm
----
Regarding the two circuits: R||L passes low frequencies and attenuates higher. R||L+R would take the response of the previous filter and add more attenuation for all frequencies.
Neither is commonly used for baffle step in the industry TTBOMK. Usually it is better to cross woofers intentionally low by using an inductor that is larger than normal in the woofer circuit. There may be some situations, driver combinations and crossover frequencies where you do have to use an additional RL ckt.
BDS is the one you are looking for:
www.pvconsultants.com/audio/frdgroup.htm
----
Regarding the two circuits: R||L passes low frequencies and attenuates higher. R||L+R would take the response of the previous filter and add more attenuation for all frequencies.
Neither is commonly used for baffle step in the industry TTBOMK. Usually it is better to cross woofers intentionally low by using an inductor that is larger than normal in the woofer circuit. There may be some situations, driver combinations and crossover frequencies where you do have to use an additional RL ckt.
rounded edges are an eye candy...
Hello,
Rounded edges are there mostly to please aesthetically in effect.
Even to smooth out some edge related second source forming ripple in midhighs we need serious radious (eg Thiel).
For ironing out some ripple at mid-lowHF, recessing the tweeter and using golden ratio for its place on baffle regarding its centre from sides is 99% enough.
Remember that rectilinear cabinet examples have achieved great smoothness (Spendor, ATC etc).
Shapes like Avalon really smooth the step transition. A sphere is not recommended because it creates another problem. Strong internal standing wave modes.
If the woofer is not positioned symetrically on the baffle you ease up the step bump, so you can compensate with more accuracy.
As for the resistor across the coil it just helps to follow the step trend more closely by changing the falling rate effect that inductance creates. It makes for more 'wooly' mids though. +- 2dB ripple is not much harmful in a speaker anyway.
Regards
Salas
Hello,
Rounded edges are there mostly to please aesthetically in effect.
Even to smooth out some edge related second source forming ripple in midhighs we need serious radious (eg Thiel).
For ironing out some ripple at mid-lowHF, recessing the tweeter and using golden ratio for its place on baffle regarding its centre from sides is 99% enough.
Remember that rectilinear cabinet examples have achieved great smoothness (Spendor, ATC etc).
Shapes like Avalon really smooth the step transition. A sphere is not recommended because it creates another problem. Strong internal standing wave modes.
If the woofer is not positioned symetrically on the baffle you ease up the step bump, so you can compensate with more accuracy.
As for the resistor across the coil it just helps to follow the step trend more closely by changing the falling rate effect that inductance creates. It makes for more 'wooly' mids though. +- 2dB ripple is not much harmful in a speaker anyway.
Regards
Salas
the edge
Edge diffraction is a specific thing. It formulates secondary sources not in phase with the original source. Hence ripple, phasey sound. You need a heavy radius to achive anything worth speaking of. Thiels do it properly. Curves as heavy as the sidepanel thickness do nothing acoustically.
Baffle step has to do with longer wavelenghts comparable with the baffle dimensions. Tappered shapes like a flat top pyramid are good cause they present a constantly changing dimension. Many midbass drivers help also because they average out the step shape due to different positions on the baffle.
Another good thing is that with many woofers you get different path lengths to floor and you end up with a less sever floor cancellation in the 100-200Hz region.
Regards
Salas
Edge diffraction is a specific thing. It formulates secondary sources not in phase with the original source. Hence ripple, phasey sound. You need a heavy radius to achive anything worth speaking of. Thiels do it properly. Curves as heavy as the sidepanel thickness do nothing acoustically.
Baffle step has to do with longer wavelenghts comparable with the baffle dimensions. Tappered shapes like a flat top pyramid are good cause they present a constantly changing dimension. Many midbass drivers help also because they average out the step shape due to different positions on the baffle.
Another good thing is that with many woofers you get different path lengths to floor and you end up with a less sever floor cancellation in the 100-200Hz region.
Regards
Salas
Re: the edge
If we take a single, full-range driver (to simplify things), the step would be nice and smooth if the driver was mounted on a sphere (larger than 4" radius), 3D ellipse or egg. It would still be there, however. The worse case would be where the driver was mounted in the middle of a circular baffle so the distance to the edge would be the same all round. In this case there would be the 6dB baffle step as always but this wouldn't be a smooth step and there would also be large peaks and troughs in the response curve.
In a room, the baffle step is compensated for at the lower frequencies by the walls. When using a compensation circuit, 6dB compensation would therefore be too much and 3dB may be preferred.
Many manufacturers compensate for the baffle step by putting the crossover point at the beginning of the step down and balancing the sound from the tweeter and woofer by increasing the woofer's relative SPL. This will not fix the comb effects, however. Sometimes it's amusing to look at manufacturers' published response curves and to realize that what they publish cannot be true unless they're using a graph recorder set to supersonic speed.
It's true that diffraction does not just apply to the baffle step response. However, the step response (6dB in the open air) is caused by diffraction. It is due to sound waves diffracting around the enclosure walls as the wavelength of the sound increases relative to the baffle dimensions. The larger the distance from the driver to the edge of the baffle, the lower the frequency of the step.salas said:
If we take a single, full-range driver (to simplify things), the step would be nice and smooth if the driver was mounted on a sphere (larger than 4" radius), 3D ellipse or egg. It would still be there, however. The worse case would be where the driver was mounted in the middle of a circular baffle so the distance to the edge would be the same all round. In this case there would be the 6dB baffle step as always but this wouldn't be a smooth step and there would also be large peaks and troughs in the response curve.
In a room, the baffle step is compensated for at the lower frequencies by the walls. When using a compensation circuit, 6dB compensation would therefore be too much and 3dB may be preferred.
Many manufacturers compensate for the baffle step by putting the crossover point at the beginning of the step down and balancing the sound from the tweeter and woofer by increasing the woofer's relative SPL. This will not fix the comb effects, however. Sometimes it's amusing to look at manufacturers' published response curves and to realize that what they publish cannot be true unless they're using a graph recorder set to supersonic speed.
monotonic
Some engineers, notably the people in Scanspeak create such IEC baffle response curves for their midbass drivers that when on a baffle in aroom they exibit a monotonic rise effect. So only a proper inductance is enough to get a flat response.
I would prefer 4dB correction in real room from 200Hz upwards. 3dB sound a bit anemic to my ears.
Yes diffraction is a general term. That is why I used 'edge diffraction' for answering about curving box sides in small amounts. The original term is 'perithlasis' it is ancient Greek and means 'flexing around'. This is the right term and the cause of the step.
It is wrongly used for discribing formation of secondary sources.
The right term is 'interference'. And this is all about ripple in the mid-highs and recessing tweeters, using porous surfaces around them etc.
Some engineers, notably the people in Scanspeak create such IEC baffle response curves for their midbass drivers that when on a baffle in aroom they exibit a monotonic rise effect. So only a proper inductance is enough to get a flat response.
I would prefer 4dB correction in real room from 200Hz upwards. 3dB sound a bit anemic to my ears.
Yes diffraction is a general term. That is why I used 'edge diffraction' for answering about curving box sides in small amounts. The original term is 'perithlasis' it is ancient Greek and means 'flexing around'. This is the right term and the cause of the step.
It is wrongly used for discribing formation of secondary sources.
The right term is 'interference'. And this is all about ripple in the mid-highs and recessing tweeters, using porous surfaces around them etc.
Some thoughts:
a) If a driver is mounted on the flat end of a cylinder, and the baffle diameter is the same as that of the driver, then the baffle step would be eliminated (room issues aside), as would baffle edge diffraction?
b) As a frequency increases, it becomes directional, which implies a step response independent of baffle?
a) If a driver is mounted on the flat end of a cylinder, and the baffle diameter is the same as that of the driver, then the baffle step would be eliminated (room issues aside), as would baffle edge diffraction?
b) As a frequency increases, it becomes directional, which implies a step response independent of baffle?
ok regarding point (a), off course I just realised that edge diffraction would occur straight away, at the driver edge....duh! but the baffle step point remains.
Which makes me wonder, if you have an array in an infinite baffle, say in wall, would it still require some type of compensation?
Is this independent or part of the same thing?Vikash said:
No. Baffle step compensation would not be required. The sound stage and imaging could be compromised however.Timn8ter said:
Part of same thing. i.e. on the basis that a cylindrical enclosure with effectively no baffle will not exhibit a baffle step response. So allthough all frequencies now radiate into full space, there may still be a step response due to higher frequencies radiating forward only.
In an infinite baffle, there is no point where field radiation is doubled, thus no baffle step as you mention Steve. But I think Timn8ter's question is better identified with the above no-baffle enclosure.
What dya think?
Cylindrical enclosure has bafflestep in spades. A cylinder with the driver on the center will have really bad ripples - the worst, in fact. Even if the cylinder is the same size as the driver, it is still there. Note that Olson's plots are for a very small driver, and larger drivers have less ripples at high frequencies due to directivity issues. All of these things can be confirmes by the software I mentioned.
Since you guys seem to want to "talk about it" rather than calculate it, here is a quick explanation.
The "step frequency" is related to the size of the baffle and the "ripples" are related to the distance from the center of the diaphragm to different parts of the baffle edges.
Both are the same effect, they are not separate.
There are several articles in the JAES:
Bews-Hawksford
Backman
Vanderkooy
Check 'em out.
Since you guys seem to want to "talk about it" rather than calculate it, here is a quick explanation.
The "step frequency" is related to the size of the baffle and the "ripples" are related to the distance from the center of the diaphragm to different parts of the baffle edges.
Both are the same effect, they are not separate.
There are several articles in the JAES:
Bews-Hawksford
Backman
Vanderkooy
Check 'em out.
Well talking about it is important. I'm after an understanding to the causes so that we can better engineer our enclosures at design time.
Now, off course it is often mentioned that a cylindrical baffle is the worst culprit due to all edges being equidistant from the driver. But there is interesting simplicity to the cause, and uniformity in the result, and I think this gives us a lot of insight to the way forward.
The scare of a cylindical baffle is misunderstood by only looking at its effects in a frequency response. The paper Understanding Cabinet Edge Diffraction - Andy Unruh is the most informative paper I've read on the topic yet (If anyone has the J Vanderkoy paper, I would love to read it).
I think we plonk a driver into a rectangular baffle, sometimes using the golden ratios for placement, with too little thought to alternative baffle shapes. By seeing that the equidistant from driver to baffle edge in the cylindical enclosure causes a transition to full space at the same point in time, and that the resulting ripples are a function of this, I think we can find better flat shapes (better than the common rectange) that will provide a more linear increase rate from driver to baffle-edge around the baffle, and thus a smoother baffle step response.
Now, off course it is often mentioned that a cylindrical baffle is the worst culprit due to all edges being equidistant from the driver. But there is interesting simplicity to the cause, and uniformity in the result, and I think this gives us a lot of insight to the way forward.
The scare of a cylindical baffle is misunderstood by only looking at its effects in a frequency response. The paper Understanding Cabinet Edge Diffraction - Andy Unruh is the most informative paper I've read on the topic yet (If anyone has the J Vanderkoy paper, I would love to read it).
I think we plonk a driver into a rectangular baffle, sometimes using the golden ratios for placement, with too little thought to alternative baffle shapes. By seeing that the equidistant from driver to baffle edge in the cylindical enclosure causes a transition to full space at the same point in time, and that the resulting ripples are a function of this, I think we can find better flat shapes (better than the common rectange) that will provide a more linear increase rate from driver to baffle-edge around the baffle, and thus a smoother baffle step response.
By how much, and why - considering the subjective reasoning I gave in an earlier post? I will try to get hold of the papers you mentioned.
Talking about it with people who probably know very little more about it than you do is pointless.
You will get a better understanding by learning what it is from an authoritative source, then calculating it and getting a feel for how changes affect the response.
If you want to learn, fine, but calculating it with a ready made program is the short cut. You can run hundreds of virtual experiments.
You will get a better understanding by learning what it is from an authoritative source, then calculating it and getting a feel for how changes affect the response.
If you want to learn, fine, but calculating it with a ready made program is the short cut. You can run hundreds of virtual experiments.
- Status
- Not open for further replies.