Some things I learned from baffle diffraction simulations

[JonW]

I'm designing my first speaker, a simple 2 way MT, and I’m currently looking into various cabinet styles. How might one best control baffle diffraction effects? There are two good free programs out there: Edge and Baffle Diffraction Simulator (BDS). I’ve been trying out various baffle designs to see what looks good. I’m new to all this so I’ll post my findings in hopes that it’s useful to some other newbies out there. For those who are experts, I suspect there is nothing new here (if I have everything correct). Of course, I don’t claim to know what I’m doing, YMMV, etc.

---

Effects of baffle diffraction on the speaker’s apparent output:
If we take a regular floor standing 2 way MT speaker with sharp edges, we see a diffraction pattern like this:

Image not available

Ugly, eh? Lots of variability, ups and downs. It might not sound so great with certain frequencies boosted, others far down relative to the boosted ones.

Change the shape of the cabinet, let’s say we use 5 sides rather than 4, adding chamfered (or faceted) edges, and offsetting the drivers from the center line will have a cabinet looking like this:

Image not available

It’s ugly, but we can fix that part later (see below). But look at how much smoother the baffle diffraction effects are:

Image not available

Smooooth. Almost soothing, when compared to the 4 sided sharp edged cabinet.

---

The software:
Of the two programs, Edge and BDS, they each seem to have strengths and weaknesses- near as I can tell from not much experience using either program. Edge is easier to use and quicker. It has more variability in the baffle shapes you can use. But there does not seem to be a way to change the edge treatments. BDS takes longer to manipulate and looks to be more precise, with the ability to change numbers exactly, not just by dragging thigns with the mouse. BDS has fewer shape choices, but does have the ability to treat the edges with chafers and roundovers.

---

Some things I learned from the simulations:
I was trying to find situations with smoother plots and minimal deviation from peaks to troughs.
-A wider baffle with a chamfer (facet) is better than a narrow baffle with sharp edges
-A chamfered edge is better than a rounded over edge (I’m still trying to understand this one fully)
-A rounded over edge is better than a sharp edge
-A 1.5” chamfer is better than a 0.75” chamfer
-A straight baffle is better than one that is slanted 5 degrees, which is better than 10 degrees
-A 5 sided (like shown at top) or 6 sided (like the remaining front after Jon cut facets in his M8ta’s) baffle is better than a rectangular 4 sided baffle. (Note the typical tweeter at top, woofer below configuration)
-There seems to be not much difference between 5 versus 6 sided
-Offset drivers help a fair amount versus centering them in the middle of the baffle- according to BDS. Edge seems to show less of a difference.
-Any of these indiviual changes seem to make small differences. The two different plots shown above are the two extreme cases I found- from the ugliest to the prettiest plots. So getting smooth plots seems to be a combination of many small effects.

---

Conclusions- What I will incorporate into my cabinets:
From what I found above, it looks like I’ll be using a wide baffle (due to 1.5” double thickness to minimize resonances/vibrations) with multiple (3 or 4) layers of BB plywood on the front baffle. In a rectuangluar box. But cut the 1.5" facets out to make the desired 6 sided part of the baffle. The 6 sides will be prettier than the 5 in the figure above. I’ll also offset the drivers from the center.

---

A question:
Sharp edges are bad for baffle edge diffraction. Chamfered edges are better, but they still have flat edges. So I would think the best thing to do is chamfer the edges. And then sand all the sharp facet edges smooth. Almost a gentle roundover of the chamfered edges Does that sound like a good idea?

---

So there you go. A new guy trying to learn about cabinet design. I hope it’s helpful to someone.

-Jon

[Black300zx]

Interesting. The next step for the cabinets I'm building is the front baffle (it will be removeable). I was planning on just rounding all 4 edges over w/ a 3/4" bit, but I think I'll have to play around with different treatments before I finish it.

[cjd]

Well, you're lumping some discoveries together here... I'm able to work at the moment though, so I'm going to take advantage of things working while I can and get back to you on this.

C

[TacoD]

Wow thanks!

I think a wide baffle (like sonus faber) is very effective, then you do not have to worry over the edges which still exists after chamfering (is that a word??). The extreme wide baffle ensures that the highs are less affected. (or am I missing a point here).

[chasw98]

Jon:
Very interesting. What did you use to get the frequency response plots of the driver(s)? Did you just take the measurements of 2 known drivers? Does this take into account a basic 2 way crossover or something else? (yeah, yeah, lots of fellow noob questions!). I am curious what the simulation would sound like in the real world if you made up some test cabinets and A/B'd the 2 cabinet designs. It seems as if you are using that razor sharp scientifically logical mind of yours for fun things now!

Chuck

[Paul W]

If you haven't already tried it, you might look at response at various angles off-axis. I use both, but generally find large roundovers to be more consistent than the stealthy facets.

[cjd]

Remember that diffraction is a function of distance (from driver to edge). This plays into two observations - one, that the odd shapes improve things, and two that a chamfer improves things. The odd shape varies this distance more randomly. And a chamfer spreads it out at an edge. A round-over on-axis IS going to be worse because there is still a greater general concentration of "edge" to at the start of the curve. As Paul has mentioned, it tends to be somewhat less consistent as you move off-axis because it's not the smooth transition.

Wide makes sense - it pushes the problem frequency down.

So, you're getting a sense of what you can do to ameliorate the effects of diffraction (and step). Now, you start to bring that into the realm of things you want to do. Please note the SCALE of those troublesome plots. Still within that standard "+/- 3dB" eh? Ouch!

Somewhere on page 15 (I think) of the RS WWMTM thread you can see some measurements where I played with heavy wool treatment at the tweeter. Another way of solving some of these diffraction troubles. dlr also has SUPERB documentation on dealing with diffraction on his site.

Also, remember that the diffraction only matters when the driver PLAYS that frequency. So, you know, that 15" woofer showing bad things up at 8kHz? Yeah, not a problem.

C

[JonMarsh]

Wide has other drawbacks, though- it moves the 2 pi frequency lower down, which means there's more of a power response discrepancy.

Better, in my opinion, with a monopole, to operate as close to monopole (if it's a free standing speaker), in most ranges. This is why you see a progressively narrowing front panel in Avalon designs (which the baffle layout above sort of mimics). Symmetrical layouts are not so good either, if you want to distribute the frequency range of othe effects- a weak response over a wider frequency range is probably less audible than a strong response in localized in some specific ranges.

Just some informed speculation. :W

[dlneubec]

I'd like to have someone take a look at my situation and perhaps offer some advice. Below are gated measurments on axis of my Mentor omnidirectional project. These are left and right speaker measurments, taken at tweeter height, from 1m from the front of the speaker plane (45 7/8" from the center of the upfiring drivers and tweeter baffle). There is some room reflection included here because I used an 8ms gated window time. The plot also includes a semi-nearfield of the RS180's, taken centered between the two woofer baffles, at the front speaker plane with a 200ms window, IIRC, and spliced into the farfield at around 300hz. The crossover between the two RS180's and RS28 is about 1450hz or so.

Notice the dip at around 3khz. I have done a lot of adjusting of the tweeter baffle size and shape, it's height, and with felt around the tweeter baffle and on the top and bottom box outside edges. You can see in the photo below the combination I found that worked best to decrease the width and depth of that dip. However, this was seat of the pants changes and then measurements, and then repeat, etc. I did not do any modeling in Edge or BDS. With the unusual layout of the drivers, I had no idea how to simulate it anyway!

So, with these givens, does anyone have any suggestions as to how to improve upon my efforts to reduce that dip? A different tweeter baffle configuration (I've probably built a good 7-8 of them so far)? A different use of felt? Or, should I just quit tweaking and say good enough?

Early in the design I did some off axis measuring and the response looked very good, however, I have made a lot of crossover design changes since then and tweeter baffle changes as well, so I really need to go back and get a set of off axis measurements to compare these to.

Alternatively, does anyone have a suggestion as to how to model something like this in Edge or BDS?

[cjd]

Dan, don't have the tweeter mounted on a nice perfect little rectangle, exactly centered...

Might be an interesting use of a dipole Neo3.

C

[dlneubec]

Dan, don't have the tweeter mounted on a nice perfect little rectangle, exactly centered...

C

Thanks.

Early on I tried some circular baffles. One had the tweeter centered and the other had it offset. The offset one measured flatter. :T

I'll try a taller baffle shape with the tweeter offset top to bottom. Left to right offset is a bit harder to pull off with the way I'm mounting it, unless I don't do the chamfer on one side. I wonder if a trapezoid shaped baffle that is wider at the bottom and smaller at the top would help?

I'll try modeling just the tweeter baffle in Edge or BDS, though I have an idea that it has a lot to do with the tweeter interaction with the woofer baffles above and below it. Maybe even the woofer cavities themselves are having an effect.

Might be an interesting use of a dipole Neo3.

C

I don't know anything about the Neo3. Can it cross as low as the RS28? I've actually thought about a rear firing tweeter as in the new Orion ++.

[cjd]

Depends on output levels I think, how low the Neo3 can go. Zaph has some numbers on it.

C

[JonW]

If you haven't already tried it, you might look at response at various angles off-axis. I use both, but generally find large roundovers to be more consistent than the stealthy facets.

Excellent point- thanks. I was only looking on axis. I’ll try off axis soon. And finding roundovers more consistent off axis makes sense. Consistent with what Chris said:

And a chamfer spreads it out at an edge. A round-over on-axis IS going to be worse because there is still a greater general concentration of "edge" to at the start of the curve. As Paul has mentioned, it tends to be somewhat less consistent as you move off-axis because it's not the smooth transition.

Makes perfect sense.

Which is why I was thinking about trying a hybrid between chamfered and rounded over edges. Basically, cut out the facets. Then sand all those sharp edges to be smooth. So no sharp edges at all. Seems to me I’d get the best of both worlds. Can’t simulate it to see, though. And I’ve not really seen many examples of speakers doing that. (The Thiel commercial speakers almost have it). The only down side I see is that the irregular 3D shape of the baffle will not allow veneering. But that’s only an aesthetic concern. So does cut-the-chamfers and smooth-the-edges sound like a wise approach?

[JonW]

Symmetrical layouts are not so good either, if you want to distribute the frequency range of othe effects- a weak response over a wider frequency range is probably less audible than a strong response in localized in some specific ranges.

… odd shapes improve things, and two that a chamfer improves things. The odd shape varies this distance more randomly. And a chamfer spreads it out at an edge.

You guys are saying exactly what I suspected prior to starting up the programs. Spread things out so no one frequency shouts at you. Nothing about these simulations is surprising. But it was a good exercise for me to do, to make sure what I would have guessed seems to be the case. Along those lines, I’m thinking (but have no sims to back it up) that facets that are sanded to be smooth, with no sharp edges, could possibly be an excellent way to go.

[JonW]

Please note the SCALE of those troublesome plots. Still within that standard "+/- 3dB" eh? Ouch!

Sorry- I’m not quite sure what you’re saying here. Is it that +/- 3 db is generally considered to be “good,” but looking at my first plot it’s actually pretty poor?

Somewhere on page 15 (I think) of the RS WWMTM thread you can see some measurements where I played with heavy wool treatment at the tweeter. Another way of solving some of these diffraction troubles. dlr also has SUPERB documentation on dealing with diffraction on his site.

Got it- page 16, post 532. I didn’t quite understand what was “better” or “worse” in terms of response vs. sound quality vs. impedance vs. …
Yeah, I’ll definitely play with felt once the speakers are actually built. But at this stage I figure I might as well try and design a cabinet to minimize diffraction effects as much as possible. And then add felt. And then, if needed, deal with it in the crossover.

[JonW]

Dan-

I can't help much, but your project looks really super. :T
The only thing that comes to mind is what Chris said- don't have your tweeter in the middle of a square like that. Make it more irregular. Maybe cut facets out on the 2 top corners? Round over the other 2? I'd think a round baffle with the tweeter in the middle would be really bad- then a given frequency would hit the edge of the baffle all around, thereby creating a really boosted signal at that one frequency. So offsetting, irregular shapes, etc. will spread things out. So, like was mentioned above, you get a little bump across many frequencies rather than a big bump at one.

Jon:
Very interesting. What did you use to get the frequency response plots of the driver(s)? Did you just take the measurements of 2 known drivers? Does this take into account a basic 2 way crossover or something else? (yeah, yeah, lots of fellow noob questions!). I am curious what the simulation would sound like in the real world if you made up some test cabinets and A/B'd the 2 cabinet designs. It seems as if you are using that razor sharp scientifically logical mind of yours for fun things now!

Chuck

Hey Chuck,

The programs are great, but they’re not THAT sophisticated. You just use a standard group of frequencies. Punch in how large of a driver(s) you’re using. That’s it, basically. Then play with the shapes.

As far as comparing the real world sound versus the simulations… I totally agree it would be interesting to try that out. But I’ve only got so much free time. So I’ll do this homework now and take my bst guess as to what will sound best. Then build that. I can email down to Florida the plans for the 100 other baffles I’d like to play with. And then the 100 different drivers to try out.

[chasw98]

I can email down to Florida the plans for the 100 other baffles I’d like to play with. And then the 100 different drivers to try out.

No problem, I will fab them up for you and get them sent back to the 'lab' for results. P.S. make sure you give me your FedEx account number to use. :B

Chuck

[JonW]

P.S. make sure you give me your FedEx account number to use.

Here, I found my FedEx account number:
337CHUCK561DRIVES966HIS285ASS372TO823INDIANA441
:

[Paul W]

So does cut-the-chamfers and smooth-the-edges sound like a wise approach?

A roundover of equal size may be slightly better, but is usually limited by bit size. A 1.25" roundover in a handheld router can be a very good workout :twisted:

Cut and smooth is a good compromise, particularly when you have room for a larger chamfer. A belt sander is very quick on MDF edges, a plane perhaps even faster (haven't tried a plane).

You can buy 3" radius quarters from Rockler or 6" quarters from Anderson International.

[JonW]

Hi Paul,

Thanks for the input.

A roundover of equal size may be slightly better, but is usually limited by bit size. A 1.25" roundover in a handheld router can be a very good workout :twisted:

Cut and smooth is a good compromise, particularly when you have room for a larger chamfer.

I’ll have room for more than a 1.25” chamfer. I’m thinking the front baffle will be 3 or 4 layers of 3/4” thick BB ply. The rest of the cabinet will likely be doubles layers (2 x 3/4”). That will allow some wide facets to be sliced out. So I can cut facets larger than I can round over.

And I must admit that a 1.25” roundover bit scares me a tad.

A belt sander is very quick on MDF edges, a plane perhaps even faster (haven't tried a plane).

I was just thinking about sanding things down by hand. So I can watch the shape change slowly. And maybe only ‘take the edge off’ rather than try and have, effectively, a full 2” smooth roundover. But I’m open to suggestions.

You can buy 3" radius quarters from Rockler or 6" quarters from Anderson International.

I didn’t know you can buy such large quarters. Thanks.

[cjd]

Sorry- I’m not quite sure what you’re saying here. Is it that +/- 3 db is generally considered to be “good,” but looking at my first plot it’s actually pretty poor?

I'm saying, you looked at it and thought it poor. Yet it's within the limits the industry makes us think of as "good"

Got it- page 16, post 532. I didn’t quite understand what was “better” or “worse” in terms of response vs. sound quality vs. impedance vs. …

The response changes on-axis are almost all due to the felt, including the amelioration of the dip in response.

C

[---k---]

And I must admit that a 1.25” roundover bit scares me a tad.

.

Sissy

Images not available

:P

Actually, a 1 1/4" in a handheld router is a little scary. I ignored those fears while working and just live with feeling stupid and glad that nothing happened.

Making a big chamfer and then running a round over over it to smooth the edge shouldn't be a problem. If you get the experiment with the height and radius, you might find something that transitions well. Just take the bearing off and do it in a table with a jig to give you the offset. I would probably do it this way so that it is uniform from top to bottom. That or just take a screwdriver to flatten the edge. I have a feeling you'll get real pissed off at yourself going back and forth sanding off a touch more here and there trying to make sure your uniform from top to bottom.

[JonW]

I'm saying, you looked at it and thought it poor. Yet it's within the limits the industry makes us think of as "good"


The response changes on-axis are almost all due to the felt, including the amelioration of the dip in response.

C

OK, got it on both accounts. Thanks. It doesn't take a genius speaker designer to not like that +/- 3 db plot. And I'll definitely play with felt when I have something built to put the felt on. I do wonder, for a faceted baffle, if there will be any benefit to having felt on the actual facet, not just the baffle front. But I guess I'll get to try that for myself.

[JonW]

Actually, a 1 1/4" in a handheld router is a little scary. I ignored those fears while working and just live with feeling stupid and glad that nothing happened.

I'm glad you can still count to 10 without having to take off your shoes. :T

Making a big chamfer and then running a round over over it to smooth the edge shouldn't be a problem. If you get the experiment with the height and radius, you might find something that transitions well. Just take the bearing off and do it in a table with a jig to give you the offset. I would probably do it this way so that it is uniform from top to bottom. That or just take a screwdriver to flatten the edge. I have a feeling you'll get real pissed off at yourself going back and forth sanding off a touch more here and there trying to make sure your uniform from top to bottom.

I wasn't thinking about anything as complicated as that. More something like in this quick picture I just made:

Image not available

So just cut the facet and then sand down the sharp transition from baffle front to facet surface. Do the same for the two side facets as well. Easy to do by hand, slowly, so I can watch the progress. And not acres of wood to take down. At the moment, it seems like a good idea to me. But they're all not winners, so someone tell me if it's not a good idea.

[chasw98]

And I must admit that a 1.25” roundover bit scares me a tad.

One of the best and possibly safest ways to use those large bits is to mount your router in a table and then hold the piece to be routed. Build a table or use a tabletop and mount the baseplate of your router in it, then just rotate the router with a bit up through the tabletop for the opening. You saw mine when you were here, not pretty, but solid and safe.

There is a package on the way to you via FedEx. Let me know when you receive it. I can't believe you gave me your account number in the open on the internet! :rofl:

Chuck

[cjd]

Ugh. I can't imagine trying to use a router table with some of the speakers I've built! I'd need a hydraulic arm to manipulate it.

[dlneubec]

Image not available

-Jon

Hi Jon,

I'm just starting to try some simulations in BDS and can't figure out how to get a response plot. I get nothing in the graphs below the driver and baffle info entry area, though I do get a plot in the graph on the top left.

Any hints?

[cjd]

You have all the right toolpacks installed? You have to hit a button to get it to run the numbers, if I remember right. Pretty sure Roman B has some nice info on using BDS on his website too.

[ch83575]

I have also been investigating baffle diffraction for my first design (a simple 2-way that has taken over a year so far) and I have noticed a few things that might be worth mentioning here.

First, I think the driver geometry has a lot to do with the diffraction. I recently switched the tweeter from the SEAS 27TBFC/G to an Usher 9930 and the diffraction effects changed drastically! Here is a measurement of the SEAS in my box (about 7" wide):

Image not available

I do not have a similar test of the Usher, but you will see in the picture below that the dip at 3k remains, but all of the other ripple (peak at 5k and dip at 8k) are not really present with the Usher when installed on the same baffle! The only conclusion I can reach is that the different dome profile and driver geometry (like how the Usher sticks out the front so much) are what are making the difference. I suppose the VC is 1mm bigger on the Usher, but I cant really imagine that making this kind of difference.


The other thing I have been investigating is the off-axis effects of baffle diffraction. Here are graphs of the Usher installed in the same speaker, normalized on the 0deg and not-normalized:

Image not available

Image not available

As you can see the effects from the diffraction really do disappear off-axis, but the window is not exactly narrow. You really need to be like 30deg off axis before they totally disappear. It is kind of a toss-up, because if you design your speakers to be listened to off-axis you miss out on a lot of the top end, but if you listen on-axis you have to deal with the diffraction being a major part of the speakers sound. I guess this is why many manufacturers that design their speakers for off-axis listening use a super-tweeter to extend the high end even off-axis.

-Chad

[JonW]

One of the best and possibly safest ways to use those large bits is to mount your router in a table and then hold the piece to be routed. Build a table or use a tabletop and mount the baseplate of your router in it, then just rotate the router with a bit up through the tabletop for the opening. You saw mine when you were here, not pretty, but solid and safe.

A router table is something I need to get (or build) some day. And another router. But for this project, I’m thinking about doing what is in the picture of post #24. And that won’t need a router table.

There is a package on the way to you via FedEx. Let me know when you receive it. I can't believe you gave me your account number in the open on the internet! :rofl:

Hey, I’m a very trusting guy. :

Ugh. I can't imagine trying to use a router table with some of the speakers I've built! I'd need a hydraulic arm to manipulate it.

I know. I’m starting to fear how heavy these speakers are going to get. All from 3/4” BB ply. Double layers all around except for the 3 or 4 layers of the front baffle. My back hurts just thinking about it. 8O

[dlneubec]

You have all the right toolpacks installed? You have to hit a button to get it to run the numbers, if I remember right. Pretty sure Roman B has some nice info on using BDS on his website too.

Thanks CJD,

I followed Romans suggestions and it's working fine now. :T

[JonW]

Hi Jon,

I'm just starting to try some simulations in BDS and can't figure out how to get a response plot. I get nothing in the graphs below the driver and baffle info entry area, though I do get a plot in the graph on the top left.

Any hints?

Hi Dan,

I don’t have the program in front of me (it’s at home). But you need to hit a button called something like “load basic frequency data.” It’s about in the middle of the page, toward the left, and has red text, I think. And after every change you make to the baffle, you have to hit the “recalculate” button in the middle of the page to update the graph. There is a long instruction manual I got from the same site I downloaded BDS. It’s a long read and mildly worthwhile. But you can figure things out just by playing with the program.




-=-=-
Chad-
Looks excellent. You’re actually using real, honest, measured data. You’ll get much farther with that than I am with simple simulations. :T My plan is to use the sims to guess at the best baffle and cabinet. Then build it and take measurements in the built cabinets.

If you haven’t already, you might want to look at the frequency response (on axis) of the various tweeters tested at zaphaudio.com. Some of them are quite flat, others less so. Good observations regarding being off axis. Like you said, it’s not an obvious thing to deal with, near as I can tell.

[dlneubec]

I know. I’m starting to fear how heavy these speakers are going to get. All from 3/4” BB ply. Double layers all around except for the 3 or 4 layers of the front baffle. My back hurts just thinking about it. 8O

Jon,

I'm curious why you want to double up all around. It seems to me that internal bracing is much more effective, keeps the weight down and allows for a smaller outside footprint. You could have a full horzontal shelf brace every 6" and still not have the exta weight of just doubling a single side. That masive baffle will make them heavy enough! :E

[JonW]

Jon,

I'm curious why you want to double up all around. It seems to me that internal bracing is much more effective, keeps the weight down and allows for a smaller outside footprint. You could have a full horzontal shelf brace every 6" and still not have the exta weight of just doubling a single side. That masive baffle will make them heavy enough! :E

Actually, I was thinking about doing both the double sides and the shelf braces, maybe 4 shelf braces in total. Do you think single thickness and 4 shelf braces is enough to really keep cabinet vibrations minimized as much as possible?

[dlneubec]

Hi Dan,

I don’t have the program in front of me (it’s at home). But you need to hit a button called something like “load basic frequency data.” It’s about in the middle of the page, toward the left, and has red text, I think. And after every change you make to the baffle, you have to hit the “recalculate” button in the middle of the page to update the graph. There is a long instruction manual I got from the same site I downloaded BDS. It’s a long read and mildly worthwhile. But you can figure things out just by playing with the program.

Thanks Jon,

I followed what Roman B. suggested here:

Using FRD tools

BTW, if you click on "Auto" under the "Optimizer" tab at the upper left, it will dynamically display response changes as you make them, at least with the settings I was using (chamfer, point, quadrangle, separate, sumdown, with advacne fequency support also loaded). Pretty cool and much easier to see differences than using "Recalculate", and faster too. :T

[dlneubec]

Actually, I was thinking about doing both the double sides and the shelf braces, maybe 4 shelf braces in total. Do you think single thickness and 4 shelf braces is enough to really keep cabinet vibrations minimized as much as possible?

IMO, yes. In my Mentor Omni prject, I have what in effect is shelf bracing every 4.5" in a 13"x13"x30" or so cabinet and in it is very dead. There is very little vibration in the sides even in the bottom 3/4, which houses a 12" sealed sub. All of my baffles are double, but that's all.

If anything, I'd ditch the double sides and up the bracing if you want more effective damping. It seems like there was some research somwhere that I read that said the side to side bracing was far more effective than any other means.

[JonW]

Hi Dan,

Thanks for the ideas. So maybe I could just go single thickness plus shelf braces. It would make things lighter and that is certainly appealing. I don't want to compromise too much on this project. Doing the braces along with the double thickness means I know I did everything that one can reasonably do. But if you say 1 layer + braces can be enough... Hmmm...

[Hank]

I almost always use 3/4" radius roundovers, but I have a big bad 1 1/4" radius bit. I use mine on my big Freud table-mounted router.

Wow- I wrote this last night and then failed to press the button - just now found it. ops: