DSP_Geek said:
Hi DSP_Geek,
Thanks for tips a, b and c. I have to admit I don’t completely understand everything that going on in your explanation. I hate to sound so green but math is not one of my strong points (long story). I learn quickly though. So I will brush up on square roots, cube roots at wikipedia. Obviously the square and cube of two are related to frequency/diffraction modes somehow …just need to make the connection.
ttruman said:
No probs, my explanation was a bit curt. I used the square roots and cube roots to generate numbers equally spaced in the logarithmic domain. All the hard work's already done and presented in the three alternatives, so you can use them as a cookbook for the dimensions you need.
Let's try a 10 inch baffle with alternative a: 35% of 10 inches is 3.5 inches, 44% is 4.4 inches, and 56% is 5.6 inches, so you'd place the tweeter center 3.5 inches from the top and 4.4 inches from one side.
For alternative b, 10 inch baffle: 4.9 inches from the top, 3.9 inches from one side.
For alternative c, 10 inch baffle: 7.0 inches from the top, 4.4 inches from one side.
ttruman said:
Cool. Let's look at what we want to do: we want to make sure that edge reflections don't reinforce each other on-axis. As others have noted, what creates a dip on-axis might very well flatten the response off-axis. Flat off-axis response is nice, of course, but I'd rather be flat on-axis than be flat off-axis and have the on-axis response look like crap.
To this end, we want to make sure the tweeter-to-edge frequencies are reasonably well-spaced. All the modes will coincide if the tweeter is at the same distance from all edges, with the unfortunate results you found.
So, how do we space the modes? Let's say we only have left and right sides to worry about, so we want the tweeter-to-left-edge mode (call it Fl) to be well away from the tweeter-to-right-edge (Fr) mode. We also know that modes occur at multiples of their fundamental frequency, so they also occur at 2*Fl, 2*Fr, 3*Fl, 3*Fr, and so on. Since a number of things, including the non-zero diameter of the tweeter, come into play, only the first few edge modes are important. Nonetheless, we want to place one set of frequencies neatly between the other set.
We want the left modes to fit between the right modes: that means we want them ordered as Fr, Fl, 2*Fr, 2*Fl, 3*Fr, 3*Fl, etc. By examination, we'd like Fl to be exactly between Fr and 2*Fr. A naive implementation would make Fl = 1.5 Fr, but that would be wrong since frequencies tend to work exponentially. You want something known as a geometric mean, which is the root of two numbers multiplied together, like so Gm = sqrt(a*b). If we apply this to Fl, we get Fl = sqrt(Fr*(2*Fr)), or Fl = Fr * sqrt(2). This means we want the distance from the tweeter to one edge to be 1.414 that of the tweeter to the other edge, so the ratios are 1:1.414. Summing the two numbers gets you a total distance of 2.414 which represents the sum of left and right distances, ie the baffle width. Creating width ratios in relation to the total baffle width: 1.414/2.414 = 58.6%, and 1.0/2.414 = 41.4%. Let's just double check: 58.6%/41.4% = 1.415, which is pretty darn close.
For three edges the concept is similar, with the mode progression going (Ft = Ftop): Fr, Fl, Ft, 2*Fr, 2*Fl, 2*Ft, 3*Fr, 3*Fl, 3*Ft, and so on. You end up getting a series of distances all separated by the cube root of 2, to wit 1, 1.26, 1.59. [1] I merely applied that to the various edges to generate the three tweeter placement alternatives. For example, alternative a has distance 1 for tweeter to top edge, 1.26 for tweeter to one side, and 1.59 for tweeter to other side. Note that the two side-to-side distances add up to 2.85, so the first side is 1.26/2.85 of the baffle width, or 44%. The distance to the top is 1/2.85, or 35% of the baffle width. The other alternatives are constructed similarly.
[1] actually 2*1.59 is 3.18, or slightly bigger than 3*1.0, but that's OK because the modes are still reasonably spaced.
DSP_Geek said:
nice post, but it got me wondering about just how many commercial offerings even vaguely begin to apply any of that?
That leads on to asking 'just how audible is an anomaly' like the one the OP posted?? (given that it usually is simply not addressed in the industry).
Sure we can see it on a microphone response, but then again we see lots of things with a mic which at the end of the day might not make a lot of difference.
Curious what thoughts people might have on all that.
It's begun. This is the setup for the tests...
I took an unused tripod and mounted a swiveling platform on it. I used the cabinet as reference to set the height so I don’t have to move the gate window when going back to measure with the cabinet.
Underneath are indicators that will allow the platform to be set at 15 and 30 deg off axis.
The MDF and cardboard test baffles can then be butt up against the platform. When flush against the front edge they will be at the same angle.
Here is ECM 8000 one meter away.
The amps …
The C23’s under test.
Pulse response.
The base line (free) on axis measurements using the rig.
I think the setup is sound and should provide good results. If you guys see anything that could skew the testing sure would appreciate the feedback.
Tomorrow night I will build the test baffles and begin the tests.
I took an unused tripod and mounted a swiveling platform on it. I used the cabinet as reference to set the height so I don’t have to move the gate window when going back to measure with the cabinet.
An externally hosted image should be here but it was not working when we last tested it.
Underneath are indicators that will allow the platform to be set at 15 and 30 deg off axis.
The MDF and cardboard test baffles can then be butt up against the platform. When flush against the front edge they will be at the same angle.
An externally hosted image should be here but it was not working when we last tested it.
Here is ECM 8000 one meter away.
An externally hosted image should be here but it was not working when we last tested it.
The amps …
An externally hosted image should be here but it was not working when we last tested it.
The C23’s under test.
An externally hosted image should be here but it was not working when we last tested it.
Pulse response.
An externally hosted image should be here but it was not working when we last tested it.
The base line (free) on axis measurements using the rig.
An externally hosted image should be here but it was not working when we last tested it.
I think the setup is sound and should provide good results. If you guys see anything that could skew the testing sure would appreciate the feedback.
Tomorrow night I will build the test baffles and begin the tests.
ttruman said:
My pleasure; I hadn't actually worked out the 3 edge problem before you asked, so (as they say in Santa Cruz) it's all good.
And as far as learning stuff, I'm still picking up _lots_ from the group here. I'd already designed a couple of well-regarded Pro Audio systems, but the learning curve accelerated seriously when I got here. My current project stalled for a while because I found an egregious error - after I'd cut the lumber, of course - and it took a year to get back on track. A shout out to LinearTeam's WinISD and Svante's Edge for a couple of outstanding software tools.
A few observations
Beware of any object that has a reflection within the time frame of your selected window. Objects behind the speaker as well as behind the microphone can influence the measurement. I had someone run my posted SPL graphs through software of his. He found a small, common reflection present in them all. This turned out to be the very small tube used to support the 1/4" pipe that my mic is mounted in. The tube is 1/2" diameter PVC and was roughly 0.25m behind the face of the mic. I used the tube to allow fine adjustment of the mic distance from the measured driver.
I had already lined the front of the mic stand (the 3-legged variety) with felt on the side facing the driver and had neglected the area directly behind the mic. I now use felt on anything in direct line-of-sight with the driver under test.
Your driver support, all parts of it, the mic stand and the mic clamp should all be damped with at east a 1/2" thick piece of felt on those areas facing the driver to minimize reflection influences.
There are also two small, distinct anomolies that I'm sure are not from the driver. These are at about 8.2ms and 9.8ms. The first one is likely from some of the support structure for the driver. You should be able to determine the origin of the second one by calculating the distance using time delta from the original impulse to that point.
Let's say it's 9.8ms - 7.4ms = 2.4ms yielding about 32" or 0.83m. Measure from this point, keeping in mind that if it's something behind the driver, such as driver support hardware, it's the round-trip distance, or 16" behind the driver. If it's the mic support, then it's that distance (32") minus the distance from driver to mic, then half the remaining for the round-trip behind the mic.
Also, forget the thin piece of foam on the floor. It's not of any use and just makes it harder to see the first reflection if it partially damps the reflection, which is all it can do.[/QUOTE]
Beware of any object that has a reflection within the time frame of your selected window. Objects behind the speaker as well as behind the microphone can influence the measurement. I had someone run my posted SPL graphs through software of his. He found a small, common reflection present in them all. This turned out to be the very small tube used to support the 1/4" pipe that my mic is mounted in. The tube is 1/2" diameter PVC and was roughly 0.25m behind the face of the mic. I used the tube to allow fine adjustment of the mic distance from the measured driver.
I had already lined the front of the mic stand (the 3-legged variety) with felt on the side facing the driver and had neglected the area directly behind the mic. I now use felt on anything in direct line-of-sight with the driver under test.
Your driver support, all parts of it, the mic stand and the mic clamp should all be damped with at east a 1/2" thick piece of felt on those areas facing the driver to minimize reflection influences.
I'd recommend some felt on the front edge of the small wood platform as well as the vertical pipe section.
Is this prior to adjustment of the window start marker? If not, that marker needs to be just before the impulse, somewhere just above 7ms. If not, you'll include all ambient room noise in the response that exists prior to the actual origin of the driver output.
There are also two small, distinct anomolies that I'm sure are not from the driver. These are at about 8.2ms and 9.8ms. The first one is likely from some of the support structure for the driver. You should be able to determine the origin of the second one by calculating the distance using time delta from the original impulse to that point.
Let's say it's 9.8ms - 7.4ms = 2.4ms yielding about 32" or 0.83m. Measure from this point, keeping in mind that if it's something behind the driver, such as driver support hardware, it's the round-trip distance, or 16" behind the driver. If it's the mic support, then it's that distance (32") minus the distance from driver to mic, then half the remaining for the round-trip behind the mic.
It would be useful to know the low end resolution of the FR. The conditions you have can't be reliable for much below about 250Hz or so. SW may show down to 100Hz, but it's not accurate at all below the low limit dictated by reflections from the floor (or furniture on the side if present).
Also, forget the thin piece of foam on the floor. It's not of any use and just makes it harder to see the first reflection if it partially damps the reflection, which is all it can do.[/QUOTE]
dlr said:
Hi dlr,
I revised the setup as advised; did the best I could with what I have available. The anamolies seen at 8.2 and 9.8ms appear to be gone...
An externally hosted image should be here but it was not working when we last tested it.
Re: A few observations
If the gate window is below 5ms I am unable to get reliable measurements. Due to my small room size this may be something I have to live with. The picture below shows the results of a gate window under 5ms. The results are typical to a width of about 5ms.
At 5.5ms things stablize and SW provides consistent repeatable measurements...
For the purposes of determining the baffle diffraction issue I don't think this applies. I am unaware of how it would affect my cross over modeling either. 250Hz would be a little over 4 octaves away @ 3000 Hz and >36db down following a 2nd order target.
Or maybe I am just missing your point.
dlr said:
If the gate window is below 5ms I am unable to get reliable measurements. Due to my small room size this may be something I have to live with. The picture below shows the results of a gate window under 5ms. The results are typical to a width of about 5ms.
An externally hosted image should be here but it was not working when we last tested it.
At 5.5ms things stablize and SW provides consistent repeatable measurements...
An externally hosted image should be here but it was not working when we last tested it.
For the purposes of determining the baffle diffraction issue I don't think this applies. I am unaware of how it would affect my cross over modeling either. 250Hz would be a little over 4 octaves away @ 3000 Hz and >36db down following a 2nd order target.
Or maybe I am just missing your point.
Re: Re: not convinced it's all baffle step
Hi Svante,
5 AM in the morning here . Been up all night to get these first tests out! So here it goes...
What happens if you put the (free) driver in a piece of cardboard of the same geometry and size as the baffle? (You dont need the lower half metre I guess. a piece of 220x500 mm would do.)
You could, just to test it, you could put the tweeter in a piece of MDF 220 mm wide, but slightly to the side and slightly lower:
Since the cardboard baffle with the matching geometry came out so close I decided to try the offset. Down 3cm - left 2.5cm.
Perhaps I need to use the MDF but it doesn't look promising. I am r-e-a-l-l-y tired now
so I am off for a short nap.
Hi Svante,
5 AM in the morning here . Been up all night to get these first tests out! So here it goes...
What happens if you put the (free) driver in a piece of cardboard of the same geometry and size as the baffle? (You dont need the lower half metre I guess. a piece of 220x500 mm would do.)
An externally hosted image should be here but it was not working when we last tested it.
You could, just to test it, you could put the tweeter in a piece of MDF 220 mm wide, but slightly to the side and slightly lower:
Since the cardboard baffle with the matching geometry came out so close I decided to try the offset. Down 3cm - left 2.5cm.
An externally hosted image should be here but it was not working when we last tested it.
Perhaps I need to use the MDF but it doesn't look promising. I am r-e-a-l-l-y tired now
so I am off for a short nap.
Interesting...
I am not at the right computer now to simulate, but at least there was some effect of moving the driver on the baffle.
Am I right in assuming that the only difference between these last cardboard baffle measurements and the ones a few posts up in the thread is the baffle?
In those there was a dip around 5 kHz, which is consistent with a smaller baffle (the driver itself), and the 2-3 kHz range is flat. This tells me that the cause is the baffle and the baffle can only contribute with diffraction.
Hmm.
Anyway, you now have a great platform for experiments, since the box and cardboard measurements were so similar.
I am not at the right computer now to simulate, but at least there was some effect of moving the driver on the baffle.
Am I right in assuming that the only difference between these last cardboard baffle measurements and the ones a few posts up in the thread is the baffle?
In those there was a dip around 5 kHz, which is consistent with a smaller baffle (the driver itself), and the 2-3 kHz range is flat. This tells me that the cause is the baffle and the baffle can only contribute with diffraction.
Hmm.
Anyway, you now have a great platform for experiments, since the box and cardboard measurements were so similar.
Am I right in assuming that the only difference between these last cardboard baffle measurements and the ones a few posts up in the thread is the baffle?
Yes.
Anyway, you now have a great platform for experiments, since the box and cardboard measurements were so similar.
Absolutely. Your efforts and suggestions have been a great help!
Yes.
Anyway, you now have a great platform for experiments, since the box and cardboard measurements were so similar.
Absolutely. Your efforts and suggestions have been a great help!
Over the past few hours I have tried numerous tests on the enclosure itself. Felt blankets, geometry changes with cardboard and simulating rounded edges using PVC pipe. Although none of the tests helped significantly what I found was that the most notable changes came from changing the geometry on the top edge of the enclosure. Just for grins I made another cardboard baffle...
I would say that is a HUGE improvement.
I got to thinking that the top of the enclosure could be modified without sacrificing the fine look of them or affecting the midrange chamber either. So this a great find. I may play around with angling the top edge back so I don't have to cut so much off but I am on a good path here.
Thanks to everyone for all your suggestions and help. Truly. They kept me motivated and moving forward towards a solution!!!!
Cheers
An externally hosted image should be here but it was not working when we last tested it.
I would say that is a HUGE improvement.
I got to thinking that the top of the enclosure could be modified without sacrificing the fine look of them or affecting the midrange chamber either. So this a great find. I may play around with angling the top edge back so I don't have to cut so much off but I am on a good path here.
Thanks to everyone for all your suggestions and help. Truly. They kept me motivated and moving forward towards a solution!!!!
Cheers
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.