Oh man, you're completely right - I had forgotten to update the ring dimensions for the large exciter in the release version of PETTaLS. It's acting like a much larger exciter. My mistake! This will be fixed in the next release.
I also don't know if the average radiation has practical value. I've seen papers where people try to derive correction filters based on the average level of radiation rather than the on-axis radiation. Correction filters are not my preferred approach, but they may be for some people!
If the anisotropy is very pure - as in there are just simply different stiffnesses in the X and Y directions, then there are different coincident frequencies in the horizontal and vertical planes. The vertical plane coincidence wouldn't really appear in PETTaLS right now.
Well, as with most things, it's pretty hard to say. Back when I was doing my PhD I never saw much error based on baffling so I just never worked on taking it into account.
Here's a picture from one of my papers with measurements from open-backed plates...
And here are corresponding simulations in PETTaLS using reasonably similar parameters:
You can see some error at low frequencies near the +/-90 positions from the open baffle, but it's not significant.
I'd expect the errors to become more significant for FFFF panels, for sure. But in order to quantify how much I'll need to do some measurements. You all are keeping me so busy already that I'm not sure how soon I can get to all of these things!
I just figured that for an open backed panel speaker, low enough frequencies would radiate into 4pi space, while sufficiently high frequencies would radiate into only 2pi space. So by assuming infinite baffle, the model would be considering all frequencies to be radiating into 2 pi space, and hence be boosting low frequency SPL by around 6 dB, compared to the open/self baffled case. But I could have that muddled up.
Eric
Happy to see the discussion comes to the baffled vs open-backed configuration. I have pointed this difference between PETTaLS and our configurations some time ago...
In my understanding, what you describe here, Eric is not correct. The main difference between baffled and open-back is while baffled, each side radiates in a 2pi space with no influence of the second one while in open-backed there are interactions, each side having access to the full space. As there are in phase opposition, depending on the observation point it can result any situation from cancellation to reinforcement.
Typically, this leads to a null a 90° for all the frequencies. The area of low SPL around 90° is wider at low frequency than at high frequency.
@EarthTonesElectronics , Dave, sorry for the question, but are you sure the plots just before are from open-backed panels? Is it with the DML alone or with some additional large baffle? I don't see on them the signature of open-backed I have in directivity measurements (panel width from 25 to 40cm). Mainly, I don't see the null at 90° neither the lobes. Or is there an other difference in the set up I don't understand?
The difference with a classical dipole with a localized source is the source locations on the panel surface are frequency dependent.
From the simulation point of view, I found papers about this topic. In both situation the Rayleigh's integral is used. In the case of a baffled panel, it is quite straightforward as it is an integration of the speed it is null at the baffle so those points are not contributors. In the case of the open-backed, it seems some iterations are needed as it is the pressure which is null in the panel plan and the speed is unknown. In our tentative of simulation with Paul maybe 2 years ago, as first step, the baffle version was implemented.
But despite it is more complex, it is important to understand this open-backed behavior as it leads to an additional 6dB roll off, peaks and dips. In my current understanding it should be also an element in the discussion about test exciter placement.
Christian
I've uploaded a new "release" to github - 1.4.1. No major fixes in this one, but some new UI features, and I added the 32mm exciter.
There's a problem in the Linux release where I can't figure out how to get the error symbol image to appear. This functionality works in the compiled binary in Windows and macOS but not in Linux. Hopefully I can find a workaround eventually. This doesn't affect the functionality of the code.
There's a problem in the Linux release where I can't figure out how to get the error symbol image to appear. This functionality works in the compiled binary in Windows and macOS but not in Linux. Hopefully I can find a workaround eventually. This doesn't affect the functionality of the code.
Thanks Christian,
I suspected I didn't have that quite right. Tell me if the following is correct. You mention only two cases, baffled and open baffle), but I think there are three (highly idealized) cases:
- Infinite baffle, assumed to be in the plane of the panel. High and low frequencies all radiate into 2 pi space, and no cancellation from front to back, and hence no rolloff at low frequencies. Here it doesn't matter if the back of the box is open or closed (at least on the front side side!), except perhaps some influence if the box is very shallow and affects the vibration of the panel significantly.
- Closed box, but no baffle other than the panel itself. Due to closed box, no cancellation from front to back (like above), but now high frequencies radiate into 2 pi and low frequencies radiate into 4 pi, with resulting 6 dB rolloff. This is the common case for conventional pistonic drivers in boxes.
- Open baffle (i.e no box, and no baffle (other than the panel itself). Here each side radiates in a 2pi space with no influence of the second one while in open-backed there are interactions, each side having access to the full space. As there are in phase opposition, depending on the observation point it can result any situation from cancellation to reinforcement. and it is important to understand this open-backed behavior as it leads to an additional 6dB roll off, peaks and dips.
Certainly oversimplified, but is that basically correct?
Eric
So it was the 25VT-4 that was using the wrong dimensions? I noticed that the OD looked large, but didn't give it much thought.
Eric
Eucy,
I assume this is the equation you are talking about, and yes it is what Pettals is using to determine fc, as far as I understand.
But the issue is that this equation applies only for isotropic materials, i.e those with the same properties in all (three!) directions. So even if you measure E and nu in each direction, you still can't calculate the two fc's. And that is true even if the properties are the same in the length and width directions of the panell. If the material is not isotropic, then the above equation simply doesn't apply. The equation might(?) represent a lower bound, but I'm just guessing there.
I think if you know the shear moduli (in addition to the tensile moduli and poissons ratios) that would be enough to calculate the two fc's, though I don't know the equation. And probably you are right that it would be easier to just directly measure those than try to figure it out by modelling the stack behaviour.
Eric
Yes there are more cases if we include the possibility of a rear enclosure, basically a closed box.
1/ => ok
2/ ok for the description and that there is no cancellation front/rear but after... in fact I don't know how a DML works with a limited rear volume. I just remember some paper or patent showing a peak in the response... For a pistonic speaker, the radiation change from 4pi to 2pi. It doesn't create a roll off in the same way as low frequency attenuation of an open baffle. Most probably 2 ranges with a difference of SPL linked by a 6dB/oct slope. A closed box with a cone has a 12dB/oct roll off in the bass not because of the radiation pattern or of the rear wave but because the cone works below its frequency of resonance. For a DML?
3/ ok except the beginning of the sentence "Here each side radiates in a 2pi space with no influence of the second one " is related to 1/.
I made tests with attenuatuation of the rear which were interesting. The few tests I did trying to close the back were not succeful. It needs to build a proper enclosure. I stopped those tests to stay concentrate on the most appreciated which is the open backed.
I would say we have 2 independent possibiliies :
1/ the baffle with in practice 2 degrees : infinite baffle (a DML flush a wall, PETTaLS) and unbaffled (also described a self baffled DM, the most common in pratice)
2/ the rear load : from no load (open baffle) to closed box (no front/rear communication) with all the possibilities in between of attenaution of the rear and of volume effect.
@EarthTonesElectronics
Dave,
Some random questions and suggestions:
Eric
Dave,
Some random questions and suggestions:
- If I'm not mistaken, Pettals calculates (or could calculate) both of the critical frequencies before running the actual simulation, but doesn't reveal them until the simulation is complete. Would it make sense to display them in some way as soon as the simulation starts? Yesterday I was trying to match both frequencies by manipulating material parameters, and it was annoying that I had to wait until the sim was completed to know if I had succeeded. I realized I could just run a short sim to say 50 Hz and get the answer quicker, but why not display them earlier? Either on the graph or in a separate text field?
- Maybe my computer is just slow, or lacking memory, but a lot of times when I try to get an impedance plot or a surface velocity map at a specific frequency, it takes a long time to come up, and I have to wonder if it is making progress, or just hung up and stopped working. It would be nice to have some indication that it is actually making some progress or not in those cases.
- I'm curious why it happens that if I run a simulation up to 20kHz, it runs through the early part of the simulation much slower than if I run it up to only say 1kHz. Obviously I understand that it takes longer to do the full sweep, but why would the early part of the sweep, say the first 3 octaves, take longer if I am doing a 20kHz sweep than it takes to get through the same 3 octaves in a sweep to 1kHz? For the longer sweep, are you adding the contributions of all the "simulated modes" regardless of how far away they are from any particular frequency? I'm assuming that there is only a limited number of modes that really contribute anything significant at any particular frequency. And maybe it depends on Q (or not?). I was thinking you had some rules for which subset of modes you bothered to add up at any particular frequency. But do you, or could you?
Eric
Oh, sure, I can definitely do that. Good idea!
This is actually a graphics problem. Matlab is always slowest when it has to open up a new window, and this has nothing to do with whether or not anything is being calculated (in this case, all of the calculations were already completed). I think I'll actually just rearrange the main window so that these plots automatically show and don't have to be brought up in new windows.
You are correct - I calculate contributions from all modes at all frequencies. I've been trying to find a way to speed up this process for a little while. It's something I'm working on as we speak... I just need to find the right approach so that the errors aren't too significant. When the modes are no longer linearly independent because of coupling through the exciters, it just becomes a little tricky.
I'm also playing around with non-modal solutions to the high-frequency response. So there will probably be some speed-up updates before too long!
For a DML, the sealed box mostly just affects the lowest mode, and has the same effect that it would with a cone driver - increasing the resonant frequency and its Q. It doesn't typically shift the higher modes much. It's also important to note that sealed boxes only really make sense with SSSS and CCCC setups!
In the picture below, blue is infinite enclosure depth, purple is one inch (probably a reasonable size for an on-the-wall speaker).
This is where theory and practice diverge, as the audio quality effects of shifting from open to closed back are huge
Eucy
Eucy
To try to give a follow up to this question is "a plastic glass Acrylic or Polystyrene ?"... in short, it might be also polycarbonate!
I had in my notes a density of 1050kg.m³ for the polystyrene and 1200kg/m³ for Acrylic (PMMA), PETTaLS gives 1180kg/m³
So I weighed my 2 samples... 1130kg/m³... so probably PMMA! Apologize. Need to update the paper about directivity and coincidence frequency
I also had a look to the DIY store close to me (Leroy Merlin, same as @RMAM )
Its catalogue shows mainly polystyrene based products. They have also a possibility of PMMA, more expensive.
On the shelves of the store, I found 2 types of materials, both labelled by the store as polystyrene. From the supplier (Sedrap) labels on the products:
- One is branded styroglass, the data sheet says extruded polystyrene, 1050kg/m³, E = 3.4GPa, G = 1.4GPa
- The second is branded polyglass, no detailled data sheet but we can read that from the Sedrap site.
So 3 possible materials with the same aspect. The possibility of confusion is confirmed by what we can read on the web... even sometimes on purpose as the cost is not the same but most probably because the confusion is easy.
I found from this document
- Polystyrene PS : density 1000 to 1100kg/m³, E 1.2 to 2.6GPa
- Acrylic PMMA: density 1200kg/m³, E 2.2 to 3.8GPa
- Polycarbonate PC: density 1100 to 1200kg/m³, E 2 to 2.4GPa
So, the density might help to identify the PS.
Other possibility : see How to distinguish the plastic materials of PP PC ABS PVC PA PS SAN POM PMMA / burning flame and smoke state! Be careful!
- PS : relatively flammable and continue to burn after leaving the fire. The flame is yellow with dense smoke.
- PMMA: light blue with flames and smells of rotten vegetables.
- PC: burns very slowly, and will slowly extinguish after leaving the fire. The flame is yellow with smoke.
Christian
Christian
Polystyrene will dissolve readily in acetone...Acrylic less so, and polycarbonate will soften and swell, but not dissolve. So no need to burn your house down.
Of course, either way, the testing is destructive.
Eucy
Haha guys, I totally agree with Eucy that there is no need to burn the your house down or risk your health!
The thing about polymers is that most any of them (that you would consider using) are pretty similar in terms of elastic modulus and density. They are all pretty close to about 3 GPa and 1200 kg/m3, or thereabout. So unless you using polymer foams, or adding carbon fiber or fiberglass, you can pretty much use any of them and get a similar result, I expect. They are different, no doubt, but not much. They may differ more in damping properties, which could be the most interesting variation between them.
Eric
@homeswinghome
Christian - I'm still browsing this matter - and I came across this early post showing no nulls @90 - https://www.diyaudio.com/community/...s-as-a-full-range-speaker.272576/post-6105174
Comments?
Eucy
Wouldn't it be more prudent at this step to make 2 categories : PMMA/PC at 1200kg/m³ and PS at 1050kg/m³ for a possible efficiency difference and see after if we get more knowledge about damping?
Yes Eucy!
It is a question of measurements. For a long time I haven't been able to detect this null and the coincidence frequency. The reason is in the use of REW.
If the observation window around the IR is longer than the time needed for the first reflection to arrive, your FR is the FR of the source plus the room. So you collect not only the sound from that direction but also of all the directions reflected by the room.
By default, this window is set at more than 100ms by REW when the time for the first reflections is between 3 to 6ms.
So to detect the null, we need to have the panel at more than 1.5m of all the surrounding walls and the ceiling. The floor is then at about 1m so some absorbent on the floor is needed.
Then the IR window of REW can be adjusted at about 5ms, just before the 1st reflections that are easy to see.
The consequence is it is not possible to measure below 200Hz (1/5ms) and in practice the precision is not that good before 500Hz.
Maybe you can have a look to my paper about how to do directivity measurements. I don't remember if it is precise enough.
Christian