Interesting!Hi Christian,
Here's the impedance and impedance phase curve of a panel with 4 drivers on it. No cross-over. Green curve is impedance, the dotted curve is impedance phase.
The important thing to note is that eventhough there are several peaks in the impedance response, they are minimal, a few ohms at most, and nothing like the massive swings presented by piston speakers and their cross-over networks.
If I had to add cross-overs to the drivers in the curves shown, then besides amplitude changes in magnitude to the individual drivers, there would be at least 90° opposing phase shifts at their associated cross-over points. Bessel or Linkwitz-Riley filters attempt to address this problem, but the over-all, relatively flat system impedance phase response would be negatively affected.
To me, it's anathema to introduce standard cross-overs into DML panels, but I do agree that there are ways of cheating a bit, and adding passive components (shorting caps!) to modify driver responses a bit without stuffing up the impedance phasing too much.
IMO Mechanical "cross-overs" (ortho-springs) are a much better way of retaining DML action while separating the panels into their own bands where they work best.
View attachment 1173770
And the parallel resistor I suggest will limit even more the impedance variation. let's suppose 6.8Ohm in parallel of this DMLwith a mini 5Ohm to max 17Ohm, it will result... oh, oh 2.9 to 4.8 Ohm. A bit too low probably (6.8Ohm was what I had close to me!).
Some simulation might say more!
Oh yeah! Sure would! Did you get a quote?
It would be very interesting to watch a frequency sweep of various materials, sizes, boundaries, etc. Or an impulse response.
That said, be aware that a close spaced mic (say 1-2 mm from the panel) acts a lot like an accelerometer. So you can use a close mic to learn a lot about how individual spots of the panel reacts at different frequencies. Unfortunately you can only look a a single point at a time.
Eric
Firstly it was cheap and available and I was curious, so I got it as a pair so I could test splitting out the panel frequency ranges with music even if they weren’t the exact right places to split.
I I did choose the one at 1000hz because I noticed a lot of the materials I was testing had a good flat response at least until there.
So It was a way to split close to a natural dip it was about to hit anyway.
The 5000 one is annoying it needs to be around 2000 or 3200 to hit second set of dips that I was seeing.
So I’ll do the crossovers to match the materials in the end, unless I manage to get a awesome material that is flat by itself until 2-3khz, then I’ll probably just do a two way 🫠
As you mentioned in the other post, I am now thinking the shape or ratio of the panel will dictate a lot of where the peaks and troughs are. So I may be able to line those up between the three panels if I don’t end up mitigating them some other way.
I also have a subwoofer for the <100hz!
Hi Christian,
Why do you want to limit impedance variations?
My natural assumption would be that a parallel load across the driver would simply present a heavier load to the amplifier.
Does it make an audible or measurable difference? I suspect that the only difference you would notice would be a slightly warmer amplifier in addition to the warmth of the load resistor.
I would be interested in seeing updated FR/distortion/spectral curves with and without the load resistors.
You're wrong. Judging by ear my panels sound *******' fantastic to me, couldn't care less how they sound to you or anyone else. Coming from a "blinkered" state of mind.
By all means, if your plan is to mass produce or use them for parties or whatever then chase that perfect straight line and get in line with the masses.
I think that's exactly what I said.
See, anybody can spend a few weeks to knock together a tin can onto an old apple crate and then try really really hard to convince themselves that it "sounds good to them", and bugger anyone else.
That's great! We need more people like that.
Last edited:
Smaller plates will have more uneven response, especially in the lower part of the spectrum.
I think 30x30cm is too small for a good response down to 100hz with any material, and will have a jagged looking lower mid.
Like others pointed out I would really try to avoid using a x-over in a DML speaker, especially in the mid range. There are many ways you can manange dips in the reponse like modifying plates and suspension or adding weights. For the last bit just use an EQ.
Getting the mid range good is the easier part any way. Bass is better with conventional sub, and some materials and exciters will struggle with the top octave, 10-20k. So a 3 way system can make sense, but in that case I would probably have the top x-over at at least 7k to get it out of the spectrum where our ears are most sensitive.
@Leob thank you that’s a good tip for the high pass being at 7khz+
When I do use the crossover it does sound good to my ears though. Which has added impetus for further discovery in that area.
The comparison being three separate panels that get no crossover filtering, or a single panel.
It seems quite common for hifi DML setups to have multiple panels, which help cover each others peaks and troughs on the spl graph, but uncommon to split the panels for frequency band responsibility.
Question to the group:
If you were going to this (in a universe where you too wanted to see the experimental outcome) what kind of crossover would you choose to go with?
When I do use the crossover it does sound good to my ears though. Which has added impetus for further discovery in that area.
The comparison being three separate panels that get no crossover filtering, or a single panel.
It seems quite common for hifi DML setups to have multiple panels, which help cover each others peaks and troughs on the spl graph, but uncommon to split the panels for frequency band responsibility.
Question to the group:
If you were going to this (in a universe where you too wanted to see the experimental outcome) what kind of crossover would you choose to go with?
- Butterworth
- Linkwitz-Riley
- Bessel
- Combination of above
- Achieve the outcome via another method, which is…
I am someone chasing that straight line right now. So I might be taking this the wrong way.
I joined this thread because I thought its purpose was to collectively work towards achieving panels that achieve a full range frequency response (which I took to mean at the time, flat like a reference).
So I think you’re right about how subjective the audio experience is, and to do what’s best for your own ear when building something for yourself.
I don’t have a definitive idea of what my best sounds like yet, I have found a few “good”’s but since I have found a few i ask myself is there more “good” sounding panels I haven’t found?
I’m still in the phase of elimination of bad sounding speakers, which is a much easier task.
Additionally you don’t see many people come to this technical deep dive thread to say, “just joined, first post actually. My speakers sound good.”, it’s hard to end up in this thread without a reason.
It’s more inline with the thread to say “my speakers sound good and this is how I achieved it. It’s an opinionated build, but you might just enjoy its sound too.”
Isn’t this the place to discuss, debate, or finesse a DML speaker build to the n’th degree using this community’s distributed knowledge to achieve insights towards a specific outcome within a specific context.
I kept calling it Trivek in my posts, it’s actually called Tyvek. Specifically DuPoint Tyvek 105gm A4 sheets.
For future readers who want to actually find it.
I seen some esoteric HiFi speakers using a DML element for the mid and regular tweeters and woofers, and also Tectonic used that approach in older models. But seems like the most common approach is to use a 2 way system with plates handling everything above sub range.
DML is a bit special in that phase issues will not be as noticable due to their diffuse nature. This means that you do not have to be as carful with overlapping frequencies, and it is less fuzzy about which type of filter you use.
I would probably try to design so that the combination of exciter and material used gives as good balance as possible to start with. But using a large exciter and a bit larger plate for the mid element, and smaller for the treble, you could probaly get a decent balance without hardly any LPF/HPF.
I considered that approach for PA applications since that would mean I could use much beefier exciters for the mid. But since a single plate sounds excellent to my ears and can achive a good response I figured it was not worth the complexity. For HiFi I can imagine that if you find a material you think sounds amazing despite not convering sufficient range it could be worth exploring, but I'm somewhat sceptical. Personally I would start with a single plate and try to get that as good as possible with regards to material, dimensions, suspension etc, before going to a more complex design.
- Linkwitz-Riley my favorite when I built 2 or 3 way loudspeakers but... keep in mind it is the targeted curve of the overall system (loudspeaker + filter). For example let say you have a driver going to 7kHz you want to crossover at 4kHz you have to take into account in the crossover that the driver is band limited. To get a final LR crossover, the real filter might not be a LR. This particularly true on the phase (important phase shift occurs before highly visible level reduction). For a DML, it is a new story you are opening.
I guess it is "DuPont" like Dupont de Nemours. Following this page (sorry it is in French), it is made fro polyethylene : Dupont Tyvek.
Christian
Hello André,
I have to admit it is opinion or let say a careful approach. I see 2 possible effects :
- On the electrical side, the local increase of the DML impedance will decrease the filter effect creating some "ripples"
- DML are sensitive to the impedance of the source. In a pure voltage control, there is a kind of auto control : at the resonance, as the impedance is high, the current is reduced. What happens with a filter?
Christian
Yes, thank you, common sense.
Yes. I've posted many times, from homemade exciters and the woes of dealing with them to my latest iteration of panels with FHE25x4 Dayton exciters. Have added pictures of how I mount different panels along with knowledge I have from these builds in order to pass along if anyone wants to try it.
I recently had my panels side by side (bass supplemented) with a set of Spendor Classic 4/5 $3k speakers and they held their own to all that listened. I don't need someone telling me that I or anyone else can't recognize a nice sound when they hear it.
Specific outcomes are not the same for everyone. Are you trying to please the masses, your significant other or yourself? Are you looking at it as a business adventure or a hobby? Not everyone has to toe the Master's line to come up with what suits their particular agenda in DML's.
Discuss, debate, and finesse is what it's all about, not being told by the "Master" that their way is THE way.
Last edited:
Good luck with the bass portion of that "full range" as they just haven't yet been accomplished, to my knowledge. Nothing wrong with chasing that flat line, it lets you know when something is wildly off but it's not the end all.
This seems to be a constant issue of contention especially in the HiFi and DIY audio community, and I should stay out of it and focus on DML speakers instead...
But I think no one is saying that it doesn't matter if something sounds good to you. In the end that is the end goal for everyone, an amazing audio experience, not seeing a flat line or a really nice RT60 waterfall.
But even if we have the same goal, a good listening experience, it doesn't mean that listening tests is the best way to guide a design. I think you need both, but I am in the "measurement trumps subjective listening experiences" camp, and the reason is that have learned how subjective our hearing is through the years working with audio in different ways (producing, mastering, sound reinforcement, sound design, etc). I don't trust that anyone, myself included, can accurately assess audio quality without a proper procedure to ensure no subjective bias. I just seen so many times people claming to hear differences that they fail to identify in blind tests.
It is not an attack on anyone or accusing them of being incapable to hear things. Sure we can all hear the difference between two very different speakers, but when taking about things like plate dimensions, exciter placement, material choice, etc, the differences when tweaking is much smaller than the ability we have to alter our subjective impression of sound.
Another issue with doing subjective judgements is that how the speaker sounds will vary a lot on source material. For example I find that a free hanging plate can add a chorus like effect that sounds amazing on some material, and just makes other songs blurry and undefined. Or an uneven frequency response can highlight interesting parts in some tracks. To get an accurate picture when making a change you need to test every change on a lot of material, which just becomes unfeasible. Even if you did take the time, your ears would fatigue before you can reach any conclusion.
I'm not saying that you should rely completly on graphs, and especially blind listening tests are very valuable, but it is a lot of work and impractical to guide a design all the way. Instead you both measure and listen and learn to correlate measurements with how it sounds, and eventually you will find that addressing eventual peaks or throughs in the graph or reducing phase/delay issues will make things sound better, and is a lot more efficient to guide your design than trying to iterate through listening tests. Not only does blind listening tests take time to perform, but without being able to correlate the listening experience to measurements it is usually harder to know exactly what the problem you are hearing is, which makes it slower to fix.
Just some friendly advice...my intention is only to help trying to avoid traps that I have myself fallen in to before, and not to put anyone down. If you prefer working with listening tests, and think chasing a flat line through measurements is folly, that is fine with me.