Hi! I'm working on a small but time-accurate speaker project, and the problem with small speakers is bass. However a speaker brand called Devialet ( Mania, Phantom, Dione are all very small without passive radiators) has small speakers but impressive bass, can someone tell me how they did it? I've tried every DSP algorithm I have ( maxxbass, Rbass, BBE phase, TruBass, dynamic bass, Dolby) it seems that TruBass works but not as expected, I tried playing a recorded track through the phantom and it worked great, (I tried it on a devialet mania subwoofer driver). I would be very grateful if someone could tell me what Devialet did with the DSP
They may have bass quantity, but the ones i heard were not very well defined.
They do it with extremely long throw drivers, huge power amplificatio, and a whole lotta EQ.
dave
They do it with extremely long throw drivers, huge power amplificatio, and a whole lotta EQ.
dave
It's a combination of a couple of things:
Excursion will help you get more output from a given cone area, which is required for...
The DSP.
A pair of 5" drivers with 8mm of Xmax will manage 83dBSPL at 20Hz, 90dB at 30Hz, and 95dB at 40Hz.
You'll probably need a large LF boost to get flat to 20Hz, and a 2nd-order-HPF-shaped-limiter would be applied to ensure the drivers don't try to produce 95dB@20Hz.
This ensures that you're always making the most of your available excursion, but at higher sound levels, you'll notice the lack of LF extension. It'll still be pretty impressive, and it'll still attempt to do something at 20Hz, but it'll just be quieter than the rest of the music.
I've listened to a few of the Devialet Phantom speakers and find them initially very impressive, but it's very easy to make the smaller ones show up their weaknesses. You can hear the LF extension disappear as they're pushed harder, and the smaller ones hit the LF limiter pretty early on. The bass "punch" disappears, and the kick drum sinks back into the bass guitar/synth.
If you'd like to create something similar, you're looking for a dynamic EQ with fast attack and release times, or an excursion limiter.
Chris
- Lots of excursion
- Clever DSP.
Excursion will help you get more output from a given cone area, which is required for...
The DSP.
A pair of 5" drivers with 8mm of Xmax will manage 83dBSPL at 20Hz, 90dB at 30Hz, and 95dB at 40Hz.
You'll probably need a large LF boost to get flat to 20Hz, and a 2nd-order-HPF-shaped-limiter would be applied to ensure the drivers don't try to produce 95dB@20Hz.
This ensures that you're always making the most of your available excursion, but at higher sound levels, you'll notice the lack of LF extension. It'll still be pretty impressive, and it'll still attempt to do something at 20Hz, but it'll just be quieter than the rest of the music.
I've listened to a few of the Devialet Phantom speakers and find them initially very impressive, but it's very easy to make the smaller ones show up their weaknesses. You can hear the LF extension disappear as they're pushed harder, and the smaller ones hit the LF limiter pretty early on. The bass "punch" disappears, and the kick drum sinks back into the bass guitar/synth.
If you'd like to create something similar, you're looking for a dynamic EQ with fast attack and release times, or an excursion limiter.
Chris
I don't find it that difficult what they did. Quality speakers, amps and DSP, and you're all set. Could be made for half price, but Devialet made it into "a posch thing", riding that wave of marketing success, so they could afford to ask much more. It's all the same physics, there is no miracle. Just specific, good product.
Yes. It's a bit like comparing the sound of an Ampeg SVT Rolling Stone Tour and a MarkBass CMD121P : both are great and interesting Bass amplifers... Each in their own way, so It depends on what kind of bass you want to hear !
T
You mean "a lot of" bass (for the size) - I'm not so sure its that good...
//
//
Thanks you very much, your information has helped me improve significantly
However I don't think their drivers have that long of a travel, it's much shorter than the drivers of JBL, Harman or B&O portable speakers, and I also have a DSP that is more powerful than any DSP on portable speakers, and with the dsp I really witnessed the magic of the TruBass algorithm is there any more magic
I independently came up with an algorithm, which should work to enhance bass and overall clarity, to some extent. But it's just a thought experiment and, so far, nobody has shown enough curiosity in it to try it out, that I know of. But it could be something to try when headroom is severely limited by battery voltage.
Basically:
1, Use active filters to split the signal into separate high-pass and low-pass paths. Preferably a 1st order or LR type that sums back together cleanly without phase distortion.
2, Pass the bass signal through a soft-clipping circuit, either digital or analogue.
3, Adjust the hard clipping level of the bass, so that the speaker and amplifier still has a little overhead (maybe 3-6dB).
4, Add the high and low signals back together. That 3-6dB of overhead is where the high frequencies go, clean and unclipped, even if the bass is so over-driven it has turned into a square-wave mess.
If you listen carefully, all the recent JBL & co. bluetooth boomboxes do this, or something very closely related, where even severe bass clipping doesn't wreck the entire signal.
Basically:
1, Use active filters to split the signal into separate high-pass and low-pass paths. Preferably a 1st order or LR type that sums back together cleanly without phase distortion.
2, Pass the bass signal through a soft-clipping circuit, either digital or analogue.
3, Adjust the hard clipping level of the bass, so that the speaker and amplifier still has a little overhead (maybe 3-6dB).
4, Add the high and low signals back together. That 3-6dB of overhead is where the high frequencies go, clean and unclipped, even if the bass is so over-driven it has turned into a square-wave mess.
If you listen carefully, all the recent JBL & co. bluetooth boomboxes do this, or something very closely related, where even severe bass clipping doesn't wreck the entire signal.
Don’t think this is very new. Even if motional feedback and a thermal guard on the voicecoil would be present. I wouldn’t be surprised either if some sort of compression would be involved too. All leading to a lot of linear distortion, just like your average radio station applies. Not my cup of tea.
devialet uses driver moving mass compensation by aligning them vice versa - gives a more silent enclosure less working under demand
for most people its doing ok. Psychoacoustically already sound mixes cut the lows in order to make music sound louder, so its a common principle to squeeze loudness out of existing drivers
for most people its doing ok. Psychoacoustically already sound mixes cut the lows in order to make music sound louder, so its a common principle to squeeze loudness out of existing drivers
Basically all speakers I listened to, give it all or nearly all from get go at supposedly low volumes, but as you up the volume into last 3-5 notches, it all falls apart, the bass is lacking, and it isn't really working. It's the wow effect of low to medium volumes. No matter how you slice it, it's the same physics and same displacement volumes for SPL somewhere in the band.
Don't let perfect be the anemone of the good! 😉
Obviously the algorithm above is just a basic concept.
Practical R&D would have to start a few steps further back, e.g.: firing up some songs in Audacity, checking the relative levels on FFT frequency plots, experimenting with crossover points, and getting a general idea of what's required to turn the volume up to 11 or 15 with minimum distortion, if '10' is the normal level at which clipping starts without any post-processing.
The soft-clipping could just as well be some smart level limiter, possibly with "read-ahead" that buffers and delays the entire signal. I see no problem, as long as the active filters and processing does not shift the relative phases.
Taking things a step further, the number of frequency bands is just limited by processing power, so you could define however many frequency bands you want to EQ, and independently limit, compress, or soft-clip just the ones that are using too much vertical bandwidth at any given moment.
The system could use a fixed table of values that defines what proportion of the available 'Xmax' each frequency band is allowed to use. E.g.: obviously 20-100Hz gets the bulk share, something like 75% of the total, and everything else gets small fractions. So there's probably no need to go crazy with 32 frequency bands.
As a refinement to that, software could dynamically allocate vertical bandwidth (max voltage levels). So if there's a loud bass thump and not much else going on, there's no reason to compress it to 75%, when it can use 100% until some other frequency bands need that space. Tonnes of options to play with...
PS, this is obviously meant for things like portable bluetooth, running on low voltage with <80dB efficient full-range speakers used outdoors, where the 4-pi environment is much more demanding, volume wise, than indoors.
The algorithms I use are based on an algorithm called "convolution" that runs on the miniDSP, techniques like maxxbass, which creates 2nd, 3rd, 4th, 5th harmonics of the low frequencies to create the illusion of low frequencies, Dolby-like algorithms are like @abstract mentioned, BBE-like algorithms delay <150hz at 2.5ms, delay 150hz -> 1.5khz at 0.5ms, and higher frequencies are not delayed, however I'm not sure how TruBass works, but it's similar to maxxbass (generates harmonics). Rbass or similar divides the available low frequencies into frequencies 1 octave lower. You can test it with jameDSP or viper4android on rooted Android or an API provider like shizuku
Tôi đã từng có quyền truy cập vào một chiếc Devialet Phantom đã được tháo rời để sửa chữa, thật không may là tôi đã không chụp ảnh vào thời điểm đó. Nhưng nó chạy 3 tas5162 song song cho mỗi loa trầm 1,8ohm (tổng cộng 6 tas5162, tôi không chắc nó được kết nối như thế nào, nhưng điều đó sẽ được thảo luận trong một chủ đề khác), tuy nhiên tôi cũng đã thử nghiệm nó trên một chiếc Devialet Mania và phải cắt hoàn toàn dải trung vì hình nón của Mania quá nặng
I once had access to a Devialet Phantom that was disassembled for repair, unfortunately I didn't take pictures at the time. But it ran 3 tas5162s in parallel for each 1.8ohm woofer (6 tas5162s total, I'm not sure how it was connected, but that will be discussed in another thread), however I also tested it on a Devialet Mania and had to cut the midrange completely because the Mania's cone was too heavy
Please post in English.Moderation team
I also apologize to the wonderful enthusiastic moderation team and members of diyaudio for my mistake😱. I used high alcohol today