Live Edge Dipoles - #1 at Parts Express 2023 Speaker Design Competition - Updated Design

Except when you're playing a turntable, this should never need to happen. I use a $99 WiiM digital streamer. I stream music (Tidal or Qobuz) on my phone to the WiiM via airplay. The Wiim has a fiber optic output which plugs right into the DSP input. The signal is never analog until it comes out of the DSP, which goes straight to the amp and then the speaker. That means all of the crossover and EQ occurs digitally where it is done as close to perfect as you can get. WAAAAY better than analog crossovers.

There is no need in 2023 to have a traditional analog preamp. You only need that for phono.

This picture is a COMPLETE bi-amplified DSP and streaming system. See the Wiim at the back top of the amp. FIber optic cable going into the MiniDSP. MiniDSP on the left feeds into the 4 channel amp (Adcom GFA2535, 60Wx4) which drives the speakers. MiniDSP remote sets volume (I mostly use my phone for volume, not the remote) and MiniDSP remote switches between analog, USB and Fiber. The MiniDSP is the new preamp.

If you have a turntable then you need a phono preamp which you run into the DSP analog input. Purists won't like that, but basically it's as good as your Analog to Digital converter. You can buy a separate A-D for phono if you think the MiniDSP's is not good enough, and use one of the digital inputs instead.

Because in my opinion the advantages of digital speakers far outweigh the disadvantage of converting A to D and then D back to A.
adcom-minidsp-wiim.jpeg

So if you have this setup above you can build any biamped DSP digital speaker. Changing speakers is as easy as downloading a new configuration file for the new crossover, which takes about 5 minutes.
 
Last edited:
  • Like
Reactions: 1 users
Member
Joined 2021
Paid Member
Yes, I meant the turntable but also my SACD player. So in my case it seems to be a no-go.
Or with an active analog xo - that was my thesis 33 years ago; but I think it would be way too complicated, if not impossible.
But the Walnut Dipoles will come :)
 
  • Like
Reactions: 1 user
It’s true that if you have a digital source then the signal remains in the digital domain until it gets converted at the output of the MiniDSP unit. HOWEVER, the Analog Devices DSP modules that are used in the MiniDSP kit perform a sample rate conversion on all incoming signals and my experiments with a 4x10HD unit certainly convinced me that the SRC degraded the sound to a marked degree. I have no idea if this is also true of the newer units.

In contrast if you use Acourate or AudioLense the source signal remains in its original sample rate and the convolution engine adapts the filter to the signal.
 
  • Like
Reactions: 1 user
My experiment was as follows. I started with this chain: digital source > DAC > amp > passive speakers. I familiarised myself with how this sounded and then inserted a 4x10HD between the source and the DAC. It was set up to leave the signal untouched so that the signal was passing straight through without any DS processing. The sound quality before and after was markedly different. That was enough for me and I concluded that the SRC was very likely the cause of the degradation. I didn’t investigate further.

I’d encourage others to try this experiment as I would be very interested to hear if they have a similar experience.
 
I too noticed that the SRC in my mini-DSP 2x4HD seemed inferior to the implemented upsampling using SOX in my RPI. Since virtually all my digital music was rendered by my RPI this was not an issue. I now am using the mini-DSP FLEX (2x4) and I am hard pressed to tell any difference. The DAC in the FLEX is also far superior to the 2x4HD. I am also using the Flex as my preamp just like Perry. The analog in works great for my TT, as a subsonic and eq on the anolog input. I have thought about using deconvolution filter on the TT to eliminate crosstalk and do phase correction, but since I don't spin that much vinyl it hasn't been worth it.
 
@Arnold Harringer as @drewmc says, the D-A on the Flex8 is much better than 2x4HD. The noise level has got to be at least 10dB less. Maybe 20. For example: On the Live Edge Dipoles (95dB spl tweeters) and Bitches Brews (100-106dB spl tweeters), if I connect the 2x4HD to a typical amp with 26-28dB gain, hiss is plainly audible from the tweeters. You have to attenuate, either by turning down inputs on the amp or inserting 12 dB attenuators etc.

That same noise is not audible from the Flex8.

The Flex Eight sounds significantly better.

Above I provide files for both 2x4HD and Flex Eight, for both Bitches Brew and Live Edge Dipoles. Files for 2x4 are not compatible with Flex Eight.
 
@perrymarshall: Perry, did you consider having the passive crossover between the mid and tweeter rather than between the woofer and mid and if so what led you to choose the latter?
To put a finer point on it - in both these dipole systems, the woofer+bass mid blends very naturally because [as discussed in detail here] a series crossover beautifully handles the amplitude issues with only one inductor and one capacitor. The only phase issue at the 100Hz xover point is the fact that the midrange driver needs to have inverted polarity and reverses phase.

The good news is that the drivers are on the same plane and there are no time delay issues relative to the long wavelength. A simple maximum phase filter in FIR fixes the phase reversal and you are back to linear phase.

If I had put the passive crossover on the tweeter, the passive CO would never be able to solve the 1" / 2-3cm offset between the midrange cone and the compression tweeter, which sits behind at the rear of the magnet. DSP applied later can sort-of solve this on paper, but it's an inevitable physical fact that if the mid-tweeter xover is passive, the two drivers are separated by 2-3cm when the wavelengths involved are 10-30cm / 4-12". Their signals arrive at different dates.

The only good way to solve this is to set the active crossover at this frequency and use time delays in the DSP to line the time arrivals exactly.

Also if you look at the drive signal curves you see that the EQ curves applied to the upper crossover are pretty complex in both systems. This is another strong reason to avoid passive crossovers for the mid-tweeter transition.

Phase and impulse response get fairly little attention from most speaker designers. Mostly those things are an afterthought. A lot of people shrug them off and dismiss them as inaudible.

A lot of purist audiophiles apply 6 dB crossovers but don’t actually know what they are doing and don’t achieve linear phase. It’s very hard to pull off.

I obsess about time and phase because they are EASILY measured. I insist they affect imaging and clarity even though listeners don't hear phase response overtly per se, the way they hear EQ.

I've always admired John Dunlavy (Duntech) and companies like Thiel and Vandersteen who made time a front-burner priority in their very disciplined, highly engineered designs. I believe such careful attention to time is a big reason why these systems sound so good.
 
Last edited:
  • Like
Reactions: 2 users
This is a Vituixcad simulation of the 18" Eminence Kappa 18LF + Radian 5208 Mid-Woofer in the Live Edge Dipoles:

live_edge_dipoles_eminence_18LF+Radian5208_vituixcad.png


The mid-woofer signal is -10dB at 70Hz and -30dB at 30Hz, which is exactly what you want to minimize excursion for the 8" Radian. The impedance peak of the Eminence works to our advantage there. The gentle 6dB/octave roll off of the 18" makes a seamless blend between the two. Since dipole loss kicks in below 200Hz for the mid-woofer, the effective acoustic crossover frequency is 200-250Hz. This is based on my measured impedance data from the drivers.
 
Last edited:
Hi Perry, Thank you for the work you share! It was your open baffle projects that got me interested to build speaker projects myself. I completed a project using Home Depot butcher block countertop that is a variation on the Lii-Song open baffle speaker designs, using the Lii-Song Fast-15 paired with their 15 inch woofer and a rear firing tweeter based on your designs. They do sound amazing, but my retirement gift will be my next project which will be the Bitches Brew live edge. I am very happy you have added the Flex8 files. I have the minidsp flex and was planning to upgrade to flex 8 for the Bitches Brew project. My retirement date will be in 2025, so will keep an eye out for shows you present your open baffle designs at and will attend to see up close, so I get everything right. I am in St. Paul, MN, so would be easy to make the road trip to Chicago Land for one of your events.
 
  • Like
Reactions: 1 user
Fantastic! So glad to hear you've gotten some inspiration here. I think Butcher Block is a great basis for a live edge design. In fact... I've never posted these before but this is a project very similar to the Walnut Dipoles from 3 years ago. These have an Eminence Alpha 15a, an Audax TW034 and a Troels Gravesen 6.5" waveguide. My carpenter Seth Cothron made this "bowling alley" flavored butcher block panel.

eminence_alpha15_audax_waveguide_open_baffle.JPG


Keep me posted on your rendition of the Bitches Brews. Have you posted any photos of your project? I'd love to see 'em.
 
  • Like
Reactions: 1 users