Seas C18EN001 coax in open baffle?

[Kjetil]

Hi everyone, I hope this is the correct forum.

I’m new to speaker building and do not really want a new hobby, yet for some reason I feel the urge to build my own speakers. I’m sure you all encounter oddities on a regular basis, with hairy ideas and little or no insight, so I’ll make a short introduction of my skillset and the research I’ve done on the subject before explaining what I want to attempt;

I’m an electrical engineer (I design/build/troubleshoot/program PCB prototypes for a living), so I’m familiar with both physics and electronics, I’ve read up quite a bit on loudspeaker physics, such as baffle step effects when a sound wave expands over baffle and if short enough wavelength it’s directed by the baffle into 2pi (half) space, otherwise it interacts with 4pi space around the baffle edge and the response dips, I understand the basics of waves, phase and interference creating peaks or dips in response, whether we’re talking waves in the water or sound waves. I’ve peaked on challenges with acoustic offsets between drivers (primarily offsets tweeter/mid due to short wavelengths = big phase error) and seen that many approaches exist; some use asymmetric filters between tweeter/mid to correct phase electrically, some tilt the baffle, listening off-axis, some (notably I’ve seen Troels’ projects) make stepped baffles with the introduction of edge diffractions to consider. At longer wavelengths (mid-to-woofer in 3-way) the phase offset resulting from acoustic offset shrinks with lower crossover frequency and is thus not as critical. OK, my problem is rather the stuff I do not know, “unknown unknowns”; I realize speaker design is an incredibly complex undertaking, more so than many outsiders appear to appreciate, so I’ve come to kindly ask advice. Reason for choosing this forum was an apparent high level of competence and lack of BS, so it seemed the right place.

The project I have in mind is a 3-way, but for now limited to the top 2/3 (woofer comes later when baffle and passive mid/tweeter crossover is sorted out);

I want to try the C18EN001 magnesium coax on open baffle, relatively narrow baffle, in a similar setup to Dan’s (on this forum) Basslines. The C18EN001 is time-aligned by design (yep, the newbie dodged a bullet), so I’ll avoid some pitfalls in lobing issues and time-alignment issues. A 3-way design is very ambitious, agreed, but this is not the worst 3-way. I have some starting points for a passive split mid/tweeter based on Troels’ work and SEAS’ proposed setup, but I know I have to get an USB based CLIO SPL measurement kit and play around in LspCad and tweak it to my baffle, or rather tweak my baffle (If I go forward, my thought is to make a very narrow baffle with flanges for detachable prototype “wings” so I can play around). No decisions is made on the woofer except it will probably be an 8 inch, and due to my non-existing passive crossover design skills, yet fairly competent PCB design skills, I’ll split woofer/mid active, either LR4 or DSP, along with necessary equalization.

My big worry is, I’m a speaker noob and I’m suggesting something I have not seen others execute properly. That is often a horrible starting point. I worry about midrange excursion in an open baffle setup, since the midrange cone acts as waveguide for the coaxial tweeter.

I’ve done a bit of calculations according to Linkwitz’ instructions, and If I understand my challenge correctly, my impression is that with a suitable baffle with sufficient Fequal (frequency where max SPL will equal that of a small closed box), and an arbitrary chosen (active 24dB or DSP) crossover point not below 200Hz, I should be able to achieve an SPL of (assuming Linkwitz’ Fequal implies I can use closed box equations at F>Fequal):

SPL=20*log10(1.18/0.00002*Sd*Xmax/sqrt(2)*2*pi*Fcross^2)

Where Sd is effective piston area of the midrange, 0.0131[m^2], Xmax is peak linear coil travel of the midrange, 0.003[m], and Fcross is the point of crossover to the woofer [Hz]. Again, if I’m not mistaken here, this is the maximum SPL I can achieve within the linear Xmax and frequency limits, implying I’m pushing the midrange excursion to its max, something I obviously do not want to do as it’s the tweeters waveguide.

So within more reasonable limits (now I’m really just guessing), let’s say solving for 1mm peak excursion on the midrange Xmax=0.001m and a crossover point of Fcross=300Hz, I get a max SPL of almost 110dB which is more than I’ll need even for HT gunfire. Solving for max linear Xmax (again, not optimal, but for reference) I get over 119dB so I should be able to use the driver on an open baffle all the way down to 300Hz, providing the SPL I need, without compromising Xmax, yes?

Any thoughts or corrections would be greatly appreciated.

Kjetil

[5th element]

https://www.linkwitzlab.com/spl_max1.xls

Is quite useful for looking at this.

Edit - With a path difference of 190mm for the baffle width and an assumed one way xmax of 3mm I get 113dB for this driver.

Due to the suspension design I would expect to see a reasonable increase in 2nd order distortion products as you approach this though. The suspension doesn't appear to have that much symmetrical throw. I wouldn't worry about it though 113dB is ridiculously loud and I am sure you'd never reach that.

[JonMarsh]

With sufficient Fequal you're going to have a rather large baffle (have you done those numbers?) which somewhat defeats the goal for using a dipole, that is, the figure 8 dispersion pattern. The most common practice these days (see Linkwitz's current projects) is to run as small a drivers a possible, and run them as much as possible in the dipole roll off region, so that the power response is consistent - but this adds up to a fair amount of EQ and a fair amount of travel required in the lower range of each driver.

To illustrate this, do the same calculations as you have above, but set Fequal to be at the top of the C18EN001's operating band for the midrange or as close as practical. Then look at those numbers.

One other point to mention, describing the Xmax as the peak linear travel is notationally convenient, but that doesn't mean you want to use the Xmax in practice, as this number is usually derived from the Bl fall off, and 70% of the centered value is considered "linear". This holds fairly true unless the driver uses an underhung voice coil (sits within a magnetic gap substantially longer than the VC length. Some drivers do this, (the bigger Accuton midranges with Neo magnet system that sell for about $900) but they are difficult to build and expensive. That C18EN001 isn't one of them.

[Kjetil]

With sufficient Fequal you're going to have a rather large baffle (have you done those numbers?) which somewhat defeats the goal for using a dipole, that is, the figure 8 dispersion pattern. The most common practice these days (see Linkwitz's current projects) is to run as small a drivers a possible, and run them as much as possible in the dipole roll off region, so that the power response is consistent - but this adds up to a fair amount of EQ and a fair amount of travel required in the lower range of each driver.

Thanks Jon, 5th,
That's a bummer. I kind of figured out something was fuzzy with my numbers when playing with the "Edge" program, as the cancellation rolloff started way higher than expected with the baffle sizes I envisioned. I'm pretty dead set on the driver itself, so with that in mind, my options are pretty much either putting it in a box, or pair it with a open baffle W22EX001 8-inch magnesium (Seas is a local company, and since the mid/tweeter is magnesium that would be my choice) and cross higher, but I'd have to add a sub as well. I guess it's back to the drawing board, if you have any thoughts on which path makes most sense please shove your $0.02 in. I'm a big fan of some open baffles I've heard, but if it doesn't make sense with the Seas coax, I'll put it in a box, end of story.

Kjetil

EDIT:

Xmax as the peak linear travel

The Xmax I used (3mm peak) was actually the linear, Xmax parameter is twice that, 12mm p-p. But with this driver, as you well know, the cone is the tweeter waveguide, so I'm not really sure I want to use linear Xmax.

[5th element]

Well in my simulation a 19cm wide open baffle would manage 113dB at 1 meter @ 300Hz. That's using the 3mm one way xmax spec.

Given that a 4th order acoustic high pass would be used and most likely of the LW variety, the target slope would be 6dB down at 300Hz. This significantly relaxes the requirements as you would only need the driver to be able to produce 107dB @ 300Hz for a 113dB nominal within the drivers pass band, then you've got bandpass gain to add into this too which will raise the final loudspeaker sensitivity by a dB or two.

Incidentally the C18 is 87dB sensitive and is a nominal 8 ohm load. With 256 watts you'll be getting 111dB out of it. This is the drivers short term maximum power handling reserved for seldom encountered transients on music with a large average to peak ratio, ie classical music. The long term max power handling is only 100 watts and even this will be for a bandwidth limited noise stimulus rather than continuous. I'd take the 100 watt figure to more relate to modern highly compressed stuff, so you're looking more like 107-108dB max nominal.

The numbers point towards the driver suffering from thermal overload before it runs into true excursion limitations. I really don't see where the problem is, the numbers suggest this would work absolutely fine.

Just go ahead and build the thing.

What driver did you have in mind to take over below the C18 anyway?

[Kjetil]

Well in my simulation a 19cm wide open baffle would manage 113dB at 1 meter @ 300Hz. That's using the 3mm one way xmax spec.

...

The numbers point towards the driver suffering from thermal overload before it runs into true excursion limitations. I really don't see where the problem is, the numbers suggest this would work absolutely fine.

Any guess to what type of tweeter distortion products a moving waveguide would induce anyway? Doppler effects?

Jon,
You have the C18 at hand no? Is there any way to measure tweeter distortion at midrange excursion, eg run the mid at some low freq high SPL sinewave while making tweeter distortion measurements?

And is there any rule of thumb max crossover point when crossing OB mid to boxed bass?

[Kjetil]

What driver did you have in mind to take over below the C18 anyway?

No clue atm. Not even sure whether it will be dipole, ported or sealed... I suppose an open baffle 10" W26 would't do the trick bass wize (as bottom end), perhaps if a single 12" driver could deliver enough bass yet handle the mids below the C18...

[5th element]

I suppose it depends on what kind of loudspeaker it is you are wanting to design, but looking at what different radiation patterns bring to the mix, at least in my opinion...

Essentially you've got three areas within sound reproduction, the bass, the midrange and the treble. This may sound ironic and rather inane, but it's the truth.

In the bass area you are plagued with room modes. These create unevenness within the bass and are local in that you can EQ the bass flat at one point, but then you'll completely screw up the bass at another. To help mitigate room modes there are two general ways to go about it. The first is to add in a number of radiation sources, a la multiple subs, with judicious use of EQ, to energise the modes of the room from many different locations and thus help to smooth them out. You can then EQ from many different points and create a nice blend that averages out well across several listening positions. Proponents of this approach usually recommend having your two main loudspeakers with reasonable bass extension, then add in 3 or 4 additional sources of bass energy. You only need one sub to be truly capable down to 20Hz however, down that low your room is not modal any more and does not benefit from the multiple sources. The other subs only need to go down as low as your room is modal. In my case that is around 40Hz. As a result this significantly relaxes the requirements placed upon the additional subs.

The main idea is to have one large sub capable down to 20Hz and run as the 'main sub' as it were, then to add in two or three more small sub units, with say 8" drivers, to help smooth out the modal region. This doesn't have to be madly expensive to do either as you could use a miniDSP to do the EQ and then buy some small powered subs from ebay. Nothing amazing is needed for the filler subs but something decent from a reputable manufacturer would be a nice idea. My ideal filler type sub would be B&Ws PV1 these are compact, sleek and cover just about the right range. Trouble is these are stupid expensive, but something like an older REL Quake would make a very good unit for this.

The other idea is to use main loudspeakers that have some kind of radiation pattern at low frequencies that is known to help energise the modal region of a room in a way that helps to control them. People use cardioids, dipoles and all sorts of other things to try and improve bass reproduction. Dipoles do indeed work very well to this end, but they have one problem, inefficiency.

One could say that the best region to have a dipole radiation pattern would be for the bass, but this is the one area that you lose the most efficiency so it's a double edged sword. You only need dipole extension down as low as your room is modal, but in most cases that usually means extension down to 35-40Hz and for enough output requires a quad of sub drivers, such as Linkwitz uses in his Orion. The caveat here is that the subs can only be used up to around 100Hz as the H or W frames traditionally used introduce their own issues much higher than this.

With the sub units only able to reach up to 100Hz and with most midrange drivers not capable any lower than around 300Hz, you've got a problem. You need to fill in that 100-300Hz range. In his Orion LW decided to go with a large diameter midrange and cross it low to a tweeter. This has its advantages as it means you can cross the large mid directly to the sub section, but it has implications further up into the spectrum if you're wanting to cross it directly to a tweeter. To get around these issues LW built the LX521.

If you were going to go for a full dipole loudspeaker with the C18, you would need a driver such as the RS225 sitting below it to cover the 100-300Hz range. That said you could easily cross the C18 and the RS225 at 500Hz, this would relax the issue on the C18 considerably and not get you into any trouble elsewhere. The RS225 is at its most linear around the 300-400Hz range so you would be doing yourself a favour. I mention the RS225 because it's a stellar unit, inexpensive with a lot of excursion capabilities and very low distortion. Okay adding in another channel of amplification and xover isn't cheap either, but the point is, the drive unit doesn't have to be.

From the top down this time you've got the treble region next. This area is particularly important because we use these frequencies for sound stage localisation and to provide us with a stable sound field. Reflections from the room can cause these to become smeared and this usually results in a lack of precision to the sound stage and can result in loudspeaker localisation. Personally I find both of these intolerable. A less than rock solid central image with sibilants bleeding through to the left and right speakers off of the central image? Yuck! No thanks.

Controlling the directivity of the treble region helps dramatically with this and also aids in widening the sweet spot if you're going with a constant directivity design. Constant directivity literally reduces the amount of output thrown out at wide angles (and thus reduces reflections and the impact of the room) and funnels it into a cone where there is a region of angles, over which, the sound emitted is even in intensity. This means that even if you go off axis, you're not going to get any fall off in upper extension and it also means, if you've got a directivity match to your mid driver, that you're not going to get any nulls appearing at the xover frequency as you go off axis either. Dipoles, with their figure of 8 radiation patterns, do obviously throw out far less sound towards the sides, but compared to a proper constant directivity wave guide, they are somewhat lacking (at least in my opinion) as I would rather not be firing sound out backwards too. Also to maintain the dipole radiation pattern, and thus directivity control, high up in frequency, you absolutely need to have a very narrow baffle around your high frequency driver. There is a very good reason why JohnK and Linkwitz have used those odd shaped baffles around their latest incarnations, that is to maintain the dipole radiation profile as high up in frequency as possible.

The original Orion did not have a dipole radiation pattern up particularly high, as when the tweeter took over it changed into a normal monopole. LW tried to get around this later by adding in rear firing tweeters, but the design was already flawed due to its dimensions and hence the LX521 was born.

Between the bass and treble region then you've got the midrange. Due to the frequencies that this covers it is less important to have a degree of controlled directivity here than with the others. There are still improvements to be had in controlling the directivity down lower than where the usual tweeter would take over (say 2kHz), but it is of less importance. Yes you are going to see gains in controlling the directivity all the way down to 500Hz, but there is usually a large price to pay for that. Efficiency and loudspeaker complexity, in the case of a dipole, and speaker size (requiring huge wave guides) and impracticalities associated there with of true constant directivity designs.

What's important to remember also is that you want to transition smoothly from one radiation type into the next. This is what a nice, closed box, constant directivity design does well. It starts out monopole radiating spherically. It then transitions smoothly through baffle step into radiating hemispherically and then smoothly transitions into its region of constant directivity if crossed over correctly to the tweeter.

Using the C18 already gives you the nice smooth transition towards constant directivity, but to maintain a nice radiation profile below this you will need to keep the baffle it is mounted on as narrow as is possible. This is why I chose 19cm, even narrower would be better. If you want to keep things dipole below the C18, then you will need something like the RS225 mounted below it on a baffle as narrow as the RS225 can fit on and crossed over appropriately. Then with a sub section below this.

The alternative, as far as I see it, would be to use the C18 down to around 300Hz in a nice low diffraction closed box, with something like the RS225 below it, covering from 50Hz up to 300Hz in a nice ported enclosure. This is far simpler than all the dipole shenanigans as it only requires a modest three way with fairly low requirements on the bass unit for excellent results.

I would then go the multiple sub route, rather than dipole, up to around 120-150Hz to give you really nice in room bass with a general freedom from room modes.

It is of course entirely up to you. Both solutions will work well if executed correctly and will end up sounding great. I just prefer the non dipole option as its less wasteful of your swept cone area, easier to design, easier to integrate into your room and is far less critical of what your room is actually like. Dipoles need a lot of space around them to work properly.

In a nutshell I would be hesitant about a design that was monopole up to 300Hz, dipole from 300Hz to 2kHz and then constant directivity from 2kHz up to 20k. You've got three radiation patterns converging at frequencies that are quite important to sound reproduction. I am sure it can be made to work but you'd need to make sure you had dotted all your i's and crossed your t's to make it do so.

[Kjetil]

Thanks Matt! I'll need to re-read that a couple of times, and do some thinking. My immediate thought was the room consideration. I'm renting apartment for the time being, but plan to buy a place with my wife in the not too distant future (we got married just a couple of weeks ago), so I have a huge problem in not knowing what future space the speakers will reside in. For this reason, i cannot guarantee the speakers appropriate space (other factors will probably carry more weight when it comes to deciding to buy a house). WAF will also be a consideration. I'll consider the multi-sub option, but regardless what path I choose I'll definitely take your advice on going for one or the other, not the boxed bass dipole top. Thanks alot for an incredible insight into factors I had not considered at all!

Kjetil

[5th element]

I figured it would be worth expanding on things rather than giving some half hearted response and not properly cover everything.

The only thing I think I really forgot to say is that with the OB you obviously don't need to design and build a really good box for the c18.

[Kjetil]

One quick question regarding gear... I’m dead set on active crossover in this project, as I don’t plan on acquiring extensive passive crossover skills. In this age of DSP, doesn’t it make much more sense with a MiniDSP than laying out a custom tweaked opamp-forest with countless revisions? Whether I’d go dipole or boxed with sub-satellites, 8x outputs would suit me nice, so I’m looking at the miniDSP 2x8 kit with the DIGI-FP SPDIF/Toslink/AES-EBU interface and a UMIK-1 for measurements. Through your audiophile eyes, does it seem like a generally recommendable kit?

Even though I’m not yet sure what I’ll be building, I thought it looked like a neat kit to acquaint myself with, since I can just hook it up to my signal generator and use my 4-channel digital 100MHz scope to verify it does what I ask of it.

[Kjetil]

I figured it would be worth expanding on things rather than giving some half hearted response and not properly cover everything.

The only thing I think I really forgot to say is that with the OB you obviously don't need to design and build a really good box for the c18.

It is really really appreciated, Matt. It's hard to find the right words to express, so much potential waste of time prevented by a proper introduction to what I'm undertaking...

If boxing in the C18, I'd probably build an aperiodic cabinet (perhaps even completely closed at the back, but I went over the angles and they do a great job bouncing stuff back and forth) like in Troels Gravesen's "Cyclop", and dampen as seen in these pictures (felt padding everywhere, with acoustilux stuffing at the back to tune),

http://www.groundsound.com/customer/ClausC/ClausC_top_1.jpg

http://www.groundsound.com/customer/ClausC/ClausC_top_2.jpg

EDIT: To quote Troels,

Why this odd interior for the middriver? Well, when SEAS has done all that possibly can be done to eliminate reflections from the driver's rear structure, the least we can do is ensure that our cabinet doesn't reflect any energy back through the cone. So, I've tried to design a box where the rear energy is absorbed as good as possible and due to the slots in the rear, we have an aperiodic system, reducing impedance peak at resonance frequency. Same as for the bass, only here managed by adding compressed damping material to the cavity.

Which is just as true for the C18 as for the M15.

[Kjetil]

Disregard the miniDSP question, the fact that you already mentioned it slipped right through when I first read your tutorial - I guess I was too focused on the technical aspect of your writings.

[Kjetil]

Just playing around a bit these late hours, this could make a killer stand mount monitor, paired with a suitable woofer for some ~20L sealed and the DSP running a Linkwitz transform to extend bass... With some decent woodwork, WAF wouldn't be a problem at all...

[Kjetil]

If I can get whats needed from a pair of stand mount monitors and have 4x DSP channels to spare (T/M crosses passive) for powered subs, it's certainly an attractive alternate to what I first envisioned.

[Kjetil]

The alternative, as far as I see it, would be to use the C18 down to around 300Hz in a nice low diffraction closed box, with something like the RS225 below it, covering from 50Hz up to 300Hz in a nice ported enclosure. This is far simpler than all the dipole shenanigans as it only requires a modest three way with fairly low requirements on the bass unit for excellent results.

The more I think about it, the more sense this approach makes, whether it be space considerations, system complexity, or WAF.

Assuming I choose this path forward, how would your recommendation on woofer setup go? Options would probably be ported, TL, or sealed with Linkwitz transform from the DSP.

What caveeats do I face if choosing a stand-mount monitor vs a floor-stander (apart from restricting volume available for ported/TL)?

Is one approach or the other more favorable when it comes to room integration or other factors?

And lastly; I'll be building the DSP based drive system as a parallel project, and I can see my budget already stretching (hardwood and CNC services also cost). I'll need four channels of amplification, and the Hypex nCore modules are a bit pricey I think. I'm thinking along the lines of reasonably powerful Class-D or gainclone amps, and if I think sealed bass / linkwitz transform I need to budget quite some watts for the bass channels. I do actually have some experience designing SMPS supplies (even mil spec EMC emission), although not in the same power league, but it would probably be alot less painful just buying them off-the-shelf if they can be had for a reasonable price.

EDIT: You did indeed write "ported" but i figured this was generally speaking rather than an actual topology recommendation, correct me if I'm wrong.

[Juhazi]

Kjetil, I to othink that dipole mid is not a wise choise in this case. Minidsp is practical and versatile.

Your scetch is pretty close to what SEAS is suggesting for C18 and close to what KEF and TAD are doing. Even better would be a floor-standing 3-way with double bass drivers (like TAD Evolution One http://www.stereophile.com/content/t...c0zARvrD1rZ.97 Then you would get less distortion and minimize floor bounce effect.

[Kjetil]

Two drivers have less distortion than one? Is that because excursion is reduced for equivalent SPL sum, that also would be achieved using a bigger driver with equivalent piston area of the pair?

Anyway, I've seen some pretty good cases for floorstanders vs stand mount monitors, and actually going down from 8" to 2x7" would make a slimmer floorstander with IMO a more attractive appearance.

Matt,
You made the case for the RS225. Is the RS180 (2x) as recommendable, ported? with the C18 on top, slim floorstander?
MTM configuration could also be an idea with that setup I guess, but it'd make a taller speaker than I want if I'm having the C18 near ear-height, so I'd rather have the C18 on top.

[Kjetil]

So I'm set with a floorstander with the C18 on top. I've dug a little into room modes and floor bounce and have run some numbers through my MATLAB. To simplify this, I'm back at one 8" woofer, probably the RS225, ported. Now, if I get this correct, the mid should be high enough above the floor so that first floor bounce cancellation occurs below the crossover frequency, and the woofer low enough so that it's first floor bounce cancellation occurs above its crossover frequency. To achieve this without a too tall speaker (WAF) I'm thinking I'll move the crossover point up to around 400Hz. What margins I will need here for best results I do not know (feel free to advise, keep in mind miniDSP XO slopes), but from arbitrary guesstimates 250cm from speakers at 95cm ear height I'd suggest having the coax (mid) at 80cm or 85cm for 314Hz or 297Hz cancellation, and the woofer at 40cm or 45cm for 611Hz or 545Hz cancellation, respectively. Are these numbers sane or should I adjust any?

And if I were to consider dual woofers, the second one would have to go below the one positioned as described above, as cancellation frequency increases with declining distance to floor.

This means the woofer would be 11.3deg off axis at 45cm from floor or 12.4deg off axis at 40cm from floor. Would I be wise to slope the front panel or even build a sloped woofer cabinet with a separate straight baffle coax cabinet on top? 400Hz is up in the mids and I'm thinking it could be a problem having it pointing at my feet...

[Kjetil]

Here's the MATLAB bounce function by the way, in case it's of use to anyone

function [ f ] = bounce( d, h1, h2 )
% Arguments:
%    d -> listening distance [m]
%    h1 -> driver height above floor [m]
%    h2 -> ear height above floor [m]
% Returns:
%    f -> frequency of cancellation

c=344; % speed of sound [m/s]

% simple Pythagoras, get distance to driver and to that of mirror image
l1=sqrt((h2-h1)^2+d^2); % hypothenus distance to driver [m]
l2=sqrt((h2+h1)^2+d^2); % hypothenus distance to driver mirror [m]

delta_l=l2-l1;

% cancellation occurs when mirror is in antiphase, thus delta_l equals half wavelength.

lambda=2*delta_l; % wavelength [m]

% λ = c / f ->
% f = c / λ
f = c / lambda; % frequency of cancellation [Hz]

end

EDIT: Spreadsheets are nice, but say for instance you had a predetermined crossover point and wanted half an octave or an octave to cancellation on either side (I don't know needed margins yet), this code could come in handy to generate the numbers, either by rearranging the code to solve for heights or by running the function as an iteration.

[augerpro]

boxycad has a floor bounce calculator that might save you some work: http://www.audio.claub.net/software.html

[Kjetil]

Thanks Brandon, that would be alot less work agreed (and I've actually played a bit with boxycad). It's more about understanding the parameters of a speaker for me at this point. For that to happen I need to chew on some of the physics :??

But I feel I'm getting a bit clearer idea of what to build now :T

[Juhazi]

Right on! Next question is the center to center distance of the woofer and mid. MATLAB should do that, more simple with XDir or alike http://www.tolvan.com/index.php?page=/xdir/xdir.php
Should not be a problem.
-
Jon Marsh here is actually up to something very similar... https://www.htguide.com/forum/showth...Besides-Troels

I found the missing plans of KINGRO4Y at Madisound http://www.madisoundspeakerstore.com...RO4Y%20Kit.pdf
and a kit with Hypex plate dsp-amp https://www.madisoundspeakerstore.co...aker-kit-each/

[Kjetil]

Right on! Next question is the center to center distance of the woofer and mid. MATLAB should do that, more simple with XDir or alike http://www.tolvan.com/index.php?page=/xdir/xdir.php
Should not be a problem.

If I'm correct we are addressing vertical lobing here? I haven't had the time to finish the final expressions, but this would be the general idea (i use tweeter in comments but it's mid in my case);

% d -> horizontal distance from mic to tweeter acoustical offset [m]
% w_off -> woofer horizontal acoustic offset, positive value adds to distance to mic [m]
% a -> angle off tweeter axis, positive angle going upwards [rad]
% h_diff -> delta height above floor between drivers (aka center to center distance) [m]
% f -> frequency [Hz]

c=344; % speed of sound [m/s]
lambda=c/f; % wavelength of our frequency [m]

% cos(a)=adj/hyp -> hyp=adj/cos(a), hyp is actual distance from virtual mic to tweeter [m].
l1=d/cos(a);

% tan(a)=opp/adj -> opp=tan(a)*adj, opp is vertical elevation of virtual mic above tweeter axis [m].
opp=tan(a)*d;

% now from woofer to virtual mic vertically we have new_opp=opp+h_diff, and due to the horizontal woofer acoustic offset we have new_adj=adj+w_off.
% the actual distance from virtual mic to woofer can then be had as the hypotenuse through Pythagoras,
l2=sqrt((opp+h_diff)^2+(d+w_off)^2)

The difference between l1 and l2 is some fraction of the wavelength lambda, from which we find phase error that can further be expressed as interference in dB, right?

[Kjetil]

I found the missing plans of KINGRO4Y at Madisound http://www.madisoundspeakerstore.com...RO4Y%20Kit.pdf
and a kit with Hypex plate dsp-amp https://www.madisoundspeakerstore.co...aker-kit-each/

I got a bit anxious as to why everything about this driver kept disappearing from the Seas website, and since I speak norwegian (I actually drive past the factory every now and then) I called the guys at the factory. They had no clue or explanation, but could inform me the driver was in regular production so it should not be an availability issue... FWIW.
EDIT: typo

[Kjetil]

I must admit this isn't a math area I excel in, and I have a hairy feeling about the simplistic approach to dB response. But I can't off the top of my head point out any error either, so this is what I've got;

% d -> horizontal distance from mic to tweeter acoustical offset [m]
% w_off -> woofer horizontal acoustic offset, positive value adds to distance to mic [m]
% a -> angle off tweeter axis, positive angle going upwards [rad]
% h_diff -> delta height above floor between drivers (aka center to center distance) [m]
% f -> frequency [Hz]
d=2.5;
w_off=-0.03; % negative value means woofer is brought forward, due to a sloping woofer cabinet (coax in separate "monitor" on top alas Tarkus speaker)
a=-10*pi/180;
h_diff=0.4;
f=500;

c=344; % speed of sound [m/s]
lambda=c/f; % wavelength of our frequency [m]

% cos(a)=adj/hyp -> hyp=adj/cos(a), hyp is actual distance from virtual mic to tweeter [m].
l1=d/cos(a)

% tan(a)=opp/adj -> opp=tan(a)*adj, opp is vertical elevation of virtual mic above tweeter axis [m].
opp=tan(a)*d;

% now from woofer to virtual mic vertically we have new_opp=opp+h_diff, and due to the horizontal woofer acoustic offset we have new_adj=adj+w_off.
% the actual distance from virtual mic to woofer can then be had as the hypotenuse through Pythagoras,
l2=sqrt((opp+h_diff)^2+(d+w_off)^2)

% the difference in driver distances is some fraction or multiple n of
% wavelength, l2-l1=n*lambda ->
n=(l2-l1)/lambda

% cos(0) = 1 ->
dB=20*log10(1+cos(n*2*pi)) % OK?? we add phase shifted cosine to a 1 (an unshifted cosine), and take dB relationship of sum/1?

% next, we do it all again in a single expression: compute an array for different angles based on the expressions
% derived above, and plot.
x=[];
y=[];
for a_rad = -pi/4:pi/64:pi/4;
    x = [x a_rad*180/pi];
    y = [y 20*log10(1+cos(((sqrt((tan(a_rad)*d+h_diff).^2+(d+w_off)^2)-d/cos(a_rad))/lambda)*2*pi))];
end
figure(1);
hold
grid on
xlabel('[deg]');
ylabel('[dB]');
plot(x,y);
hold

This is my plot from -45deg to +45deg:

[Kjetil]

Same plot of my current speaker drawings. w_off=-0.06 h_diff=0.4m, at 2.5m listening distance on horizontal tweeter axis.
Green is 200Hz, blue is 400Hz (crossover point) and red is 600Hz. As tweeter (and mid) sits at 84cm in my current setup, I'd be listening slightly above axis (about 1.5 degrees), so if I've got both the math and the understanding right on this, it doesn't look bad at all:


The current state of the plans look like this now (only internal air volumes and geometry sketches drawn):


But I'll verify the math with the Tolvan software before celebrating :roll: but probably not tonight.

[Kjetil]

BTW, the drivers in the sketch are perfectly time aligned to listening position (not considering individual acoustic offsets, but drivers are same depth and wavelengths fairly long at 400Hz). Cancellation frequency (floor bounce) from the mid as drawn is 300Hz, and for the woofer the number is 544Hz, XO at 400Hz.

[5th element]

Crikey tons to respond too! Better get on it.

Through your audiophile eyes, does it seem like a generally recommendable kit?

Even though I’m not yet sure what I’ll be building, I thought it looked like a neat kit to acquaint myself with, since I can just hook it up to my signal generator and use my 4-channel digital 100MHz scope to verify it does what I ask of it.

It is decent and the flexibility gained from using a decent DSP unit far outweighs some of the downsides. One of these is the way the miniDSP controls the volume, or the way in which you will control the volume. If using an analogue or digital input you're going to want to keep the signal level high, that is you do not want to control the volume prior to the DSP. The same is true internally to the DSP too, you don't really want to be adjusting the volume using the DSP unit itself. This is one of the major flaws with the miniDSP stuff, what you really want is a multi channel analogue volume control solution after the DAC to maximise performance.

For the best performance with the miniDSP you are going to want to use its digital input.

The other issue with the miniDSP is that the ADC/DACs used are not great. They are by no means bad, but they could certainly be better. I see you are interested in buying the digital expansion board which is a great idea as using a digital input makes a lot of sense. So would using a digital output to a nice multichannel DAC, something using an ES9018 for example at a later date as an upgrade path.

The ES9018 actually has a 32 bit, bit depth in built digital volume control so you do not lose quality when you use this.

The more I think about it, the more sense this approach makes, whether it be space considerations, system complexity, or WAF.

]

Dipoles are great at certain things, but do have many quite considerable caveats.

Assuming I choose this path forward, how would your recommendation on woofer setup go? Options would probably be ported, TL, or sealed with Linkwitz transform from the DSP.

The only reason I mentioned ported was because the RS225 works very well in decent sized ported cabinets. It gives you great extension and will give you a lot of output. There are more ways to skin the cat though. Sealed will require a smaller cabinet and does give you the advantage of using a LT. If you're using a LT then you can go as small as you want, but only within reason. The smaller you make the box and the more EQ you need to apply, the less efficient the system becomes. This requires more amplifier power and places greater thermal demands on the driver. The maximum SPL that any driver can reach in a sealed cabinet, for a given frequency, is exactly the same regardless of the box volume. What changes is the amount of power required to get you there.

What caveeats do I face if choosing a stand-mount monitor vs a floor-stander (apart from restricting volume available for ported/TL)?

Not being able to attempt to counter floor bounce and/or increase bass system efficiency by getting some coupling to the floor from the bass driver.

I'll need four channels of amplification, and the Hypex nCore modules are a bit pricey I think. I'm thinking along the lines of reasonably powerful Class-D or gainclone amps.

TI have recently released a new class D chip amp called the TPA3251D2. This is absolutely awesome and comes close to rivalling the ncore in certain areas. The one thing it does not have is the same power output. Supply is an issue with these currently, but in a few months we're likely to see modules using this chip popping up. I don't know what your time frame is here but it's worth considering. To that end though you could always use some of the many TPA3116 board available on ebay. This are cheap as chips and I've use the same chip myself to great success. It sounds very good, certainly not an ncore or TPA3251D2, but as a means of getting you started that might be worth considering too.

Two drivers have less distortion than one? Is that because excursion is reduced for equivalent SPL sum, that also would be achieved using a bigger driver with equivalent piston area of the pair?

Effectively, yes. You can also get some improvement by mounting bass drivers push/pull but I doubt SWMBO would approve. You basically have the second driver mounted magnet out. This can provide a useful amount of 2nd order harmonic distortion cancellation.

Anyway, I've seen some pretty good cases for floorstanders vs stand mount monitors, and actually going down from 8" to 2x7" would make a slimmer floorstander with IMO a more attractive appearance.

It would indeed but choose your drivers carefully!

Now, if I get this correct, the mid should be high enough above the floor so that first floor bounce cancellation occurs below the crossover frequency, and the woofer low enough so that it's first floor bounce cancellation occurs above its crossover frequency.

Correct.

Cancellation frequency increases with declining distance to floor.

You can mount one driver right next to the floor and this effectively removes floor bounce from its pass band entirely.

If I'm correct we are addressing vertical lobing here? I haven't had the time to finish the final expressions, but this would be the general idea (i use tweeter in comments but it's mid in my case);

You're losing me with all the maths and coding I usually use ready made programs/spreadsheets to calculate these things as they are fuss free and usually fault free too. It's worth standing on the shoulders of giants sometimes! Although I do realise that nowadays people are getting lazy about being able to do all of this themselves, but what's technology for if you can't use it to minimise time to 'market' as it were?

And yes we are talking vertical lobing. With a design aimed at reducing floor bounce you have to put distance between the bass and the midrange unit. The further they are apart the lower you would ideally need to cross them over.

I got a bit anxious as to why everything about this driver kept disappearing from the Seas website, and since I speak norwegian (I actually drive past the factory every now and then) I called the guys at the factory. They had no clue or explanation, but could inform me the driver was in regular production so it should not be an availability issue... FWIW.
EDIT: typo

That's great news, that puzzled me too!

BTW, the drivers in the sketch are perfectly time aligned to listening position (not considering individual acoustic offsets, but drivers are same depth and wavelengths fairly long at 400Hz). Cancellation frequency (floor bounce) from the mid as drawn is 300Hz, and for the woofer the number is 544Hz, XO at 400Hz.

Time alignment is a moot point when you've got a DSP behind you as you can add in the delay required to arrive at the same thing. Therefore going with a slanted cabinet would only really be for aesthetics.

You are most certainly on the right path with all of this though and I'm impressed with the attention to detail you are giving to the various aspects of the design.

[Kjetil]

Crikey tons to respond too! Better get on it.

Thanks again, Matt :T

It is decent and the flexibility gained from using a decent DSP unit far outweighs some of the downsides. One of these is the way the miniDSP controls the volume, or the way in which you will control the volume. If using an analogue or digital input you're going to want to keep the signal level high, that is you do not want to control the volume prior to the DSP. The same is true internally to the DSP too, you don't really want to be adjusting the volume using the DSP unit itself. This is one of the major flaws with the miniDSP stuff, what you really want is a multi channel analogue volume control solution after the DAC to maximise performance.

For the best performance with the miniDSP you are going to want to use its digital input.

Indeed, no point in having a DAC in front of the miniSDP feeding analog audio into the ADC for processing and output through yet another DAC. Better feed it digital data in the first place! To begin with I'll probably devise a output buffer with some digital pots (controlled by an MCU with rotary encoder and/or infrared) and TI's new series of high-performance audio opamps. Thanks for pointing this out, or I'd probably do it on the DSP loosing SNR from the DAC.

The other issue with the miniDSP is that the ADC/DACs used are not great. They are by no means bad, but they could certainly be better. I see you are interested in buying the digital expansion board which is a great idea as using a digital input makes a lot of sense. So would using a digital output to a nice multichannel DAC, something using an ES9018 for example at a later date as an upgrade path.

The ES9018 actually has a 32 bit, bit depth in built digital volume control so you do not lose quality when you use this.

Does something like that exist off the shelf, and has anyone walked that path yet with the miniDSP? My fear when treading new terrain like that is always the software side of things, the learning curves can be pretty steep.

Not being able to attempt to counter floor bounce and/or increase bass system efficiency by getting some coupling to the floor from the bass driver.

These are reasons good enough for me. It'll be a floorstander, and most likely a ported one. I'm a bit undecided on the woofer, between the RS225 and the W22EX001 (that would be all-magnesium, all-local drivers, and I don't have to pay shipping, plus they'd make a nice visual match I think), but it'll be a single woofer I'm fairly certain.

TI have recently released a new class D chip amp called the TPA3251D2. This is absolutely awesome and comes close to rivalling the ncore in certain areas. The one thing it does not have is the same power output. Supply is an issue with these currently, but in a few months we're likely to see modules using this chip popping up. I don't know what your time frame is here but it's worth considering. To that end though you could always use some of the many TPA3116 board available on ebay. This are cheap as chips and I've use the same chip myself to great success. It sounds very good, certainly not an ncore or TPA3251D2, but as a means of getting you started that might be worth considering too.

TI has certainly done many great things for audio. My time frame is very flexible, if I'll have to wait for them I could just try something else temorarily. Thanks for the tips!

You're losing me with all the maths and coding I usually use ready made programs/spreadsheets to calculate these things as they are fuss free and usually fault free too. It's worth standing on the shoulders of giants sometimes! Although I do realise that nowadays people are getting lazy about being able to do all of this themselves, but what's technology for if you can't use it to minimise time to 'market' as it were?

And yes we are talking vertical lobing. With a design aimed at reducing floor bounce you have to put distance between the bass and the midrange unit. The further they are apart the lower you would ideally need to cross them over.

I do use those spreadsheets, but found myself in a situation where I realized I couldn't quite process their output data, due to a lack of understanding. Now, when I try to solve the numbers myself, I have to learn as I go, so it's really more of a learning exercise for me.

Time alignment is a moot point when you've got a DSP behind you as you can add in the delay required to arrive at the same thing. Therefore going with a slanted cabinet would only really be for aesthetics.

Yes, because the DSP can simply delay one output, of course! But will the slanted baffle pointing the woofer upwards towards the listener provide a better performance in the lower mids than the off-axis straight baffle would, think 400Hz XO with the woofer center at 45cm above floor?

You are most certainly on the right path with all of this though and I'm impressed with the attention to detail you are giving to the various aspects of the design.

Thanks, that's truly a relief to hear. The devil is in the details, right? :twisted: It's going to be alot of work and money spent regardless, so I'd rather do it right or not do it at all.

[Kjetil]

With all this said, there are many DIY projects actually already using the ES9018 with the miniDSP. With the I2S, you can do that with very little knowledge (i.e. flywire I2S from the kit) and existing DIY kits (e.g. Buffalo). No need to re-invent the wheel...

DevTeam

http://www.twistedpearaudio.com/digital/buffalo8.aspx

I'd have to look at what outputs are available on the miniDSP pin headers, but I could just layout an adapter-PCB between the two without too much effort. I suppose the ES9018 volume functions are controllable through some SPI or I2C interface (I haven't read the DS yet), so I'd probably have to put an MCU/encoder on the adapter board as well, but that kind of programming/interfacing is a walk in the park. 8)

[Kjetil]

From the looks of it, I could possibly get away with the basic miniDSP 2x4. It's product brief reads:

Expansion connector signals:
2ch x I2S in/ 4ch x I2S out / Analog INx2 / Analog Out x 4/

But pinout diagram shows 4 I2S output data lines.

The 2x8 documentation does not have a pinout diagram (of the I2S expansion connector, that is), but the product brief reads:

I2S in&out expansion connector compatible with miniDIGI I/O card

Digging up the miniDIGI manual, a pinout diagram is shown on page 2, that sports more I2S input data lines but the same amount of output data lines (four lines, I2S has a bit clock, and a word clock that splits the data into left/right channels, so four data lines equals eight audio channels)

Now, I'd probably have to verify with the miniDSP guys that I will not run into software limitations with the 2x4 board, but it seems to me it has all the hardware I need if I'm using the ES9018 through I2S. I might even put a Toslink/SPDIF input to I2S on the adapter board to save the space of a basically redundant miniDIGI module.

EDIT: I'll also be looking a bit at the open source FreeDSP using standard ADI SigmaDSP tools, I might prefer that one to miniDSP, and the Arduino Nano needed for programming is in a bin on my lab somewhere I really like having eagle files for the board as well :T

[Kjetil]

From first impression looks I absolutely love the freeDSP! :sn I might love it so much I'll build a custom board from the freeDSP core and the ES9018, and EAGLE is my own layout software of choice too so I can use the freeDSP files as a starting point. I have to think a bit about this, it would probably delay the speaker itself 8O

[Kjetil]

The Arduino Nano needed to program the freeDSP could surely also control volume features of the ES9018 and read an encoder or IR sensor, so basically I wouldn't even have to add another micro either :

[Kjetil]

Just to get a mental picture of the freeDSP circuit, I picked everything apart. Removed analog output buffers, replaced audio input connectors with a 3.5mm jack (think I'll keep it just to support analog) and spaced things apart. The components to the left are Arduino, 3V3 voltage regulator, USBi connector and EEPROM. The top components are analog in with voltage selectors (will remove), and the encircled components are DSP, reset button, and I2S connector. The DSP part is so simple I think I'll just make my own SMD layout from scratch here (not a big fan of through hole components). Think I'll use an existing ES9018 board to begin with though, but make a DSP board with SPDIF/Toslink and an I2S connector matching whatever DAC board I go for. Not sure if I'll keep the Arduino though.

[Kjetil]

Those ESS fellas with the Sabre DAC sure ain't very public with their data nor their distributor chain. An octopart search returned nothing(!) Seems Shaw Electronics is the only distributor on the block. Any 8ch "mainstream" parts that can match it? Closest contenders I found was TI PCM1690 and some Cirrus products.

[Kjetil]

In any case, since this is my profession I always like having my back cleared when making a design, so I based my circuit not on the freeDSP circuit but the functionally equivalent demo circuit from Analog Devices (only difference is really a transistor in the 1V8 regulator for digital circuitry, freeDSP chose a bigger type easier to solder with crude tools, but their design is probably based on the same demo circuit). That way, I wouldn't necessarily have to publish schematics down the line if I were to integrate an 8xDAC into the design and sell it, but by all accounts it'll likely stay open hardware regardless. I just like having options. But I did adopt the I/O configuration of the freeDSP so the hardware will be 100% compatible. Next up will be adding an SPDIF interface, probably a WM8804 SPDIF transceiver with PLL. This is the schematic of my circuit so far;

[Kjetil]

Actually, the ADAU1446 looks much more promising. Much more power and a built-in SPDIF transceiver.

[Kjetil]

ADAU1446 can output separate I2S streams as well (up to 9 output streams each with separate clocks it appears, not counting configurable GPIO resources), and has 16 asynchronous sample rate converters available. I'm inclined to use it with four Texas Instruments PCM179x rather than use an octal DAC chip made of unobtainium. Not sure yet if I'll adopt the Arduino programming method of the freeDSP. There are no audio ADC/DACs on this chip. I may suspend this thread for a while and start a thread in the digital forum when I get a bit further on this.

[JonMarsh]

Well, at least you've got a solid handle on the issues and some ideas about how you want to proceed- that will be interesting to follow.... :T

[5th element]

Well actually I built my own DSP unit too using an ADAU1445. I started out using pcm1792s but could not for the life of me get them to perform as TI specify that they should, even after like 20 different PCB attempts.

I then contacted my local ESS distributor asked them for a data sheet, which they provided and then I bought one and gave it a go. Awesome chip! Using it's own internal jitter suppression gizmo with a nice low jitter reference, it sounds really nice and measures excellently too.

I know getting data on these is less ideal than you'd like and buying them isn't as straightforward as getting other silicon but I can't speak highly enough about the performance and sonics, both objectively and subjectively.

[Kjetil]

Well actually I built my own DSP unit too using an ADAU1445. I started out using pcm1792s but could not for the life of me get them to perform as TI specify that they should, even after like 20 different PCB attempts.

That is quite worrying news. My take on this might be plainly on the SPDIF to I2S side and use an external DAC board. I've been looking at the even faster ADAU1452 that comes with a reasonably priced evaluation board and have some jitter supression gizmos on its own, but I'm not quite sure it's my chip yet, the ADAU1446 looks like an easier animal to get started with for my purpose (see TDM comments below).

I plan to omit using a microcontroller and instead use my BeagleBone Black running Linux Ubuntu (thus making the board a BeagleBone "cape") for control tasks. This way I could use either SPDIF interface from external devices, or put a HDMI screen/TV and USB wireless keyboard on it and play music from the BeagleBone itself using its I2S interface. A great thing about this is I can disable the BeagleBone onboard MCLK oscillator (its enable pin is tied to an IO), and instead feed a low-jitter MCLK from my DSP board with the upside of having clocks synchronous.

With this tight integration, I felt the measurement mic input should be part of the equation, feeding back to the BeagleBone (taking up an output line from the DSP), and with the ADAU1452 I'd only have 3 output lines left, forcing me to do Time-Domain Multiplexing (TDM) in order to have enough outputs (8 analog channels, a design requirement), which will probably take even longer to setup properly, debug and get running. With this said, all I'm going to do with this on the hardware side for now is reserve a mic I2S input to the DSP, and make sure the DSP has an I2S output line tied to the BeagleBone.

To keep things simple I think regardless of which DSP I use, the wise opening move is probably to use external I2S ADC/DAC boards, and focus on the digital side of things such as configuring the DSP and modifying the BeagleBone AudioCape drivers to work with my DSP. As a starter try to record an input source to the DSP (like SPDIF in) on the BeagleBoard, as well as outputting audio signals to the DSP for SPDIF out playback. I know the audio chain on Linux is a bit messy, so getting this to work will probably not be a walk in the park at all. When the SPDIF playback works, I can bring in the I2S ADC/DAC boards and play for real.

[5th element]

That is quite worrying news. My take on this might be plainly on the SPDIF to I2S side and use an external DAC board. I've been looking at the even faster ADAU1452 that comes with a reasonably priced evaluation board and have some jitter supression gizmos on its own, but I'm not quite sure it's my chip yet, the ADAU1446 looks like an easier animal to get started with for my purpose (see TDM comments below).

The performance wasn't terrible, but certainly not what the datasheet implies. Beyond a certain drive level distortion shot up dramatically. Ironically I could circumvent a large portion of this by using a TSOP to DIP adaptor. Don't ask me why I have no idea. I tried all sorts of things, talked to TI directly, but no cigar.

I plan to omit using a microcontroller and instead use my BeagleBone Black running Linux Ubuntu (thus making the board a BeagleBone "cape") for control tasks. This way I could use either SPDIF interface from external devices, or put a HDMI screen/TV and USB wireless keyboard on it and play music from the BeagleBone itself using its I2S interface. A great thing about this is I can disable the BeagleBone onboard MCLK oscillator (its enable pin is tied to an IO), and instead feed a low-jitter MCLK from my DSP board with the upside of having clocks synchronous.

The SigmaDSP chips have what's known as two clock domains, the input and output domains. The chips can be run in either master or slave mode. In master mode the chip generates all of the clocks required for operation, ie if you want to have an ADC at the front and DACs at the end. All these clocks will be synchronous and hence you end up with seamless operation without the use of the chips internal ASRCs.

If you are going to use the internal S/PDIF receiver or want to use your own, such as the WM8804, then the input side of the Sigma needs to be configured as a slave. In this respect data is clocked into the DSP core based on the clocks generated by the recovered clocks from the S/PDIF line. This is absolutely fine, but if you do this then you cannot use the SigmaDSP chip in master mode for the clocks on the output to a relevant DAC. If you do this then the input and output domains will not be synchronous and you will end up with samples being lost as one clock inevitably runs faster than the other. This creates an audible click every time this happens and dramatically increases jitter. For this reason the Sigma chips contain ASRCs.

If you use the ASRC on the output side of the chip then it takes the data out of the DSP core at the rate set by the input domain but resamples it to the rate set by the master set output clocks. This completely solves the problem but I'd prefer not to have my data going through lots of sample rate conversions if possible.

If you were running with the input side as a slave to the WM8804 then you could also run the output side as a slave to the WM8804 and this would completely solve the problem too.

I use an I2S input myself and slave both the input and output domain to it and get around all of those problems.

One thing to keep in mind is that the coefficients used to program the filters are sample frequency dependent. That is, if you have the DSP core programmed with coefficients to operate at 192kHz, then you cannot use these with 96kHz. The only way to do this is to update the coefficients. I keep mine running at 192kHz for this very reason, but it's worth keeping in mind. If you were going to use the WM8804 as an input device, to which the DSP core was slaved, then you would need to make sure that your sample rate is fixed from whatever your source is, or keep tabs on what the sample rate is and update the DSP coefficients whenever the sample rate changes.

In a pair of DSP active loudspeakers I made, I just put a SRC4192 inside to get around this as its a headache I didn't want to deal with and the on board DSP there didn't have ASRCs inside.

I really like the idea of using the beaglebone to provide the I2S data and supplying it with a nice quality master clock, from which the other clocks are derived from frequency division. You may have to supply it with two master clocks, one congruent with 44.1k source material (and its frequency multiples) and one with 48k source material. But once you've got that you can slave the input and output domains of the DSP to those clocks. The Sigma line of chips do not require a master clock themselves and only need the bit clock and LR clock as these are what's necessary to clock data into the device. The DAC however, whichever you choose, will require a master clock. The jitter attenuation device within the ESS DACs though is very nice because providing your data is bit perfect and not resampled in any way before hand, will virtually eliminate any problems with jitter that may have arisen downstream of it.

With this tight integration, I felt the measurement mic input should be part of the equation, feeding back to the BeagleBone (taking up an output line from the DSP), and with the ADAU1452 I'd only have 3 output lines left, forcing me to do Time-Domain Multiplexing (TDM) in order to have enough outputs (8 analog channels, a design requirement), which will probably take even longer to setup properly, debug and get running. With this said, all I'm going to do with this on the hardware side for now is reserve a mic I2S input to the DSP, and make sure the DSP has an I2S output line tied to the BeagleBone.

This is why I didn't change to the ADAU1452. I need a few input and many more output lines as I run multichannel with mine. Incidentally the ADAU144x series have plenty of processing power for my fairly complex system. I run 4 way active for my mains and have 4 surround channels to. This uses extensive DSP processing for pretty much all channels and I've still got plenty to spare. I only use 192kHz too and you lose half your processing power per doubling of sample rate.

Do keep in mind that TDM will most likely limit the maximum sample rate that you can use. Usually its limited to around 48kHz, so do check on that.

To keep things simple I think regardless of which DSP I use, the wise opening move is probably to use external I2S ADC/DAC boards, and focus on the digital side of things such as configuring the DSP and modifying the BeagleBone AudioCape drivers to work with my DSP. As a starter try to record an input source to the DSP (like SPDIF in) on the BeagleBoard, as well as outputting audio signals to the DSP for SPDIF out playback. I know the audio chain on Linux is a bit messy, so getting this to work will probably not be a walk in the park at all. When the SPDIF playback works, I can bring in the I2S ADC/DAC boards and play for real.

For I2S at a distance you can use high speed LVDS transceivers and receivers with a standard network cable. This works really well and it's how I get my source out of the PC!

Given how knowledgeable you are with the hardware side of things + being able to code, this will help you tremendously in getting something excellent going. Building my own DSP was the reason why I finally took the plunge and taught myself how to program micro controllers. It was a huge learning curve for the first couple of days, especially as I was having to learn how to configure the micros as well as learning how to code, but was so worth it!

And yes you can control the internal volume of the ESS DACs with I2C (you can also change its internal FIR filters too!). That's another nice thing, it isn't SPI. Although I don't have a specific issue with SPI, I do prefer working with I2C in applications that don't necessitate high speed as its protocols are so clearly defined.

The only reason why you may want to use the higher power SigmaDSP chips is if you want to go with FIR filters. I don't, so hardly see the point. But the FIR filters require significantly more processing power to realise.

[Kjetil]

The performance wasn't terrible, but certainly not what the datasheet implies. Beyond a certain drive level distortion shot up dramatically. Ironically I could circumvent a large portion of this by using a TSOP to DIP adaptor.

Most peculiar. Obviously, I would think the TSOP adapter made everything worse due to improper layout and placement of decoupling caps, etc. I have no good suggestions either, at least none you wouldn't have tried after your number of layout attempts...

I'm still trying to wrap my head around the clocking structure... I'm trying to figure out how to interface the DSP (it will be an ADAU144x) with the McASP0 of the AM3359 (BeagleBone Black processor). The signals available to me from BeagleBone pins not used by essentials such as HDMI is;

• AFSR (receive frame sync)
• AFSX (transmit frame sync)
• ACLKR (receive bit clock)
• ACLKX (transmit bit clock)
• AXR0 (transmit/receive data pin)
• AXR2 (transmit/receive data pin)
• AHCLKX ("High speed transmit clock", aka MCLK)

In the BeagleBone design, the on-board oscillator for audio (that I will disable and feed externally) feeds AHCLKX (not the expected AHCLKR, thus AHCLKX receives the clock), which may be explained by the AM335x Sitara Processors Technical Reference Manual (Rev. L) on page 4556 :Z

22.3.5 Clock and Frame Sync Generators
The McASP clock generators are able to produce two independent clock zones: transmit and receive
clock zones. The serial clock generators may be programmed independently for the transmit section and
the receive section, and may be completely asynchronous to each other. The serial clock (clock at the bit
rate) may be sourced:

• Internally - by passing through two clock dividers off the internal clock source (AUXCLK).
• Externally - directly from ACLKR/X pin.
• Mixed - an external high-frequency clock is input to the McASP on either the AHCLKX or AHCLKR
  pins, and divided down to produce the bit rate clock.

Same page states,

In operation, the transmitter uses AFSX and the receiver uses AFSR. Optionally, the receiver can use AFSX as the frame sync when the transmitter and receiver of the McASP are configured to operate synchronously.

And on page 4554,

In operation, the transmitter uses ACLKX as the serial clock, and the receiver uses ACLKR as the serial clock. Optionally, the receiver can use
ACLKX as the serial clock when the transmitter and receiver of the McASP are configured to operate synchronously.

When I look at the interface used in CircuitCo:Audio Cape RevB it corresponds with the above when transmitter and receiver is configured to operate synchronously (refer the BeagleBone Black schematic). It does not use (A)FSR nor ACLKR pins.

What I'm trying to determine is if I should tie (A)FSX/(A)FSR together and ACLKX/ACLKR together in order to have bidirectional sync/clock signals, or if I should simply operate the transmitter and receiver of the McASP synchronously without (A)FSR or ACLKR signals.

Now to wrap my head around what synchronous operation means to me, the MCLK oscillator will feed the AHCLKX, and it will generate ACLKX and (A)FSX output signals. On one data line, I should be able to playback music from the BeagleBone, but for the mic input on the other data line going the other way (DSP->BeagleBone), would I resample the mic input on the DSP? (sample rate and sync will have to match the data going the other way) And would I feed the ADAU144x the (external) BeagleBone MCLK oscillator at all or would I operate it independantly from its own clock?

[Kjetil]

What I'm trying to determine is if I should tie (A)FSX/(A)FSR together and ACLKX/ACLKR together in order to have bidirectional sync/clock signals, or if I should simply operate the transmitter and receiver of the McASP synchronously without (A)FSR or ACLKR signals.

Obviously I would have to set it up for synchronous operation since I have only one McASP and bidirectional data. So no (A)FSR or ACLKR. Just master clock oscillator to AHCLKX/MCLK, and let the McASP feed the DSP FSX/LRCLK and ACLKX/BCLK along with data. I guess what I'm really trying to ask is how to set this up clockwize (hah) with the DSP feeding data back in sync, with regards to resampling, etc.