TOPIC: Re: Digital crossover for my WMTMW speakers

I'm looking on these sentences from the link I posted above:

Why must the phase be zero on the first and last samples? What to do if it isn't? Any ideas?

This is related to the discrete time signals you're dealing with. A discrete time signal corresponds to a periodic repetition in the frequency domain. Sample zero will be inherently connected to sample N-1. If sample zero and or sample N-1 are not zero (magnitude and phase) the arising discontinuity will lead to artefacts.

@Crumboo

I´ve downloaded scilab tonight and wanted to try to see if I could create some filter, but ran into major "convergence" issues when trying to create a target filter So I decided to locate my companys very old matlab (I think 4.0) and see if I have any luck in creating a target and correction filters. I´d still rather work in scilab (pricewise, not performance wise - I´d much prefer matlab from what I´ve seen so far). I´ll keep you posted on progress, but spare time is short these days.

Have you started working in scilab??

Just a few short comments from previos posts.

1) sample rate depends on what you intend to use brutefir for. If you only play 44.100hz audio CD´s, I´d prefer that sample rate and there would be no need to any sample rate conversion, but if you play a number of different samplerates (music/video), an alternate sample rate might be just as feasible (on win xp soundcards had a tendency to output everything at 48khz, regardless of materials original 44.100hz - don´t know if thats an issue still).

I´d prefer an ASRC on the input to ensure that no matter what material you throw at it, it´ll still be processed right. If it´s possible to use the miniDIGI as SPDIF channel selector, Sample Rate Conversion and SPDIF output it would be a nice solution (you can use MiniDIGI as input selector with a knob, not via IR remote). Ideally I would probably go for a 96khz sample frequency since you have a lot of filtertaps available, but haven´t seen this product (haven´t really looked though).

2) volumen control seems fine (it´s an 8 in - 8 out solution, right?). Ugly remote though Like kamakaze says volume control should happen after filtering. I´m not that scared of doing digital volume control but I know some are... anyway if you´re using analog out on the soundcard in the Brutefir PC, then analog is the way to go.

@Jakster

Glad to hear that you are looking into this as well! I'll be happy to follow your progress.

No, I haven't used Scilab yet. I have a fairly new version of IDL, that I used in my former work (I don't really like it though - there are lots of bugs that forces me to close and restart the program). Hope to find some time to have a look at Scilab soon, and I hope it won't be too hard to get used to. So far, I created the target filter using measured data and the equations from Keele's and Horbach's work in MS Excel and then imported this "ideal" filter response in IDL to do the IFFT and windowing in order to get a usable filter. I'm not really finished with this, there a still some issues to solve...

Thank you for the comments!

Yes, I believe I stick to 44100 Hz as sample rate. At the moment I use two signal sources: a CD-player and Apple Airport Express (which both have analog and optical outputs).

miniDIGI sure looks like a good solution, and it I believe it can be used as channel selector from what I read in the manual. I'm also thinking about parting it with a miniDSP to get an analog input as well and the possibility to do room corrections in a more convenient way than using bruteFIR for this.

Yes the volume control is 8 in-8 out. I agree that the remote is ugly, but the thing has a "learning" feature so it should be possible to use my remote for the CD player I suppose that I'll use analog output from the sound card, otherwise I need a 6ch DAC....

Had another look on the filtering and it seems that I'm getting somewhere. I made a try on the LP-filter, and it came out as this:



Top: Desired filter response
MIddle: Filter curve (Hamming window, 999 filter coefficients - note that it is not symmetric!)
Bottom: The filter response curve

Haven't verified everything yet, and didn't have time to look at the phase response. Seems promising though.

Great project and great progress! Sorry for being disconnected, we're just back from Chinese New Year and just have a backlog of emails/things to do. We're keeping a keen eye to your progress and great for sharing all your info.

The coming weeks will see the release of more info for the OpenDRC and miniSHARC board.
- OpenDRC is the stereo version (i.e. Digital in/out). It will allow room correction or to be used as a convolution engine. Prelim info was already posted but more to follow...

- miniSHARC is the PCB only processing board with I2S in/out and with the FIR filter banks. With the same format as all the miniDSP boards, it will indeed allow you to load your custom FIR filters. I think that this could be a good fit for your project. i.e. miniSHARC + miniDSP 2x8 for all 8 outputs.. This is a bit down the line because it will require some new firmware/new software. But we'll release some beta version of the code later on.

Once you have some clearer idea of your filter size, we can see what can be done but I think that it will be more than sufficient for multiway crossover. Currently allowing up to 12k taps as a start. i.e. could do multiway easily.

DevTeam

Oh, this sounds very very promising

I am really looking forward to learning more about miniSHARC! Great news!

@devteam

Thank you for your kind answer! I'm looking forward to hear more about this new product! If the card allows 12k taps, it will mean 2k taps for each channel in my stereo 3-way system. I think it will be enough, but I have to do some more calculations...

I assume this setup would do:

miniDIGI (allowing spdif inputs) -(IS2)-> miniSHARC (multiway FIR crossover) -(IS2)-> miniDSB 2x8 (for more more room correction/eq if necessary + providing multichannel analog outputs)

?

@Jakster

Have switched to Scilab now

@Devteam
Awesome!! – Can´t wait for additional information

From what you write I can´t see whether your mini Sharc is stereo in, stereo out or if you are going in a direction of multiway??

My “dream system” would be something like this:



The Sharc Receiver board is really just a miniDIGI with a buffered I2S output (and I added IR commands going to the mini Sharcs). This is followed by 1-4 mini Sharcs depending on crossover needs (for a two way system two mini sharc boards are used, all boards are identical). On the mini Sharc jumper settings should set the ASRC (I would use two in cascade to ensure sample rates as low as 12kHz at all input frequencies) to filter frequency sample rates fs=12 or 24 or 48 or 96kHz. Master/slave clock´s should be used to synchronize mini Sharc boards (+ an option for using an external low jitter clock). Output from Sharc´s is I2S (FIR filtered, volume and delay processed) goes to a third board combining outputs to I2S multichannel (8 channels), which will allow interface to all multichannel DAC´s. Optionally a multichannel DAC could be included on this board (though I personally wouldn´t use it, since I would go for the ESS9018 “super” dac.

System pro´s:
- R&D:PCB layout is easy to develop (since you already have the OpenDRC), only one program is needed for all mini Sharcs (which is almost the same as the one you have running on OpenDRC). Depending on the setting of the ASRC the filter is longer or shorter (ranging from 3072 at 96khz to 24576 at 12kHz).
- Flexible, user buys the amount of mini Sharcs needed in his/her setup
- Room correction would be possible at low frequencies (frequency resolution 0,5Hz at 12kHz)
- Use the cheapest ADSP-21369 available (filter length would suffer slightly, I know)
- External low noise clock for I2S would reduce jitter
- Interface to external multichannel DAC could create a full on high end system
System Cons:
- Price, Sharc DSP´s is somewhat expensive (20-30$), but R&D would be less
- Possibly having to change USB to each of the Mini Sharcs when loading filters
- Maybe just one buyer (me )

I probably can´t influence you anyway but hey… at least I tried

Thanks again for awesome products.

@Crumboo

Good – I like the idea of using an open platform (Scilab) for this – It will hopefully allow others to follow your example and build filters for FIR convolution. I wish I could tell you that I´ve made some progress in scilab but honestly, this half hour is the only time I have had to myself all week (busy job, wife pregnant, 1 year old son, renovation of my house). So far all I have done is load a tweeter measurement and created impulse and frequency/phase response, next in line is creating and arbitrary linear phase target filter.

Wow Jakster, I certainly would like that setup!! I really really hope that the developers will give this a serious thought (and I can assure that there will be buyers...)

Can you tell me more about the ESS9018 by the way? Is there any commercial products for 8 channel D/A that uses it or do you plan to build something on your own?

@ Jakster: Your proposal looks great I would also be a potential buyer

What I would really appreciate is multiple samplerates for multiway system (could be as low as e.g. 2kHz for subs). That way maybe even one miniShark could handle a complete multiway setup with sufficient frequency resolution. Should generally be possible to do SRC in DSP code?

@crumboo: several multiway ESS9018 dacs are available for DIY, one beeing [url=http://http://www.twistedpearaudio.com/digital/buffalo.aspx]Buffallo III by Twisted Pear Audio[/url], a complete setup with 8 channel analog output modules (missing on Buffalo II board) will be quite expensive though. Seems not that easy to source sabre DAC chips, so complete DIY might be difficult.

By the way: I am in a quite similar situation as Jakster timewise, but I will also try to find some sparetime and install Scilab and would like to join the "development team"

Anyway, looking forward to hearing more about those awesome new miniShark boards

I like the flexibility and simplicity of the system. I don´t know how many units miniDSP usually sells of a product, but I would guess that it isn´t great quantaties, which makes R&D a rather large product cost (this is all speculation), so it may be almost just as cheap to do it this way (using multiple Sharcs). Also if you wanted to do a 6 way speaker with added woofers you could buy 8 Mini Sharcs and do a 8 way system (you would need 2 I2S to I2S multichannel boards though).

@Crumboo

Curryman beat me to it

Regarding ESS chip, I´d personally go for twistedpear´s solution (Buffalo III):
twistedpearaudio.com/digital/buffalo.aspx
www.diyaudio.com/forums/twisted-pear/

It´s costly but as I remember the chip alone costs around 100USD. You can find other boards, ESS has a development board, Scomed99 has a solution on this site, 2 channel though (www.minidsp.com/forum/6-diy-hifi-project...-8x8-kit--hi-end-dac)

and maybe others. I like twistedpear for the flexibility (2,4,6 or 8 channel output). They don´t have an 8 channel I/V converter, but maybe that´s being developed.

@Curryman

I agree, one Mini Sharc would be able to do most if not all signal processing tasks, but I can´t get my head around how to construct a program that will allow multiple "arbitrary" sample rates (as low as 2kHz) for a 2-5 way system. I could probably do it for my specific need (3 way, fs at 6, 20 48 khz) but not for arbitrary fs´s. Sample rate conversion isn´t very hard to do in code (don´t know about Asyncron though), basically just lowpassfilter, downsampling, upsampling with polyphase filter (as I remember it). Cost´s instructions though.

Awesome!! - it´ll be some time before miniDSP releases new products so we should have some time.

Guys, for minisharc maybe better to post here? www.minidsp.com/forum/3-suggestion-box/4...arc-not-yet-released in this theme it's offtopic.

Hi again,

Had a look on the Twisted Pear Audio homepage - very interesting products! I saw that they have SPDIF input selectors also, which can be a good alternative in the setup? A main problem that has to be solved if one want to use the 8 ch DAC is combining the outputs from several miniSHARCs to multichannel I2S - do anyone see a solution for this?

The problem of filtering for room correction in the sub bas range (which would require lots of taps if multirate filtering can't be used) maybe can be solved by adding a standard miniDSP (with IIR filter) in the chain? Something like this:

SPDIF reciever -> miniDSP --> miniSHARCs -> I2S to I2S multichannel converter -> multichannel DAC

I'm getting somewhere with Scilab - I'll get back soon!

@Kamikaze
While I don´t really agree that posts regarding Mini Sharc´s are off topic in this thread, I´ll agree that it fits better in the suggestion box. I might duplicate my post there later.

@Crumboo
I also like twistedpear, well thought through products for "semi diy" people like me (and lots of support at diyaudio).

Not really - In the system I drew a few days ago I synced the clocks of mini Sharc together so they were in sync and could be combined (see clk in/out, clk master/slave). I don´t know any standalone solution that combines I2S signals - would probably require asyncron sample rate convertion of each output from the miniSharcs.

Hardware wise, it´s really up to you, how you want to do it (and what products miniDSP releases). If they release a three way mini Sharc (one chip) and you decide to do room correction using IIR filters, I would flip the order of miniDSP and minisharc (provided miniDSP 8x8 can do I2S in) because the miniDSP is analog out, which means system wouldnt require an 8 channel DAC afterwards. I would use miniDIGI - not Twistedpear solution.

Scilab: Great.. I ran into problems last night with my arbitrary target filter - maybe I´m using to steep slopes. Anyway target filter didn´t go all the way to 0 out of band, so I have very high gains in the inverse filter a DC and Fs/2... from 2000hz-20000hz hz it looks ok though.