TOPIC: Re: Digital crossover for my WMTMW speakers

Thanks curryman,

I'm not sure I understand you right about the target phase. If I design all filters so they results in a zero phase response from the speaker, then it'd just be a matter of adjusting the delays in order to get the speakers to sum correctly?

maybe thats just too simple

Hehe...yes maybe, things usually aren't as simple as one expect!

Hello Crumboo

this is what the FIR filter corresponding to output.txt looks like in ArtemiS:



The limited dynamic results probably from txt format export/import (integer format vs. floating point in Scilab? and ArtemiS).

Thanks a lot curryman!

The magnitude response looks exactly right, but there still seems to be something wrong about the phase response. It should look like the lower left section in the plot I posted along with the .txt-file, but seems to be zero phase in ArtemiS. I double checked and calculated the FFT of the filter in Scilab, but the phase response still differs from what you got in ArtemiS. Strange it is....any ideas? Maybe someone using Matlab can check the response of the filter also?

I'll ask my colleagues tomorrow

Hi Crumboo,

I discussed the ArtemiS results with my colleagues and now I am a bit wiser

There are different thinks to take care of:
First of all the import: the imported impulse response starts at time 0 seconds, so the resulting filter has a group delay of 1/2 filter length. When I correctly remember your code you compensate for this by shifting your inpulse by 1/2 length.
A second pitfall was related to the filter editor in ArtemiS: I can load the impulse response to the filter in ArtemiS and it will be used as expected (in this case including the group delay ). However if I choose to edit the filter (it can be very handy to be able to "draw" filters by hand ) and I have initially loaded an impulse response, ArtemiS will only use the magnitude with 4096 points -> Thats the reason for the zero phase! If I load a complex spectrum, it will not be changed
So, to summarize all this: When doing it "correctly" in ArtemiS, I get the same result as you And by the way, Matlab also shows the same (man, at the end it's just a FFT ).

Have a nice weekend!

Wow, you're being very helpful curryman - lots of karma to you!

Nice to hear that the response looks as expected! Since the code now seem to work I'll be happy to share it. It's a bit of a mess though, so I'll clean it up a little and write some comments first.

Have a nice weekend!

yes share - if you don´t mind me stealing your code and eventually using it - though it looks like it will be a while before I start creating my own system (around a year).

How far along are you?? Have you done measurements and created your filters and are waiting for mini-sharc news, or??

A cheap and fast way of listening to your effords would be to install fddshow with convolver and mixer, setup your mixer so that you have 3 channels for each speaker and use convolver whichs loads a wave file with filter coefficients. You can use zoomplayer (with fddshow). This is all meant as a temperary solution until you decide which hardware platform to use. Since I´m in no hurry with my hardware platform, I´ll wait and see what miniDSP comes up with - I just hope it´s something that will give me the frequency resolution I want (fft convolution, lower fs or something completely new). The other platform I´m thinking of is a PC with linux - running jriver or mythtv (front end media center programs) with jack audio and brutefir.

Anyway, I can´t contribute much - but I am eager to hear your progress - a big thanks to you (and Curryman) for the work so far.

/Jakster

@Jakster

I have been quite busy lately so there isn't much news. I guess it'll be a month or two before putting this into action. Meanwhile I'm waiting for news about the miniSHARC!

Hopefully, it will not be too far away in the future! It sure seems to have the features I need (8 ch I2S in/out, load custom filters), and the frequency resolution will probably be good enough for me (I've simulated the LP-filter section using 2000 taps and the result seems to be good enough, although it really wouldn't hurt with more taps...). I'll probably use IIR PEQ for room correction in the low frequency range. If this feature isn't included in the miniSHARC (it is in the OpenDRC-DI), maybe I'll just add a 2ch miniDSP for this. Also, I'm quite sure that I'll use Buffalo III as DAC and miniDIGI as input/source selector. Should be fun to install all these cards in a nice looking box.

As soon as all part are in place my plan is to hook everything up and do some (gated) measurements for each speaker element group and use this for the filter calculations (with the Keele-Horbach crossover curves as targets).

About the code: I'll get back as soon as I've had another look at it. If you are eager to to have a look I can email you the "messy looking" uncommented code.

Best regards

Hi Crumboo

I´ve started to look at the whole FIR crossover based system again - and found your old posts. Hope you don´t mind that I post in this thread instead of your minisharc post.

I can see that you have purchased both minisharc and ministreamer and I am wondering what kind of system you are planning at the moment?

Is it BruteFIR with ministreamer or are you going the minisharc route??

Have you finished the keeler-horbach scilab code? (and if so - can I have it?)

Myself - I´m probably leaning strongly towards Jriver (under windows) which allows me to custumize long FIR filters (including crossovers), surround decoding, roomcorrection and different filters for different sample rates. Also Jriver can change the video/audio delay so the filter group delay in long filters doesn´t result in sound and video being out of sync. Right now the system I want to use is an intel NUC with ministreamer (provided it works with jriver) and and 8 channel dac (like buffalo).

/Jakster

Hi Jakster,

I recently picked up this project again after a long time. I had an idea about using BruteFIR under Linux using USBStreamer, but I didn't get USBStreamer to work on Linux. Then I lost insterest for a while. So recently I got miniSharc instead and made a prototype unit (more info on this here: www.minidsp.com/forum/opendrc-projects/9...lo-iii-8ch-prototype).

Yet I've only been using it for some basic 2ch dsp and FIR phase correction of my speakers but the plan is to use it as digital FIR crossover (my speakers are 3-way) as soon as the power amps are in place.

The good side of using miniSharc instead of the computer based setup is that it will all fit in a convenient box without having a computer. The downside is the number of taps available. I think however that the processing power of miniSharc will be enough for my needs at this time. Hopefully the devteam will give access to the ASRC in the future, giving the possibility to downsample the low frequency channels (please devteam!).

Because of the processing power, a computer will of course be able to do so much more in terms of processing so maybe I'll have another look at this in the future.

About the code: well...I haven't been working on that for a long time. I could send you the old files, but maybe you should check out rePhase instead?

Best regards,
Mattias