dual-mono dsp

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
A dual-mono dsp is on my wishlist :santa2:
Hypex dlcp is if I'm not misstaken capable of this but I would like to use
another solution for dac's and buffers. It would suprise me if I'm the only
one that has been thinking of this solution.
Two "Minishark's" might work. Comments anyone ?

dsp_slave-master.png
 
HP filter before each amp to avoid the full modulation at the input to be played at the output (befote the amp if each of one is playing in its own frequency range) = to allow each amp to output at its best ?

10 m ? symetric transmission + FIFO between each DAC ?

What is the streamer ? a pc ?

PC (soundcard internal or external) to I2S or USB ????
 
It might BE easier to have 6 x 10 m of flat wire for the speakers !

Is it possible to split the output from a single MiniShark in two I2S Signal (instead two in your road map) ?

Is it really for a 2 speakers (not a two rooms systems with 2 speakers in each room ?

Teleporter Digital Transfer Module : it's I2S output, but I don't know if it's glue to your source ?
 
Last edited:
A dual-mono dsp is on my wishlist :santa2:
Hypex dlcp is if I'm not misstaken capable of this but I would like to use
another solution for dac's and buffers. It would suprise me if I'm the only
one that has been thinking of this solution.
Two "Minishark's" might work. Comments anyone ?

dsp_slave-master.png

This is similar to what I am doing in my current active loudspeaker endeavors - I have one Raspberry Pi in each speaker and stream audio to each over WiFi (it's compressed audio, however, with hardware you could stream uncompressed audio I believe). There are one or more DACs on each Pi, which are connected to the onboard amplifier(s). I process the audio stream using software on the Pis (using Linux DSP crossover software, part of which I wrote) and output the resulting channels to the DACs, amps, and drivers.

The cost of this is quite affordable and the performance seems pretty good to me. I don't believe that I will ever buy dedicated DSP hardware again when I can do more using less expensive general computing hardware.

Unless you are stuck on high rate audio, there is no need for cables and the range is only limited by your signal strength.
 
Hi Charlie,

Tell us more please !

What is the streamer solution used to sent the data on each RasberryPi ? From a PC ?

EQ : FIR ? IIR ?

The implementation is explained here and there in this thread:
http://www.diyaudio.com/forums/pc-based/274331-ladspa-plugin-programming-linux-audio-crossovers.html
The crossover/EQ is implemented using IIR filters in software. There is also multichannel fractional delay capability separate from the filtering. Works very well. There is a link to download the software package, and a link to a tutorial (not mine) on the subject, in my signature.

To get the audio to the Pis in the loudspeakers I am streaming audio at 48kHz after compressing it to 320k MP3 using the LAME encoder. I'm trying to conserve bandwidth because I stream over WiFi. I haven't explored higher sample rates or bit rates. Part of the reason is that 48kHz is easily achievable using inexpensive hardware (Pi, DAC, etc.) but above this things start to become more challenging (and expensive).

I use the freely available VLC software as streaming server and client, with a couple of additional things that make it easy to manage everything.

I reported some issues with delay between left and right channels in the thread I linked to above. After adding a RTC to each Pi and using NTP to get the timebases closely synchronized I'm no longer experiencing these problems. The way I have it set up, if there is something that is not right I can reset it all from the server with a click of the mouse.

With this approach, I would like to build a distributed audio system, for instance stereo main speakers plus multiple subs each operating independently. Everything should operate together as if they were physically connected. This offers freedom in placement of the subs, and should make it possible to add more speakers in other areas that are covered by the LAN and remotely turn them off and on as needed.
 
This is my setup. Right now I'm running (3x8m) signalcables under the floor.
No electronics between the speakers is a major priority.
setup.jpg


The teleporter was a useful link, thanks. :)

A dsp solution is what I'm aiming/hoping for and low cost is not a priority.
I have a few friends that are more than interessted in a similar solution
that also have need for the added cpu recources that a dual dsp would supply.
That's 5-way filters with high order xo's and a lot of different compensation points.

I have the Najda dsp today and one way of getting closer to a dual mono
would be to separate the dac's and buffers from the present chassie and run
3x spdif instead of analog signal cables under the floor.
 
The spidf signal is not made for long distance... at least with the coaxial one but maybe the symetrical or even better fiber (TOSlink) to avoid the ground loop could work ? But because the jitter a FIFO to reclock the jittered signal bedore each DSP (Your Master & Slave) must be usefull ! Look at for here FIFO II solution from IanCanada!

Or a MiniShark from Minidsp : one I2S input & 2 I2S output : putt in front of each I2S output a Twisted Pair Transporter to link the two DSP above (canal R & L) : while the MiniShark can do 32 bits room correction, the two other DSP manage the XOs of each channel ?.... In theory !

Nice speakers :) B&G planar close to array of bass drivers à la StiigErik member ?
 
@ CharlieLaub,

I really believe we need an all in one solution like Jriver or AUdirvana able to FIR, IIR and XOs in a lot of channels for driving indidually many drivers (multi amp)!

To transport each channel near the dac/amp/driver = speaker, the best way is TCPIP as you know already.

We need multiple transporters à la TwitedPear but each having its own channel to have the right XO & EQ data because one transporter in front of each dac/amp/driver !

A network could do that ! Putt an adress to each transporter et voilà...you just need the PC to be a server to route the right datas to each transporter! It would be a TCPIP to I2S stuff instead a mono point solution like tha TwitedPear Transporter...

Does it not exist in the Pro Sound engineer world : sort of soundcards hubed with RJ45 plugs ?

The soft should be a Convolver network playback streamer (tcpip) on a pc (so rj45 output or wifi) with a cheap hub between it and each "TCPIP to I2S" stuff

A little card with a class C adress to I2S should cost Nothing : less than multiple standalone 32 bits DSP everywhere! the unique DSP should be 64 bits into the PC like you do with a soft solution!
 
Last edited:
Theroically a 35 euros Rasberry is near such a stuff, but here it needs a linux client and the I2S plugs are not usefull for our needs like uf-l plugs are (while some guys like Ian make small pcb to link I2S plugs from a Pi to a pcb 3 channels I2S with uf-l )....

Question remains : which soft to route the EQed, XOed channels to each RasberryPi with TCPIP ?

Channel 1 treble Left to Rasberry 1 (192.168.xxx.yyy)
Channel 2 mid Left to Rasberry 2 (192.168.xxx.yy2)
.
.
etc

To my poor knowledge the actual soft convolvers just output on an embeded soundcard into the pc ! Such a stuff with just a programmable class C adress from a PC could be a killer ! Not sure MiniDSP would apreciate it as they have the inverse philosophy ! Imagine than Audirvana or Jriver, Sygnalist, or any Linux team work on an option to encapsulate the convolved data into TCPIP packetts ;)

So instead a RasberryPi with complex distro : just a rj45 to I2S hardware board where the adress are remotly manageable via a pc; after two solution : multiple Fifo to synchronise the outputs before the dacs or one little FIFO on each of those cards with a link to an unique Master clock card (like a Clock 2 from Ian but with more outputs)... IanCanada Helpppppppppp please :) !
 
Last edited:
@ CharlieLaub,

I really believe we need an all in one solution like Jriver or AUdirvana able to FIR, IIR and XOs in a lot of channels for driving indidually many drivers (multi amp)!

To transport each channel near the dac/amp/driver = speaker, the best way is TCPIP as you know already.

We need multiple transporters à la TwitedPear but each having its own channel to have the right XO & EQ data because one transporter in front of each dac/amp/driver !

A network could do that ! Putt an adress to each transporter et voilà...you just need the PC to be a server to route the right datas to each transporter! It would be a TCPIP to I2S stuff instead a mono point solution like tha TwitedPear Transporter...

Does it not exist in the Pro Sound engineer world : sort of soundcards hubed with RJ45 plugs ?

The soft should be a Convolver network playback streamer (tcpip) on a pc (so rj45 output or wifi) with a cheap hub between it and each "TCPIP to I2S" stuff

A little card with a class C adress to I2S should cost Nothing : less than multiple standalone 32 bits DSP everywhere! the unique DSP should be 64 bits into the PC like you do with a soft solution!

You could check out the "VBAN" feature of Voicemeeter, a pro audio mixing/routing software. The next version will have this feature, but there is a pre-release beta available now:
VB-Audio VoiceMeeter Banana
https://twitter.com/vb_audio

It seems to be able to send 16-bit audio streams over TCPIP in between instances of that software. Here is a screen shot:
CHols60WIAAUM1Q.jpg:large
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.