4 different ways to obtain a "pure digital DSP"

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
4 and half different ways to obtain a "pure digital DSP"

If you think a modular conception is better (to follow the progress of DAC without changing the DSP... or the contrary ;) ),
and it is obviously better (and so more elegant for the mind)
to separate DAC from the digital filtering part that is a nest of electric perturbations,
then to my knowledge there are four different ways to goal this aim
(three one I get thank to nice poeple I "met" on the net).


SOLUTION 1 : thanks to pguerin of a French forum ---------------------------------------------------------------------------------------------------------------
For 150€ (+66€ taxes !!!) the miniDSP NanoDIGI : nanoDIGI 2x8 K | miniDSP
No facade (needs a PC to communicate, but small is beautyfull ! so cute !!! )
You have got 2 more output than a DCX (interesting for 2x3ways+subwoofer+bodyshaker)

SOLUTION 2 : thanks to www.dcx2496.fr ( and his kindly answer to my mail )
-----------------------------------------------------------------------
For 25€ (if you have already a Behringer DCX at 230€ !), and you are a bit used to electronic,
you can tweak the DCX to shunt the DAC quite easily (3 DIT4096 found on ebay cost 5€ for each)

SOLUTION 3 : thanks to Almat of the same French forum -------------------------------------------------------------------------------------------------------
For 1000€ to 1500€
An externally hosted image should be here but it was not working when we last tested it.
there is a lot of professional (that is a proof that "full digital" DSP is a good idea)
Xilica XD, Electrovoice DC one, DBX, ...
It really expensive but it's allowed FIR filtering

SOLUTION 4 : my dream, and the future, in my opinion
-----------------------------------------------------------------------------------
several DAC directly linked to digital outputs of a PC.

SOLUTION "5" : not exactly a "pure" DSP because it uses resistors to make an addition on the different signals...
------------------------------------------------------------------------------------------
... but a funny and amazing experience made by abraxalito : a"home made DSP" built with lots lots of DACs
http://www.diyaudio.com/forums/digital-line-level/195791-open-source-dsp-xos-16.html#post2982124
 
Last edited:
It's a small world :) :) :) ! Fortunately, I was only writing nice things about you :D ...

In fact, I asked to the guy of the Filtrage actif numérique et très haute fidélité
if I knew Dominique's forum, and if I could advise me another forum...

Inconvenient : it's bigger, here. Advantage : it's bigger, here.


Erin : it was about activ crossover. But I presume a DSP or a PC could have a fonction of digital EQ too ???
 
Last edited:
...and it is obviously better (and so more elegant for the mind)
to separate DAC from the digital filtering part that is a nest of electric perturbations,

I dunno :) I'm not at all convinced that its better to separate the DAC from the XO. What about if you could buy a DAC (at a reasonably cheap price) where you could customize the frequency response yourself? Would that be interesting in any way ? its the idea I'm playing with right now, btw :cool:

Also what if the XO had no DSP in it at all - that neatly circumvents the 'nest of vipers' (sorry, 'electronic peturbations') ?
 
Yes, I am a bit lazy :)
But it's not so easy too for me to read a long text in English, looking for informations...

What about if you could buy a DAC (at a reasonably cheap price) where you could customize the frequency response yourself? Would that be interesting in any way ? its the idea I'm playing with right now, btw :cool:
first reaction : bad idea. Modularity is better for a "hi-end" system.
second reaction : it could be a really nice idea, for a light and smart system
IF IT WORKS ONLY ON RIGHT OR LEFT SIGNAL (choosed with a jumper for instance)


...................................................................../ left < 2 ways
source - duplication (just a wire with SPDIF?) <
.....................................................................\ right< 2 ways
 
first reaction : bad idea. Modularity is better for a "hi-end" system.

Interesting - I was thinking modular at first, but I agree that not having modular saves money by elimination of connectors, multiple boards etc. So I am building two prototypes - one modular, one with a fixed length filter.

second reaction : it could be a really nice idea, for a light and smart system
IF IT WORKS ONLY ON RIGHT OR LEFT SIGNAL (choosed with a jumper for instance)

So you'd like to be able to configure a stereo board to work only in mono? I was thinking of that too as it happens - the mono configuration would be balanced output, using the two channels in opposite phases on either R or L.

P.S. Thanks for your PM alerting me about your reply - the system sends me notification anyway :)
 
A few strange comments above. Crossover is just an equalization function with a bit or routing. If done digitally, it is DSP. If no DSP, it is analog.

Most interested in UE and other products. I have one really big problem. Windows in not a real time OS and it has very poor thread timing management. Streaming applications and business machines is not a good match. Now, before I hear arguments, yes the basic horsepower is so great, it probably works fine most of the time. My problem is that digital system fault behavior is very hard to predict. If the threads running the xover were interrupted by background threads, swap delays, network hic-ups, etc, what protection is there that some very bad words are not sent to the DAC? What happens when the PC crashes? Power hit? We know how analog behaves. ( not always well either) Windows XP runs about 300 threads by itself. The attraction of a music server that was also the DSP engine is a good argument, but that precludes paring down the OS. Not as easy as Unix. Does this ever happen? Yes. Where is True64 Real-time-option now we need it? Oh yea, HP bought DEC and dumped all the best technology.

My next wish is that the results from UE could be exported to a mini PC running RAM and PROM, no fan, so it could turn on and off like an appliance, not a #@*&! Blu-Ray Java-based patience-testing .... " Please Wait"... Reading Disk"... There is no boot time for my passive crossover. I guess Mini-DSP is as close as it gets right now.
 
So you'd like to be able to configure a stereo board to work only in mono?
no,no,no. I meant, instead of using one stereo DAC for left and right,
I was thinking of using 2 DAC :
- one stereo DAC in the left loudspeaker, for tweeter and bass+medium (instead of left and right)
- the other DAC in the right loudspeaker, for tweeter and bass+medium (instead of left and right).
tvrgeek : If done digitally, it is DSP. If no DSP, it is analog.
abraxalito : My approach uses DACs and a digital delay line
In fact I thought abraxalito was trying to use a sort of very "light" DSP, only able to separate 2 ways tweeter/bass+medium.
But I don't understand how to do this only with a delay line ???
 
Ah so you were proposing to split up the channels - have one board handle only L, another for R. This would make sense for putting the electronics inside each speaker, however I don't favour doing that so the idea doesn't tempt me :)

OK, some more explanation. The DACs and delay line create what's called a 'transversal filter' which is an FIR (finite impulse response) filter built in hardware.

If you think about what an FIR filter does, it sums an array of data, each data element multiplied by a certain coefficient and this weighted sum becomes the output. The data array is simply the most recent N samples of the waveform where N is the filter length. This is done in a DSP (software) using a 'MAC' instruction - for multiply-accumulate. In my hardware FIR filter, each DAC is its own multiplier and the 'accumulate' is achieved by summing all the currents from the DACs together. Each DAC's current needs to be multiplied by a fixed coefficient and this is handled by a weighted resistor array. My blog has an LTSpice schematic showing this. Each tap of the delay line feeds into its own DAC - I'm finding 20 or so DACs (and therefore around 20 samples of delay line) works quite well to implement a 2X OS filter.
 
I like a lot the idea :) (if I understood it...), because it's really original, smart, and quite funny some way.

But of course, I presume the ratio price/performance is quite bad (20 DACs !), compared to a DSP.

Is there a link between the number of DACs and the frequence of the crossover ?


PS tvrgeek : in fact, 'cause the addition made by resistors, we could say it's a hybrid analog and numeric crossover
 
I chose very cheap DACs to work with so that the price would end up quite reasonable - they're TDA1387 which I think get recovered from old 'Soundblaster' PC audio cards. I pay about $0.08 each, so a 21 DAC design costs under $2 in DACs. You can buy a DSP for $2 but there's not a huge choice at that price :D

I have yet to build a crossover this way, but I have similated one - on the thread called 'Open Source DSP XOs'. The number of DACs relates to how steep the crossover needs to be and how much stop-band rejection is needed. The sim I showed on that thread was just a first stab, not particularly well optimized and using 48 DACs.

<edit> Attached a pic of the beta prototype I'm building up right now :)
 

Attachments

  • betaproto.jpg
    betaproto.jpg
    116.2 KB · Views: 123
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.