DSP Xover project (part 2)

I'd like first to briefly mention that the term 'asynchronous' doesn't fit nicely the context of I2S coms. Indeed, I2S is inherently 'synchronous' in the sense that both endpoints use the same clock transmitted along data and synchronously to it.

So this is it: there are 2 endpoints (external DAC and DSP board) but only one clock, so we must decide whether the clock comes from the DSP or from the DAC. We'll call 'Master' :D the endpoint supplying the clock and we'll call 'Slave' :( the endpoint sync'ed with the Master's clock.

I said here in a previous post that the expansion port that has 6 extra outs will only work in Master mode. This actually means that data on this port is at the system's central rate, which is derived either from the onboard PLL or from the external clock supplied via the other expansion port (future software revision as discussed above).

You understand now that if you want this port to work in Slave mode (i.e. clock supplied by your DACs), then you're introducing a 3rd clock in the system. :mad:
We have to set now who's the 'boss of it all', as would say Lars Von Trier. Do you want the master clock to be generated from the onboard PLL? Do you want that clock to be supplied by the USB receiver? Or do you want the master clock to be supplied by the DACs? This is the question you must answer.

My guess is: just feed your DACs with the system's central clock - which can be either PLL generated or supplied by the receiver. For allowing this, you set your DACs in Slave mode and let the DSP board be the Master of this I2S com. :)

Thank you Nick!!!, this is a very good ans simple explanation....

As you can see some of us need simply one DSP with digital inputs and outputs without limitation for use with our USB and DACs devices, maybe in the future you can do a simpler product for this purpose...

Regards

Alberto
 
As you can see some of us need simply one DSP with digital inputs and outputs without limitation for use with our USB and DACs devices, maybe in the future you can do a simpler product for this purpose...

Sorry for the complex answer - but Dalhberg has complex questions :)

The board actually can't be any simpler than it is now.

It does have digital ins and outs so that everything we said above applies only to I2S and is merely optional.
You can indeed simply plug in your equipment and not worry about clocks at all.

Again, sorry if the post above was confusing.

Thanks a lot, most appreciated :)

You're welcome ;)
 
Possible way Vs Best way

Hi Nick,
Your quote " You can indeed simply plug in your equipment and not worry about clocks at all. "
For me this is 100% the problem!
ie Its of great concern to me that there are many ways to connect DSP/ DAC's / USB IO / Clocks together and they all "work" ie produce sound.
But I dont know which is the BEST way to do this...?

What I (and many others?) am seeking is a powerful DSP board with digital in / digital out ( with various USB / I2S / SPDIF / Balanced / optical options ) and the ability to connect our own modules for DAC's / A to D / and power supplies.
I personally would be very grateful if you would please explain in a simple ( laymans terms!) block / flow diagram of your methodology and then maybe a few of the other top digital guys can add their comments?
Thanks in advance
Derek.
 
Quote: What I (and many others?) am seeking is a powerful DSP board with
digital in / digital out ( with various USB / I2S / SPDIF / Balanced / optical options )
and the ability to connect our own modules for DAC's / A to D / and power supplies.



Please don't pester "Chapark" with questions about a whole other idea of how
to make a dsp xo in this thread. There is another one and if he wants to join in
he is of cource most velcome but respect that the man probably has little or no
time to spare these days.

Maybe I'm barking up the wrong tree since you are on a first name basis and
I'm sorry if that's the case :)
 
Last edited:
Ok. The 5V supply is only used to power digital sections. Some chips use this 5VD as supplied, others require 3.3V and/or 1.2V. The latter are derived from 5VD with onboard regulators.

There's a separate 5VA (analogue). It's derived from the +12V supply. That's why you must always provide +/-12V.

If you plan to use an analogue volume chip, CS3318 needs +/- 8 to 9V and CS3308 needs +/- 5V. These 2 supplies are again derived from +/-12V with onboard regulators.


1A is safe. Power lines are also available on the expansion ports so you can draw current here for your devices attached.

If 1A is the consumption of the 5v line,what is the draw of the +/- 12v line?
 
I wouldnt assume we wanted to slave the dacs at all, that would actually be the worst choice, to slave the DSP would be much preferred, the clock for the dac outputting sound should have its clock as local as possible, particularly as there is no controlled impedance method to supply clock to anything, or clock buffer/s
 
It does have digital ins and outs so that everything we said above applies only to I2S and is merely optional.
You can indeed simply plug in your equipment and not worry about clocks at all.

And over the spdif outputs the dacs are clocked by the onboard clock and there an is alternative to reclock it or..... ?

From the Buffalo3 se spec
High-precision, ultra-low-phase noise clock and integrated reclocking
 
I wouldnt assume we wanted to slave the dacs at all, that would actually be the worst choice, to slave the DSP would be much preferred, the clock for the dac outputting sound should have its clock as local as possible, particularly as there is no controlled impedance method to supply clock to anything, or clock buffer/s

And over the spdif outputs the dacs are clocked by the onboard clock and there an is alternative to reclock it or..... ?

No worries: there's a master clock input. If you select this option, it will clock everything on board - including SPDIF outs.
 
nice, thanks it sounded like thats what you were suggesting (that this would be possible) its always best to have the masterclock close to the final stage, then you have less issues with jitter. it demands a certain speed of clock though still yes? in order for the filters to be correct?
 
- Analogue audio will be AD converted using that external clock.
- SPDIF will be rate converted to the system rate
- I2S input data will bypass the sample rate converter because it's expected to be at the rate of the supplied external clock.

And how will this masterclock affect the usb reciver ? Will it in any way limit the usb recivers input compatibility ?
 
Last edited:
Let me rephrase that last question.
Lets say that I will go for 24/192 on the digital outputs, that's the systemrate right ?
Since the samplerateconverter for I2s will be bypassed am I then forced to (software)
resample all to 192khz or is this (usb reciver output samplerate) set by the new master clock ?
It doesn't seem that Wave IO resamples anything by itself but if I'm reading it right the dsp
will be in need of a fixed samplerate that's the same as the systemrate.
Or is the systemrate something else then the set output samplerate and
that's set by the master clock?
 
Put me down for 1 board with volume chip.

That's cool, cheers! :)

it demands a certain speed of clock though still yes? in order for the filters to be correct?

There's actually some flexibility in this regard for dividing/multiplying the supplied clock, more details to come when this function is implemented.

One sure thing is that if you need a sampling frequency multiple of 44.1 kHz, then your external clock must be related somehow to 11.2896 MHz.
Likewise if you need a sampling frequency multiple of 48 kHz then you must supply an external clock related to 12.288 MHz.
You were probably expecting this.

I will go for the onboard clock to start with anyway and most likely it will work just fine. Any external clock will
be in a far away future anyway.

That's cool, thanks Bengt. Any other question please ask.
 
I found another one :D
How about "Double precision mode", how will this work ?
You wrote earlier that if needed you can switch from 24bit to 48bit precision.
Is this something that will be by user choise and how will that effect other
aspects of the XO ?
As an example, I could imagine that there could be a reduction in available
eq-bands or something like that.

Comments ?