Behringer DCX2496 digital X-over

Jump to Latest
#4,002
torgeirs said:
To BenToronto, off topic: Maybe you can use the
https://dspconcepts.com/st program and a STM32F746 Discovery when the Board Support Package is released?
I think it can be tuned to short delays.
Click to expand...
Feels like I dropped into the middle of a conversation on that website! But looks interesting.

Pity I am a Mac partisan (Apple customer since 1979).

But at the risk of being totally off-topic on this thread, I have been trying (unsuccessfully) to learn about using Aduino or Raspberry Pi as the brains inside a motional feedback system. Likewise miniDSP components might be helpful.

Anybody have some serious empirical information about my original question about the latency of the DCX2496? (BTW, is the "LE" any different interiorly?)

B.
 
#4,004
Davey said:
Did you click the link in my previous post?? It contains the serious empirical information you are looking for.
Click to expand...
Dave -

Thank you for re-helping me. I the link in your earlier post didn't stand out on my screen until you reminded me to look closely for it. Sorry.

But, again for a DIY forum, just posting ".8 ms..." is more tantalizing than definitive. Was that measured on a scope? What kind of processing load being handled? Regardless of audio frequency? What tasks push it slower or faster?

Again, appreciated.

Ben
 
#4,005
bentoro - i dont own a DCX - just playing with a ADAU1701 processor on a freedsp.

There is a constant delay ~1ms @ 48kHz given by the antialiasing filters in the AD and DA conversion stages. In my understanding this value will be halved by doubling the samplerate but i did not check this. On digital in- and outputs the constant delay is only one or two samples.

The processing load doesnt influence this because every sample runs through the complete processing chain bevor the next one arrives - it just has to be ready at this time.

There is a group delay of the filters you implement but this ist just a characteristic of the filter himself - not of the processing. And of course you can add extra delay by your wishes.

Maybe the DCX is not the optimal platform. Have a look at the freedsp or at sure's really cheap signal processing board. Both are programmable by the graphical development tool Sigmastudio from Analog Devices and give you the complete freedom to create any processing you may wish.
 
#4,006
ropf said:
bentoro - i dont own a DCX - just playing with a ADAU1701 processor on a freedsp.

There is a constant delay ~1ms @ 48kHz given by the antialiasing filters in the AD and DA conversion stages....
Click to expand...
Couldn't ask for more smart content. Thanks.

Since my second DCX2496 is on the way to me, guess I'll have to see.

For sure, delays are of no importance for the usual feedforward music system but matter a lot for my application.

You mention ADC delay. The speaker end is, of course, analog. But possibly working with a Class D amp, I might save processing time if the feedback loop DCX outputted digitally instead of analog. (I presently run my music system with AES input to Jack A using a USB to S/PDIF converter from the digital music file. That eliminates one DAC conversion in my chain.) Or just going to another processor platform, as has been suggested.

With motional feedback, several kinds of correction (EQ, passband) are needed. Since cone subwoofers are stupid engineering (a big resonance smack in the middle of the audible band), very tricky handling feedback.

But that means all kinds of phase/delay nonsense to address with .8 ms latency in the DCX2496 being just one headache in the feedback loop.
 
Last edited:
#4,008
ropf said:
Maybe you find this helpful...
Click to expand...
Thanks again.

Yes, the Dutch have a national commitment to using accelerometers for motional feedback going back to the Philips speakers. I have the same ACH-01 accelerometer on my bench ready for testing.

With the DCX, the balanced input can be immensely valuable for working with a sensor with a low voltage output. And the good S/N helps too. But still the accelerator output is very low (and unbalanced, I believe). So I'll likely need an op-amp before the DCX.

But.... when I test a Wheatstone Bridge as a motional feedback sensor, the balanced input might be perfect since it is a bridge. Again, pretty low voltage.

Anybody have thoughts about this circuitry?

Ben
 
#4,009
That is definitive. I measured with a dual-trace scope using a dirac pulse with no crossovers/EQ/delay/etc/etc of any kind active. One channel monitoring my source, the other channel monitoring the path through the DCX.

The 0.8mS is the inherent latency of this device.

Dave.
 
#4,014
Hi!

I can run my unit with usb-adapter. But suddenly i can not recall stored memorys and save from remote. I can do all tweaking and calibrating tough.

Can i "wake up" my unit by install a new firm on the dcx or is it a risky procedure?

Any suggestions?
Thanks!
 
#4,015
Davey said:
That is definitive. I measured with a dual-trace scope using a dirac pulse with no crossovers/EQ/delay/etc/etc of any kind active. One channel monitoring my source, the other channel monitoring the path through the DCX.

The 0.8mS is the inherent latency of this device.
Click to expand...
Finally got home and fired up the old 'scope.

Photo below is 100 Hz square wave, with audio signals in and out (haven't tried AES input to DCX). One cycle (or is that one Hertz??) is 10 boxes across. As you see, the DCX2496 (my new one, with no PCMIA slot, but not LE) is about half a box late or .5 ms.* Could be the "true" latency is somewhere between .5 and .8 ms.

Didn't see, to change when I loaded the DCX with various computing tasks.

You can fool with the time delay and see the latency shift about too, as expected. Hey, the DCX makes a great tool for working with audio because you can delay signals like when you are trying to superimpose an input and an output wave to see if they are identical, like with motional feedback.


Funny thing, I now realize that I have no idea what the DCX "phase" adjustment does (on the window that also inverts the signal). Can somebody tell me, please.

Ben
*which means the delay inside the DCX is not long enough to corrupt a motional feedback signal, if working only below, say, 150 Hz.
 

Attachments

  • 100 Hz sq wave DCX.jpg
    100 Hz sq wave DCX.jpg
    173.3 KB · Views: 377
Last edited:
#4,016
No, it's not "could be somewhere between 0.5mS and 0.8mS." It IS actually 0.8mS. 🙂
I'm not sure why it's so difficult to take my word for it on this. I performed this measurement many years ago. It hasn't changed. 🙂

Regards the phase adjustment query. That question has already been answered a few years ago in this very thread as well. 🙂
The phase adjustment implements a 2nd-order all-pass function keyed to the turnover frequency you have set in the crossover function. It works regardless if you have the crossover active or not.

Cheers,

Dave.
 
You must log in or register to reply here.
Top Bottom