Behringer DCX2496 digital X-over

I considered that...
knowing you i'm sure you did... but I wanted to know what made you decide against it, since i'm aware of a couple of projects you've got going that would seem to fit quite well with just starting from scratch and then using something like allocator, or leverage the Sandybridge chipset (Intel i3, i5, i7 etc) you probably have somewhere already for something like this

I just happen to think that software or DSP accelerated software is much more powerful these days than the aging DCX. the case.... well we both have a problem there so i'm not saying anything ;)

for tying into the sharc, maybe you could use a low ohm (20-47ohms) or perhaps even zero ohm SMD resistor network soldered right on the edge of a PCB which has your interfacing stuff on and then use solderpaste and hot air to mount that vertically. thats about the neatest way I can think of. of course it would need to be supported.
 
Last edited:

opc

Member
Joined 2004
Paid Member
While we're on the topic of mediocre performance, I've attached all my "baseline" measurements below. This is a bone-stock DCX2496 measured on an APx585.

Pretty much everything is buried in there somewhere. I didn't do THD vs. Frequency but I'll run those a little later before I modify anything.

I'm going to start by tearing out the entire SMPS module and replacing it with a fully linear supply using a separate PCB mount Amveco toroid for the digital and analog sections. I'll be using TI's TPS7A4700/3301 ultra low noise regulators pretty much everywhere. I'll make a pair of little break-off PCB's to replace the two on-board linear 5VA regs, and replace the 3.3VA reg with the same sort of thing but soldering that might be a little more tricky since it's derived from an SMD transistor.

I expect the above to clean up the noise floor by a large margin, along with moving the AES/EBU input away from the analog signals and direct wiring it to the isolation transformer.

I'll report back with progress.

Cheers,
Owen
 

Attachments

  • FFT Spectrum Monitor - midrange.png
    FFT Spectrum Monitor - midrange.png
    131 KB · Views: 412
  • Signal to Noise Ratio (2VRMS).png
    Signal to Noise Ratio (2VRMS).png
    55.4 KB · Views: 106
  • FR - 6 channels no XO.png
    FR - 6 channels no XO.png
    165.9 KB · Views: 87
  • FR - 6 channels 200Hz 1500Hz.png
    FR - 6 channels 200Hz 1500Hz.png
    257.7 KB · Views: 90
  • FR - 6 channels 200Hz 1500Hz (zoom).png
    FR - 6 channels 200Hz 1500Hz (zoom).png
    147.2 KB · Views: 87
  • FFT Spectrum Monitor digital input - idle.png
    FFT Spectrum Monitor digital input - idle.png
    141.4 KB · Views: 89
  • FFT Spectrum Monitor - zoom in no xo.png
    FFT Spectrum Monitor - zoom in no xo.png
    239.2 KB · Views: 386
  • FFT Spectrum Monitor - woofer.png
    FFT Spectrum Monitor - woofer.png
    131.6 KB · Views: 388
  • FFT Spectrum Monitor - tweeter.png
    FFT Spectrum Monitor - tweeter.png
    127.7 KB · Views: 390
  • FFT Spectrum Monitor - no XO.png
    FFT Spectrum Monitor - no XO.png
    170.6 KB · Views: 400

opc

Member
Joined 2004
Paid Member
It may be that something like the soon to be released miniSharc from minidsp might do what you're after Owen ... miniDSP - December 2012 Newsletter | miniDSP

Wow! That's pretty much exactly what I would like to start with. It looks like someone already did all the work that I have no idea how to do :)

hmm... definitely something to think about. I'll wait for it to release and see how the SW interface looks.

Thanks for the link!

and... a few more measurements as I hit the limit on the last post.
 

Attachments

  • THD+N Ratio - 9.1VRMS.png
    THD+N Ratio - 9.1VRMS.png
    50.5 KB · Views: 92
  • THD+N Ratio - 2VRMS.png
    THD+N Ratio - 2VRMS.png
    50.2 KB · Views: 99
  • Signal to Noise Ratio (9.1VRMS).png
    Signal to Noise Ratio (9.1VRMS).png
    53.9 KB · Views: 113
opc, i'm not sure ordinary quite covers it....

hochopeper thanks also for the link, looks like they finally removed enough stuff I dont want lol. the last version still had just a bit too much I didnt need and not enough power. I do wish they didnt force 96x sampling for everything...
 
unfortunately IMO they reinforce my point more than anything....

opc with his renowned analogue design chops and with the DSP and software side taken care of for him, if he can be arsed, stands to end up with a significantly better unit by going his own way.

afaik hes working with the ESS ES9102 ADC as well

nice job on the mods though! they really are pretty well thought out and well executed.
 
I might try it with my Amanero USB-I2S board this evening on the input side... that would allow a nice USB connection directly on the back of the DCX-2496 for digital input, which runs directly into the SHARC. No need for on-board ASRC, clocks, etc...

Please let as know! Im very interested. Actually im quite surprise that people who are making new digital input - didn't solved one very important thing. To get rid of ASRC. This is possible only via asynchronous USB or via slaving source to DCX dac clock /in fact not that difficult with sound card as Esi juli that can be slaved/ Im all for USB input to DCX! Maybe we can ask Amanero to write special version drivers with only 24/96 Khz enabled just to be on safe side. /in win7 you can force system to resample everything to desired FS/ Maybe this is worth of new thread.
 

opc

Member
Joined 2004
Paid Member
Might be these measurements are helpful for you:
Ergo Audio :: Stock vs. modded

Holy Jitter Batman!

That last measurement is unbelievable... Perhaps that helps explain why the stock unit gets on my nerves so badly.

I'll repeat a jitter test on mine as well before making any changes.

This is part of the reason I'm worried about IC19. If it's responsible for that level of jitter, then I get the feeling it's easier to just start from scratch with something like the miniDSP SHARC.

oettle:

What exact changes would you say were responsible for the reduction in jitter? You have improved it by a large margin, but I would still say that it could use further improvement.

From your listening, was the jitter reduction a significant improvement sound-wise? I'm not of the mind that every ps of jitter is audible, but when jitter artefacts clearly extend up into the audible range as they appear to in your last measurement, I would expect them to be audible.

Also, it appears from our measurements that the old PS was quieter than the new PS. Our stock measurements vary by quite a bit.

Cheers,
Owen
 
Does anyone have any information on exactly what IC19 does? It's a Quick Logic QL3004 but I can't seem to find any data on it.

It looks like either a level translator or some sort of re-clocking or switch interface.

QL3004 is a strange FPGA device (unless all FPGA's share the same I/O level flexibility). It's datasheet tells:

Advanced I/O Capabilities
• Interfaces with both 3.3 V and 5.0 V devices
• PCI compliant with 3.3 V and 5.0 V buses
for -1/-2/-3/-4 speed grades

It's got actually two supply inputs: Vcc which is Supply Voltage, min=3.0V, max=3.6V. So this is also 3.3V (nom) device. However, it's got another supply called Vccio meaning "I/O input tolerance voltage" which can vary from 3.0 to 5.25V in "commercial grade" devices. So it has some special logic I/O ports which might be programmable for different I/O voltages or just have such parameters for Vih, Vil, Voh (whatever are applicable) that the different I/O levels just work with the device with the set Vccio.
 
What exact changes would you say were responsible for the reduction in jitter? You have improved it by a large margin, but I would still say that it could use further improvement.

Here the list of mods again. The red ones are the more important ones.

qusp: I lived for about a year in Sydney. Very friendly and helpful people there. Small little world, isn't it.
I made the experience that understanding of basics isn't a disadvantage.
 

Attachments

  • DCX table.doc
    151 KB · Views: 159
by level shifter, maybe i've used some incorrect terminology, logic circuitry is not my thing, I mean shifting voltage compliance for the i2s to allow communication between 3v3 based and 5v based chips.

as for the rest....what? haha i'm at a loss to pull the basics from the pile of condescension. the DCX is a trainwreck before and after the mods compared to what can be done IMO, you made some improvement, which you should be congratulated for, so I did. more in the jitter than the rest, but 3db improvement overall is still significantly less than 16bit performance, no need for subtlety there.

i'm sure it can still be used as part of a good system, but no way would I expend the effort some are going to with the options that are currently available

first time i've ever heard the people of Sydney called friendly =) (Aussie in joke)

anyway i'll leave you guys to turn water into wine, godspeed!
 
Last edited:
I'm using 3 of these trainwrecks and I’m pretty happy with them. To my experience there is nothing wrong with the DSP, the ADC and DACs. The rest can be modded. So you end up with a price which is about twice the original price of the DCX which is still very cheap. There are much more expensive solutions out there which are not necessarily better.
 
I picked up some cheap ebay dac kits (one built, one truly a kit) and they are setup with the behringer now in some really deluxe mounting ;)

for diy guys, having spdif out really is a fun thing to have. I can try these dacs for a bit, hack on them, maybe try something else later if I want.

(these are dual-mono wm8741 with wm8805 receivers, all underneath the white boards).

I will do a measurement of these, later on, but sonically they blow the onboard analog-out system (including its stock psu) away.

the dual mono config of these ebay dacs is also kind of a fun detail, to me. each dac chip now handles its own speaker. left/tweeter gets its wm8741 chip, right/tweeter gets its own. no one has to share ;)

the wolfson chips have a 2db cut feature so you can't overdrive them when 'near 0db' levels come in (which happens hourly on many internet radio stations that I listen to). I don't miss 2db, but I do hate harsh clipping (a lot of cheaper modern dac chips do NOT handle gibbs issues well).

its also really nice that I no longer have to deal with the ultra hot 22db signals that come from the behringer analog-outs. those were really too strong for the cirrus vol control chip I use and I hated having to atten so much just to interface with them.
 

Attachments

  • 8275382706_a44a4a9d0c_b.jpg
    8275382706_a44a4a9d0c_b.jpg
    228.7 KB · Views: 304
  • 8275383160_b0e6749294_b.jpg
    8275383160_b0e6749294_b.jpg
    314 KB · Views: 306
Well done liuxworks. Congratulations!

I think your solution is a good approach to replace long analog cabling between DACs and power amps. Also the onboard clock oscillator might have a bit better jitter values than the poor DCX crystal circuitry.
For short cabling it is worth to think about passive analog outputs or Jan’s active outputs and about a low jitter clock instead.

Keep on modding - it’s fun!
 
but with the 'dac-sink' (half dac, half heatsink, lol!) sitting on the floor, I've swapped long dcx analog cables for long dac analog cables. heh.

I still think the cirrus transmitters would be easier to use since they don't need any local crystals. otoh, if removing the tanks from the wolfson still lets it work, then it would be an even draw between them.

the 8420 will have to be looked at next, I guess. everyone complains about it, so perhaps it does have to go ;)