Using 16/44.1 audio data it’s best to set the digital output of your PC to 24/44.1. So there is no quality loss by a software SRC. Quality loss can be significant. Typical values of software SRCs are between -50 and -180 dB THD.
If you use digital volume control or you are not sure your software does set your digital output to 24 bit. That’s because 16 bit data is equivalent to a dynamic range of about 110 dB. With e.g. a -40dB volume set up the dynamic range is reduced to 70 dB. If your digital output supports 16/44.1 only avoid digital volume control and use an analog volume control between DACs and power amps.
The internal data rate of the DCX is always 24/96. So 24/44.1 kHz data needs to be up sampled by an ASRC. The THD of an AD1896 is about -140dB. To improve this value it is possible to up sample data by e.g. a -170dB software SRC. But you have to replace the digital input (PLL + ASRC) of the DCX by an USB to I2S converter which is synchronous to the DCX clock. If this converter has its own clock which is not synchronous to DCX there is an additional ASRC necessary. So total THD would get worse.
Because nobody is able to hear the difference between -140dB and -170dB this USB mod isn’t a high priority mod. More important are other components of your audio chain like DCX crystal oscillator (-70dB), DACs (-100 dB), analog outputs (? dB), power amps (-60 to -100 dB) and speaker drivers (very good ones -60 to -70dB). Nevertheless THD is added. So it’s a matter of choice where to start with modding.
If you use digital volume control or you are not sure your software does set your digital output to 24 bit. That’s because 16 bit data is equivalent to a dynamic range of about 110 dB. With e.g. a -40dB volume set up the dynamic range is reduced to 70 dB. If your digital output supports 16/44.1 only avoid digital volume control and use an analog volume control between DACs and power amps.
The internal data rate of the DCX is always 24/96. So 24/44.1 kHz data needs to be up sampled by an ASRC. The THD of an AD1896 is about -140dB. To improve this value it is possible to up sample data by e.g. a -170dB software SRC. But you have to replace the digital input (PLL + ASRC) of the DCX by an USB to I2S converter which is synchronous to the DCX clock. If this converter has its own clock which is not synchronous to DCX there is an additional ASRC necessary. So total THD would get worse.
Because nobody is able to hear the difference between -140dB and -170dB this USB mod isn’t a high priority mod. More important are other components of your audio chain like DCX crystal oscillator (-70dB), DACs (-100 dB), analog outputs (? dB), power amps (-60 to -100 dB) and speaker drivers (very good ones -60 to -70dB). Nevertheless THD is added. So it’s a matter of choice where to start with modding.
Once again:
A SRC (sample rate CONVERSION) is always a conversion of time domain. Otherwise it is not a conversion at all. A SRC is NEVER LOSSLESS.
A SRC is like converting digital data to analog and back to digital again. But instead of a DAC and ADC there is a DSP with mathematical algorithms.
OK DOES ALL THIS Go out the window say when u put franks clock int he dcx and feed a direct 110ohm aes/ebu signal straight to the dcx main board at the transformer ??
would it matter feeding from a src or asrc device?
say a musical fidelity v link.. i got to think the aes connection is far superior to the spdif
at least in term of digital signal reflections.. which can cause jitter at the clock???
would it matter feeding from a src or asrc device?
say a musical fidelity v link.. i got to think the aes connection is far superior to the spdif
at least in term of digital signal reflections.. which can cause jitter at the clock???
asnswer V link 192~ it will output native input upsample(software) before you input the to the v link and presto feed the dcx 24/96 native.. much much better
go over to computeraudiophile.. there are some completely loss-less src converters..ie null test from one upscale to back to original and signal completely nulls..I said some.. no free one I have seen is completely lossless. that i have have found yet.. but it's better than letting the dcx upsample.
go over to computeraudiophile.. there are some completely loss-less src converters..ie null test from one upscale to back to original and signal completely nulls..I said some.. no free one I have seen is completely lossless. that i have have found yet.. but it's better than letting the dcx upsample.
ok just been following the threads over at computeraudiophile(phool) I wish i could find the thread but in compares a lot of the SRC's and does a lot of testing showing which ones will null and some wont..if it nulls on a different SRC than the one it was upsampled on pretty good bet it's inaudible that's why I say loss-less i e upscale SRC "a" downscale on SRC "b" and you get a null.
but only a few do that no free ones.. the worst sounding ones are the one that lose phase linearity.
but only a few do that no free ones.. the worst sounding ones are the one that lose phase linearity.
Last edited:
Summing up
Dear Friend,
I tried to read the 3D from the beginning...I stopped on message #500... I con't do it! I've to give up, it's too huge and it's hard to get from the overall discussion just what I need.
I got a DCX for 100 euros and I want to start to mod it. I replaced the psu with a linear high quality regulated psu. The improvement is there, but I suppose that the most important think is about output stages.
I have some questions, and I would be delighted if somebody can answer very quickly or pointing me to the heart of the 3D where specific discussion has been made. I also used the search tools, but the 3D is astronomically huge so it's very hard to filter out.
The DCX has a nice AKM voltage DAC. My experience is that the best thing for an oputput stage in this cases is either tubes or transformers. I've always been happy with transformers in high frequencies, on the bass they always lack the kind of drive that tube have (it may be lack of current). Anyway.
Questions:
1. What's the peak-2-peak and RMS voltages at 0dBFSat the DAC legs
2. I found some reference to output tubes and output transformers, could somebody point me to the most relevant posts on this side.
3. How the DCX handles the clock over the incoming SPDIF? I've seen that the DCX does resampling, is this done in an ASRC fashion? Does it pay to reclock the SPDIF?
Best
Pierre
Dear Friend,
I tried to read the 3D from the beginning...I stopped on message #500... I con't do it! I've to give up, it's too huge and it's hard to get from the overall discussion just what I need.
I got a DCX for 100 euros and I want to start to mod it. I replaced the psu with a linear high quality regulated psu. The improvement is there, but I suppose that the most important think is about output stages.
I have some questions, and I would be delighted if somebody can answer very quickly or pointing me to the heart of the 3D where specific discussion has been made. I also used the search tools, but the 3D is astronomically huge so it's very hard to filter out.
The DCX has a nice AKM voltage DAC. My experience is that the best thing for an oputput stage in this cases is either tubes or transformers. I've always been happy with transformers in high frequencies, on the bass they always lack the kind of drive that tube have (it may be lack of current). Anyway.
Questions:
1. What's the peak-2-peak and RMS voltages at 0dBFSat the DAC legs
2. I found some reference to output tubes and output transformers, could somebody point me to the most relevant posts on this side.
3. How the DCX handles the clock over the incoming SPDIF? I've seen that the DCX does resampling, is this done in an ASRC fashion? Does it pay to reclock the SPDIF?
Best
Pierre
Hi, I'm doing the single ended passive output mod as detailed on the dcx2496.fr site, and instead of cutting into the ribbon cable I am making a daughter board with both connectors on so I do not have to cut into the ribbon cable, similar to this:
Now as I am going single ended, would it be worth shorting the -ve (cold) analogue outputs from the DAC to ground or would it be better to leave them floating?
Thanks
Edit: Pierre77, I'll dig out a photo of the Pin out of the connector with the DAC analogue output pins, that way it's easy to measure the voltage.
I think that the peak to peak output voltage may be 5v, but you may want to check out the data sheet to be sure
Now as I am going single ended, would it be worth shorting the -ve (cold) analogue outputs from the DAC to ground or would it be better to leave them floating?
Thanks
Edit: Pierre77, I'll dig out a photo of the Pin out of the connector with the DAC analogue output pins, that way it's easy to measure the voltage.
I think that the peak to peak output voltage may be 5v, but you may want to check out the data sheet to be sure
Last edited:
1.7V RMS, I think. But I can measure for you if you like.
There are plenty of posts. You can search inside this thread. Me, I use transformers and love them. I've heard other types of output stage, but transformers sound best to me.
Yes, the ASRC does the reclock. It all get bumped to 96KHz. Some say it pays to reclock or resample before the DCX. The receiver chip can be unreliable from time to time. I have no problems driving it via AES from a DEQ2496.
Output stage stuff I have not tried
I have wondered this aloud before but someone thought it would be inferior:
Use the SSM2141 IC for differential to SE conversion per band/channel.
The other thing I have hoped to try is the Broskie Unbalancer x 3 boards, implemented with 6n16b-v tubes , THAT should work. There is prior art. If I did that I'd drop in AK4396 DACs.
And an upgrade idea I had recently that would go pretty fast if it worked?
Run the I2s heading for each AKM DAC out to ESS ES9023 DAC boards while leaving the AKM DACs in place. I am not an engineer but I have seen folks use some nominal resistor value on the I2S lines, I have seen everything from 100 to 50 to 36 to even 33 ohms used on the I2s lines. If there something about how the DCX is made that would NOT allow this? The MCK could come right from the DCX as usual? Or from a shared 50.000 MHz asynch clock for the ES9023s? I would probably feed them with a shared pre-regulator at 5-6Vdc, then have 3.6V [LT1761 or Micrel 5205 is it?] regulators for each DAC, a regulator for the optional 50MHz master MCK Oscillator. ES9023 DAC outputs could be buffered, or, not. Am I making this sound easier than it would be? I'd end up with stock differential outs, and some simple single-ended outs that would sound pretty darned nice. I have the DAC boards to try this. I ALSO have Frank's SRC and DCX clock upgrade board to install -- maybe the ES9023 clocking could come from there directly?
Background: I would only be using the DCX for playing FLAC via USB, bi or tri-amping, so, I also have interest in getting USB audio into the DCX the very best way, ok, easiest way, in combination with the above. I do not know if anyone has succeeded in getting a USB direct I2S input happening with the DCX yet? I would not care about rates other than 24/96. I would NOT care about any types of input but USB {FLAC}. Yes, I could use a USB to S/PDIF interface but that seems dumb.
I might or might not replace the DCX PSU. If I did so I'm just make a box outside the DAC with transformers and regulation/filtering.
I have wondered this aloud before but someone thought it would be inferior:
Use the SSM2141 IC for differential to SE conversion per band/channel.
The other thing I have hoped to try is the Broskie Unbalancer x 3 boards, implemented with 6n16b-v tubes , THAT should work. There is prior art. If I did that I'd drop in AK4396 DACs.
And an upgrade idea I had recently that would go pretty fast if it worked?
Run the I2s heading for each AKM DAC out to ESS ES9023 DAC boards while leaving the AKM DACs in place. I am not an engineer but I have seen folks use some nominal resistor value on the I2S lines, I have seen everything from 100 to 50 to 36 to even 33 ohms used on the I2s lines. If there something about how the DCX is made that would NOT allow this? The MCK could come right from the DCX as usual? Or from a shared 50.000 MHz asynch clock for the ES9023s? I would probably feed them with a shared pre-regulator at 5-6Vdc, then have 3.6V [LT1761 or Micrel 5205 is it?] regulators for each DAC, a regulator for the optional 50MHz master MCK Oscillator. ES9023 DAC outputs could be buffered, or, not. Am I making this sound easier than it would be? I'd end up with stock differential outs, and some simple single-ended outs that would sound pretty darned nice. I have the DAC boards to try this. I ALSO have Frank's SRC and DCX clock upgrade board to install -- maybe the ES9023 clocking could come from there directly?
Background: I would only be using the DCX for playing FLAC via USB, bi or tri-amping, so, I also have interest in getting USB audio into the DCX the very best way, ok, easiest way, in combination with the above. I do not know if anyone has succeeded in getting a USB direct I2S input happening with the DCX yet? I would not care about rates other than 24/96. I would NOT care about any types of input but USB {FLAC}. Yes, I could use a USB to S/PDIF interface but that seems dumb.
I might or might not replace the DCX PSU. If I did so I'm just make a box outside the DAC with transformers and regulation/filtering.
DCX MODS
unclear on the model; I probably have two of them, and 1-2 SRC2496 and a DEQ2496, all packed away. I know ONE of the SRCs is newer, not much on the output board TOP-SIDE, probably very busy UNDER, but THAT is off topic. The PLAN has been to use 2-3 matching amps and bi/tri-amp open-baffle arrays [the design of which keeps changing]. We got some cats so the Magnepans HAD to go away. So, there is a remote chance I'd want to input USB somewhere ELSE in that classic SRC DEQ DCX chain, but the bare minimum would be fix up the DCX and sell the rest.
At the time I ordered Frank's sweet upgrade board I had never touched anything SMD, now I have dragged myself into it and gotten some great tips from folks at DIYAUDIO {they are reading I'm sure, THANK YOU}. I ordered that pre-built, and for the record it comes with everything, I mean you could be trapped on an island and install the kit with what it comes with {well...}. WoW! A+ on that!
I now also have hot air technology... and have done some fairly fine work...and mods, but too much caffeine = not good. I suspect I could still wreck a DCX -- so part of me thinks it's less likely if I do not have to YANK any ICs. I can tack wire-wrap wire to ICs that size.
*If* it is as simple as snagging the I2S heading for the AKM DACS and sharing it, I could have a great single-ended conversion with minimal grief.
I HAve noted some ES9023 distortion on really "hot" recordings even running it at 3.6V and 100K pin#6 to ground, but a slight touch of the Windows or Mac volume slider {down} always cures it. I assume the DCX processor does it's magic and IS using I2s to the DACs? Somewhere I have a DCX PSU schematic, but from memory, that should not be hard to make a linear version of. Frank's board is pretty much a must [and I own it]. Hopefully I will leap_frog some classic problem situations.
unclear on the model; I probably have two of them, and 1-2 SRC2496 and a DEQ2496, all packed away. I know ONE of the SRCs is newer, not much on the output board TOP-SIDE, probably very busy UNDER, but THAT is off topic. The PLAN has been to use 2-3 matching amps and bi/tri-amp open-baffle arrays [the design of which keeps changing]. We got some cats so the Magnepans HAD to go away. So, there is a remote chance I'd want to input USB somewhere ELSE in that classic SRC DEQ DCX chain, but the bare minimum would be fix up the DCX and sell the rest.
At the time I ordered Frank's sweet upgrade board I had never touched anything SMD, now I have dragged myself into it and gotten some great tips from folks at DIYAUDIO {they are reading I'm sure, THANK YOU}. I ordered that pre-built, and for the record it comes with everything, I mean you could be trapped on an island and install the kit with what it comes with {well...}. WoW! A+ on that!
I now also have hot air technology... and have done some fairly fine work...and mods, but too much caffeine = not good. I suspect I could still wreck a DCX -- so part of me thinks it's less likely if I do not have to YANK any ICs. I can tack wire-wrap wire to ICs that size.
*If* it is as simple as snagging the I2S heading for the AKM DACS and sharing it, I could have a great single-ended conversion with minimal grief.
I HAve noted some ES9023 distortion on really "hot" recordings even running it at 3.6V and 100K pin#6 to ground, but a slight touch of the Windows or Mac volume slider {down} always cures it. I assume the DCX processor does it's magic and IS using I2s to the DACs? Somewhere I have a DCX PSU schematic, but from memory, that should not be hard to make a linear version of. Frank's board is pretty much a must [and I own it]. Hopefully I will leap_frog some classic problem situations.
CHECK HIS SOLDERING VERY GOOD ON THE CLOCK
both of mine would not work for crappy solder work.. the resistor network had solder bridges between the pins..
also because of poor engineering the clock will not fully seat on the dcx.. it hit the caps it sits over and eventually will work loose. replace the caps with shorter ones and glue the clock in place! or make sure you bend the pins and solder the connector up high enough on the main board.( he dorfed mine)
the stock dcx ps suck. esp the newer ones...
the caps are shitty and some of the regs are way out of spec..
check that first..esp if you have front panel issues.. or intermediate dropout issues.. the stock ps is fine if you spend $10 on parts..i'll try to find the link to a stock PS upgrade.. don't waste the the money on the linear PS unless you upgrade the Output section then i's still use the dcx to runt he digital side an make a separate linear to run the analog side..
the older dcx the copper traces are much more robust.. use chip quick to remove the chips..
also could check the ribbon cable if you use them
they come loose easily and the ribbons lose connection in the connector..
that;s why behringer glues the crap out of everything
both of mine would not work for crappy solder work.. the resistor network had solder bridges between the pins..
also because of poor engineering the clock will not fully seat on the dcx.. it hit the caps it sits over and eventually will work loose. replace the caps with shorter ones and glue the clock in place! or make sure you bend the pins and solder the connector up high enough on the main board.( he dorfed mine)
the stock dcx ps suck. esp the newer ones...
the caps are shitty and some of the regs are way out of spec..
check that first..esp if you have front panel issues.. or intermediate dropout issues.. the stock ps is fine if you spend $10 on parts..i'll try to find the link to a stock PS upgrade.. don't waste the the money on the linear PS unless you upgrade the Output section then i's still use the dcx to runt he digital side an make a separate linear to run the analog side..
the older dcx the copper traces are much more robust.. use chip quick to remove the chips..
also could check the ribbon cable if you use them
they come loose easily and the ribbons lose connection in the connector..
that;s why behringer glues the crap out of everything
http://www.diyaudio.com/forums/digital-line-level/231297-huge-issue-didden-dcx-mod.html
sounds likes crap I wouldn't wast my time with this POS!
sounds likes crap I wouldn't wast my time with this POS!
