Check there: Tweaking the Behringer DCX2496, audiophile digital active crossover
Everything's explained and illustrated.
Everything's explained and illustrated.
I know your awesome work on the dcx, no offense meant ;-).
Thierry Martin's solution seemed to correspond to kalmi79's approach (quick and cheap parts swap)
I consider adapting the relay/mute circuit of your passive solution to his tweak. The on/off thump being THE drawback of his approach, even though I can easily live with it.
As you're around, would you be so kind as to "decode" Oetlle's earlier remarks? I'm missing something I guess (if this has already been addressed, a link to the answer would be nice.I've read most of this huge thread but have skipped pages...I shouldn't have I guess)
I just received two freerider PCBs and two SRC AD1896 which you can get for free on Analog Device website (they sent me two samples totally free of charge). I also wonder on the TCXO clock. Wouldn't a 33.8688 clock allow true 192 compatibility?
Thierry Martin's solution seemed to correspond to kalmi79's approach (quick and cheap parts swap)
I consider adapting the relay/mute circuit of your passive solution to his tweak. The on/off thump being THE drawback of his approach, even though I can easily live with it.
As you're around, would you be so kind as to "decode" Oetlle's earlier remarks? I'm missing something I guess (if this has already been addressed, a link to the answer would be nice.I've read most of this huge thread but have skipped pages...I shouldn't have I guess)
I just received two freerider PCBs and two SRC AD1896 which you can get for free on Analog Device website (they sent me two samples totally free of charge). I also wonder on the TCXO clock. Wouldn't a 33.8688 clock allow true 192 compatibility?
I found the answer on pages 124 and 142 (I wish I could edit my previous post):
"I myself use an ultra low jitter 24.576 MHz oscillator, which supports digital inputs up to 96 kHz. The advantage is that PLL; SRC, DSP and DACs run absolutely synchronous for best sonical quality. You can also assemble a 28 or 30 MHz oscillator to support 192 kHz but then you have to stay with the existing high jitter 24.576 clock on the DSP board or you have to add an additional low jitter clock. Regarding sound quality it doesn’t make sense for me to down sample a 192 kHz input to the 96 kHz internal frequency of the DCX."
The selectronic solution does indeed do both, providing a replacement 24.576 clock replacing the original dsp one and adding a 28000 TCXO to support 192Khz. The debate around the audible degradation OR enhancement of the signal through upsampling is still open (last time I tried, I didn't hear a difference)
Wouldn't the use of a 24.576 clock on the freerider board (plus removing R34 + C77 +strapping c77) work as a general clock replacement?
"I myself use an ultra low jitter 24.576 MHz oscillator, which supports digital inputs up to 96 kHz. The advantage is that PLL; SRC, DSP and DACs run absolutely synchronous for best sonical quality. You can also assemble a 28 or 30 MHz oscillator to support 192 kHz but then you have to stay with the existing high jitter 24.576 clock on the DSP board or you have to add an additional low jitter clock. Regarding sound quality it doesn’t make sense for me to down sample a 192 kHz input to the 96 kHz internal frequency of the DCX."
The selectronic solution does indeed do both, providing a replacement 24.576 clock replacing the original dsp one and adding a 28000 TCXO to support 192Khz. The debate around the audible degradation OR enhancement of the signal through upsampling is still open (last time I tried, I didn't hear a difference)
Wouldn't the use of a 24.576 clock on the freerider board (plus removing R34 + C77 +strapping c77) work as a general clock replacement?
The Freerider solution solves the 'dull sound' problem but there is no sonical improvement (jitter). Also the Oettle design has an improved SRC/PLL design (supply, layout and cabling) for more accurate PLL operation.
The Selectronic solution (152 Euro) is similar to the Freerider one but it has an optional additional 'non low jitter' temperature controlled oscillator (TCXO). The TCXO doesn't make sense at all because your ear isn't a frequency counter. That means we can't here absolute frequencies but we can hear relative frequencies very well.
So important is the phase jitter RMS of the oscillator. The Oettle design (80/140 Euro) uses a 1 ps ultra low jitter oscillator. But the best oscillator won’t help if it doesn’t has a high PSRR (ripple rejection) and low noise power supply. This page (Tent) might help you understanding: Low jitter clocks
Also cabling is very important for low jitter because 1 ps is an equivalent frequency of 1000 GHz! So we are talking about every single millimetre but not centimetres.
You can measure the improved jitter. Ergo did it (link: The improvement …) SRC/Clock
I and also others could hear this difference.
You can also replace the existing S/PDIF transformer with a Scientific Conversion one (independently of SRC/PLL mod). My experience is that there is NO sonical difference at all because the digital input is reclocked anyway by the SRC to the internal 96 kHz sample rate of the DCX.
So the different designs seem to be more or less the same but they aren’t! The Oettle design is an improved Freerider solution plus a low jitter Tent clock.
My understanding is that unless the implementation is pathological (and the DCX doesn't seem to fit that category), jitter is not a problem and is inaudible.
Was your listening test(s) done double blind?
Yes, and the earth is not a globe because all Australians otherwise would fell into space. Nevertheless I was sure this is true I travelled to Sydney and lived there for about a year. So meanwhile I have the experience Australians won't fell into space. And this is for sure because my wife and my two children were part of this double blind test.
So, why don’t you buy a mod kit and share the experience of this sonical improvement with meanwhile a lot of other DCX users. I promise you there is absolutely no risk to make your own experience.
Btw: Sydney was a great experience for me and if it wouldn’t be so far away I would like to be there more often. So enjoy living there and perhaps you will also enjoy listening to my mod soon.
Hmm, thanks for the suggestion but I don't see a lot of audio engineering research support for jitter mods so I might just pass on that.
BTW, I really enjoyed Germany when I visited there in my younger days - but that was before the fall of the wall!
Thanks for the answer! As for Jan, I don't want my remarks to be taken as "attacks", I'm just a DIYer with a couple of Dcxs trying to sort it out (one for my active three way front speakers and the other for the remaining channels of my HT...a third one is coming for a further bass expansion). I will certainly end up with your SRC/clock in one and selectronic's in the other (I will measure and report here for everybody's sake of course). I don't mean to rain on anybody's parade and I know that differences are audible when the less-than-optimal power supply and clock circuits are replaced.
I disagree on the "non low jitter"...the selectronic clock is <10ps and yours is <2ps
and there is a sonic improvement with their design*. (no reason not to trust Thierry Martin and other French diyers who, like you, have no commercial interest in saying so) Their clock is an improvement on Tent labs X02 on the same lines as you describe and is meant to be connected to their low noise/high PSSR power supply. Your power supply is integrated (hence a cheaper solution) yet their power supply also provides the Dacs with cleaner current)
I can understand the point you make about cable length: your solution does indeed provide the shortest path.
To summarize:
Freerider : removes original SRC bug. Improves jitter rejection (claimed <5ps could be optimized by reduced cabling ? and power drawn from selectronic low noise linear module instead of IC1 pads?) 192Khz compatibility
no reclocking (which could be DIYed IMHO since it's done from the same pins on the other two solutions but at the expense of 192Khz).
Oetlle: best jitter improvement <2ps. Full sync reclocking 96Khz. Integrated optimized power supply. (Already improved Freerider solution)
Selectronic: jitter improvement <10ps. 192Khz compatibility and Reclocking (optimized from Tent Labs x02) Draws power from external low noise power supply (which also regulates DACs supply). Most expensive solution but includes scientific conversion transformer for true spdif & AES impedance matching (this can be important depending on the source/connection).
It comes down to a matter of price/ease of installation and a somewhat audiophile audibility frontier (Is a less than 10ps jitter difference audible...I doubt it). I do believe that the price of each solution does correspond to what they offer.
Thanks everybody for the fascinating work!
*The fact is that Thierry's site is much more complete in French (and justifies the power supply changes and other design choices) and that some other personal sites (also in French) do tell of comparative listening.
I disagree on the "non low jitter"...the selectronic clock is <10ps and yours is <2ps
and there is a sonic improvement with their design*. (no reason not to trust Thierry Martin and other French diyers who, like you, have no commercial interest in saying so) Their clock is an improvement on Tent labs X02 on the same lines as you describe and is meant to be connected to their low noise/high PSSR power supply. Your power supply is integrated (hence a cheaper solution) yet their power supply also provides the Dacs with cleaner current)
I can understand the point you make about cable length: your solution does indeed provide the shortest path.
To summarize:
Freerider : removes original SRC bug. Improves jitter rejection (claimed <5ps could be optimized by reduced cabling ? and power drawn from selectronic low noise linear module instead of IC1 pads?) 192Khz compatibility
no reclocking (which could be DIYed IMHO since it's done from the same pins on the other two solutions but at the expense of 192Khz).
Oetlle: best jitter improvement <2ps. Full sync reclocking 96Khz. Integrated optimized power supply. (Already improved Freerider solution)
Selectronic: jitter improvement <10ps. 192Khz compatibility and Reclocking (optimized from Tent Labs x02) Draws power from external low noise power supply (which also regulates DACs supply). Most expensive solution but includes scientific conversion transformer for true spdif & AES impedance matching (this can be important depending on the source/connection).
It comes down to a matter of price/ease of installation and a somewhat audiophile audibility frontier (Is a less than 10ps jitter difference audible...I doubt it). I do believe that the price of each solution does correspond to what they offer.
Thanks everybody for the fascinating work!
*The fact is that Thierry's site is much more complete in French (and justifies the power supply changes and other design choices) and that some other personal sites (also in French) do tell of comparative listening.
To be a bit more precise: The Freerider solution has no relevant jitter improvement at all. It might be that the AD1896 (which is used in all 3 designs) is more tolerant to jitter on the digital input than the CS8420 but that’s unimportant as long you do not have a very worse digital input signal. I assume that this 5 ps value is coming from a wrong reading of the AD1896 data sheet?
The standard Selectronic solution (135 Euro) also has no jitter reduction at all but the optional one (152 Euro) provides a TCXO to improve jitter. So far I haven’t seen a TCXO with specified phase jitter RMS. So I would repeat my opinion that the Selectronic oscillator is not a special low jitter type. But probably it is better than the existing crystal solution of the DCX.
The oscillator used on the Oettle (80/140 Euro) design is a special ultra low jitter type (phase jitter RMS typ. 0.5ps max. 1ps (12kHz-20MHz)) which is very difficult to get. This oscillator is comparable with the Tent oscillators. The oscillator has a separate ultra low PSRR (min. 110 dB) and ultra low noise (18uV) power supply which is located directly beside the oscillator. Standard linear supplies have a 50dB PSRR and much worse noise. With such a standard supply even the best oscillator isn’t low jitter any more. That’s similar with a low noise supply (and the Selectronic isn’t to my understanding) – with a 10 cm cabling even the lowest noise supply isn’t any more low noise.
I would disagree that the price/performance ratio of the three SRC solutions is similar. If you want to get rid of the ‘dull sound’ problem only but don’t need sonical improvement the Freerider solution might be the cheapest one (if you get the ICs for free ).
Where does this information come from that the original DCX transformer doesn’t match S/PDIF specification? The input impedance is realized by a simple 110 ohm resistor which can easily be changed. I even doubt that a lower value resistor will improve something except for very long cabling (several 10 metres).
Although the hugh cap lovers will tell you the opposite:
I don’t want to say that the switched power supply of the DCX is the very best. There is a lot of noise at about 50 kHz. But on the other side a linear supply has a lot of noise at 50 Hz. My ears can hear 50 Hz but not 50 kHz. So to my opinion a linear supply is worse than a switched one. So far I haven’t seen any sonical measurement of a linear supply which improves sonical performance. Some of the so called linear supplies even use a DC/DC converter for the +5V rail because of thermal problems. To my knowledge a DC/DC converter is a switched supply.
If you are looking for a real low noise power supply for the DACs – here you are: Vreg
Regarding the input sample frequency it is best to use the sample frequency of the data source. So if your source is e.g. a CD use 44.1 kHz. That’s because the best sample rate converter is the one which isn’t there, which means avoid unnecessary conversions.
The standard Selectronic solution (135 Euro) also has no jitter reduction at all but the optional one (152 Euro) provides a TCXO to improve jitter. So far I haven’t seen a TCXO with specified phase jitter RMS. So I would repeat my opinion that the Selectronic oscillator is not a special low jitter type. But probably it is better than the existing crystal solution of the DCX.
The oscillator used on the Oettle (80/140 Euro) design is a special ultra low jitter type (phase jitter RMS typ. 0.5ps max. 1ps (12kHz-20MHz)) which is very difficult to get. This oscillator is comparable with the Tent oscillators. The oscillator has a separate ultra low PSRR (min. 110 dB) and ultra low noise (18uV) power supply which is located directly beside the oscillator. Standard linear supplies have a 50dB PSRR and much worse noise. With such a standard supply even the best oscillator isn’t low jitter any more. That’s similar with a low noise supply (and the Selectronic isn’t to my understanding) – with a 10 cm cabling even the lowest noise supply isn’t any more low noise.
I would disagree that the price/performance ratio of the three SRC solutions is similar. If you want to get rid of the ‘dull sound’ problem only but don’t need sonical improvement the Freerider solution might be the cheapest one (if you get the ICs for free
).Where does this information come from that the original DCX transformer doesn’t match S/PDIF specification? The input impedance is realized by a simple 110 ohm resistor which can easily be changed. I even doubt that a lower value resistor will improve something except for very long cabling (several 10 metres).
Although the hugh cap lovers will tell you the opposite:
I don’t want to say that the switched power supply of the DCX is the very best. There is a lot of noise at about 50 kHz. But on the other side a linear supply has a lot of noise at 50 Hz. My ears can hear 50 Hz but not 50 kHz. So to my opinion a linear supply is worse than a switched one. So far I haven’t seen any sonical measurement of a linear supply which improves sonical performance. Some of the so called linear supplies even use a DC/DC converter for the +5V rail because of thermal problems. To my knowledge a DC/DC converter is a switched supply.
If you are looking for a real low noise power supply for the DACs – here you are: Vreg
Regarding the input sample frequency it is best to use the sample frequency of the data source. So if your source is e.g. a CD use 44.1 kHz. That’s because the best sample rate converter is the one which isn’t there, which means avoid unnecessary conversions.
Your information that the Selectronic clock is an improvement on the Tent labs one made me more curious. Unfortunately I can’t speak French and the pictures of the Selectronic SRC mod are pretty tiny.
1. My guess is that they use a LM1117 for power supply? The PSRR for this device is about 65 dB @20kHz and 40 dB @100kHz getting worse at higher frequencies. The output noise is 10 mV. Tent uses a very good shunt regulator. Compared to the Oettle values (110 dB and 18uV) the Selctronic supply is about 100 to 1000 times worse.
2. The used TCXO might be a 10ps jitter device (Is this value based on the range 12kHz to 20MHz ? Unfortunately there is no clear data.). The Tent one specifies 3ps jitter from 10Hz onward which is equivalent to 1ps (12kHz to 20MHz).
3. The TXCO oscillator seems to be a 3.3V output type although the input device on the DCX board is a 5V type.
If my above assumptions are right I can’t understand why the Selectronic solution should be better compared to the Tent solution. (Sorry, I even would call it a POOR design!)
AES/EBU specifies 110 ohm impedance and S/PDIF 75 ohm. The Scientific Conversion and the existing DCX transformer are both specified for 75 to 110 ohm. So for S/PDIF use it might make sense to change the onboard 110 ohm resistor to 75 ohm. Also the existing DCX cabling between the input connector and the transformer could be improved but I can’t see any sonically relevant difference between the two transformers.
So the only true information seems to be that the Selectronic mod is the most expensive (if my assumptions are right).
Thanks for the precisions!
Most of my following answers to your questions are summaries from Thierry Martin's sites
He had mounted three Tentlab XO2 clocks before he put the first selectronic clock mode which, according to him, goes much further in terms of jitter reduction and its positive impact on musicality and microdetails. The new module provides the same, adding an improved src and a second clock. That's why I was confused by your remark since their claims and Thierry's report are almost identical to yours: freerider debug + ultra low jitter clock.
Here are the links to their clock's specs (not much information actually):
http://www.dcx2496.fr/docs/7082-3.pdf
http://www.dcx2496.fr/docs/tcxo.jpg
The cabling of the CS8420 doesn't match the specifications of the datasheet for consumer interface. Changing the resistor to 75ohms solves half the problem. This is only relevant for the installation of a spdif input to the CS8420): RXN should be connected to ground through a 0.01uf cap
Explanations are here: Entrée SPDIF pour le Behringer DCX2496
For the linear supply, I know only one measurement here: AW DIY - DEQ/DCX PSU advantages
Those are from Stef1777 who announced his project on this very thread (around p100)
As for selectronic's power supply, I found no measurement but a couple of reviews including that one: TWEAK DCX DIY. On this page, the diyer summed up his impressions about the various enhancements of his dcx. Details about it are given here: alimentation convertisseurs Behringer DCX2496.
As I have read any posts from both you and Thierry Martin on this very thread, and not being an historian of the thread (not yet...) I wonder...haven't you two directly discussed the differences of your respective mods? (which point toward the same direction but through different means).
For other freerider experimenters: everybody can get TWO SRC AD1896 for free as sample on Analog Devices website. You just have to register (for free)
Most of my following answers to your questions are summaries from Thierry Martin's sites
He had mounted three Tentlab XO2 clocks before he put the first selectronic clock mode which, according to him, goes much further in terms of jitter reduction and its positive impact on musicality and microdetails. The new module provides the same, adding an improved src and a second clock. That's why I was confused by your remark since their claims and Thierry's report are almost identical to yours: freerider debug + ultra low jitter clock.
Here are the links to their clock's specs (not much information actually):
http://www.dcx2496.fr/docs/7082-3.pdf
http://www.dcx2496.fr/docs/tcxo.jpg
The cabling of the CS8420 doesn't match the specifications of the datasheet for consumer interface. Changing the resistor to 75ohms solves half the problem. This is only relevant for the installation of a spdif input to the CS8420): RXN should be connected to ground through a 0.01uf cap
Explanations are here: Entrée SPDIF pour le Behringer DCX2496
For the linear supply, I know only one measurement here: AW DIY - DEQ/DCX PSU advantages
Those are from Stef1777 who announced his project on this very thread (around p100)
As for selectronic's power supply, I found no measurement but a couple of reviews including that one: TWEAK DCX DIY. On this page, the diyer summed up his impressions about the various enhancements of his dcx. Details about it are given here: alimentation convertisseurs Behringer DCX2496.
As I have read any posts from both you and Thierry Martin on this very thread, and not being an historian of the thread (not yet...) I wonder...haven't you two directly discussed the differences of your respective mods? (which point toward the same direction but through different means).
For other freerider experimenters: everybody can get TWO SRC AD1896 for free as sample on Analog Devices website. You just have to register (for free)
Hello Frank,
Really glad to see you back here. Just to add, regarding the digital transformers I cannot say what is the difference in sound one vs. other since I did that conversion with all other improvements, but...
Jon Paul of Scientific Conversion fame is my good friend and we did the measurement to compare his transformer vs. the one in Behringer. SC outperforms it in tenfolds. Particularly in capacitance measurement which greatly influence the shape of the signal that is passing through. Now could we hear that I do not know, but I am sure we are better with intact signal than with the one that is changed in shape due to not as good transformer.
As I mentioned numerous times, your board brings DCX's performance to a completely different level that is clearly audible and in my mind presents dramatical improvement equal to eliminating opamps in output. The very same excitement about the board I had was confirmed by very advanced users; to mention few - Jan Diden and Ergo. We spoke about this in this thread on numerous occasions, and we measured and posted results. If none of that was not helpful for someone to have a sure path to a success than nothing will.
Really glad to see you back here. Just to add, regarding the digital transformers I cannot say what is the difference in sound one vs. other since I did that conversion with all other improvements, but...
Jon Paul of Scientific Conversion fame is my good friend and we did the measurement to compare his transformer vs. the one in Behringer. SC outperforms it in tenfolds. Particularly in capacitance measurement which greatly influence the shape of the signal that is passing through. Now could we hear that I do not know, but I am sure we are better with intact signal than with the one that is changed in shape due to not as good transformer.
As I mentioned numerous times, your board brings DCX's performance to a completely different level that is clearly audible and in my mind presents dramatical improvement equal to eliminating opamps in output. The very same excitement about the board I had was confirmed by very advanced users; to mention few - Jan Diden and Ergo. We spoke about this in this thread on numerous occasions, and we measured and posted results. If none of that was not helpful for someone to have a sure path to a success than nothing will.
Hello Vladimir,
Thanks for your help. I just try to sort out the features of the different mods which seems to be a tough job because there is a lot of wrong and missing information. But the biggest problem is that there is a lot of believing: E.g. it’s better because it’s more expensive. Also there is a lot of changing but not improving because e.g. it’s cheap … But nobody is born with all this knowledge. So I try to share my little knowledge with others.
Regarding the Scientific transformers you are right - it’s better than the original DCX one although the difference isn’t such dramatic as the Scientific web site tries to tell us. I myself have several of them. Based on my measurements and listening there is no sonically improvement. That’s because the jitter on the S/PDIF line is to some extent unimportant because the signal is reclocked by the SRC anyway. It’s only important that the PLL is able to lock properly on the incoming signal. But depending on the source and cabling in some application the Scientific transformer might be an improvement. Nevertheless I wouldn’t start with this mod.
Kind Regards,
Thanks for your help. I just try to sort out the features of the different mods which seems to be a tough job because there is a lot of wrong and missing information. But the biggest problem is that there is a lot of believing: E.g. it’s better because it’s more expensive. Also there is a lot of changing but not improving because e.g. it’s cheap … But nobody is born with all this knowledge. So I try to share my little knowledge with others.
Regarding the Scientific transformers you are right - it’s better than the original DCX one although the difference isn’t such dramatic as the Scientific web site tries to tell us. I myself have several of them. Based on my measurements and listening there is no sonically improvement. That’s because the jitter on the S/PDIF line is to some extent unimportant because the signal is reclocked by the SRC anyway. It’s only important that the PLL is able to lock properly on the incoming signal. But depending on the source and cabling in some application the Scientific transformer might be an improvement. Nevertheless I wouldn’t start with this mod.
Kind Regards,
Thank you Beelde for the additional information. Here are my comments based on datasheets and real measurements. I DO NOT comment believing, opinions or rumours.
Thierry seems to compare a former version of the Tent clock About upgrading
with the Selectronic ‘ultra? low jitter clock part kit’ (49 Euro) SELECTRONIC ::: L'univers électronique :::
Selectronic claims that its clock is better than the Tent one. This might or might not be true regarding the power supply.
The Tent oscillator is definitely better regarding phase jitter (3ps @10Hz onwards). The Selectronic datasheet does NOT specify jitter at all but it shows that it is a 3.3V type although the input device on the DSP board is a 5V type. The usage of a temperature compensated oscillator (TCXO) shows me that the designer didn’t understand the sonically relevant nature of jitter. TCXOs are used to build frequency counters. Meanwhile Tent sells a new clock with a shunt regulator Low jitter clocks
I guess the new power supply of this Tent clock is better than the Selectronic one.
But if I understood you correctly you want to buy the Selectronic digital input / clock (152 Euro) SELECTRONIC ::: L'univers électronique :::
This design is based on two simple 3.3V linear regulators. Unfortunately I can’t read the exact type. So I guess it’s a LM1117 or a LM2936 (as used in the Freerider design). For the PLL and also the clock oscillator high PSRR and low noise are very important for stable PLL operation and a low jitter clock.
Here are some values to compare the different designs: LM1117: noise 10000uV, PSRR 60dB@20kHz, 35dB@200kHz, LM2936: noise 500uV, PSRR 50dB@20kHz, 30dB@200kHz. The two supplies (3.3V and 5V) on the Oettle SRC/clock mod are based on a double stage regulation: noise 18uV, PSRR 110dB@20kHz and 200 kHz.
Also cabling has an impact on noise and jitter values.
S/PDIF defines a cheap consumer interface and a professional isolated one based on a transformer. If your S/PDIF source has the same grounding as your DCX and if there is a very short S/PDIF cable it might be an advantage to omit the transformer. For most applications it’s better to stay with this transformer because of galvanic isolation.
To evaluate the impact of a linear power supply I would measure the noise floor of the DCX analog outputs. Obviously you haven’t seen such a measurement too. My guess is that there isn’t a real improvement but perhaps at 50/60Hz even a worsening.
I’m aware of the scope picture AW DIY - DEQ/DCX PSU advantages
Reading this picture my estimation is that the 50kHz noise of the SMPS is reduced to about 1/3 with the linear supply. So we are talking about a 10dB noise reduction at 50 kHz. Not pretty much. What the picture doesn’t show is what happens at 50/60 Hz. As already mentioned my ears can hear 50Hz but not 50kHz.
A very simple and cheap solution to improve the existing DCX SMPS would be to add CLC filters on the +9V and +/-15V rails (before the LM 78/7915 regulators) similar to the +3.3 and +5V rails. http://www.awdiy.com/uploads/pdf/DEQ2496-PSU-1.5.pdf
I’m aware of 3 linear supply mods. Jans (which is the most sophisticated) replaces the DCX supply completely. The Selectronic one replaces the analog rails +9V and +/-15V only. Steff’s one uses a DC/DC converter for the +5V rail because of thermal problems. Don’t ask me about technical details. I’m using none of these supplies.
If you are concerned about ADC and DAC supply noise I wouldn’t do that alimentation convertisseurs Behringer DCX2496 but I would replace the two 7805 regulators on the DSP board by this Vreg
That’s really low noise!
I DO NOT comment this TWEAK DCX DIY
My conclusion is if you want to do the SRC/clock mod and you have already the AD1896: Talk to Ward perhaps he is selling you the Oettle mod without this device for e.g. 70 Euro. If this is too expensive and you want to get rid of the ‘dull sound’ problem only use the Freerider solution.
Also a very important mod regarding sound quality is the output stage. If you have high impedance amplifiers with less than about 1 m cabling use this DCX passive upgrade
Otherwise or if you need volume control use that DCX 2496 active upgrade
When you have done these two mods and you have some money left you can also do these mods (but there is no guarantee for big sonical improvement if at all):
1. Replace DACs by AK4396
2. Replace 7805 Regulators at least the one for the DACs
3. Use linear supply
4. Replace S/PDIF transformer by Scientific one.
OK, that’s most of my knowledge regarding the DCX. Now it’s getting difficult. Whom should you believe? Good luck!
Thierry seems to compare a former version of the Tent clock About upgrading
with the Selectronic ‘ultra? low jitter clock part kit’ (49 Euro) SELECTRONIC ::: L'univers électronique :::
Selectronic claims that its clock is better than the Tent one. This might or might not be true regarding the power supply.
The Tent oscillator is definitely better regarding phase jitter (3ps @10Hz onwards). The Selectronic datasheet does NOT specify jitter at all but it shows that it is a 3.3V type although the input device on the DSP board is a 5V type. The usage of a temperature compensated oscillator (TCXO) shows me that the designer didn’t understand the sonically relevant nature of jitter. TCXOs are used to build frequency counters. Meanwhile Tent sells a new clock with a shunt regulator Low jitter clocks
I guess the new power supply of this Tent clock is better than the Selectronic one.
But if I understood you correctly you want to buy the Selectronic digital input / clock (152 Euro) SELECTRONIC ::: L'univers électronique :::
This design is based on two simple 3.3V linear regulators. Unfortunately I can’t read the exact type. So I guess it’s a LM1117 or a LM2936 (as used in the Freerider design). For the PLL and also the clock oscillator high PSRR and low noise are very important for stable PLL operation and a low jitter clock.
Here are some values to compare the different designs: LM1117: noise 10000uV, PSRR 60dB@20kHz, 35dB@200kHz, LM2936: noise 500uV, PSRR 50dB@20kHz, 30dB@200kHz. The two supplies (3.3V and 5V) on the Oettle SRC/clock mod are based on a double stage regulation: noise 18uV, PSRR 110dB@20kHz and 200 kHz.
Also cabling has an impact on noise and jitter values.
S/PDIF defines a cheap consumer interface and a professional isolated one based on a transformer. If your S/PDIF source has the same grounding as your DCX and if there is a very short S/PDIF cable it might be an advantage to omit the transformer. For most applications it’s better to stay with this transformer because of galvanic isolation.
To evaluate the impact of a linear power supply I would measure the noise floor of the DCX analog outputs. Obviously you haven’t seen such a measurement too. My guess is that there isn’t a real improvement but perhaps at 50/60Hz even a worsening.
I’m aware of the scope picture AW DIY - DEQ/DCX PSU advantages
Reading this picture my estimation is that the 50kHz noise of the SMPS is reduced to about 1/3 with the linear supply. So we are talking about a 10dB noise reduction at 50 kHz. Not pretty much. What the picture doesn’t show is what happens at 50/60 Hz. As already mentioned my ears can hear 50Hz but not 50kHz.
A very simple and cheap solution to improve the existing DCX SMPS would be to add CLC filters on the +9V and +/-15V rails (before the LM 78/7915 regulators) similar to the +3.3 and +5V rails. http://www.awdiy.com/uploads/pdf/DEQ2496-PSU-1.5.pdf
I’m aware of 3 linear supply mods. Jans (which is the most sophisticated) replaces the DCX supply completely. The Selectronic one replaces the analog rails +9V and +/-15V only. Steff’s one uses a DC/DC converter for the +5V rail because of thermal problems. Don’t ask me about technical details. I’m using none of these supplies.
If you are concerned about ADC and DAC supply noise I wouldn’t do that alimentation convertisseurs Behringer DCX2496 but I would replace the two 7805 regulators on the DSP board by this Vreg
That’s really low noise!
I DO NOT comment this TWEAK DCX DIY
My conclusion is if you want to do the SRC/clock mod and you have already the AD1896: Talk to Ward perhaps he is selling you the Oettle mod without this device for e.g. 70 Euro. If this is too expensive and you want to get rid of the ‘dull sound’ problem only use the Freerider solution.
Also a very important mod regarding sound quality is the output stage. If you have high impedance amplifiers with less than about 1 m cabling use this DCX passive upgrade
Otherwise or if you need volume control use that DCX 2496 active upgrade
When you have done these two mods and you have some money left you can also do these mods (but there is no guarantee for big sonical improvement if at all):
1. Replace DACs by AK4396
2. Replace 7805 Regulators at least the one for the DACs
3. Use linear supply
4. Replace S/PDIF transformer by Scientific one.
OK, that’s most of my knowledge regarding the DCX. Now it’s getting difficult. Whom should you believe? Good luck!
Last edited:
Thank you...actually, I am not set on any mod but as I have two dcxs to mod (bought as defective units and fixed...and waiting for a third) I did a lot of research about various solutions. I've been familiar with the different mods for quite a while and seeing you and Jan active on this thread was the perfect occasion to get first hand experience about the different solutions. The last pages might save a lot of time to some diyers (who, like me, had better spend time soldering and listening to music)
I'll definitely try your src mod (I have plenty of arguments and testimonies to try it). I might try selectronic's on the second (if my budget backs up my will to experiment at the time...but I'm often more curious than rich). The Freerider boards might come in handy for a third dcx managing less critical channels.
As for the output stage,
I have a selectronic I/O board and plan on using a passive upgrade on my second DCX. I won't need volume control as I my current diy project is a 4 channel UcD 180HgxHr amp with stepped attenuators (and a system to switch between balanced, unbalanced and attenuator bypass to allow me to do A/B comparison) The passive mod will certainly be Thierry's cap coupling with a DPDT switch to choose between unbalanced and balanced (Hypex UcD are fine with both but as my first full balanced amp modules, I feel like trying it first hand). I consider adding Jan's relay based mute circuit to it.
Now I feel like I can make an educated guess regarding the different solutions. It's always nice to have a designer explaining and presenting his own decisions.
It's not really getting difficult now. Jan's solution is well documented, Selectronic's as well by Thierry but I didn't have as many details for yours...which is now corrected. I can stop researching for a while and move to the testing phase. Great! Thanks again!
By the way, funding Gumyoko montessori school is a really nice and worth supporting gesture
I'll definitely try your src mod (I have plenty of arguments and testimonies to try it). I might try selectronic's on the second (if my budget backs up my will to experiment at the time...but I'm often more curious than rich). The Freerider boards might come in handy for a third dcx managing less critical channels.
As for the output stage,
I have a selectronic I/O board and plan on using a passive upgrade on my second DCX. I won't need volume control as I my current diy project is a 4 channel UcD 180HgxHr amp with stepped attenuators (and a system to switch between balanced, unbalanced and attenuator bypass to allow me to do A/B comparison) The passive mod will certainly be Thierry's cap coupling with a DPDT switch to choose between unbalanced and balanced (Hypex UcD are fine with both but as my first full balanced amp modules, I feel like trying it first hand). I consider adding Jan's relay based mute circuit to it.
Now I feel like I can make an educated guess regarding the different solutions. It's always nice to have a designer explaining and presenting his own decisions.
It's not really getting difficult now. Jan's solution is well documented, Selectronic's as well by Thierry but I didn't have as many details for yours...which is now corrected. I can stop researching for a while and move to the testing phase. Great! Thanks again!
By the way, funding Gumyoko montessori school is a really nice and worth supporting gesture
I read this web site once more AW DIY - DEQ/DCX PSU advantages
I have to admit I didn't read it carefully enough the first time. The headline claims the scope pictures show the 'Ground floor noise at DCX ouput'.
That would mean the output noise floor is about -50dB for the original and -60dB for the modded (linear supply) DCX. If that would be true you could hear both noise levels clearly.
But Behringer specifies a max noise level of -112dB and Ergos measurements http://www.pilghamaudio.com/_data/docs/pdfs/jitter.pdf show a noise level of about -130dB (yellow: unmodded) and -135dB (green: modded analog output and Oettle SRC/clock) both with the original PSU.
So my conclusion is that the scope pictures can't show output noise level and that there isn't a real noise problem at all even with the original PSU.
You are right, most probably the scope pictures show "noise at output of supply" of one of the PSU rails but not analog output.
If so another interesting question would be if you see only the noise measured at the probe tip. My assumption is that the SPMS measurement shows additional noise coming from mains. So real noise level might be lower for the SMPS.
If so another interesting question would be if you see only the noise measured at the probe tip. My assumption is that the SPMS measurement shows additional noise coming from mains. So real noise level might be lower for the SMPS.