You're welcome - looking forward to hearing your report on how it sounds 🙂
I'm working on the filter for the 12 chip version, will post it up in a while.
BTW - when piggy-backing chips, bend pin7 outwards on the chips to be added as there's nothing to gain from connecting it. Only the remaining 7 pins need to be connected.
I'm working on the filter for the 12 chip version, will post it up in a while.
BTW - when piggy-backing chips, bend pin7 outwards on the chips to be added as there's nothing to gain from connecting it. Only the remaining 7 pins need to be connected.
Last edited:
Oh my!
Abrax, may I order The Kubelik R3 or it will be obsolete in the next three minutes?
🙂
George
Abrax, may I order The Kubelik R3 or it will be obsolete in the next three minutes?
🙂
George
Filter components for 12 chip version :
L9,L10 become 10mH Bourns RL181S-103
C12,C18 become 10n
C13,C19 become 1n8
C14,C20 become 5n6
C15,C21 become 820p
R40,R62 become 910R
Caps are NP0/C0G 50V
L9,L10 become 10mH Bourns RL181S-103
C12,C18 become 10n
C13,C19 become 1n8
C14,C20 become 5n6
C15,C21 become 820p
R40,R62 become 910R
Caps are NP0/C0G 50V
Haha George - its only right that the DIY variants improve on the stock offering 🙂 I'm not going to make the Kubelik kits into 12 chip versions because piggy-backing chips is a little bit too tricky for a mainstream DIY effort. Also getting rid of the extra heat is an added complication.
By all means send me PM if you're up for building a Kubelik kit 🙂
By all means send me PM if you're up for building a Kubelik kit 🙂
Hi abraxalito,
I am in awe of what you do.
(although i don't understand the words 🙂 )
Thank you for doing it and making it available.
I am pondering your kits.
I have read quite a lot, but my knowledge is on the level of my skill which are both WAY below my level of enthusiasm 🙂
I missed out on the original set of 10 that you put out, that looked fun too.
I have never done any smd (well, I haven’t done much through hole either actually)and it looks like this would give me an interesting reason to work towards it.
Can I check I am thinking (from what i have read) along the right lines please?
Am I right in thinking that the Kubelik is a step above the Lingdac and it seems to me like that is just amazing value.
But they are not as good SQ wise when compared to your DecaDac (which you are now looking to upgrade.)
I am a bit lost as to whether the limitation to 16 bit will affect me.
I am wanting to rip CDs to the best quality I can in the future.
I have a khadas tone board and Ian Canada’s Dual mono (not tested the latter properly yet)
I am thinking of at least a couple of Kubelik for a couple of different things.
It would be great to sell off my KTB 🙂
So, with the phidecadac, the 3rd order and 7th order do two different tasks, right?
The 7th is more ‘invisible’ and the 3rd is ‘flavoured’ for vocals right?
Oh, boy, how to choose 🙂
So, what we are really saying is we all need a few kubeliks for general use and a couple of phideca (a 3rd and a 7th) to compare against each other 🙂 on your main system right?
I am thinking a kubelik for my librelec pi and pc to my monitor.
Or do you think the 3rd order perhaps?
Then the 7th order for my main…...... or maybe test them both vs .....
The new upgraded Decadac is likely to be more expensive but improved sound floor, right?
which means quieter at lower volumes, right?
I am in awe of what you do.
(although i don't understand the words 🙂 )
Thank you for doing it and making it available.
I am pondering your kits.
I have read quite a lot, but my knowledge is on the level of my skill which are both WAY below my level of enthusiasm 🙂
I missed out on the original set of 10 that you put out, that looked fun too.
I have never done any smd (well, I haven’t done much through hole either actually)and it looks like this would give me an interesting reason to work towards it.
Can I check I am thinking (from what i have read) along the right lines please?
Am I right in thinking that the Kubelik is a step above the Lingdac and it seems to me like that is just amazing value.
But they are not as good SQ wise when compared to your DecaDac (which you are now looking to upgrade.)
I am a bit lost as to whether the limitation to 16 bit will affect me.
I am wanting to rip CDs to the best quality I can in the future.
I have a khadas tone board and Ian Canada’s Dual mono (not tested the latter properly yet)
I am thinking of at least a couple of Kubelik for a couple of different things.
It would be great to sell off my KTB 🙂
So, with the phidecadac, the 3rd order and 7th order do two different tasks, right?
The 7th is more ‘invisible’ and the 3rd is ‘flavoured’ for vocals right?
Oh, boy, how to choose 🙂
So, what we are really saying is we all need a few kubeliks for general use and a couple of phideca (a 3rd and a 7th) to compare against each other 🙂 on your main system right?
I am thinking a kubelik for my librelec pi and pc to my monitor.
Or do you think the 3rd order perhaps?
Then the 7th order for my main…...... or maybe test them both vs .....
The new upgraded Decadac is likely to be more expensive but improved sound floor, right?
which means quieter at lower volumes, right?
Last edited:
Hi Grarea
Given that your enthusiasm level is high I'd say why not have a stab at building Kubelik? You could order two - practice on the first one knowing its OK to make mistakes and learn by doing. Knowing you have a spare takes the pressure off getting things right first time. Then if your mistakes on the first one aren't fatal ones (like lifting too many tracks through multiple re-soldering) you'll end up with two working DACs. Thing is - take it nice and slow and double (or triple) check each component and its orientation before applying solder.
You are right in thinking Kubelik is a step above lingDAC in all practically all respects. There is an aspect that Kubelik beats Deca DAC in - that's mid to high frequency transparency (perhaps some would say 'resolution'). There's better focus in the highs. The Deca though wins out on sheer dynamics in the bass, it has better boogie factor. In comparison Kubelik is more polite, which to my ears is more DSD-like. In my next creation I hope to combine these positive attributes in a single DAC, still working away on that one.
As regards the 16bit limitation - its no limitation at all when your source material is 16bit. Which all CD-sourced recordings are. Rip your CDs to FLAC, that's lossless compression and you'll lose nothing.
On Deca DAC the two filters tend to suit different kinds of music. The 3rd order being less dynamic suits compressed music and the 7th order I find more suited to higher dynamic range material - classical orchestral for example. The 'flavouring' wasn't the intent in designing the filters, I wanted a simpler, lower cost option so I made the 3rd order filter board, not knowing what it would sound like.
As for general use DACs, Kubelik tends more towards the less intrusive kind of sound, what I might call 'elevator music'. But its only a slight tendency its still exciting to listen to, just not quite as demanding of attention as Deca DAC. So for general use, Kubelik may be the more suitable choice.
The upgraded Deca DAC I couldn't answer on whether it'll be quieter at lower volumes. I only listen with digital volume and find both Deca and Kubelik plenty quiet enough at lower volumes. But the in-progress DAC does have the overall aim of lower noise so it might well be possible to notice an improvement.
Given that your enthusiasm level is high I'd say why not have a stab at building Kubelik? You could order two - practice on the first one knowing its OK to make mistakes and learn by doing. Knowing you have a spare takes the pressure off getting things right first time. Then if your mistakes on the first one aren't fatal ones (like lifting too many tracks through multiple re-soldering) you'll end up with two working DACs. Thing is - take it nice and slow and double (or triple) check each component and its orientation before applying solder.
You are right in thinking Kubelik is a step above lingDAC in all practically all respects. There is an aspect that Kubelik beats Deca DAC in - that's mid to high frequency transparency (perhaps some would say 'resolution'). There's better focus in the highs. The Deca though wins out on sheer dynamics in the bass, it has better boogie factor. In comparison Kubelik is more polite, which to my ears is more DSD-like. In my next creation I hope to combine these positive attributes in a single DAC, still working away on that one.
As regards the 16bit limitation - its no limitation at all when your source material is 16bit. Which all CD-sourced recordings are. Rip your CDs to FLAC, that's lossless compression and you'll lose nothing.
On Deca DAC the two filters tend to suit different kinds of music. The 3rd order being less dynamic suits compressed music and the 7th order I find more suited to higher dynamic range material - classical orchestral for example. The 'flavouring' wasn't the intent in designing the filters, I wanted a simpler, lower cost option so I made the 3rd order filter board, not knowing what it would sound like.
As for general use DACs, Kubelik tends more towards the less intrusive kind of sound, what I might call 'elevator music'. But its only a slight tendency its still exciting to listen to, just not quite as demanding of attention as Deca DAC. So for general use, Kubelik may be the more suitable choice.
The upgraded Deca DAC I couldn't answer on whether it'll be quieter at lower volumes. I only listen with digital volume and find both Deca and Kubelik plenty quiet enough at lower volumes. But the in-progress DAC does have the overall aim of lower noise so it might well be possible to notice an improvement.
Right, PM sent, dercision made.
Going for three kubeliks.
At that price, it gives me a chance to play with them.
I def want one for my libreelec pi, then one in case i mess up while playing. Then I will still have a couple to play with the balanced thing if I fancy it.
Plus a couple of friends are very curious about all of this shenanigans, I could do with one as a lend out.
Then, once proven to myself I can solder these things, I can launch at the phideca dac.
(as tempted as I am right now, I shall do the sensible thing.....)
Going for three kubeliks.
At that price, it gives me a chance to play with them.
I def want one for my libreelec pi, then one in case i mess up while playing. Then I will still have a couple to play with the balanced thing if I fancy it.
Plus a couple of friends are very curious about all of this shenanigans, I could do with one as a lend out.
Then, once proven to myself I can solder these things, I can launch at the phideca dac.
(as tempted as I am right now, I shall do the sensible thing.....)
Basically I have cleared Step 2 ( Blue stage) and hope to complete with Step 3 soon. I know the SOT resistors may not be necessary, but would it really make some minute level of difference if I were to try trim offset voltages to closer to mid-rail?
I see that the SOT resistors are in parallel with R2 and R6. Can you advice on the adjustment approach?
Last edited:
Having done the adjustments on a couple of DACs, I haven't noticed any difference in how they sound.
The assumption I made (based on the build of several prototypes) was that the offset would be a negative one initially. Putting in R69 and R70 can only move the offset higher - if its initally too high then R2 and R66 need to be raised in value first (try 130ohm).
The procedure is as follows - measure TP4 (-ve probe) with respect to TP1 (+ve probe). The voltage is then divided by 2000 (the I/V resistor) to calculate the offset current. Measure the voltage across R2 and divide that voltage by the offset current. This gives the resistance needed for R69 - choose the next highest value. Repeat for TP5, R66 and R70.
The assumption I made (based on the build of several prototypes) was that the offset would be a negative one initially. Putting in R69 and R70 can only move the offset higher - if its initally too high then R2 and R66 need to be raised in value first (try 130ohm).
The procedure is as follows - measure TP4 (-ve probe) with respect to TP1 (+ve probe). The voltage is then divided by 2000 (the I/V resistor) to calculate the offset current. Measure the voltage across R2 and divide that voltage by the offset current. This gives the resistance needed for R69 - choose the next highest value. Repeat for TP5, R66 and R70.
Kubelik
First of all, thank you to Richard for sending me an already built-up Kubelik. I have been listening to it for the last couple weeks, and it is incredible what he can offer at this price point. If you are on the fence about it, I would recommend just to go for it.
For the sake of comparison, I had been listening to an SMSL DAC (Sanskrit 10th MKII). Kubelik blows it out of the water. While the SMSL sounds clean, I feel that Kubelik reveals much more in the way of quality of the music. For example, I never realized there could be detail in the decay of a crash symbol (each hit sounding different).
One thing I really appreciate is the sound stage, it is realistic and has a well defined center (something the SMSL was severly lacking in). Also it handles music with a lot going on (a lot of instruments or a lot of distortion) very well, it never feels like it gets bogged down or you need to turn it up louder to get more detail.
I also have a DecaDAC (both 3rd and 7th order filters). DecaDAC with 7th order filter has an even bigger soundstage and improved bass (probably my favorite so far). But having listened to all 3 for a couple weeks each, I have never felt anything lacking after the intial getting used to stage (unlike SMSL).
I hope this doesn't sound too sales pitchy, but I am extremely impressed with Richard's work and what he can offer at a very reasonable cost. I wish I had tried one of his DACs before I had purchased the SMSL (a little intimitaded by the SMD soldering and didn't know where to get a power supply). Soldering the DAC chips went much smoother than I anticipated, and I'm using one half of a VRDN for PS.
Side notes: I was initially concerned that these DACs are designed with NOS droop compensation. My only music source is Amazon music, which I run in exclusive mode because I don't trust Windows to keep it's nose out of the music. So the DAC is likely fed 192Hz 32bit most of the time. I have not found this to be any issue, it handles it perfectly (I haven't run into the tape hiss problem). If you are putting the DAC in a chassis, look for "panel mount USB" (I initally used Neutrik style, but recently found more convenient and cheaper style on AliExpress).
Thanks again Richard!
First of all, thank you to Richard for sending me an already built-up Kubelik. I have been listening to it for the last couple weeks, and it is incredible what he can offer at this price point. If you are on the fence about it, I would recommend just to go for it.
For the sake of comparison, I had been listening to an SMSL DAC (Sanskrit 10th MKII). Kubelik blows it out of the water. While the SMSL sounds clean, I feel that Kubelik reveals much more in the way of quality of the music. For example, I never realized there could be detail in the decay of a crash symbol (each hit sounding different).
One thing I really appreciate is the sound stage, it is realistic and has a well defined center (something the SMSL was severly lacking in). Also it handles music with a lot going on (a lot of instruments or a lot of distortion) very well, it never feels like it gets bogged down or you need to turn it up louder to get more detail.
I also have a DecaDAC (both 3rd and 7th order filters). DecaDAC with 7th order filter has an even bigger soundstage and improved bass (probably my favorite so far). But having listened to all 3 for a couple weeks each, I have never felt anything lacking after the intial getting used to stage (unlike SMSL).
I hope this doesn't sound too sales pitchy, but I am extremely impressed with Richard's work and what he can offer at a very reasonable cost. I wish I had tried one of his DACs before I had purchased the SMSL (a little intimitaded by the SMD soldering and didn't know where to get a power supply). Soldering the DAC chips went much smoother than I anticipated, and I'm using one half of a VRDN for PS.
Side notes: I was initially concerned that these DACs are designed with NOS droop compensation. My only music source is Amazon music, which I run in exclusive mode because I don't trust Windows to keep it's nose out of the music. So the DAC is likely fed 192Hz 32bit most of the time. I have not found this to be any issue, it handles it perfectly (I haven't run into the tape hiss problem). If you are putting the DAC in a chassis, look for "panel mount USB" (I initally used Neutrik style, but recently found more convenient and cheaper style on AliExpress).
Thanks again Richard!
Progress towards the next generation DAC...
Kubelik's array of six DACs was limited as much by heat as board space. Linear regulation down from a 20V supply to the DAC's supply (3-5V) is very wasteful. Increasing the number of chips beyond the 10 of Deca DAC really needs an alternative solution for supply regulation.
I did come up with a heat-saving solution of running the DACs in series with the analog stages rather than in parallel as now. It worked fine in prototype but it means the currents for DACs and analog circuits have to be about the same. Going to 20 DACs means the DAC supply requirement goes well beyond the analog one - in this case a switcher looks to be the best solution.
After a bit of basic math I figured the switcher should be a buck one generating the DAC supply rather than a boost for the analog as the analog stages are more power-hungry. I have a few modules from Taobao based on MP1584 and have my first prototype using one of these playing music now. MP1584 has the advantage that its fixed frequency and you get to select that with a resistor, I've gone for the minimum of 500kHz. The module's inductor is 4.7uH which is probably great for currents above 1A but here I need in the region of 100mA so I have increased its value to 47uH with the aim of decreasing EMI.
Kubelik's array of six DACs was limited as much by heat as board space. Linear regulation down from a 20V supply to the DAC's supply (3-5V) is very wasteful. Increasing the number of chips beyond the 10 of Deca DAC really needs an alternative solution for supply regulation.
I did come up with a heat-saving solution of running the DACs in series with the analog stages rather than in parallel as now. It worked fine in prototype but it means the currents for DACs and analog circuits have to be about the same. Going to 20 DACs means the DAC supply requirement goes well beyond the analog one - in this case a switcher looks to be the best solution.
After a bit of basic math I figured the switcher should be a buck one generating the DAC supply rather than a boost for the analog as the analog stages are more power-hungry. I have a few modules from Taobao based on MP1584 and have my first prototype using one of these playing music now. MP1584 has the advantage that its fixed frequency and you get to select that with a resistor, I've gone for the minimum of 500kHz. The module's inductor is 4.7uH which is probably great for currents above 1A but here I need in the region of 100mA so I have increased its value to 47uH with the aim of decreasing EMI.
Attachments
Phi Deca DAC upgrade to MkII status
Using what I've learned since the design of the original Deca DAC, I've decided to implement a small upgrade. From now on any Deca DACs will be MkII variants and "MkII" will be visible on the PCB silkscreen. Those who have original Deca DACs can upgrade their own if they feel up to the job of replacing a few SMD components. Subjectively the upgrade moves the Deca DAC much closer to the Kubelik's level of mid-HF transparency, without loss of the dynamics or bass drive intrinsic to Deca DAC.
The main part of the upgrade is lowering the impedance of the components in the droop-correcting LP filter. Lower impedance results in lower noise and even though the noise from the original stage is below the ~-92dB RBCD noisefloor, the improvement in transparency is noticeable. The I/V stage needs its output bias current increasing so as to drive the lower impedance input and I've also taken the opportunity to raise the DAC supply voltage a little to squeeze the most current out of the chips. I'll post up new schematic shortly - the PCB layout had to change a little bit to accommodate a new component too.
For those who'd like to modify their existing Deca DAC's droop filter, here are the changes to the components around U17. All footprints are 0805.
R39,R57 -> 750R
R40,R58 -> 820R
R41,R59 -> 330R
R42,R49 -> 1k6
C15,C23 -> 4.7nF NP0
C16,C24 -> 22nF NP0
C17,C25 -> 10nF NP0
C18,C26 -> 2nF NP0
Current source:
R37,R47 -> 56R
If you want to lower the noise still further you can substitute the ICL28210 (U17) for an OPA2209 (the opamp I'm using in Kubelik).
Using what I've learned since the design of the original Deca DAC, I've decided to implement a small upgrade. From now on any Deca DACs will be MkII variants and "MkII" will be visible on the PCB silkscreen. Those who have original Deca DACs can upgrade their own if they feel up to the job of replacing a few SMD components. Subjectively the upgrade moves the Deca DAC much closer to the Kubelik's level of mid-HF transparency, without loss of the dynamics or bass drive intrinsic to Deca DAC.
The main part of the upgrade is lowering the impedance of the components in the droop-correcting LP filter. Lower impedance results in lower noise and even though the noise from the original stage is below the ~-92dB RBCD noisefloor, the improvement in transparency is noticeable. The I/V stage needs its output bias current increasing so as to drive the lower impedance input and I've also taken the opportunity to raise the DAC supply voltage a little to squeeze the most current out of the chips. I'll post up new schematic shortly - the PCB layout had to change a little bit to accommodate a new component too.
For those who'd like to modify their existing Deca DAC's droop filter, here are the changes to the components around U17. All footprints are 0805.
R39,R57 -> 750R
R40,R58 -> 820R
R41,R59 -> 330R
R42,R49 -> 1k6
C15,C23 -> 4.7nF NP0
C16,C24 -> 22nF NP0
C17,C25 -> 10nF NP0
C18,C26 -> 2nF NP0
Current source:
R37,R47 -> 56R
If you want to lower the noise still further you can substitute the ICL28210 (U17) for an OPA2209 (the opamp I'm using in Kubelik).
I am absolutely in admiration for your relentless developments and kind sharing with the community. Bravo!
It shall be very interesting to see "what" wins at the end of day: improved Deca Dac, or "advanced enhenced Kubelik", as both basis seem to benefit from each other's development, which is great.
Whatever the winner will be in a few month's time, I will go for it as I will need a new DAC for my secondary system. It shall also be interesting to hear how that refined product, whatever it will be then, will perform against the D70S... and even my own reference DAC.
Many thanks for all this and again well done
Claude
It shall be very interesting to see "what" wins at the end of day: improved Deca Dac, or "advanced enhenced Kubelik", as both basis seem to benefit from each other's development, which is great.
Whatever the winner will be in a few month's time, I will go for it as I will need a new DAC for my secondary system. It shall also be interesting to hear how that refined product, whatever it will be then, will perform against the D70S... and even my own reference DAC.
Many thanks for all this and again well done
Claude
Thanks @Claude - well the winner at the moment is the one with 20 paralleled DACs, its noise floor is the lowest of all of the DACs so far. PCB layout for that is about half complete, I'll post up an image when its completed. I'm planning to call this one 'Celibidache' who is one of my all-time favourite conductors. But I've also had the suggestion to call it the 'Double Decker DAC' 😀