Kubelik DAC PCB to manufacture
After a few years of publishing DAC designs I have decided to make a break with the past in terms of nomenclature. UK high-end DAC manufacturer dCS has named their DACs after composers for some time, I decided a DAC is more like a conductor than a composer in that it doesn't originate what it produces but interprets a 'score' (the digital file). So from now on my DAC designs are going to be named after conductors. The first one in the new naming scheme is named after Rafael Kubelik - here is the PCB layout.
The design is fairly close to the schematic I showed a week ago except I have abandoned the MFB filter due to noise concerns. S-K filters don't have frequency independent input impedance (they look more like a capacitor than a resistor to the driving stage) so the I/V stage has been reinstated. There are two stages to the analog chain : I/V, then droop correction/LPF, necessitating 4 opamps in total. The classA output buffers from Phi Hex and Deca are implemented on both stages.
After a few years of publishing DAC designs I have decided to make a break with the past in terms of nomenclature. UK high-end DAC manufacturer dCS has named their DACs after composers for some time, I decided a DAC is more like a conductor than a composer in that it doesn't originate what it produces but interprets a 'score' (the digital file). So from now on my DAC designs are going to be named after conductors. The first one in the new naming scheme is named after Rafael Kubelik - here is the PCB layout.
The design is fairly close to the schematic I showed a week ago except I have abandoned the MFB filter due to noise concerns. S-K filters don't have frequency independent input impedance (they look more like a capacitor than a resistor to the driving stage) so the I/V stage has been reinstated. There are two stages to the analog chain : I/V, then droop correction/LPF, necessitating 4 opamps in total. The classA output buffers from Phi Hex and Deca are implemented on both stages.
Attachments
Sure, we'll make the gerbers available so you can order your own and take advantage of the PCB houses' cheaper shipping arrangements.
As for the parts I think the only tricky to source parts will be the inductors. But perhaps I can adapt the PCB to use Fastrons like these : https://www.mouser.com/ProductDetail/Fastron/07MFG-273J-50?qs=3%2FphUcBLEfZu3pw%2F8xurOg%3D%3D If they do indeed come back into stock.
As for the parts I think the only tricky to source parts will be the inductors. But perhaps I can adapt the PCB to use Fastrons like these : https://www.mouser.com/ProductDetail/Fastron/07MFG-273J-50?qs=3%2FphUcBLEfZu3pw%2F8xurOg%3D%3D If they do indeed come back into stock.
Last edited:
Even better, fantastic!
Parts availability has become a big problem, many out of stock parts with long lead times 🙁
Parts availability has become a big problem, many out of stock parts with long lead times 🙁
Last edited:
Or you can offer us a super funny kit!
I like your kits!No tricks, just fun.
Thanks for your work. I read about an high output dac, it could be perfect for a Mofo amp.
Ciao.
I like your kits!No tricks, just fun.
Thanks for your work. I read about an high output dac, it could be perfect for a Mofo amp.
Ciao.
I was indeed planning to make kits (or rather, ask my wife to make them!). But recently most people have been asking for fully built DACs so I have been wondering if kits are really that desirable.
You can modify the Kubelik to give higher output (just a couple of resistor changes) if you give up the headroom for inter-sample overs. But I guess even 4VRMS isn't going to be enough for a MoFo amp. I think the 36 DAC design will be best suited to that - there's (in theory) up to 36V peak to peak, balanced.
You can modify the Kubelik to give higher output (just a couple of resistor changes) if you give up the headroom for inter-sample overs. But I guess even 4VRMS isn't going to be enough for a MoFo amp. I think the 36 DAC design will be best suited to that - there's (in theory) up to 36V peak to peak, balanced.
Kits are always nice 🙂
It would also possibly be nice to have an overview of the various DACs you can offer today (PhiDAC or whatever it will be named 🙂, 36V variant, etc.) and their coming developments...
Just thinking out loud
Claude
It would also possibly be nice to have an overview of the various DACs you can offer today (PhiDAC or whatever it will be named 🙂, 36V variant, etc.) and their coming developments...
Just thinking out loud
Claude
OK we definitely won't rule out kits then.
DACs that I offer at present are only the Phi Deca DAC (10 chips) with a choice of filters (3rd or 7th order). In a few weeks (hopefully) we'll be offering Kubelik, no options to choose from there other than kit/built. I don't plan to stop offering Deca DAC when Kubelik arrives but I do have a plan to upgrade it to include my most recent discoveries but as yet can't quite see how to bring that about.
The 36 chip DAC won't feature on this thread, its on my 'Multibit adventures' thread and if there is interest there from potential customers I'll introduce it on its own commercial (Vendor's Bazaar) thread.
DACs that I offer at present are only the Phi Deca DAC (10 chips) with a choice of filters (3rd or 7th order). In a few weeks (hopefully) we'll be offering Kubelik, no options to choose from there other than kit/built. I don't plan to stop offering Deca DAC when Kubelik arrives but I do have a plan to upgrade it to include my most recent discoveries but as yet can't quite see how to bring that about.
The 36 chip DAC won't feature on this thread, its on my 'Multibit adventures' thread and if there is interest there from potential customers I'll introduce it on its own commercial (Vendor's Bazaar) thread.
Many thanks for the kits abraxalito!
And very clear answer...
How do you rate the sonic differences between Phi Deca DAC (current stand) and Kubelik?
Do you believe Phi Deca DAC has the better "over all / final" potential, or is it more complicated than that?
I am trying to understand how the latest developments translate to in terms of sonic signature...
Many thanks again for developing all this!
Claude
And very clear answer...
How do you rate the sonic differences between Phi Deca DAC (current stand) and Kubelik?
Do you believe Phi Deca DAC has the better "over all / final" potential, or is it more complicated than that?
I am trying to understand how the latest developments translate to in terms of sonic signature...
Many thanks again for developing all this!
Claude
Reliable sonic impressions are a bit tricky to tease out at this early stage. Mainly because I'm still playing with different options in the design. The most recent variant of Kubelik I've been listening to has a 5th order filter and ISL28210 opamps, its otherwise broadly the same as the version on the PCB layout.
Overall I have slightly preferred this version of Kubelik to the standard issue Deca DAC. Its more transparent in the higher frequencies at the expense of solidity in the bass. I am finding that sonic impressions correlate fairly closely with noise in opamps (provided the DAC chips themselves aren't the limitation noise-wise) and in this respect the LT1028 in Deca DAC is a bass monster with extremely low levels of noise in the lower frequencies. None of the JFET or CMOS input opamps have come close to how good the LT1028 sounds in the bass.
In the higher frequencies though the MFB filter in the Deca DAC isn't the best choice which is why I've swapped to S-K in Kubelik (together with lower R values) and this gives a really 'pristine' feeling to e.g. violins. For quite some time my wife hasn't been 100% happy with violin sound on Deca (and previously, hex) but wasn't completely sure it was the DAC as some recordings of violins sounded considerably better than others. With Kubelik though the violin sound got the thumbs up!
Does any of this help much?
Overall I have slightly preferred this version of Kubelik to the standard issue Deca DAC. Its more transparent in the higher frequencies at the expense of solidity in the bass. I am finding that sonic impressions correlate fairly closely with noise in opamps (provided the DAC chips themselves aren't the limitation noise-wise) and in this respect the LT1028 in Deca DAC is a bass monster with extremely low levels of noise in the lower frequencies. None of the JFET or CMOS input opamps have come close to how good the LT1028 sounds in the bass.
In the higher frequencies though the MFB filter in the Deca DAC isn't the best choice which is why I've swapped to S-K in Kubelik (together with lower R values) and this gives a really 'pristine' feeling to e.g. violins. For quite some time my wife hasn't been 100% happy with violin sound on Deca (and previously, hex) but wasn't completely sure it was the DAC as some recordings of violins sounded considerably better than others. With Kubelik though the violin sound got the thumbs up!
Does any of this help much?
Many thanks, that helps a lot indeed.
It might be difficult to find bass heavy AOPs that are neutral otherwise, per definition. I don't expect OPA1656 to give you what you want, but who knows. AD825 biased in Class A could on the other hand, at the cost of extreme transparency.
I shall watch all your future developments with a lot of interest
Thanks again and good luck
Claude
It might be difficult to find bass heavy AOPs that are neutral otherwise, per definition. I don't expect OPA1656 to give you what you want, but who knows. AD825 biased in Class A could on the other hand, at the cost of extreme transparency.
I shall watch all your future developments with a lot of interest
Thanks again and good luck
Claude
Hi Richard,
I replaced the output cap (wima) on my phidecca with a clarity cap bypassed with 0.47 if film and foil. This change resulted in a clearly audible improvement in the mids and highs.Seriously tight fit though, had to sand down the cap leads to reduce diameter to get them to fit the pad holes.
Paul
First PCB prototype of Kubelik
PCBs came rather quickly this time around, yesterday I built up one and today did a quick comparison with the original PhiDAC. This is using OPA1642 opamps as I'm curious to compare them with OPA1678 (which I'll use on the next one).
The most noticeable difference on connecting PhiDAC after listening to Kubelik for a while was that the depth collapsed rather, giving a flatter presentation. Kubelik has more sense of space, not quite as good as Deca DAC though as ISTM the most solid soundstage illusion needs very low noise in the lowest frequencies. Transparency overall is improved too which gives a 'cleaner' feel to some instrumental timbres and voices too. As one example, Michala Petri's recorder sounds more flute-like on PhiDAC, as if some noise is added. My overall impression is this DAC will appeal to those who put a lot of weight on mid-range delivery.
BTW - unlike the original PhiDAC, Kubelik has no on-board regulation, so far I'm only using a zener assisted LM317 20V regulator fed from a regular linear wall-wart.
PCBs came rather quickly this time around, yesterday I built up one and today did a quick comparison with the original PhiDAC. This is using OPA1642 opamps as I'm curious to compare them with OPA1678 (which I'll use on the next one).
The most noticeable difference on connecting PhiDAC after listening to Kubelik for a while was that the depth collapsed rather, giving a flatter presentation. Kubelik has more sense of space, not quite as good as Deca DAC though as ISTM the most solid soundstage illusion needs very low noise in the lowest frequencies. Transparency overall is improved too which gives a 'cleaner' feel to some instrumental timbres and voices too. As one example, Michala Petri's recorder sounds more flute-like on PhiDAC, as if some noise is added. My overall impression is this DAC will appeal to those who put a lot of weight on mid-range delivery.
BTW - unlike the original PhiDAC, Kubelik has no on-board regulation, so far I'm only using a zener assisted LM317 20V regulator fed from a regular linear wall-wart.
Attachments
Last edited:
Interesting potential, well done!
Who knows how better it might sound with improved PS and OPA1656...
Who knows how better it might sound with improved PS and OPA1656...
This seems like a Japanese approach with the constant refinement of circuit based around a key component that does not change.
Interesting, I've not come across that approach before - does it have a name or a proponent behind it?
Does it still count as an unchanged component if there are multiples of it?
Does it still count as an unchanged component if there are multiples of it?
