Hi Chanh,
just to make sure, I installed the 3,3 volt shunts directly at the chip. There were no couple of 47 uF in the way.... Are you still planning to do listening tests with and without the 3,3 Volt shunts as only changed parameter? That would be the real honest feedback. Changing several things at the same time will never make 100% clear what really made the difference.
I am really looking forward to more people doing this test with only one change at a time. Last week I had independent feedback from Marcel for Switzerland, who removed the 3,3 volts ten shunts and replaced them with LDOs again. He is using the 8 volts at the analog side of the dac and is also pleased with this. I think there is no discussion about this improvement anymore...
don't get me wrong, this is not a discussion who is wrong or who is right. I just reported back what I heard and what some else has done. I know you spent a lot of money on the dac and specially on the tent shunts. for that reason it would make most sense to this in two steps, wouldn't you agree? if you install all in one step, you for sure will have a hell of a dac, but knowing what made the true difference is something I am sure also you are interested to know, not only me ;-)
look forward to your reports !
Yes, finally a post with some measured sanity.
We have had endless long posts here of various DDDAC changes and mods with no real logical approach or explanation.
To me they just read like a DDDAC sales brochure but offer virtually zero real information WRT mods that might work in
another application.
As the DAC's designer you well know that there are many details that must be addressed, separately to fully and fairly evaluate any
design change such as this case, a shunt reg.
For evaluation of a shunt reg here's an example of things I would be testing in isolation:
a/ is the shunt reg super low impedance but low bandwidth?
b/ is the shunt reg a higher impedance but very wide bandwidth
c/ what bypass caps were used
d/ were there combinations of bypass caps, IOW say a film cap in parallel with an elecro. This can have a big effect
e/ where is the shunt reg located? Is it right on the DAC or further away?
f/ what sort of grounding is the shunt reg tied to? For a DAC like 1794 there is lots of HF content so is there a ground plane, how good is it?
g/ Does the bypassing cap location make a difference?
Most tinkerers don't have a really good understanding of these finer points and just proceed into having fun modding their DAC's - which is
great in the spirit of DIY But as such the results often don't often translate reliably to something else.
Another example is the multiple paralleling of 1794 DAC's. This technique is very well known and has been used for many years to, for
example get lower noise from transistors etc.
However for me again there has been no real scientific evaluation of the advantages - a lot of effort and cost goes into stacking boards up.
Looking at the distortion results of DDDAC, they are very poor and don't come close to getting what the chip is capable of. So in the
case of designing a current output DAC with a passive OP stage we have compromised the DAC's linearity by a huge margin. You can use
one DAC and get 20 x lower distortion without having to use opamps. Maybe the DAC sounds good -because- of this higher distortion,
it would not be the first time.
Maybe the multiple parallel DACs sound better because of higher OP current drive - but then we can get huge current drive with one
DAC and a simple OP circuit.
For me, the first thing I would do is evaluate what the change in sound is with 1794 OP voltage swing. The 1794 does not like to have
OP voltage swing, it causes a lot of distortion. So I would make a test jig up and vary the OP resistance and see how this affects sound.
I can almost guarantee that the more voltage swing on the DAC OP will translate to a richer and more musical sound, because of the added
distortion.
Once I had done this I would then evaluate the optimum driving impedance for the OP transformer. Transformers work best with a certain
drive impedance and also they usually need a snubber on the OP. Looking at the square wave performance of yours, it has overshoot so
needs to have a snubber installed. You should be able to get "perfect squares" from a transformer with no overshoot.
You might well find that a single well optimised 1794 driving a transformer that is also optimised will sound better than a stack
of 16 that are not optimized.
Food for thought.
cheers
Terry