'Best' power regulator for DAC

[studiostevus]

All,
I am in the process on building a TDA1541a dac, and on this quest I am wondering about your experiences with different regulators to power a dac.

Currently I am using 4 Salas shunt regs for all the supplies which are:
-Analog TDA1541 (-15V)
-Analog TDA1541 (-5V)
-Digital TDA1541 (+5V)
-Digital reclock circuit (+5V)

As far as I could find, the Salas shunts have good performance, both on noise, and impedance up to 200KHz. To me, they also sound nice.

-For the analog lines, this should be fine. There may be better-sounding alternatives though (are there?)

-For the digital lines, I am concerned that the rapidly increasing impedance beyond 200KHz will be problematic, as the clock signals are 2.8MHz and 11MHz.

What are your experiences?

 

[pidesd]

i ve tested a few solutions and so far i found salas 1.1 preceded by heavy filtering with inductances and different sort of cap for each stage ''the best''. heavy bias helps with impedance. of course battry operation is better than mains if you can cope with the cost and complexity. one thing i d like to try is cap multipliers before salas regs. heavy bias helps with impedance.

also, recently i experimented with a copper wire wound in an non inductant pattern which acted as Rset (in salas reg) and found significant improvement. makes me wonder what Rref from non inductive wire would do....

 

[pidesd]

also wonder how this reg would perform in the first post...

http://www.diyaudio.com/forums/digi...ding-ultimate-nos-dac-using-tda1541a-369.html

 

[Ken Newton]

Active current-source fed shunt regulators ( such as the Salas) are my favorite, but require individual setting of each current-source.

In addition to the shunt versus series regulator question you raise there is the feedback versus non-feedback topology question. Regulators are D.C. amplifiers and as such have many of the same concerns of any linear amplifier regarding the affects of feedback upon sound quality. Feedback regulators feature far lower output impedance than non-feedback regulators, and also feature a more exact and stable D.C. voltage setting. However, that lower output impedance usually quickly rises with frequency (due to insufficient gain-bandwidth product of most IC regulators, and many discrete ones too), eventually becoming pretty much ineffective for digital clocks and other high frequency signals, and that more exact and stable D.C. voltage setting usually doesn't matter for audio amplifier circuits. It's usually a regulators A.C. performance that really matters for linear audio applications.

The regulator's output stabilization capacitor, or the powered circuit's bypass capacitor, is what provides quiet power for circuits drawing high frequency current from the regulator, such as typically encountered from digital circuits. The output impedance of that capacitor is equal to their reactance = 1 / (2 * pi * f * C). So, for example, a small 100nF supply bypass/decoupling capacitor on a 10MHz oscillator looks like a regulated voltage source having a 0.160 ohm output impedance (ignoring device package parasitic inductance) to the 10MHz dynamic load of the oscillator.

 

[UV101]

SPowers...........no competition!

I'll be selling my built Salas shunts on here because I've gone back to internal Dac running SPowers again ;-)

Btw, the system I used to work this all out was my ears!!!

 

[KlipschKid]

http://www.diyaudio.com/forums/part...rs-different-manufactures-they-all-equal.html

http://www.diyaudio.com/forums/parts/192155-replacement-regulators-7805-7815-7915-a.html

http://www.amb.org/audio/sigma22/

 

[studiostevus]

i ve tested a few solutions and so far i found salas 1.1 preceded by heavy filtering with inductances and different sort of cap for each stage ''the best''. heavy bias helps with impedance. of course battry operation is better than mains if you can cope with the cost and complexity. one thing i d like to try is cap multipliers before salas regs. heavy bias helps with impedance.

also, recently i experimented with a copper wire wound in an non inductant pattern which acted as Rset (in salas reg) and found significant improvement. makes me wonder what Rref from non inductive wire would do....

Check out the output impedance of the Salas 1.1.... This doesnt look good for digital...

 

[qusp]

it will be very difficult to find a regulator to work into the very high frequencies of modern digital, but it really doesn't need to be, at these frequencies ANY regulator will be relying on the quality of the localised decoupling caps at the load to do the heavy lifting

 

[qusp]

btw i had the same concerns exactly, check the builds and fairytales thread for conversation on this topic, even incredibly high speed regulator controllers that use high speed gets designed to supply modern PC CPU clock loads will rely on quality ceramic decoupling directly at the load pins. i prefer lifepo4 for clock loads anyway

 

[SoNic_real_one]

Power supply 101 - decoupling close to the pins of digital circuits!
Only that will provide the necessay DC "storage" with low impedance in HF. No reg will be able to do that, no matter how it is build.
Same goes for audio signals - just the capacitors need to be higher.

 

[studiostevus]

Power supply 101 - decoupling close to the pins of digital circuits!
Only that will provide the necessay DC "storage" with low impedance in HF. No reg will be able to do that, no matter how it is build.
Same goes for audio signals - just the capacitors need to be higher.

Sure... I am now running Salas shunts, with decoupling done by 33uF BG STD's millimeters from the DAC chip, bypassed with 1uF acrylic caps.

 

[SoNic_real_one]

In audio freq range, because of the lower value of freq (10-20Hz), the by-pass caps would have to be too large. So the regulator output impedance still counts (being in parallel with those caps).
In HF (over 100kHz), the power lines itself start to act like impedances (inductive), "separating" the regulator from the load. So, the only think to do for HD loads, is to improve the quality of the bypass capacitors. Installed closer to the pins to minimize the series impedance of the power lines.

 

[qusp]

imo neither your 33uf blackgate, nor acrylic (especially if they PTH) caps are suitable. bulk solid polymer (should be already in place in most quality dacs) and np0 ceramic or perhaps surface mount pps film is superior. i don't know why people think blackgate are the bees knees for everything, often even removing ceramic caps to replace them with.