Buffalo Tweaking

[glt]

Those opamps have PSRR > 100 dB I think. The opamps are configured as voltage followers so whatever noise in the input will show up in the output (I think that's why there is a pi filter in front of it). Thus a clean supply at the inputs of the opamps will give you "best-bang-for buck" And this is what has been proposed to use with the "Vref" mod. There is also a circuit based on the Vref, but it is a regulator (not just a reference). This "Vref-reg" will be useful to power the clock.

[glt]

Originally posted by jkeny
Diagram of PS mods points:

I would suggest one other hugely important mod - feed the clock with a very low noise separate 3.3V supply - I don't know if anybody has done this & can report?

I don't think you can provide power with a reference. Better provide the reference to the opamp and the opamps provide the power.

[jkeny]

Glt,
As usual, be careful with these PSRR figures. The neg supply is worse than the pos & at 20kHz this is down to about -65dB - not so great after all! This can be easily bettered by discretes!

Leo, noticed significant differences with different supplies to VA!

Why the necessity for the voltage buffer? Is it just a cost cutting exercise that was on the original Sabre eval board?

I don't think you can provide power with a reference. Better provide the reference to the opamp and the opamps provide the power.

I don't know what you mean by this 3.3V is the reference, what power are you talking about?

edit: glt, can you post the Vref regulator circuit here?

[jkeny]

An interesting idea was just posted on DiyHiFi by wildmonkeysects in which he surmises that MAYBE the sweet spot is to use V out & ensure your AVCC PS is

a really, really clean, preferrably open loop Avcc. No overshoots, no close phase margins if using feedback, no cheezy ceramic caps, no deriving the reference from Vdd

And this

internally, there is probably a sweet spot for the external load where the Sabre's thermal modulation is minimal. Not convinced it would be all I or all V. Dunno yet. Dunno the internal layout.

In other words forget about I to V conversion & try V out with a clean AVCC - what's there to lose trying this? this could be why Rolls likes the transformer IV solution so much?

Just some things to consider!

[glt]

Over at the other thread, there were two designs: A "Vref" that provided a very clean 3.3 voltage reference and a Vreg based on the Vref that could be used as a regulator. The Vref was a replacement of Va at the input of the opamp. Most people experimented adding Vreg to the input of the opamp, in effect cleaning the voltage noise. You could use the Vreg design to power the analog section directly bypassing the opmap, but I don't think anyone reported doing that.

[jkeny]

But a vref is a voltage regulator ie. if the input supply changes, the output stays fixed at 3.3V.

[Spartacus]

Hi John,

Dan here, from the SKA forum. This is what I'm using for the Analog supply to my Buffalo:



Q2 is an MPSA14 darlington. Way, way better than LM317s. This is currently feeding Greg's Mini-reg to provide a nice and clean 3.3V Vref.

The clock: a separate supply with an AD797 based reg feeding a 66MHz oscillator from Guido Tent.

I'm using a Twisted Pair DIR to convert S/PDIF to I2S. Didn't make a jot of difference to the sound, but enabled 192K to work.

Analog stage is an SKpre symmetrical discrete opamp for balanced to SE conversion. The output offset is trimmable, so no coupling caps needed. Superb.

Next up ..... bypassing the LM4562 buffer and feeding the analog supply directly from the Mini-reg. Not at all convinced this will help, but worth trying.

Dan

[jkeny]

Ah Dan,
I never knew your username here - I'm easy to trace between forums - I use the same username!

Nice! Does this pre-regulation into the Mini-reg improve it by much?

Is that clock supply a PFMflea?

The SKPre is just used for BAL to SE, then IV stage or what?

I'm not sure why the LM4562 buffer is there? Is it to isolate the left from right AVCC?

[Spartacus]

Originally posted by jkeny

Nice! Does this pre-regulation into the Mini-reg improve it by much?

Well originally it was pre-regulated by an LM317 as per the standard Twisted Pair setup. I changed to a design similar to the above, but using a Zener diode, which was an improvement. Changing to the LT1031 - plus a couple of refinements - was a further improvement.

Originally posted by jkeny
Is that clock supply a PFMflea?

Similar, I use a JFET current source instead of the 7812, plus a couple of other minor changes.

Originally posted by jkeny
The SKPre is just used for BAL to SE, then IV stage or what?

Just balanced to SE - no I/V needed.

Originally posted by jkeny
I'm not sure why the LM4562 buffer is there? Is it to isolate the left from right AVCC?

The LM4562, together with a reference, forms a regulator. But if your reference is able to supply sufficient current the opamp isn't needed. You need to bypass the 1K + 1uF low-pass filter though.

[jkeny]

Thanks Dan,
So Buffalo operating in voltage mode output into SKpre - no I to V conversion?

Ah I understand now the role of LM4562 and understand what glt was saying! Looks like 25mA current is plenty for each channel's AVCC.