OP + buffer

Usually you will connect the feedback resistor to the buffers output, but depending on eventual difference in rise-time, bandwidth it is wise to connect a small capacitor between the OP-amps output and the feedback input of the OP-amp. This will prevent high frequency ringing and possibly oscillation. If the feedback resistor is around 10000 ohm a 47pF cap would work fine setting -3dB at 338 KHz.
To further prevent stability problems a 47 ohm series resistor can be placed at the buffers output that isn't included in the feedback path.
This assuming that the load you will drive is 600 ohm or higher and less than 1000pF.

I agree, this is the classical solition, also suggested in the corresponding appnote from BB.
But that doesn't hit the point: For the BUF634 is a closed loop circuit, it doesn't seem to be necessary to include it into an overall loop. You are of course right, stability problems may occour if the driving OP is much faster then the BUF.
So, my question is: Wouldn't it be a better solution to close FB loop just behind the OP, doing some RC-LP-filtering behind it, then entering the BUF?
I don't like those FB loops too much that contain more stages than necessary.

No, not RIAA. The circuit should serve as an output stage for a CD-Player (Teac VRDS-25x).
Why RC filtering? Well, I'm using an extremely fast OP (LM7171) and I don't want to "override" the buffer in terms of frequency - you know, TIM and all that sort of things.
I am a great fan of Spectral-like designs - means: using very high bandwidth circuits an doing a dedicated amount of passive low pass filtering between the stages, not that "braking circuits by feedback"-thing.

I haven't verified for myself, but what I've heard that sounds reasonable to me is:

1. put the buffer inside the global feedback loop for the lowest THD performance

2. don't include the buffer in the glodal feedback loop to maintain highest stability with wierd loads (for example, long, high capacitance interconnect to a power amp).

then I assume it only will drive a preamp input of 10000 ohms or above and no capacitive load.
If this is the situation I would skip the buffer and only use the OP-amp which I assume will give you the best sound.

I don't know what kind of output the DAC in your CD-player
have. Current or voltage ?
If voltage, are the low pass filter included in the DAC ?

Well... I haven't been too glad with "just an OP" in the output so far, and the BUF634 (within the feedback loop)definitely is a step ahead.
The Teac's DAC configuration is four times current-out AD1862 (fully balanced), followed by 4* OPA134 for I/V-conversion.
That's where I leave the original board and feed the signals into four 12 dB-Bessel filters, followed by the stage to be discussed here.
BTW.: That's only the balanced signal path, for the unbalanced output, I did this all again and added an INA103 instrumentation amplifier for the generation of a non-compromise-signal. If you have 4 DACs, use them...
But after the INA the next buffer-problem occurs..

I have tried the ZAP filter, and it did not perform as well the DIY solution - less stability, less dynamics, but by far better than the player'S original output stage. At that time my own setup worked with OPA627 and BUF634.
There is a new ZAP out now, perhaps I will try that (I know the german distributor quite well), but for now pushing the limits of my own solution is more interesting.
BTW.: LClock XO ??? Great Stuff!! I use two of them in that player!

Regarding the I/V converter you could try a simple resistor.
I'm sure it will be better than your OPA134.
With careful selection of components in your 12 dB Bessel filter it can be driven directly by a resistor loaded DAC.
You will have less output but that can be fixed in the output stage. I use BurrBrown PCM63 which can accept 0,4 volts before the protecting diodes starts to conduct.
I don't know how AD1862 is built, so I can't say how much voltage you can get from them with a resistor I/V-convertor.

Have you looked at the ADSL OP-amps from Analog Devices ?
Less open loop gain and a open loop -3dB point above 100KHz.
Might be a competitor to LM7171 in your circuit.

Thought more about your feedback ideas, but by intuition the global feedback solution gives me less shrills.
But since it is so easy try both, please do.

I have Tried every Combonation of Opamp Buffer one can try and what works best from my point of view is the Multiloop feedback technique as discribed by walt jung in the December 1998 EDN Walts Tools and Tips. http://www.penton.com/ This satisfies the Stability as well as suppling a constant amount of feedback over the Aduio bandwidth. Regarding the BUF-634 This is an Open Loop Device not a Closed Loop one and IMHO sounds Less than Ideal when operated outside the Loop in Addition to having about 35 Mv of DC Offset that the Feedback loop would normaly Correct.