Why might a diamond buffer sound better?
Posted 25th May 2015 at 07:00 AM by abraxalito
Here's an interesting post by Charles Hansen (of Ayre fame) about his preference for diamond buffers in amp output stages - https://audioasylum.com/forums/amp/me...18/185962.html
He's saying that diamond buffers sound better but that he has no idea why this would be.
Simulating the PSRR of the diamond vs a traditional EF2 reveals a significant difference - about 6dB better PSRR for the diamond. Could this explain the SQ improvement? Charles designs his amps without GNFB so you'd tend to think that his OPS PSRR is really rather critical.
Since making this discovery I've been on the look-out for opamps with diamond buffer OPSs - OP260 was one I found (courtesy of Esperado) but there are a few from Linear Technology which I've ordered up a few of. In particular, LT1886 and LT1723 look very interesting as potential amp/buffer stages in DACs. LT is fairly unique in that they publish a simplified internal schematic in all the DSs I've looked at (not the LT1115 though, but I understand its basically an LT1028 which also has a diamond OPS).
He's saying that diamond buffers sound better but that he has no idea why this would be.
Simulating the PSRR of the diamond vs a traditional EF2 reveals a significant difference - about 6dB better PSRR for the diamond. Could this explain the SQ improvement? Charles designs his amps without GNFB so you'd tend to think that his OPS PSRR is really rather critical.
Since making this discovery I've been on the look-out for opamps with diamond buffer OPSs - OP260 was one I found (courtesy of Esperado) but there are a few from Linear Technology which I've ordered up a few of. In particular, LT1886 and LT1723 look very interesting as potential amp/buffer stages in DACs. LT is fairly unique in that they publish a simplified internal schematic in all the DSs I've looked at (not the LT1115 though, but I understand its basically an LT1028 which also has a diamond OPS).
Total Comments 2
Comments
-
The economy of the diamond buffer is in the way the bias for the output pair is set by the Vbe of the input set. The driver and Vbe multiplier are one and the same functional unit.
The downside is the bias current in the driver and output pairs must be equal unless different transistors are used or small resistors on the driver emitters are added. Both options are sub-optimal. Much as I like working with the Diamond Buffer circuit, I'm not convinced the design has any intrinsic merit sonically speaking. It is a relatively simple circuit element with high PSRR and low distortion. That's the sum total of the attraction as far as I am concerned.
The rationale Charles Hansen gives for it sounding better, viz. having no Vbe multiplier to make the input signals "unequal" is bogus, due credit that he himself admits as much, and for the correct reason. In my opinion if you have to add a third pair of transistors in front of the buffer in order to drive it you've negated the advantage of using a diamond buffer in the first place. The diamond buffer is now nothing more than a compound transistor for current gain. You might as well capitulate and use a Sziklai arrangement.
P.S. It's not clear whether his power Diamond buffers are intended for Class A or Class AB operation. The Diamond circuit works fine in class A, no feedback required. Class B is a different story!!Posted 26th May 2015 at 12:22 AM by rjm
Updated 26th May 2015 at 12:27 AM by rjm -
I agree Charles' 'explanation' doesn't make sense but then he's maintaining the 'mystique' perhaps for marketing reasons. Likewise his 'argument' for why no feedback doesn't make sense to me either whereas I can quite see why he avoids it from my pov. As far as I know he's not using his in classA, must be AB.Posted 26th May 2015 at 01:56 AM by abraxalito
