MyRef RevD - successor to RevC

[linuxguru]

Here's the schematic and PCB layout for a successor to the MyRef RevC - it's basically very similar to the MyRef RevC, with a minor change derived from an idea used in the Musical Fidelity MVT.

Basically, the idea is to bypass the darlington LTP of the LM318, and use a lower-noise (and maybe faster and more linear) discrete LTP instead. I'm suggesting NEC 2sc1845 with 390 ohms degeneration on each leg, and a 33k resistor to set the tail current (which should improve H2 compared to an active current source) - these values are just placeholders, which will be tweaked after prototyping and auditioning. I expect that the change will result in darker silences and improved dynamic range over the RevC, which should be discernible on quiet passages.

[ppcblaster]

Any chance in a future MyRefC to use the LM4780 to increase power?

[AndrewT]

Hi,
how does the 4780 allow increased power?
As I see it, 4780 has half the heatsink interface area allocated to each chipamp channel.

[regiregi22]

Originally Posted by ppcblaster

Any chance in a future MyRefC to use the LM4780 to increase power?

Having in mind that a 4780 equals two 3886, it will be needed to bridge or parallel two of them, which will make a new complete design.
Or you can experiment with two equal MyRefC modules by paralleling or bridging them, but I am not in favour of making experiments without deeply knowing the design.

Regards,
Regi

[linuxguru]

It's possible to use 2x LM3886 or 1x LM4780 as two Howland current pumps in parallel, with current sharing at the output through the 0.47 ohm resistors. In my sims, I found (quoting from memory) that the R5, R6, R8, R9 resistor networks have to matched across the two parallel Howlands to about +/- 0.01% to get the current sharing to match to about 95% (at 20 kHz).

It's feasible, but it doesn't buy much. The voltage rails are still constrained to about +/- 35V, so about all that the parallel Howlands succeed in obtaining is the current-drive to drive 4-ohm loads. Compensation still has to be tweaked for the 4-ohm load, but that's probably manageable.

You could probably get 80 to 100W into 4 ohms. It's still going to remain 40 to 50W into 8 ohms.

[ppcblaster]

Something like Twisted Pear's Sympatico,

"Sympatico is a single small module per channel with one LM4780 dual amplifier chip in a balanced, bridged configuration" "100 + Watts/Channel into 8-ohm loads (conservative)"

The thought came from the Sympatico design,

[linuxguru]

A bridged topology is different from a parallel Howland. In a bridged configuration, each chipamp is driven 180 degrees antiphase, while in a parallel Howland, both are driven in-phase.

A bridged MyRef is somewhat more complicated - it will require an additional LM318 to drive the second chipamp antiphase to the first. This LM318 has to have its own complete GNFB loop, along with all compensation elements, with adjustments to take care of the phase reversal on one half of the bridge.

A bridged-parallel MyRef topology is also possible, with 4x LM3886.

[Pixo]

if you had two identical my-ref modules, will it be possible to bridge them by driving one of the inputs out of phase with the other and connecting the outputs of those two amps to a single 8 ohms speaker?

[regiregi22]

My intuition follows to say yes, but the teachers must have their say

[cent88]

Is it necessary to have low noise resistor(which is connected from signal path to ground that works together with input coupling capacitor)?