My_Ref Fremen Edition - Beta build/Fine tuning

[bcmbob]

Leave me alone Andrew, I'm one of those sensitive musician types and I might go over the edge.

[AndrewT]

I've heard that some of the best musicians live on the edge. Why should you be any different from the geniuses?

[bcmbob]

That makes me feel much more betterer

[soongsc]

Originally Posted by bcmbob

In the third grade Mrs. Haffy told me to put more time on my spelling workbook. Sometimes we see what we think and not what we type. Those da*m spell checkers can't read my mind either.

You are lucky they don't try to change your typing. The computer tries to correct my Chinese, I have a hard time checking and correcting it.

[KSTR]

Originally Posted by ClaveFremen

R11:
It's needed and improves noise rejection from GND by several tens dB.

Every mV accross R11 appears 1:1 at the output (it easy to see the circuit is a voltage follower for signal appearing accross R11). What R11 does, though, is reducing shield current in the interconnect when both source and amp are earth-grounded at different points or have impressed current from other low impedance mechanisms (ground loops from improper high loop area cabling). Since LF drops along the shield's ohmic resistance are input signal, too, plus then undergo some 30x of amplification, a net reduduction of hum/buzz may be achieved in some special situations. The typical "it depends" kind of thing. A good way to fix this is using an balanced cable together with a true differential input.


-------:-------

As for speaker return current minimized loop area, see pic for the strategy (which is simple : keep send and return lines together). This is done by wiring at the bottom side of the PCB, no PCB change is necessary. Now it is drawn as a dead short, of course you must tee off the speaker send and return -- as a twisted pair -- at some point (but it does not matter where).
There is something non-perfect there, still : the run allong asymetric lengths of V+ and V- lines, as well as the too big loop area of the GND plane vs supplies itself.
Exact position would have to be found experimentally**) by injecting a magnetic field then moving the wire until lowest output is achieved.

But to fix this on the PCB, it would be a major workaround, a complete rewrapping of signal and power paths... resulting in a short but wide PCB where power/relay stuff is left of the LM's and signal is right of them, strictly adhering to the star GND principle as well....

**) speaking of which, I'm OK again and will continue working on the PCB tomorrow....

[soongsc]

You really have big hopes that this improves the listening experience?

[KSTR]

Well, "big hopes" is a strong term but striving for layout/wiring which gives lowest possible interference (which can be quantified by measurement) has little chances to degrade perceived performance and is a prerequisite before any tweaking with boutique parts (and resistor orientation and similar things) makes sense. Some aspects of that are somewhat interconnected there, of course. And striving for the best circuit comes even before the layout/wiring, but in this thread we take the ciruit design as a given.

But that's only me, of course.

[bcmbob]

KSTR, I have often wondered if balanced cables would make an audible difference with this or any MyRef/LM3886 design. Hopefully that subject will get some attention as things develop.

[ClaveFremen]

I've started to test the C9/R10 mod and for sure it sound different.

With the position reversed (C9 to ground) there seem to be a more contrasted, direct sound with wider and deeper soundstage and tighter bass.

The classic arrangement seems more 'soft', maybe confused.

So far I would say Klaus'mod is effective in improving sound but I'll decide after listening again tomorrow... I don't want another ceramics debacle...

Originally Posted by KSTR

Every mV accross R11 appears 1:1 at the output
(...) The typical "it depends" kind of thing. A good way to fix this is using an balanced cable together with a true differential input.

But we don't have a differential input here...

Originally Posted by KSTR

As for speaker return current minimized loop area, see pic for the strategy (which is simple : keep send and return lines together). This is done by wiring at the bottom side of the PCB, no PCB change is necessary. Now it is drawn as a dead short, of course you must tee off the speaker send and return -- as a twisted pair -- at some point (but it does not matter where).

We could test it with RC boards but I really don't think this is a major issue.

Originally Posted by KSTR

There is something non-perfect there, still : the run allong asymetric lengths of V+ and V- lines, as well as the too big loop area of the GND plane vs supplies itself.
Exact position would have to be found experimentally**) by injecting a magnetic field then moving the wire until lowest output is achieved.

This sort of things can be addressed only using a quite bigger board, I fear.

Limits given I don't feel it's a bad layout, is it?

Originally Posted by KSTR

**) speaking of which, I'm OK again and will continue working on the PCB tomorrow....

Sorry, I can't understand to what are you referring to...can you clarify?

I'll read with interest your tests results

Originally Posted by KSTR

Some aspects of that are somewhat interconnected there, of course. And striving for the best circuit comes even before the layout/wiring, but in this thread we take the ciruit design as a given.

Well, in the core part (the amp one) circuit is given, since I'm designing a My_Ref variant that try to be, nevertheless, faithfull to Mauro's design (some mods are inspired by the Evolution too).

Though it doesn't mean that the amp section can't be modified or improved.

In fact:

• an alternate compensation is offered (it modify not by a small margin how the amp works)
• LM3886 grounding is different.
• A voltage limiter has been introduced to permit higher rails on LM318
• Your C9/R10 mod is going to be part of it.

The PS section, also, is quite different from a My_Ref...

[KSTR]

Bob,
A MyRef, as well as any other OpAmp design, could be modified to be a differential input (even and especially for a unbalanced source output). There are some issues though (like you need to know the source impedance which must be fixed, and for the MyRef you'd need a permanently attached cable).
It would probably be easier and more generally useful to put a high impedance differential input/buffer ahead of a MyRef (which could be either the subtracting circuit itself or the input circuit does the subtraction). Many ways to skin a cat here....