While a 'tree' ground would work it's much better a true star ground approach.
Are you aware that you're contradicting yourself?
i.e. the star ground center...AKA, as Andrew already pointed, Main Audio Ground.
I was thinking the same thing first .
The difference here is: in the described case, you only have one point to set a shared reference, but the current to and from each ground would be minimal (as small as you can get).
A shared 0V line means, that different ground share their 0V lines, with, for example, larger currents for the lm3886 and smaller currents for the lm3886. In the Myref-amplifier, this is the case, from the "tree-ground" center between smoothing caps all the way to the transformer, or wherever to you connected it.
Did I make the difference clear?
This is true for every star ground... a supplementary transformer doesn't change anything in this regard...
The transformer does not connect to the MAG. The PSU Zero Volts connects to the MAG.
Not electrically. The grounds are all on the same 0V.
But the shared 0V lines (and hence the induced noise of each supply to the other) are minimized and finally galvanically insulated at each transformer.
I think this approach could be beneficial when low power and high power 0V lines would be shared, like it is the case for the Myref.
Of course, as stated a few posts before, I am aware that this maybe contradicts the myref principles of simplicity and cost efficiency.
You would be surprised how noisy the power ground is. In the EMC test lab, it's a 20+db difference whether I connect the MAG to safety earth ground or not. Also with the commonly used PS cap configuration, sending 1KHz into the circuit raises the noise level as well. I have tried this with different amps.
I'm sorry but if you connect the 0V lines to the star ground there is no galvanic insulation...to have it you should have a transformer between them, which is not the case...
George, it's not clear... the +20db noise is measured when MAG is connected to safety ground or vice versa?
Just to rule out a misunderstanding here. The PSU Zero Volts is the point between the smoothing caps, isn´t it?
+20db noise when the MAG is connected to safety ground. All measurements were above 50KHz due to equipment limitation, so I do not know what happens below this range.
It confirm my experience with the Fremen Edition with PGND connected to safety ground it sound more closed, soundstage collapses, and it's not as detailed as with PGND unconnected.
Yes, that was stupid and irrelevant writing.
I meant isolation from the mains power, which is clear anyway.
I think this all boils down to not use shared 0V lines for high and low power supplies, from my side.
This can be achieved locally, by using separate stargrounds, (or treegrounds, or groundplanes...) and connect them all at one point. From there you, would of course share 0V lines.
Or, if you would go all the way, use separat transformers, in this case you would only have them connected at one point and not share 0V lines.
I guess a second transformer is not really necessary, some circuits might be more sensitive to this than others, but maybe I will try it when I'm bored.
At first, I will design and etch a small plug-in board, to retrofit the better shunt supply for the LM318. When that is done it should be easy to connect a second transformer for this one.
This is where the big bag of worms begin. I could go into a two line design, but then I cannot find a power cable of adequate size that is already certified in different countries. Since cable size is also critical to sound, the only choice is to go with an appropriate PS design with line filter that keeps the sound quality up, yet meets test requirements. Pretty much a redesign of everything. DIY does not have these restrictions.
This reminds me that most of my listening locations do not have earth ground. Seems like I need to investigate this issue.
There are some subtleties here that aren't obvious when you look at the schematic, but will become clearer when you look at the layouts.
Until the V1.2 Twisted Pear PCB, there was single star ground (PGND) located between C3 and C8. It also coincided with the transformer 0V connector on the board. The GND for the LM318 shunts was referenced to this, as well as the 1 ohm resistor (R11) that isolates SGND (the small-signal ground plane).
From V1.3 onwards, I made a small change to isolate both the zener shunts and SGND, from PGND, more effectively. The high-current star ground remained at the same location between C3 and C8, but I introduced a small-signal star ground at one of the terminals of C11. The zener shunts are referenced to this, as well as R11. This small-signal star ground and most of the local traces to it are mostly *outside* the large-current charge/discharge current-loops associated with the main bridge rectifier, C3, C8, the speaker return current trace and the traces leading to C1, C2.
The two star grounds are connected by a single trace, which is not going to see much current - it's there just as a DC reference. The few nH inductance on this trace actually helps the isolation further.
The advantage of this arrangement is slight, but it keeps the relatively small zener currents and C6, C11 circulatory currents in tighter loops which are less likely to pick up interference from the large-current loops which touch the PGND node. SGND (the small-signal ground plane) also benefits by being tied to the quieter ground.
It certainly hasn't hurt the sonics, and I might even say that the V1.3/1.4 grounding actually contributes to darker sonics than the V1.2 grounding. One experiment that hasn't been tried yet: Use a small ferrite bead or inductor between the two star grounds - there shouldn't be any signal current on this trace anyway, so it shouldn't hurt...
Edit: The traces from C3, C8 to R1, R4 originate within the high-current loops, but that can't be helped. I made them cross out of the loop perpendicularly - i.e. R1, R4 are perpendicular to the large current flows between C3, C8 and C1, C2.
Last edited:
Did a bit study on the power supply theory vs real power supply. It seems that in power supply design we talk a lot about regulation, supply impedance, etc. However, there is no study on handling the return current that goes through the ground plane. This is a critical aspect that needs to be looked into. What we want to do is to retain the return current locally. Additionally, when the return current charges the main filtering caps, it also effects the current flow in the positive side. This is about as much as I can tell so far.
I have not read of any detailed discussions, but it seems the main return should be the power supply. It seems that if it somehow tends to flow towards the earth ground, then it is not as ideal. Most interconnecting equipment seem to connect via power ground, if return somehow gets to the power ground of other equipment, then it is also not ideal.