Ground cleanliness for linear power supply

[sahko123]

I can personally think of only 2 ways to keep the ground clean and its to have a ground plane or star grounding. My question is is the any other ways to improve ground noise and potentially lower ground noise. the current application is for a low noise linear power supply. Currently have a problem with weird voltage spikes under load. (around 90mVpp at 2 amps) and seemed to have tracked it down to the ground which seems to have a similar spike which i believe affects the zener reference which then exaggerates the spike through the error amplifier. Could this also be a case of poor CMRR for the discrete opamp that is use as the error amplifier?

[TimFox]

When you see voltage spikes on "ground", what are you measuring the voltage with respect to?

[sahko123]

The oscilloscope ground lead as close to where the ground originates at the center tap of the transformer and the load is a pair of 10 ohm resistors in parallel

[TimFox]

I don’t understand your answer:  can you post a drawing showing exactly where the probe point and probe ground are connected to your circuit?  With the large circulating currents in a rectifier circuit, this is very important.

[David Hess]

A ground plane is not a replacement for a star or single point ground.

Additional noise control can be implemented using circuits which reject common mode noise like instrumentation amplifiers, current instead of voltage signaling, differential signaling, and common mode chokes.

[exe]

90mV seems a lot. I don't think zenner can "exaggerate", it actually does the opposite.

Can you please post the schematic, the picture of the power supply itself, and the oscilloscope shot? What is the load. My best guesses are: 1) the power supply oscillates 2) the power supply goes in current limiting mode.

[Phil-GS]

Agree with sahko123
One of the first things I learnt from a very experienced designer was to use star grounding.
The other bit of advice was don't use stainless steel fixings - too resistive.

[sahko123]

There is no current limit besides the transformer diodes or pass element having limited current handling so I assume the power supply is oscillating. Power supply schematic to follow soon

[sahko123]

This here is the schematic for the current power supply. It isn't the best schematic but should be easy enough to read

[TimFox]

It would be better to draw the circuit with explicit wires, instead of a generic ground symbol.  Specifically:
1.  The center tap of the transformer should go directly to the filter capacitor, and another wire to the “low” side of the load, to keep the pulsating current out of the load.
2.  If you use a star ground connection, drawing it explicitly will help you wire it correctly.
3.  My original question:  exactly where did you connect the oscilloscope?

[sahko123]

At the end of the connector for the centre tap of the transformer

[bob91343]

The basic problem here is defining the term ground.  It's a common and disastrous mistake made, assuming ground is ground.  It's not.

Ground loops are a frequent problem, caused often by not returning signals and power to their low end sources.  A ground plane should not have current flow in it, so it has to be connected at one point only.  Of course if its impedance is negligible, there won't be a problem, but therein lies the trap.  Just because two points are connected does not assure they are at the same potential.

[TimFox]

At the end of the connector for the centre tap of the transformer

There is a large pulsating current flowing back to the transformer center tap.
Please answer fully: where did the probe end connect and where did the probe ground connect?  Voltage must be measured between two points.

To bob...
Before I retired, when mentoring junior employees, I called it “the G word”, and insisted that more careful language be used: chassis, power supply common, protective earth, the green wire, coax shield, low, etc. depending on the actual connection or node.  Spice doesn’t help by assuming everything connected to node 0 is at the same potential.

[sahko123]

because the circuit is currently only bread-boarded the transformer is connected via alligator clips the grounds of the 4 probes are all connected to the same alligator clip. So the transformer center tap is connected through an alligator clip to the 4 probes ground which also has another alligator clip going to the ground of the circuit. The actual probes are connected to the Emitter of the pass element, the unregulated +18v at capacitor bank, the base of the TIP3055 and the last to ground at the furthest point from the star(ish) ground.

Im also trying to rewire to reduce distances between points of high current like ground and Vcc and am also gonna try sensing not directly from the emitter but nearer to the load (which wasnt connected to the start ground point but closer to the center tap of the transformer).

My main reason for posting was less about the ground but more about understanding where the oscillating is coming from which at first i thought was ground loop stuff or something else to do with where the circuit references its ground(power supply common) from

[TimFox]

Does that mean that the length of wire from the transformer CT connection to the "star-ish" collection of alligator clips is along the voltage being measured?  Remember that there is a lot of pulsed current flowing through that wire, compared with the load current on the power supply.
In general, to get really low noise on a linear supply, you need to be extremely careful with the exact topology, especially of the "ground-like" wires.

[sahko123]

what do you mean by along the voltage being measured?

The ground is being 'measured' not at the star grounding point (which is the reference for all of the probes) but as far from that point as possible

[TimFox]

It would be more clear if you indicated these points on your drawing.
It is possible that you have formed a loop with substantial physical area in the probing and are picking up by induction.
To get a really good measurement, build your circuit properly (not on a breadboard) and place a BNC jack where you want to measure and connect that directly to the oscilloscope with a coax cable.

[CatalinaWOW]

What is implied in several of the answers is that you should think of ground in real devices (not schematics or simulations) as a circuit, not a node.  Each piece of wire, each connection, each chunk of metal is one or more components in this circuit.  If you are careful, and use a Kelvin setup (4 point resistance) you can measure the values in this circuit.  They are usually measured in small fractions of an ohm, but when amps are flowing and you are worried about millivolts those small fractions are important.

Star connections simplify the circuit, making it easier to understand and deal with.  Each leg of the star is just a series connection of impedances.  They aren't absolutely necessary but it is easy to overlook something or make other mistakes in other configurations.  When measuring resistances (or impedances) in a non star configuration remember that you are measuring two terminals of a black box which has unknown internal configuration and very possibly other terminals connected to other points in your ground circuit.

[sahko123]

Ill try improve the breadboard circuit first and if failing that ill build it up on strip-board i just dont have that much of it left so i want to conserve it but would you say that according to the circuit there shouldn't be any reason for the power supply to oscillate? Also i failed to mention that the oscillation is also in time with the power supply ripple even at very low current draws

[TimFox]

Assuming you are seeing spikes at twice the line frequency (coherent to the line), it could be an excitation (evanescent oscillation) of the transformer winding (and other capacitances) due to diode recovery.  There is a large literature on this aspect of power supplies, which might be useful.

[sahko123]

 And what sort of capacitor (value) should i put in at the base of the darlington pair?

[sahko123]

roight, after re-cajiggering the whole thing to have shorter connection paths for a more clean prototype i realized that when i moved the scope ground reference i could change the regulated output ripple reading by insane amounts pictures to follow

[sahko123]

The first image is the probe ground clip connected to where the CT of the transformer comes into the circuit at a star ground point and the second about 2 inches or so from that same point along the same breadboard trace. The voltage is being read from the same place in both pictures at the emitter of the TIP3055 pass transistor. Please note the voltage ranges in the bottom left of the screen. (second photo on next page)

[sahko123]

this is the second photo for file size reasons

[sahko123]

The second picture and the first are measured at 1.2-ish amps. The second photo is showing the waveform with an improperly grounded probe. And no the only reference to mains earth is the scope probe. so the insane ripple is down to improper probing nad probably some from no cap no base and no B-E resistor