I use a combination of both grounding schemes. The last project I did was organized into four subsystems: positive DC rail, negative DC rail, active screen regulator, and preamp/driver. Each was built on its own circuit board with the "modified dead bug" method. Except for etching connection pads, the copper was left intact, and used for making the ground connections.
When installed on a steel chassis, each circuit board was insulated from the chassis. The circuit board ground planes were connected by wire to the central DC neutral, each wire going from the ground plane to neutral, and not "daisy chained" from one board to another. (That might have been more aesthetically pleasing, but sets up a nasty pick-up loop.) The DC neutral was connected to the main AC ground by means of a high current (50Vprv / 25A) integrated bridge wired with the diodes in anti-parallel ("69" style), with a 10R / 10W resistor, and a 0.1uF / 150Vdc capacitor in parallel, so that the ground connection wouldn't be lost in case a fault was severe enough to poof the diodes.
This eliminated any trace of AC ripple from the output, which showed a 60Hz sine wave (not the expected 120Hz pulses from reservoir capacitor charging) at about 6.0mVp-p, and barely audible at the woofer.
When installed on a steel chassis, each circuit board was insulated from the chassis. The circuit board ground planes were connected by wire to the central DC neutral, each wire going from the ground plane to neutral, and not "daisy chained" from one board to another. (That might have been more aesthetically pleasing, but sets up a nasty pick-up loop.) The DC neutral was connected to the main AC ground by means of a high current (50Vprv / 25A) integrated bridge wired with the diodes in anti-parallel ("69" style), with a 10R / 10W resistor, and a 0.1uF / 150Vdc capacitor in parallel, so that the ground connection wouldn't be lost in case a fault was severe enough to poof the diodes.
This eliminated any trace of AC ripple from the output, which showed a 60Hz sine wave (not the expected 120Hz pulses from reservoir capacitor charging) at about 6.0mVp-p, and barely audible at the woofer.
Probably not a bad approach. With careful ground return placement I've never had an issue with ground plane, one amp so quiet no sound at all was audible from tweeter or woofer. One advantage of ground plane, with foresight in layout and assuming you're using a PCB you have the option of scoring the copper plane with a razor and converting it to very low impedance star.
I usually have a "star" on PCB with rectifiers and voltage regulators. The "star" is a big pad with holes where all wide traces from last filter caps go. All filter caps have "inputs" and "outputs", they are separate. Such a way currents from everywhere to corresponding filter caps are isolated from each other.
Attachments
I think where some of the confusion between star and ground plane circuits come from the fact that they both work well some times, poorly others.
I would think star would be the way to go for audio frequencies. You do not want to mix currents if you do not have to.
Ground planes are better at isolating and absorbing high frequency energy, rf amps and high speed digital circuits fall in this domain. And it makes sense if you think about it. At audio frequencies the wavelength of the frequencies are huge, in the MHz to Ghz ranges they shrink to almost nothing.
Now what happens when you try to transmit electrical energy using an antenna that is way too short? Pretty much nothing. As you make the antenna larger it gets more efficient. When you match it to the wavelength by 1/2, 1/4... wavelength you have a better impedance match and it behaves like a skipping rope oscillating. A lot of motion (energy) for little input.
At high frequencies ohms law does not cut it because the electrical energy follows wave theory more and because of that they have their own methods of slaying a beast. Ground planes are used to short out radiation by all thoes antennas we have operating. The size and shape of the conductors are also taken into account to reduce their radiated energy or limit their capacitive or inductive effects.
None of this really applies up to the hundreds of kHz we want to play with unless you do something really stupid (Run small signal wires next to large signal ones).
At least this is how I see it.
I would think star would be the way to go for audio frequencies. You do not want to mix currents if you do not have to.
Ground planes are better at isolating and absorbing high frequency energy, rf amps and high speed digital circuits fall in this domain. And it makes sense if you think about it. At audio frequencies the wavelength of the frequencies are huge, in the MHz to Ghz ranges they shrink to almost nothing.
Now what happens when you try to transmit electrical energy using an antenna that is way too short? Pretty much nothing. As you make the antenna larger it gets more efficient. When you match it to the wavelength by 1/2, 1/4... wavelength you have a better impedance match and it behaves like a skipping rope oscillating. A lot of motion (energy) for little input.
At high frequencies ohms law does not cut it because the electrical energy follows wave theory more and because of that they have their own methods of slaying a beast. Ground planes are used to short out radiation by all thoes antennas we have operating. The size and shape of the conductors are also taken into account to reduce their radiated energy or limit their capacitive or inductive effects.
None of this really applies up to the hundreds of kHz we want to play with unless you do something really stupid (Run small signal wires next to large signal ones).
At least this is how I see it.
I do think wiring over a shield plane or PCB traces over a shield plane would be a good alternative where shielded wiring would otherwise be used, e.g. low level high impedance input circuits. I would still terminate the circuit signal and power returns in a star and connect the shield plane to the star at one point. I would avoid connecting local signal returns to the ground plane.
The idea of splitting a PCB ground plane to multiple islands, connected together at a star point is worth looking into. Each island should probably be dedicated to a single circuit return path.
I think even with line level and preamps, it's still important to keep the current loops separate, even though the currents are smaller.
Cheers,
Michael
You might want to check Analog Devices Analog Dialogue "ask the applications engineer #12" as he goes into some of the grounding issues of digital and analog.
I used to do video design and what is video but audio on STEROIDS. You do 20-20,000 Hz, I do 20Hz to 30 MHz (hi def). Ground planes are definitely way better for controlling noise. Someone mentioned power supply noise from the rectifiers and caps and that sounds like it would be true. The cards I did plugged into a backplane frame assembly with external regulated supplies. You definitely want a star system getting power to the ground plane board. The ground plane will average/randomize the power supply noise from chips delivering pulses. Other good things were leaving holes in the plane beneath chips and the '-' input to reduce capacitance where it isn't wanted. I've toyed with the idea of a 4 layer board preamp to see how good it could get but am too lazy, my old Hafler DH-101 is fine with line level and I don't do vinyl anymore.
Your focus on grounding is excellent as the signals are only as good as the power and ground system.
Things you DON'T want to do particularly with video is use a shared ground wire on the inputs and outputs. 2 volts into 150 ohms times 6 outs is significant current and WILL screw up the input when it has the same shared ground. Star wiring that completely eliminated the interaction and no, it wasn't my design but it's good to learn from others mistakes.
G²
One famous now retired designer of computer video interfaces Lazar Lvovsky once looking at my PCB said, "Anatoliy, your ground wire is too thin!" 
I tried to explain him basics about currents on my PCB, but he would not listen: the habit is the second nature, as a saying says.
Each and every cubicle has own culture.
Sometimes it is good to share beliefs and knowledge through walls of cubicles.
I tried to explain him basics about currents on my PCB, but he would not listen: the habit is the second nature, as a saying says.
Each and every cubicle has own culture.
Sometimes it is good to share beliefs and knowledge through walls of cubicles.
Dear Michael,
What you mean is, run separate ground wires from components to star ground on one layer, And have a ground-plane shield on one other layer, and connect tis shield to the same star ground?
I was playing with the same idea. But what about the capacitance and inductance between the layers with amplifiers?
With kind regards
Bas
If you have to ask whether you need a ground plane, then you don't want it. Controlling where you want return currents to flow is really the goal here, not possible with a solid ground plane. Ground planes are more for low impedance signals were minimizing lead inductance is key. It's not helpful or necessary for higher impedance tube circuitry at audio. Think like an electron and use return wires twisted with their sources.
If you are doing point-to-point wiring then how are you going to make a "ground plane"? Aren't you basically going to have a star ground by definition (unless you do some really sloppy wiring)?
That's easy: it's called "dead bug construction", so named for the appearance of an eight pin DIP IC turned upside down. I've used this method for different kinds of solid state designs: audio, RF, and digital. In dead bug construction, you wire the circuit point-to-point while using the copper cladding for grounding. For mechanical reinforcement, you solder a large resistor (2M2 / 0.5W) to the ground plane, and use the opposite end as a tie point. So long as the resistor is very much larger than the impedance, it has little effect on circuit operation. It's the preferred method for RF construction since the insulating medium is air, and not a plastic board with its higher permittivity and stray capacitance. It's also friendlier to junk box parts, and you can complete the circuit faster than you can etch and solder a PCB.
There is also a modified dead bug method that uses etched contact pads as the tie points. This is often a better alternative for UHF circuits, as you can minimize lead lengths. I also use it for hollow state designs since the voltages run a good deal higher than those encountered in most solid state work.
The first real solo job I had as an engineer was to get a consumer unit for WLL through EMC test. In the nature of the thing it had a 2.4GHz CDMA radio/antenna with an HDLC link to a second box with 8 ringing SLICs to drive POTs handsets, a 200MHz microcontroller with RTOS in flash, a custom FPGA and a battery-backed PSU. The box was recyclable plastic.
The 2 main boards were a 7-layer stack with dedicated analog and digital ground planes, split at the CODEC chips, so that on one side you had the audio signals, and on the other the digital busses. We had a tool that was supposed to help with analysis of the PCB. It was a nightmare. The unit worked but no way was it going to meet CISPR/B. A bunch of 8 TI DSPs which were doing the LD-CELP voice compression with their clocks all hammering away in unison did nothing to help.
On the 3rd. iteration and well behind schedule it achieved compliance.
This would have been impossible without close attention to the design of the layout, tracking and power planes.
If you're designing something with digital and analogue systems on the same PCB, such as a DAC, then ground planes should be in your thoughts (of course dependent to a degree on the frequencies involved). You may only have 2 layers available to you.
If you're building a valve amp, point-to-point wiring with a star ground is easy to keep track of.
In between these extremes combinations and hybrids of the two are acceptable, the real effort is in understanding and visualising the current flow in the circuit.
w
The 2 main boards were a 7-layer stack with dedicated analog and digital ground planes, split at the CODEC chips, so that on one side you had the audio signals, and on the other the digital busses. We had a tool that was supposed to help with analysis of the PCB. It was a nightmare. The unit worked but no way was it going to meet CISPR/B. A bunch of 8 TI DSPs which were doing the LD-CELP voice compression with their clocks all hammering away in unison did nothing to help.
On the 3rd. iteration and well behind schedule it achieved compliance.
This would have been impossible without close attention to the design of the layout, tracking and power planes.
If you're designing something with digital and analogue systems on the same PCB, such as a DAC, then ground planes should be in your thoughts (of course dependent to a degree on the frequencies involved). You may only have 2 layers available to you.
If you're building a valve amp, point-to-point wiring with a star ground is easy to keep track of.
In between these extremes combinations and hybrids of the two are acceptable, the real effort is in understanding and visualising the current flow in the circuit.
w
Oscillations (unintended) are another question really, usually involving parasitic reactances where there is excess gain. If you rely on a ground plane to solve them, then you are counting on accidental trace bypass capacitance coupled to the plane. A strategically placed discrete bypass will mostably solve it more reliably. Remember using multiple high impedance signals that are coupled to a single ground plane, then you need to account for crosstalk using these same "accidental" capacitance's to a common point.
Not exactly. Vector of gain through the loop that includes parasitic coupling is more than 1 by modulus and is positive. But oscillations mean an extremal case, unwanted couplings may cause extra distortions. Or they may reduce them, that's why the same topology works differently laid out differently.
Yes if you want to build it as a sustained oscillator, but it's possible to have oscillations without the exact criteria. Under damped resonance with neg gain can be excited to higher levels as well. Accidental or unwanted topologies is a problem too when building LF circuits with super wideband construction ie microstrip/stripline,
ie amplifier as a colpitts et al.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.