Hi there
There're several ways to approach this subject.
One is to let the groundpath follow the signalpath on the PCB, and then, from the output, connect the ground to a ground-starpoint.
The other way is, as You mention Your self, to make a ground-layer, or several ground-layers, wich will give a better protection against HF-interferrence, BUT also a lower capacitance from any signal to ground, wich in worst case will lower the bandwith of the amplifier.
Wich is the best is a never ending story, and depends on the situation, especially the sensitivety of the circuit.
There're several ways to approach this subject.
One is to let the groundpath follow the signalpath on the PCB, and then, from the output, connect the ground to a ground-starpoint.
The other way is, as You mention Your self, to make a ground-layer, or several ground-layers, wich will give a better protection against HF-interferrence, BUT also a lower capacitance from any signal to ground, wich in worst case will lower the bandwith of the amplifier.
Wich is the best is a never ending story, and depends on the situation, especially the sensitivety of the circuit.
My personal opinion (for what it's worth) is to combine both methods.
Use a double-sided PCB. In the low power section of the PCB, use the top as a ground plane. This will improve noise rejection. It will also lower the resistance and impedance to ground. Care must be taken to ensure that the components can't accidently become shorted to the ground plane. This also makes it fairly easy to run ground paths parallel to sensitive signal lines, to reduce the noise picked up. This might slightly lower the high-frequency response, but I don't think that dropping the -3dB point from 270KHz to 250KHz will cause anyone to complain.
In the high-power section of the PCB, duplicate the traces from the bottom of the PCB on the top. This will double the current carrying ability of the PCB, which is VERY important near the output transistors. Plated through-holes are useful here, but not required.
If there are a lot of output transistors (more than 4), it might be useful to solder a 14 gauge wire to the top of the PCB from near the power bypass capacitors for the output transistors to near the connection for the star ground. This will help to maintain a low resistance and low impedance path for the power bypass.
The ground on the PCB (near the power bypass capacitors for the output transistors) should be connected to a star ground for the amplifier with heavy wire (minimum possible resistance and impediance).
One of the things I've seen done is to connect the ground of the input signal (single-ended) through a small resistor (20-50 ohms) to the star-ground. I think that this is supposed to help with ground loops.
Use a double-sided PCB. In the low power section of the PCB, use the top as a ground plane. This will improve noise rejection. It will also lower the resistance and impedance to ground. Care must be taken to ensure that the components can't accidently become shorted to the ground plane. This also makes it fairly easy to run ground paths parallel to sensitive signal lines, to reduce the noise picked up. This might slightly lower the high-frequency response, but I don't think that dropping the -3dB point from 270KHz to 250KHz will cause anyone to complain.
In the high-power section of the PCB, duplicate the traces from the bottom of the PCB on the top. This will double the current carrying ability of the PCB, which is VERY important near the output transistors. Plated through-holes are useful here, but not required.
If there are a lot of output transistors (more than 4), it might be useful to solder a 14 gauge wire to the top of the PCB from near the power bypass capacitors for the output transistors to near the connection for the star ground. This will help to maintain a low resistance and low impedance path for the power bypass.
The ground on the PCB (near the power bypass capacitors for the output transistors) should be connected to a star ground for the amplifier with heavy wire (minimum possible resistance and impediance).
One of the things I've seen done is to connect the ground of the input signal (single-ended) through a small resistor (20-50 ohms) to the star-ground. I think that this is supposed to help with ground loops.
Choices...
With a ground plane one can guarantee a low impedance to earth for all returns, but one has no control over the current flow, which could give rise to problems. It's an easy solution, and frequently used at RF.
Personally I'd use the star technique, since you then have absolute control over current paths, and where signals are referenced to. Local ground planes can be useful for reducing noise, particularly from digital circuitry though.
If the amplifier runs from a dual rail supply (+ve, -ve), make a sturdy connection between the PSU reservoir capacitors at the earth point, and return all earths to this point. This is a sonically optimal solution.
There should be no need to raise signal earths using resistors, if careful consideration is given to the earthing of the system as a whole.
Andy.
With a ground plane one can guarantee a low impedance to earth for all returns, but one has no control over the current flow, which could give rise to problems. It's an easy solution, and frequently used at RF.
Personally I'd use the star technique, since you then have absolute control over current paths, and where signals are referenced to. Local ground planes can be useful for reducing noise, particularly from digital circuitry though.
If the amplifier runs from a dual rail supply (+ve, -ve), make a sturdy connection between the PSU reservoir capacitors at the earth point, and return all earths to this point. This is a sonically optimal solution.
There should be no need to raise signal earths using resistors, if careful consideration is given to the earthing of the system as a whole.
Andy.
Thoth said:
OOPS. I should have said that I saw this in some fairly good commercial amps ($1800 list, 250WPC). I think that this was done, as they had no control over what the amp might be connected to, and wanted some assurance that a ground loop wouldn't be created.
If you have COMPLETE control over the grounding in your system, this is probably not needed. I feel that it's cheap insurance, with minimal sound degradation. If you wish, you may consider me to be paranoid
.
If you want explain on: gourn layer = low current
The magnetism born with current bu not with voltage.
Don't forget that parralels track on PCB with high power on long track, can produce some capacitance.
For the art of PCB my professor always says that the diffrent layers must be rotate of 90°.
The up left*right
The down up*down.
You can see it automaticly in good PCB tracers.
The magnetism born with current bu not with voltage.
Don't forget that parralels track on PCB with high power on long track, can produce some capacitance.
For the art of PCB my professor always says that the diffrent layers must be rotate of 90°.
The up left*right
The down up*down.
You can see it automaticly in good PCB tracers.
My X-100 amp uses orthogonality
whenever possible, but no groundlayer. In fact, I have dispensed with Ground altoghether ... I do have a ground-layer which is completely away from signal area.
Check out the separate thread and photos (X100)
Petter
[Edited by Petter on 08-03-2001 at 06:32 PM]
whenever possible, but no groundlayer. In fact, I have dispensed with Ground altoghether ... I do have a ground-layer which is completely away from signal area.
Check out the separate thread and photos (X100)
Petter
[Edited by Petter on 08-03-2001 at 06:32 PM]
Re: Choices...
Hi Andy,
'D. Self' makes the suggestion that if you are going to do use this kind of grounding scheme that it is best to make the star ground connection on a stub off to one side of the heavy trace between the supply caps not to the heavy trace itself. His point is that there can be extremely large currents flowing in the trace between the two caps and by moving the star ground off to one side you eliminate the possibility that one ground trace feeding into the star ground might see a different ground reference than another.
Phil
ALW said:
Hi Andy,
'D. Self' makes the suggestion that if you are going to do use this kind of grounding scheme that it is best to make the star ground connection on a stub off to one side of the heavy trace between the supply caps not to the heavy trace itself. His point is that there can be extremely large currents flowing in the trace between the two caps and by moving the star ground off to one side you eliminate the possibility that one ground trace feeding into the star ground might see a different ground reference than another.
Phil
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.