Hello,
I have successfully build an audio pre-amplifier based on Self Douglas préAMP 2012 (2), to the exception that I removed all the TONE controls section, the MC section, the Subsonic Filter section, the IEC amendment section, as well as the variable gain section from jumpers.
Now I want to make a new guess including the MC section, the Subsonic Filter section, the IEC amendment section and the gain section. My question is regarding the ground plane. Below I share a capture of the solder side where the audio ground traces are highlighted. TP1 is my HQG (High Quality Ground) where the analog ground join the Power Supply. R717, 718, 817 and 818 are connected from the components side!
I tried to create a ground plane on the components side, excluding the digital ground traces that goes to the relay and it's transistor. You'll find below another capture, again with the ground traces and ground plane highlighted. Notice that the ground plane join the Power Supply ground at the HQG. Also of course, the R* above have their traces irrelevant with the ground plane!
We can observe that the digital ground traces are excluded from the ground plane to assure that the digital ground goes directly to the big cans of the power supply. All the donuts that are highlighted on the ground plane of the picture above are actually making a second connection to the analog ground via thermal brake (cross) like these ones here:
So my question(s) is regarding the possible ground loops that this ground plane can create.
I admit that this kind of situation has always intrigue me...
Thanks, in advance...
I have successfully build an audio pre-amplifier based on Self Douglas préAMP 2012 (2), to the exception that I removed all the TONE controls section, the MC section, the Subsonic Filter section, the IEC amendment section, as well as the variable gain section from jumpers.
Now I want to make a new guess including the MC section, the Subsonic Filter section, the IEC amendment section and the gain section. My question is regarding the ground plane. Below I share a capture of the solder side where the audio ground traces are highlighted. TP1 is my HQG (High Quality Ground) where the analog ground join the Power Supply. R717, 718, 817 and 818 are connected from the components side!
I tried to create a ground plane on the components side, excluding the digital ground traces that goes to the relay and it's transistor. You'll find below another capture, again with the ground traces and ground plane highlighted. Notice that the ground plane join the Power Supply ground at the HQG. Also of course, the R* above have their traces irrelevant with the ground plane!
We can observe that the digital ground traces are excluded from the ground plane to assure that the digital ground goes directly to the big cans of the power supply. All the donuts that are highlighted on the ground plane of the picture above are actually making a second connection to the analog ground via thermal brake (cross) like these ones here:
So my question(s) is regarding the possible ground loops that this ground plane can create.
- Should I forget the components ground plane and stick with the original analog ground solder traces?
- Or delete all the analog ground solder traces and keep the components ground plane?
- Or keep both of them without worrying?
- Or either exclude all the analog ground donuts (cross) from the ground plane to assure that the ground plane doesn't create a second electrical ground connection, i.e. a loop?
I admit that this kind of situation has always intrigue me...
Thanks, in advance...
The usual relatively simple and effective method is to provide a star ground , where each point of earthing connects not to each other, as seen in your images, , but to the actual power supply earthing point, with then individual connections from where a connection to earth is needed. Noting you can also star positive as well, which helps, once visualized how star grounding is effective. Your circuit traces referring to the branch of 4 just above the supply appear to have commenced, for star earthing, but then strangely abandoned that guiding principle.
Last edited:
Star grounding applies when you're talking different pieces of equipment located in opposite ends of a building. It's been misconstrued to mean that the ground return for each and every current on a PCB must only meet at one start point. You'd be far better off with a solid ground pour connecting all grounds.
If you have digital circuitry on the board, keep it separate so its return currents won't flow through the analog part of the ground plane.
Henry Ott, "Electromagnetic Compatibility Engineering" is a good place to start. I seem to recall that it's Chapter 16 that deals with mix signal PCB layouts. Zach Peterson of Altium Academy talks about it a great deal on his YouTube channel as well: https://www.youtube.com/@AltiumAcademy
But... If you really-really want to keep your star ground thing, here's how you do it: First pour a ground plane. Then for each pad that you do NOT want to connect to this plane, click the pad, select the Connections tab, and change the connections to the plane to 'none'. You can then copy-n-paste that pad's attributes to other pads. I'm assuming you use KiCAD here. If you're using something else, there's usually a way to force the pads not to connect to the plane.
Tom
If you have digital circuitry on the board, keep it separate so its return currents won't flow through the analog part of the ground plane.
Henry Ott, "Electromagnetic Compatibility Engineering" is a good place to start. I seem to recall that it's Chapter 16 that deals with mix signal PCB layouts. Zach Peterson of Altium Academy talks about it a great deal on his YouTube channel as well: https://www.youtube.com/@AltiumAcademy
But... If you really-really want to keep your star ground thing, here's how you do it: First pour a ground plane. Then for each pad that you do NOT want to connect to this plane, click the pad, select the Connections tab, and change the connections to the plane to 'none'. You can then copy-n-paste that pad's attributes to other pads. I'm assuming you use KiCAD here. If you're using something else, there's usually a way to force the pads not to connect to the plane.
Tom
Definitely, and I wish we could debunk this myth in the hobbyist community. Professional engineers all know better,
but there aren't enough of us around here to address this every time it comes up.
And there's a bonus, routing to a ground plane is trivial and extremely low inductance, unlike long, multiple return traces
to one "ground" pad. Some might like to look at these videos for further elucidation. I hasten to add that these effects
are present in pcbs starting in the low kHz region. They are not just RF considerations.
but there aren't enough of us around here to address this every time it comes up.
And there's a bonus, routing to a ground plane is trivial and extremely low inductance, unlike long, multiple return traces
to one "ground" pad. Some might like to look at these videos for further elucidation. I hasten to add that these effects
are present in pcbs starting in the low kHz region. They are not just RF considerations.
Yes much better. For a comprehensive look at both star and larger ground planes and other forms : https://www.analog.com/en/resources/analog-dialogue/articles/staying-well-grounded.html "Having separate power supplies (and, thus, separate ground returns) for the various portions of the circuit. For example, separate analog and digital supplies with separate analog and digital grounds, joined at the star point, are common in mixed-signal applications"
Looking good!
Thicker traces wouldn't hurt. Also, consider separating the centre tap charging/discharging currents from the smooth current at the DC side... not many do... a tiny cut is enough.
Quite a few HiFi manufacturers use star earthing on their PCBs as well as DIY / hobbyists. Naim for one. I also prefer ground planes, but perhaps there is more to it that we're missing?
I could see an argument (for example) for keeping high current ground paths like speaker returns in a power amp totally separate from delicate input signal grounds.
The other problem with star grounds on PCBs is the ground to signal loops are invariably bigger than just a ground plane on its own. On a power amp a ground plane often does not make sense, so you have to focus on the loop areas and not mix signal and power returns.
Yes this is exactly what I do. The only digital ground trace is from the U3 +5 VDC regulator that power the relay and it's transistor. It's ground trace goes directly to the two big capacitors ground on the components side, without any other analog connection.
While reading the link proposed by Chris Daly above (post #6), I realized that you were probably talking about the Ground Plane as in the link figure #3 ! ;-)
Henry Ott passed away in 2018. Then his web-page went dark.
But it has been recovered.
https://hott.shielddigitaldesign.com/
Proper layout is the best way to view it, and currents can be encouraged to flow in a small, tight loop.
The currents will naturally do the right thing if you allow them to do so.
Smaller HiFi mfrs are often "old school" amateurs lacking formal (or any) education in this specialty area.
Or they just "star" as a sales bullet point.
For my own designs,I tend to use multiple separate ground planes.my DAC, for example, has a ground plane for the noisy digital bit and a separate plane for the last stages which, while still digital, are timing and voltage sensitive.
I take speaker returns straight back to the reservoir caps in power amps, so that's star earthing of a sort.
So basically I agree - use everything where its most appropriate!
I take speaker returns straight back to the reservoir caps in power amps, so that's star earthing of a sort.
So basically I agree - use everything where its most appropriate!
"I take speaker returns straight back to the reservoir caps in power amps, so that's star earthing of a sort."
Actually that is minimizing the loop area for the load current, as should be done.
Most "star earthing" enlarges the currents' loop areas considerably.
Actually that is minimizing the loop area for the load current, as should be done.
Most "star earthing" enlarges the currents' loop areas considerably.
His book is still in Bezos' Bookstore. Just search for ISBN 978-0470189306.
I simulated the effects of that. Taking the speaker return all the way back to the power supply makes performance worse. Recall that it's the voltage across the speaker that the amplifier is trying to control. So the feedback ground and the speaker ground should be connected. More here: https://neurochrome.com/pages/grounding
Tom
Yep. Which maximizes the inductance in the feedback path. That also maximizes the potential for an error current to create an error voltage across that feedback (and ground) line. That's exactly what you're trying to avoid.
I suggest setting up a circuit simulation where you introduce a small error current (or voltage) in the ground or feedback path and then play with taking the feedback reference at various points. You can estimate the inductance of a PCB trace as about 1 nH/mm of trace length. You can calculate the trace resistance from the sheet resistance of copper.
Tom
Good tip - thanks. I'll have a play at rerouting some grounds and see what differences I can hear.