I have a few ideas about how to do the ground connections on an OPA827+BUF634 board but I am not entirely sure what is the optimal way of doing it.
Note : For all of them the 3 pole 0.1" inch header in the bottom is for connecting a volume pot.
It is a 4 layer PCB with Internal Power and ground planes.
Idea 1 :
Input/feedback/Pot ground island(all 4 layers), connected through a wire from the input header back to the star point at the PSU. The 0V return wire also goes to this star point. Input trace has to cross the cutout between power ground plane and input ground plane. Also Input, output, feedback pads/traces has the power ground plane under them.
Idea 2 :
Input/feedbak/Pot/0V return wire ground island(all 4 layers), connected through a wire from the output header back to the star point of the PSU. Ground island is enlarged to give some shielding to the sensitive input,output, feedback pads/traces to and from the OpAmp. However, the output trace from the buffer has to go over the power ground plane most of the way and cross the cutout between power ground plane and input ground plane
Idea 3 :
To hell with it, currents are relatively small, just connect everything to the power ground plane and be done with it. Everything is connected to ground through the power supply ground return wire.
I am leaning towards option 2 as the best way to do it.
Note : For all of them the 3 pole 0.1" inch header in the bottom is for connecting a volume pot.
It is a 4 layer PCB with Internal Power and ground planes.
Idea 1 :
Input/feedback/Pot ground island(all 4 layers), connected through a wire from the input header back to the star point at the PSU. The 0V return wire also goes to this star point. Input trace has to cross the cutout between power ground plane and input ground plane. Also Input, output, feedback pads/traces has the power ground plane under them.
Idea 2 :
Input/feedbak/Pot/0V return wire ground island(all 4 layers), connected through a wire from the output header back to the star point of the PSU. Ground island is enlarged to give some shielding to the sensitive input,output, feedback pads/traces to and from the OpAmp. However, the output trace from the buffer has to go over the power ground plane most of the way and cross the cutout between power ground plane and input ground plane
Idea 3 :
To hell with it, currents are relatively small, just connect everything to the power ground plane and be done with it. Everything is connected to ground through the power supply ground return wire.
I am leaning towards option 2 as the best way to do it.
Attachments
Last edited:
Idea 3 IS an option, Opc does the same thing with his "The Wire" headphone amplifier and it seems to work ok.
But it just feels horribly wrong to it that way, you now have your input/feedback ground and 0V reutrn wire on the samme plane as your power plane and as we all know, the power plane isnt the cleanest ground around.
However, with the relatively small currents we are talking about in a headphone amp(28mA Peak for 100mW into a 250 Ohm headphone) the power ground plane on the headamp PCB board should not be too bad when it comes to noise.
But still, you have those nasty unlinear ground currents going though the power ground plane when leaving class A of the output buffer and you now have to rely on the caps charging and recharging to deliver the necessary current to the output buffer which you do not want mixed up with everything else.
Then again, the power ground plane around the input/feedback and 0V return wire should be relatively clean since the main caps and small SMD caps for the buffer is placed on the other half of the board, and the input opamp is not much of an issue since the current is has to deliver is well below the quiescent current of the opamp so the current draw into the opamp supply pins and current into the ground through the caps is more or less stable.
Decisions, decisions....
Idea 1 has the headphone 0V return wire connected to the power supply star point.
But if using idea 3 you would still have input/feedback/Pot ground connected to the power ground plane even if you made the 0V return wire connect to the power supply star ground.
Problem with Idea 1 is, as you said, the input trace has to cross the cutout between the two ground planes, creating a large current loop area (the return current won't follow closely the signal track, and will take a big detour, all the way through the supply ground wire, back to the input ground wire). Idea 2 is a bit better. Also, plane cutouts create a capacitance effect, but it should be negligible at this size.
Idea 3 doesn't have the same problem, but as you said, there is a risk that the supply ground current pulses contaminate the signal ground.
The best idea IMHO is to think in the opposite direction: contain the supply ground to a controlled area, adopt a "star on star" layout, with geographic separation of clean signals/gnds and dirty signals/gnds, that way you can control the current loops areas, and minimise their surface.
Idea 3 doesn't have the same problem, but as you said, there is a risk that the supply ground current pulses contaminate the signal ground.
The best idea IMHO is to think in the opposite direction: contain the supply ground to a controlled area, adopt a "star on star" layout, with geographic separation of clean signals/gnds and dirty signals/gnds, that way you can control the current loops areas, and minimise their surface.
Last edited:
Then how exactly would you do it?
I have seen the layout you did for the Marsh headphone amp, it was a very nice layout. You took and connected everything to the same ground, just as I was suggesting in my idea 3. It looked like your board had great performance, however that was with a high bias class a output stage which in most cases means that you do not have much of a problem with supply ground current pulses.
Last edited:
Well I would make a star-like ground connecting, with relatively large traces, the supply gnd pin and the gnd poles of the filter caps (and make sure the + and - traces are very close-by), and then connect it to the signal ground plane.
In my case, it was indeed a bit different, as the amp works in class A, so I favored minimal impedance (and also visually pleasing appearance
) over ground trace control , but still you can see that kinda-"star on star" arrangement, and the long board allowed me to keep the supply ground currents as far from the signal traces as possible, each current loops resolved locally, and not spreading to unwanted territory.
In my case, it was indeed a bit different, as the amp works in class A, so I favored minimal impedance (and also visually pleasing appearance
) over ground trace control , but still you can see that kinda-"star on star" arrangement, and the long board allowed me to keep the supply ground currents as far from the signal traces as possible, each current loops resolved locally, and not spreading to unwanted territory.A nice illustration of what I mean (I think) is Putzeys' preamp board layout, published in Linear Audio Vol. 5, which you can see here: http://www.linearaudio.net/images/bp v5 f28c bottom.png
the difference being that the star ground configuration is only applied to the first filtering stage (diodes+smoothing caps) after the main AC supply.
the difference being that the star ground configuration is only applied to the first filtering stage (diodes+smoothing caps) after the main AC supply.
One thing I forgot to mention that is will be a design with seperate 0V return wires, from headphone transducers all the way back to the power supply star ground/ground on the amp board, depending on what groung scheme I end up choosing.
Output terminal will be 4-pole XLR.
Cross talk should be low.
Output terminal will be 4-pole XLR.
Cross talk should be low.
Last edited:
It doesnt hurt and down the road if you want to go balanced your headphones has already been recabled for it.
Unless it is a really really bad design, crosstalk is not that big a deal but I sleep better knowing that I have done something to improve it.
Idea 4 :
A cross between Idea 2 and 3.
One single ground plane as in idea 3 but I kept the input/feedback/Pot/0V return wire ground as an isolated island but connected to the rest of the power ground plane. This connection is done where the output from the buffer goes across the cutout in idea 2.
Basically, some "clever" partitioning of the power ground plane means you can get a quiet ground island, shielding for the input, output and feedback nodes of the opamp, no signals are crossing a cutout equals no large current loops, you can connect 0V reurn wire directly to that quiet ground island and the whole power ground plane only needs to be connected to the power supply through the power input terminal ground wire.
Cant see any big issues here.
Any potentially major issues I have overlooked?
A cross between Idea 2 and 3.
One single ground plane as in idea 3 but I kept the input/feedback/Pot/0V return wire ground as an isolated island but connected to the rest of the power ground plane. This connection is done where the output from the buffer goes across the cutout in idea 2.
Basically, some "clever" partitioning of the power ground plane means you can get a quiet ground island, shielding for the input, output and feedback nodes of the opamp, no signals are crossing a cutout equals no large current loops, you can connect 0V reurn wire directly to that quiet ground island and the whole power ground plane only needs to be connected to the power supply through the power input terminal ground wire.
Cant see any big issues here.
Any potentially major issues I have overlooked?
Attachments
Last edited:
A better look at it.
On the bottom where the opamp is mounted there is a cutout in the ground plane around the opamp.
The ground island is on all 4 layers, the power ground plane is on 3 of the layers(not on the power plane layer). Vias should be used to stich the 4 ground island planes together as well as stitch the 3 power ground planes together.
It should end up being a very low impedance plane with good shielding of the important parts of the circuit and hopefully low overall distortion and noise.
On the bottom where the opamp is mounted there is a cutout in the ground plane around the opamp.
The ground island is on all 4 layers, the power ground plane is on 3 of the layers(not on the power plane layer). Vias should be used to stich the 4 ground island planes together as well as stitch the 3 power ground planes together.
It should end up being a very low impedance plane with good shielding of the important parts of the circuit and hopefully low overall distortion and noise.
Attachments
I can not see much issue in leaving the ground return wire where it is. It makes wiring your amp easier and I do not think it will make much of a difference.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.