The cap is there to provide coupling between the feedback ground and the ground side of the zobel (8R+100nF) at the amp output. The zobel (boucherot cell) goes straight from the amp output to the board local decoupling, which satisfies the first requirement for stability. The second requirement is that the feedback ground is not too far from the zobel ground. The second requirement is not hard to satisfy normally.
The capacitor couples the zobel ground to the feedback ground at RF, because in moving the feedback ground off-board we separated the two points by a significant stretch of wire.
The capacitor couples the zobel ground to the feedback ground at RF, because in moving the feedback ground off-board we separated the two points by a significant stretch of wire.
One way of thinking about this that may help is that instead of allowing PSU returns to cause noise voltages at points along the ground bus, we move the feedback loop off board so that the local ground differences become subject to the amplifier's PSRR as though they were ripple. Why shouldn't PSRR apply to ground too? Ground noise is far smaller than ripple so the amp should swallow it up no problem.
Last edited:
To get a working understanding you really need to see various noise scenarios in action and which steps were taken to fix them. I remember a powerpoint somewhere like that.
Here is a good one:
https://centralindianaaes.files.wordpress.com/2012/09/indy-aes-2012-seminar-w-notes-v1-0.pdf
Here is a good one:
https://centralindianaaes.files.wordpress.com/2012/09/indy-aes-2012-seminar-w-notes-v1-0.pdf
You normally run a 2-5 nF cap from the incoming signal ground at the input connector directly to the chassis. This makes the screen and the housing a single enclosure at RF.
I don’t see how the 100N in your diagram can have a lower inductance than the piece of wire already connecting the two. If it does, I would urge you to look again at the physical layout and reduce the loop areas.
Further, if you have any OPS instability, you could inject that into the input assuming the wire connection XL was higher than the the 100N Xc
I don’t see how the 100N in your diagram can have a lower inductance than the piece of wire already connecting the two. If it does, I would urge you to look again at the physical layout and reduce the loop areas.
Further, if you have any OPS instability, you could inject that into the input assuming the wire connection XL was higher than the the 100N Xc
That’s why you use a star or T ground system and never mix signal and power grounds. No need to move the feedback off the board!
It doesn't have to have a lower inductance, it just has to be more conductive than the 10R resistor at the important frequencies. When I said "not too far from zobel ground" that was not quite meant literally, just that the voltages should not be too different at RF. The capacitor is really just an AC link. It doesn't carry significant current.
The feedback components don't really need to be moved off the board (that's not what I meant), the routing works like in my diagram in post 34. That way the 100nF cap has a very short connection to feedback ground.
The feedback components don't really need to be moved off the board (that's not what I meant), the routing works like in my diagram in post 34. That way the 100nF cap has a very short connection to feedback ground.
Last edited:
He has not shown any RFI protection. However, even in a practical implementation of his layout, you would connect a small value cap from the input screen to the chassis right at the input. This would make sure that no matter what the inductance to the system ground was, at RF it would be a short to the chassis, so maintaining the single enclosure principle.
Better to start with a physical implementation (Self for example) and try to evolve it to -110 dB.
The problem is you cannot sim the loop areas - it’s better to build and experiment from there if you want to develop a new approach.
Would your cap solve anything that a good layout using any of the existing techniques would not?
Even Self’s solution will only work (as is the case with any other) if the loop areas were well Controlled.
The problem is you cannot sim the loop areas - it’s better to build and experiment from there if you want to develop a new approach.
Would your cap solve anything that a good layout using any of the existing techniques would not?
Even Self’s solution will only work (as is the case with any other) if the loop areas were well Controlled.
In trying to follow along here, I have one simple question --
Is Rod Elliot's ground loop isolator/Nelson Pass's diode bridge with CL-60 considered an HBR?
Which seems to mean that the zero volt line from the CT ( and power supply common ) should be connected to the earth side of this device and the amp grounds on the other. Is that so?
Is Rod Elliot's ground loop isolator/Nelson Pass's diode bridge with CL-60 considered an HBR?
Which seems to mean that the zero volt line from the CT ( and power supply common ) should be connected to the earth side of this device and the amp grounds on the other. Is that so?
Have you looked at Bonsai's article here? http://hifisonix.com/wordpress/wp-content/uploads/2019/02/Ground-Loops.pdf
The ground loop isolator breaks the loop between earthed pieces of equipment. The HBR discussed in this thread is to break cross-channel ground loops, both are explained in the article
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.