Op Amp Bypassing

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

bluebird

Well-known member
Joined
Jun 11, 2004
Messages
1,070
Location
Los Angeles
Talking about NE5532's lately, reminded me that Douglas Self recommends connecting the V+ with V- via a 0.1uF cap rather than using two caps from V+ to GND and V- to GND.

Something about powering up the circuit, I can't remember. Have I been wasting one 0.1uF cap for years? Anyone know whats up with this?
 
to me using a single cap from V- to V+ is being cheap, its been long recognized that a 100nF cap in each rail is the way recommended by most manufacturers, some opamps need another cap in parallel to become stable. The only benefit I see from using a cap from V+ to V- is that you are not dumping dirty current to ground.
 
bluebird said:
Talking about NE5532's lately, reminded me that Douglas Self recommends connecting the V+ with V- via a 0.1uF cap rather than using two caps from V+ to GND and V- to GND.
NE5532 is not a high speed op amp and so bypass caps are not as critical. If your ground is low impedance and quiet and the amp is not sourcing / sinking a lot of current, you actually don't need any bypass caps at all with most older amps. But adding one cap between the rails could be an effective compromise to add a little stability.

Note that at vaguely high frequencies (like > 4kHz), ground currents follow the supply currents where possible. This is counter intuitive. You might think that return currents would simply follow the shortest path to the supply bypass cap ground. If they did, no bypass caps at the OA would be necessary. But high frequency AC currents actually reduce the impedance of the ground plane directly underneath the supply trace. And because current always follows the path-of-least-resistance, high frequency ground currents will return directly underneath the supply traces. This slight increase in AC impedance in ground can cause instability in high speed op amps. That is why becomes necessary to add local bypass capacitors. If you used only one cap between supply lines, it would not have the same effect because it would have no impact on the impedance of local ground.
 
There is an old article from 90s in wireless world about the 1 cap solution.  That article suggested the noise is injected into ground needlessly when 2 caps to ground is used .  1 cap across v+  and v- creates less noise in the amp and accomplished better results. 
 
Note that at vaguely high frequencies (like > 4kHz), ground currents follow the supply currents where possible. This is counter intuitive. You might think that return currents would simply follow the shortest path to the supply bypass cap ground. If they did, no bypass caps at the OA would be necessary. But high frequency AC currents actually reduce the impedance of the ground plane directly underneath the supply trace.

Interesting, thank you.  Do they have cad software that during layout, return paths of A/C can be simulated?
Curious what the shift looks like around 4kHz.  Assuming a circuit made for audio, how best should signals 'shift' between path of least resistance / drain follows source? Is it desired that the change be a narrow-banded one, dependant on layout?  Can one mitigate/optimise such things???  I realise this is a bit off the subject, my bad BB/SW.

Edit: btw I get there are techniques such as no greater angles trace cornering than 45deg. But you've got me imagining there's some kind of no-man's-land where overall parts placement and the shape of floods become critical. 

 
Not that I'm aware of. At low frequencies the return path is just the shortest path. Higher than ~4kHz return currents travel directly under supply lines. So you could make a point of not breaking the ground plane under the trace because that would force the return current to go around and possibly pollute some other part of the circuit. But if you use local bypass capacitors it probably wouldn't be a problem regardless. There's a great video on youtube about it but I have no idea as to how to go about finding it now. I'm not familiar with any issues regarding traces with angles >45deg.
 
squarewave said:
Not that I'm aware of. At low frequencies the return path is just the shortest path. Higher than ~4kHz return currents travel directly under supply lines. So you could make a point of not breaking the ground plane under the trace because that would force the return current to go around and possibly pollute some other part of the circuit. But if you use local bypass capacitors it probably wouldn't be a problem regardless. There's a great video on youtube about it but I have no idea as to how to go about finding it now.

Not audio specific but this is likely the video you are thinking of - https://www.youtube.com/watch?time_continue=1880&v=1xicZF9glH0&feature=emb_logo

30 mins but well worth watching.

 
That's actually not the one. He's using a 1M crystal. The one I'm referring to used like a PWM chip to make a relatively low frequency pulse into a 50 ohm load. But the board layout with the signal going around the edge is the same. Very similar but not the same.
 
That's actually not the one. He's using a 1M crystal. The one I'm referring to used like a PWM chip to make a relatively low frequency pulse into a 50 ohm load. But the board layout with the signal going around the edge is the same. Very similar but not the same.

This video?
 
This video?
Nope. But it's basically the same. I'm not sure he's 100% correct about it though. It took a while for him to get to the point but ~5:30 is where he describes high frequency noise following the path under the trace to the bypass cap at the chip because the ground plane under the trace is lower inductance at 1kHz+. But then he says that without that cap the currents would follow a path back around the edge of the board anyway? That would not be my prediction. I would think that without the bypass cap, the high frequency currents are going to go straight up to the electrolytic and leave an unbalanced current at the measurement point and that's why there's a difference in noise.

Nice scope though. Makes my scope look like pong on an atari.
 
Interesting, thank you. Do they have cad software that during layout, return paths of A/C can be simulated?
I wish... Decades ago I killed a lot brain cells thinking about all AC current flows within and into/out of a single console input strip. In concept keeping these currents netted out to zero should help manage sources of crosstalk. After several beers I gave up and moved on.

JR
 
You could use an EM field solver for simulation, they can provide a heat map of current densities. But they are very expensive, outside the realm of typical DIY.
 
Back
Top