parasitic inductance problem my PCB design

Thread Starter

DaveH

Joined Jan 1, 2009
53
I'm looking for some guidelines.

I have 15 mil trace that runs parallel to another 15 mil trace for about 1 inch and there is only about 10 mil separation between them.

One of these lines is power rail which has about 25mA through it, but according to my simulation it's oscillating by a fraction of a milliamp at a few kilohertz. Simulations are one thing but I bet for real the oscillation is more severe.

Unfortunately the other line is a signal line, which is a disaster! I understand the theory pretty well, the flux lines will be normal to it, it depends on distance and the flux density rate of change (and loop area I think), but anyway, my PCB is no good and has to be redesigned because the parasitic coupling is really bad.

What do the guys think here when you're doing a line level audio pcb, what clearances between these kind of tracks do you observe in your layout etc.

The thing is this kind of thing you can't spot the problem in breadboarding and sure won't show up in your SPICE, you can only spot and avoid stuff like this with experience.

When I redesign my PCB I'm not sure how far apart to move the power rail from the signal rail and I'm wondering what if the power rail runs close to just one through hole soldered joint which is of course just a small circle of 3 mm, I wonder if that will pickup the contamination too.

I can only tell by redoing the PCB and then even my next PCB may still have the problem. I'll be happy to hear any tips from you all.
 

KL7AJ

Joined Nov 4, 2008
2,229
I'm looking for some guidelines.

I have 15 mil trace that runs parallel to another 15 mil trace for about 1 inch and there is only about 10 mil separation between them.

One of these lines is power rail which has about 25mA through it, but according to my simulation it's oscillating by a fraction of a milliamp at a few kilohertz. Simulations are one thing but I bet for real the oscillation is more severe.

Unfortunately the other line is a signal line, which is a disaster! I understand the theory pretty well, the flux lines will be normal to it, it depends on distance and the flux density rate of change (and loop area I think), but anyway, my PCB is no good and has to be redesigned because the parasitic coupling is really bad.

What do the guys think here when you're doing a line level audio pcb, what clearances between these kind of tracks do you observe in your layout etc.

The thing is this kind of thing you can't spot the problem in breadboarding and sure won't show up in your SPICE, you can only spot and avoid stuff like this with experience.

When I redesign my PCB I'm not sure how far apart to move the power rail from the signal rail and I'm wondering what if the power rail runs close to just one through hole soldered joint which is of course just a small circle of 3 mm, I wonder if that will pickup the contamination too.

I can only tell by redoing the PCB and then even my next PCB may still have the problem. I'll be happy to hear any tips from you all.
Hi Dave:

Parasitic inductance MAY not be a problem if you thoroughly bypass the LOAD end of the trace.

Another alternative is to FEED power to the trace from a different location than normal! This isn't always feasible, but sometimes you can "massage' troublesome oscillations to a location where they aren't so troublesome by such methods.
 

bountyhunter

Joined Sep 7, 2009
2,512
I'm looking for some guidelines.

I have 15 mil trace that runs parallel to another 15 mil trace for about 1 inch and there is only about 10 mil separation between them.

One of these lines is power rail which has about 25mA through it, but according to my simulation it's oscillating by a fraction of a milliamp at a few kilohertz. Simulations are one thing but I bet for real the oscillation is more severe.

Unfortunately the other line is a signal line, which is a disaster! I understand the theory pretty well, the flux lines will be normal to it, it depends on distance and the flux density rate of change (and loop area I think), but anyway, my PCB is no good and has to be redesigned because the parasitic coupling is really bad.

What do the guys think here when you're doing a line level audio pcb, what clearances between these kind of tracks do you observe in your layout etc.

The thing is this kind of thing you can't spot the problem in breadboarding and sure won't show up in your SPICE, you can only spot and avoid stuff like this with experience.

When I redesign my PCB I'm not sure how far apart to move the power rail from the signal rail and I'm wondering what if the power rail runs close to just one through hole soldered joint which is of course just a small circle of 3 mm, I wonder if that will pickup the contamination too.

I can only tell by redoing the PCB and then even my next PCB may still have the problem. I'll be happy to hear any tips from you all.
The only concern is capacitive coupling between traces. If you leave at least .050" or more air gap, I believe the capacitance will be negligible.

Trace inductance is reduced by making the traces as wide as possible.
 

studiot

Joined Nov 9, 2007
4,998
PCB inductance at audio frequency at 25mA current?

I go with BH thinking capacitive.

You could try breaking the power track, and reconnecting through a choke wound with thick wire and a ferrite.
 

Thread Starter

DaveH

Joined Jan 1, 2009
53
I'd like to thank you all for your tips. I don't know why on earth (no pun intended) I started to think it was inductive coupling, something was wrong with my thinking.

After a lot of work it definitely looks like capacitive coupling for sure. In fact in my pcb I've unwittingly made some pretty damn good capacitors between my negative power rail and signal lines. I was stupid enough to have long parallel signal traces running next to the negative rail with probably 6 mil gap - I've learnt my lesson.

Luckily for my +ve power rail which also has some bad audio frequency transients, I designed a guard rail trace (on the same net as earth) that protected all my signal traces, when checked with the scope, though the positive rail had some bad audio frequency transients it was not capacitively coupling with signal traces at all, thank god

This guard rail technique I picked it up on some consultants web page and I think it's very good idea, easy to implement and save your pcb design being scrapped otherwise. Does anyone else use that?

Also studying my circuit closely for a long time, I've learnt that though there are bad audio frequency transients on the +ve and -ve power rails, these transients are not necessarily contaminating the output from the op amps that are powered by these rails. The op amps seem happy to tolerate bad behaviour from the power supply rails, the output doesnt seem to reflect that - luckily for my design. What the op amps can't do is stop the their audio signal output from being contaminated when the problem is that the input signal gets capacitively coupled due to a terrible pcb design.

so there you go!

And when somebody here suggested sticking inductors on the power rails, I thought that's a great solution. However when I did a simulation run I found that to get the transients down by a factor of 10, I'd need an inductor of 30 mH (pretty damn big). Now these things are available, I even found 60mH and 100mH in small 5mm pitch radial cans but the thing with these size of inductance is that they have high dc resistance of > 100 ohms and also that can't really cope with a 25mA constant current - no use for power rail. I've decide I'm gonna leave my power rails alone even though they have these spikes.
 
Last edited:

John P

Joined Oct 14, 2008
2,025
If you really did your board right, you'd use multiple layers with ground and power planes.

As a next-best, you can use the biggest ground plane you can, and put all the sensitive parts of the circuit over it, and have short branches bringing power to the places where you need it, with the basic power conductors elsewhere. Filter capacitors should have traces of minimal length connecting to the ground plane through vias.
 
Top