Hi,

For various gain clone-like circuits it is noted that you should use an as short as possible feedback loop to prevent oscillation. In my case I have an LM1875 and the application note suggests to put some filter capacitors as close as possible to the supply pins. When I do this I get an awkward PCB layout but when I look at the example layout in the application note the designer was pretty lenient about the 'as close as possible' bit yielding an overall much nicer looking layout.

Is the oscillation mechanism that is to be prevented in the feedback loop and near the supply pins the same? Is it caused by the trace capacitance and/or impedance or something different? I'm interested in a formula that roughly models this effect so I can get a feel for what happens with for example trace lengths of 1, 10, 100 mm respectively.

 

Hi,

For various gain clone-like circuits it is noted that you should use an as short as possible feedback loop to prevent oscillation. In my case I have an LM1875 and the application note suggests to put some filter capacitors as close as possible to the supply pins. When I do this I get an awkward PCB layout but when I look at the example layout in the application note the designer was pretty lenient about the 'as close as possible' bit yielding an overall much nicer looking layout.

Is the oscillation mechanism that is to be prevented in the feedback loop and near the supply pins the same? Is it caused by the trace capacitance and/or impedance or something different? I'm interested in a formula that roughly models this effect so I can get a feel for what happens with for example trace lengths of 1, 10, 100 mm respectively.

Did you include the 0.1uF caps in parallel with the electrolytics?

 

Here's a calculator for trace inductance.

One good layout technique for decoupling capacitors is to place the pad for the capacitor in series with the trance to the power pin on the IC, not offset from the power trace. That minimizes the capacitor stray inductance.

 

Here's a calculator for trace inductance.

One good layout technique for decoupling capacitors is to place the pad for the capacitor in series with the trance to the power pin on the IC, not offset from the power trace. That minimizes the capacitor stray inductance.

The TS needs to get the 0.1uF caps in close to the pins, the electrolytics are a little more forgiving.

 

Did you include the 0.1uF caps in parallel with the electrolytics?

I did and these are the caps that the datasheet talks about.

Here's a calculator for trace inductance.

So if I make an unreasonably large feedback loop but with no ground plane in sight there wouldn't be any problem? Specifically these two oscillation problems can be avoided by not having parallel traces (if possible)?

 

I think you will get a much more insightful opinion of you ask this question of an applications engineer at TI.

 

Audio power amp chips are crazy-sensitive about power supply decoupling. All of the advice above is good to follow. Also, know that app note pc board layouts are tweaked to make the chip perform at its best. When in doubt, copy them down to the 0.1 mm.

ak

 

First, if you connect one side of a decoupling cap to a power pin of an IC and connect the other side to a thin ground trace you have done nothing to solve your noise or an unstable power problem. The first thing you have to have is a low impedance (low inductance) ground plane. Without a ground plane, decoupling capacitors will do you little good, no matter how close they are placed to the pin.

 

Audio power amp chips are crazy-sensitive about power supply decoupling. All of the advice above is good to follow. Also, know that app note pc board layouts are tweaked to make the chip perform at its best. When in doubt, copy them down to the 0.1 mm.

ak

BTL amplifiers draw supply current in the same direction regardless which way its passing through the load, so they can use less supply decoupling - but you still need close attention to the small low ESR capacitors in parallel with the electrolytic.

 

So if I make an unreasonably large feedback loop but with no ground plane in sight there wouldn't be any problem? Specifically these two oscillation problems can be avoided by not having parallel traces (if possible)?

I see that this is all wrong because a ground plane would mitigate inductance problems. After playing with some of the formulas I can see how you might create an 30 nH inductor when traces are too long. Here it's stated that an inductance of this magnitude could cause oscillations.

When I look at the self-resonant frequencies of various ~30 nH inductors they are in the high MHz to GHz range. Might this be the frequency where the oscillation will occur if the op-amp has sufficient gain at that frequency? Or is my guess work pure nonsense?

 

Audio opamps can't process frequencies in the multi-MHz, but that doesn't stop the little critters from trying. One of the possible causes is positive feedback from the load through the power pins. These chips can move a lot of current, which can cause small IxR voltage drops in the traces going to the power pins. Combined withhigh gain-bandwidth products, this makes then extra sensitive to poor decoupling techniques. It is called de-coupling for a reason, as in preventing coupling of the load current to the source voltage.

ak

 

Audio opamps can't process frequencies in the multi-MHz, but that doesn't stop the little critters from trying. One of the possible causes is positive feedback from the load through the power pins. These chips can move a lot of current, which can cause small IxR voltage drops in the traces going to the power pins. Combined withhigh gain-bandwidth products, this makes then extra sensitive to poor decoupling techniques. It is called de-coupling for a reason, as in preventing coupling of the load current to the source voltage.

ak

When I rewired a TV sound sub-panel to be a 6MHz FM transmitter, I forgot a shunt capacitor to keep the RF out of the input to the audio power chip that I'd repurposed to drive the bank of varicaps. The chip got seriously hot and cracked in two.

Eventually I moved on to using a 74LS629 VCO to produce FM, it didn't need a power amplifier to drive the frequency control pin.