Amplifier radiative feedback - not mentioned anywhere?

[Fulcrum]

Hi everyone.
I was thinking about unwanted sources of instability in amplifiers. One source that I cannot seem to find any sources on (hah) is radiative feedback, i.e. when the amplifier output is picked up by the input via electromagnetic radiation. This could easily happen if the gain is large, traces long, and frequencies high, and I have had problems with it before. But a lot of googling later, nothing is showing up? Is this just completely ignored in amplifier design guidelines or have I just missed it? If anyone has some good sources I would love to see them.

[chris_leyson]

I guess it's not mentioned under amplifiers as it's all down to good layout and wiring practice. I once had to fix a pcb that was used to drive diesel injectors in a lorry and at some speeds, RPM from a tachometer, the thing would go crazy and keep on supplying more and more fuel. It turned out that the input track from the tachometer was running parallel to the output track going to the injector driver and was picking up false tachometer pulses, had to drop the slew rate on the output track to fix it.

[Fulcrum]

I guess it's not mentioned under amplifiers as it's all down to good layout and wiring practice. I once had to fix a pcb that was used to drive diesel injectors in a lorry and at some speeds, RPM from a tachometer, the thing would go crazy and keep on supplying more and more fuel. It turned out that the input track from the tachometer was running parallel to the output track going to the injector driver and was picking up false tachometer pulses, had to drop the slew rate on the output track to fix it.

I feel like it's something that should be mentioned more, with guidelines as to how to avoid it. It's a shame if you design a high-gain high-frequency amplifier and it oscillates wildly because you didn't know how to estimate the radiative feedback.

[DimitriP]

You are looking for something like this one: 
Signal Integrity Issues and Printed Circuit Board Design by Douglas Brooks
ISBN-13: 978-0133359473
ISBN-10: 0133359476

for about $100

[Circlotron]

Would have been much more of a problem in the days of valve amplifiers, with several inches of output stage anode swinging hundreds of volts, and low level circuitry at generally a much higher impedance than solid state stuff, so much more likely to couple a signal from output back to input.

[tszaboo]

Because there is no significant electromagnetic radiation, at least not with the sub 10 Mhz region. That radiation is generated by electric currents and how much current flows into an amplifier? Or through the ~kiloohms of resistances around it? Almost nothing. What you want to search for is capacitive coupling, ground currents and leakage currents.

[dmills]

There speaks someone who clearly does not play with modest power amplifiers in the microwave bands....

With multilayer boards and the wide use of microstrip and stripline this is not the pain it once was, you can do VHF and even low microwave without needing loads of silver plated boxes and plumbing fittings, time was that would just not work.

The real gotcha is that it can become a problem a long time before it causes things to burst into oscillation, particularly in filter designs, if you are trying for 100dB or so of stopband attenuation then poor layout can stop you getting there at just a few tens of MHz. Build some receivers with lowish first IFs if you want the full experience. 

I do not miss the need for neutralisation caps around 4CPX250s and the like.

Regards, Dan.

[T3sl4co1l]

I've probably got an instance here,



It's a ~300MHz BW amp in three stages, but it sings at ~1.8GHz.  Holding my hand over it seems to quench it.  It could be near-field capacitance coupling, or it could be a kind of evanescent (surface) wave carried over the board, who knows.  Probably not in-board coupling (the ground plane is solid and well stitched, and the supply is well filtered).

Circuits are usually small enough that true radiation isn't applicable.  Rather, it's better expressed as electric or magnetic coupling.  As you push into the GHz, it does get harder to keep the circuit in its own near-field (or, say, into the 100s of MHz with ye olde tubes), and you'll see more standing wave and radiation effects.

Tim

[VK5RC]

@TeslaCoil, re the circuit, the main trace layout looks a bit like a microwave stripline filter. All those (presumed) caps should kill it though.

[KJDS]

The only time I've ever had it happen was a multistage amp at 35GHz.

[Seekonk]

In college I used to sell and repair stereo equipment.  Someone brought me a Southwest Technical amplifier to repair. If you are old enough you remember Southwest Technical kits. They used whatever parts they could get cheap. It never worked right after he built it.  I used my FM tuner as an audio source and noticed the S meter would peg.  They had a ground trace in a C around the outside of the board.  Input was at one end and output at the other, Ground for everything was in the middle.  Perfect oscillator coil.  Moved the ground and it was just fine, It was where the kit instructions said to put it. They must have gotten a buy on some hot RF transistors.

Was working at a R&D center and they had a FET accelerator amp they built. This was back when a FET cost $30.  It oscillated and they had to hang a .47uf on the output to common to get it to stop. I lifted the FET gate lead and put a 100 ohm resistor in series with it, problem solved.

[Fulcrum]

Lots of good replies here.
I started thinking about this type of feedback coupling due to a circuit I built that would oscillate strongly based on the orientation of a photon detector (SiPM, 6x6mm, 1280pF terminal capacitance) on the first amplifier input. The amplifier was a ~60dB 2. order narrow bandwidth filter centered around 80MHz. It would oscillate wildly at ~650MHz when the detector was normal to the amplifier output  trace (which was a good 3 cm away from the first amplifier stage and no longer than 6mm). When it was parallel to the trace, the oscillations would stop.
The only thing I could think of that could cause this was some sort of radiative feedback.

[tszaboo]

There speaks someone who clearly does not play with modest power amplifiers in the microwave bands....

Yes. But at least I can read the entire post that I'm replying to.

[G0HZU]

A simple way to demonstrate this effect is to make a multistage amplifier that provides lots of gain over a very wide bandwidth. Often the amplifier will behave itself until it is put in a box and the lid fitted.
The lid can help support a reflective feedback path from output to input and the oscillation frequency will depend on how the lid is moved or flexed. A quick and dirty fix here would be to stick a thin slice of some modern RF absorber material inside the lid. However, the best performing absorber material can be very expensive. Sometimes it can cost much more than the whole PCB/components if you need something that covers a lid area of (say) 5cm x 10cm.