oscillation - RF Cafe Forums

 Post subject: oscillation

Posted: Fri Jan 29, 2010 12:22 am 

I have been testing an L-band front-end with a 2 port RF switch followed by a lna, phase shifter and attenuator. I get the gain that I expect out of the board, but when I hook it up to a spectrum analyzer (with the input of the board terminated in a load), I see some low level oscillations (<-75dBm) starting at a few KHz, all the way up to 2-ish GHz and then goes quiet after that. I've been trying to find the source of the oscillation and havent yet found it. When I switch ports on the RF switch, the amplitude of the oscillation (and some of the peaks) change. The only time the oscillation goes away is when I remove the dc blocking cap between the switch and LNA, leaving the input of the LNA open. I've played around with the bias & blocking caps and inductor on the LNA and even bypassed the on-board voltage regulators/power conditioning to rule noise on the power lines out. I am using 2 pins on a PIC microcontroller to directly toggle the switch control lines.

I would appreciate any guidance on what to do next.

Karthik

 Post subject: Re: oscillation

Posted: Fri Jan 29, 2010 4:32 pm 

Low-level and covering a large range of frequencies:

1. Doesn't sound like LNA instability - that would almost certainly be bigger in amplitude, unless your LNA has an unbelievably bad IP3.

2. Disappears when RF switch is disconnected - can't be bias related, as the capacitor blocks DC.

3. RF switch is driven from a PIC - which continues to run.

Sounds like you're getting a signal coupled into the RF path from the PIC via the switch.

First thing to do is to put a scope probe on the control leads of the switch, AC coupled, and set to the highest possible sensitivity. Any noise, and you've found your culprit.

To test this hypothesis in-circuit, disconnect the two leads from the PIC to the switch, drive them to the desired voltages <i> with a clean source </i> (not the PIC supply voltage!), but leave the capacitor between the switch and the LNA.

Good luck!

 Post subject: Re: oscillation

Posted: Fri Jan 29, 2010 4:44 pm 

Hi fred,

just did that...cut the traces from the pic and soldered jumpers hooked to a power supply...did not change a thing. Im beginning to wonder if the ground pad on the lna under the chip isnt soldered properly. But then, i have a hundred boards, and all seem to be behaving similarly. Is there any way to check for a floating ground without pulling everything apart?

Karthik

 Post subject: Re: oscillation

Posted: Fri Jan 29, 2010 4:51 pm 

Is there any chance that noise could be coupled in via the power pin of the RF switch?

(I'm assuming you have a solid ground plane under the whole thing.)

Another thing to try would be lifting the power to the PIC on your test board. (That's a subset of "try to find where the source of the noise is"...) Its on-chip oscillator may be affecting things.

As far as the ground is concerned, you haven't given many details, so I can't help much there.

good luck - and keep us informed, this sounds like it might get interesting!

 Post subject: Re: oscillation

Posted: Fri Jan 29, 2010 5:00 pm 

The only power going into the switch is through the control lines. As for the ground plane, the LNA is a qfn package with a ground paddle on the lower surface...I dont know how to verify if that ground connection has been made. The board layout does have that pad connected to gnd, but.....

So far, the only time the oscillation/noise died away was with the dc blocking cap disconnected.

 Post subject: Re: oscillation

Posted: Fri Jan 29, 2010 5:08 pm 

Where I was going with removing the power from the PIC is that the current leaving the ground pin of the PIC might be coupling into the switch via a common impedance in the ground return.

Paddles on the bottom of a chip are sometimes a problem - but if you have multiple vias on the PCB under the paddle, I think you don't need to worry about the ground connection. Still, if your LNA chip has a pin or pad which isn't the paddle, you should check there. (Not that I think that's likely - you would already have checked that, I think).

 Post subject: Re: oscillation

Posted: Sat Jan 30, 2010 9:08 pm 

fred,

As you suggested, I hooked the oscilloscope on the switch control lines, and there was noise on the lines. Here is the part that I dont have an answer for. When I disconnected the switch control lines from the PIC and left the lines floating, there was noise on the output dc blocking cap of the LNA! When I connected the control lines to a 'clean' source, the noise went away - both on the output of the LNA and the 'oscillation'. Now, this worked on one board. I need to repeat this on another board/boards to make sure this is in-fact the solution..

Thanks for the help..Truly appreciate it

 Post subject: Re: oscillation

Posted: Mon Feb 01, 2010 12:59 pm 

Hi Karthik,

When you say "noise on the lines" going to the RF switch, is that random noise or a periodic signal?

Just curious!

 Post subject: Re: oscillation

Posted: Mon Feb 01, 2010 1:32 pm 

It looks like a 1KHz sawtooth riding on top of the control pulse. And it is 180 deg out of phase between the 2 control lines (though they are on different outputs of the PIC). Other output lines on the PIC are clean. And this noise is visible on the PIC even when the switch is not connected to it....might need either a pullup on the PIC or a cap to filter it out...atleast that is the thinking right now...till im proved wrong

 Post subject: Re: oscillation - update

Posted: Fri Feb 05, 2010 10:21 pm 

I've spent the last week running more tests. I find that cleaning up the noise on the switch control lines removed most of the 'oscillation' on the RF side. Most of it, not all. I eliminated the noise using a series resistor and a cap to ground on each control line. I have 2 switches hanging on the line with the control lines Tee'd off from the PIC to each switch. The cap to ground is on each control line of each switch, while the resistor is common to both switches on each line. The control lines on the 2 switches are interchanged. By that, I mean, the control lines on switch 1 is connected as A-B and on switch 2 it is B-A. That way, when I send A=1, B=0, switch 1 looks at port A and switch 2 looks at port B and vice versa. Now, after cleaning up the control line (with a series resistor and a cap to ground), A=1,B=0 (switch 1 looking at port A, switch 2 looking at port B) is clean, but A=0,B=1 (switch 1 looking at port B, switch 2 looking at port A) is still noisy on both switches. Both switches are looking at identical loads. When I remove the input cap (on the noisy port) on the switch and place a 50ohm resistor to ground, the noise goes away. When I put the cap in, the noise returns. Is this an impedance mismatch issue? If it is, why is it oscillating on one port and not another? The microstrips are as identical as they can be on both switch ports. Why would this happen? The noise on both switch line look identical (in the sense that it does not appear as though there is periodicity on either line) and it is fairly small <5mVp-p. Any suggestions will be greatly appreciated