Amp is oscillating Help

I designed this circuit board and was warned the 5 Ghz smt transistor might oscillate. It does! I thought I was doing the right thing when I opened up the ground plane under the transistor. Now I find that a wire across the opening I made stops the oscillation. I just layed the wire across the open area, I haven't soldered the wire yet. ( I hope that doesn't bring a surprise) Here's the PCB (Top left), red circle shows where smt transistor is mounted. (Top right) shows opening I made in ground plane. (Lower Right) shows single wire across open area that stops oscillation.

formatting link

Any one want to tell me what I might have done different, and what I should do now. Oh, I thought putting it in the metal box would stop the oscillation, it got worse. Hmm.. maybe my (wire) will not solve the problem when I put it in the box. (Only need 10 Mhz band width.) Thanks, Mikek

Reply to
amdx
Loading thread data ...

Could you post a schematic as well? Or at least tell us what configuratoin you're using the transistor in? (And we're talking BJT here, right? -- Not MOSFET?)

I'm guessing the added wire gets you enough extra capacitance from base to ground to drop the loop gain below one, but who knows? I suggest putting a chunk of copper tape over the area and chunk to make sure that's stable as well.

I'm not surprised the metal box actually made things a bit worse! If you line it with microwave absorber material, it might help some -- but since it oscillates in free space, it probably wouldn't stop it altogether.

---Joel

Reply to
Joel Koltner

Why the heck did you think that? I assume you were trying to increase the bandwidth instead of increasing stability?

Reply to
Jeff Johnson

Never, ever split a ground plane unless safety regs force you to. That's what I've been preaching for over two decades now :-)

Well, why put a 5GHz transistor in there then? Isn't a 300MHz or maybe

800MHz type enough? Through-hole parts with such long traces also present a risk here. These transistors can oscillate at frequencies you might not even see and appear to otherwise function. Until somebody else notices ...
--
Regards, Joerg

http://www.analogconsultants.com/
 Click to see the full signature
Reply to
Joerg

Slotline resonator. Cool.

Looks like the emitter is not grounded, which encourages oscillator demons. The schematic would be helpful.

Use a slower transistor, or a base resistor.

John

Reply to
John Larkin

Oh, if you're holding the wire down, maybe it's your finger that's killing the oscillation.

John

Reply to
John Larkin

Ya, silly me second picture down.

formatting link

Yes, transistor. Mikek

Reply to
amdx

I should restate that, I designed the PCB not the circuit. I don't know why the designer used this transistor, probably what he had. Mikek

Reply to
amdx

No, but that's how it started, I noticed the signal cleaned up when I put my finger on it. Then I layed a piece of wire on it and it looked good. I did it several times, because it didn't seem like it would be enough material to stop the oscillation, I'm not sure it is, once I put it in a box. Mikek

Reply to
amdx

BTW, I have not built the fet input circuit yet, just the parts to the right of the shielded area on the schematic.

formatting link
Mikek

Reply to
amdx

Sorry I missed that sermon, do you consider my opening a split ground plane? Mikek

Guess I just wasn't smart enough to make the change. Any reason a 2n2222 or a 2N4401 wouldn't work? Mikek

Reply to
amdx

Yep. It would be ok if the traces weren't so long but nothing you can do about that right now.

A 2N4401 should be fine. Or use another of those BC547 you've go in the output stage.

--
Regards, Joerg

http://www.analogconsultants.com/
 Click to see the full signature
Reply to
Joerg

Or another BC547A as long as was already using one. Mikek

Reply to
amdx
[...]

Yep. I guess we wrote that at the same time :-)

--
Regards, Joerg

http://www.analogconsultants.com/
 Click to see the full signature
Reply to
Joerg

I went ahead and soldered a couple of wires across the open area, and that stopped the oscillation. However when I put the PCB in a metal box the waveform distorts and depending an position in box it may oscillate. Unless I get another fix, I'll change the transistor. Mikek

Reply to
amdx

Old adage: "All the amplifiers I build oscillate. All the oscillators I build don't."

Most broadband circuits use substantial amounts of degeneration. (Not just for stability.) Check out e.g. low-capacitance input stages in scopes. Or google "video amplifier".

Tim.

Reply to
Tim Shoppa

Change that transistor to something less hot. Even if it stops, you won't know how close it is to the edge.

--
Regards, Joerg

http://www.analogconsultants.com/
 Click to see the full signature
Reply to
Joerg

After reading the responses, I'll be putting in a slower transistor. The amp was spec'ed at a voltage gain of 17, I'm getting about 13. I'll swap the transistor, and recheck the gain.

Is lowering the collector resistor value the best way to correct this?

As I understand it, the Voltage gain is RL/RE and I have two 47 ohms in parallel on the emitter, so I would expect 20.

What factors did the author use to calculate 17?

The voltage at the collector of T2 is 7.43 volts with V+ at 12 volts. Does this seem a little high?

Thanks, Mikek

Reply to
amdx

There is a 39ohms in the output, plus 10ohms in the emitters of T3 and T4. Assuming you terminate with 50ohm you'll lose 6dB there, so the gain would be more like 20dB, or a factor of 10.

Nah, about right. The emitter wants its taxes :-)

--
Regards, Joerg

http://www.analogconsultants.com/
 Click to see the full signature
Reply to
Joerg

Mmm... sounds good to me (...although of course you'd want to raise the collector resistor's value if you want to raise the gain).

I haven't seen that arrangement with the two paralled 47ohm resistors before. I'm thinking he's trying to establish a local "ground" (on the left resistor) that's somewhat decoupled from the overall ground -- kinda the reverse of sticking a small resistor and then a capacitor to ground in the collector line, that is seen commonly? Anyone know?

His transistor's bias current is ~10mA, so that gets him an emitter resistance of 26mV/10mA=2.6ohms, so even 470ohms/(47ohms/2+2.6ohms) gets you down to a gain of 18. The drop to 17, then, is probably due to the loading on the collector -- dropping that 470ohm collector resistor to an effective value of

444ohms gets you down to a gain of 17; it takes a collector load of only 7.9k to end up there. Indeed, ignoring the output transistors, it looks to me like the collector's loading is something less than (6.8k+3.3k)/2=5.05k.

By the time you're hitting 10MHz the various parasitics are probably not entirely negligible either.

Not to me (in fact, it sounds quite reasonable).

---Joel

Reply to
Joel Koltner

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.