Help! uWave xistor bias oscillates!

are the holes for the emitter grounds right near or under the emitter lead...

you need to have basically NO inductance in series with the emitter

also try adding a ferrite bead in on the leads feeding the Dc to the base and the collector, place the beads directly at the microstrip.

I agree, you have a RF oscillation that is "pumping". You need to stop the RF oscilation., It may help to identify the RF oscillation frequency with a spectrum analyzer.

Mark

There is a full ground plane underneath. Vias connect the emitter strips to the ground plane. ( I drilled holes and soldered wire. Spaced the holes close to each other.) Also, the ends of the emitter strips are connected by foil to the bottom ground plane.So both emitter strips are

That's EXACTLY what I see, a sort of pumping! Maybe I should move the

22 uf to the other side of the 1K?

The emitters are directly bonded to the ground plane, right? Emitter followers can oscillate if there's any inductance in the base lead.

I first heard of that phenomenon in Pease's book on analog troubleshooting, but I'm sure it's well-known and/or obvious to many. There's a nice explanation at

. Could that be what's going on?

Apart from that: 47 uH sounds like a heck of a lot of inductance for microwave work. Those bias chokes may have significant parasitics. Who knows... they might act like coupled resonators at certain frequencies. At a minimum they should be oriented perpendicular to each other. Can you try different choke values to see if the oscillation moves in frequency?

-- jm

------------------------------------------------------

Note: My E-mail address has been altered to avoid spam

------------------------------------------------------

Sorry, each port is connected to the network analyzer (50 ohms) thru a DC block. Also, there is a ground plane underneath. The emitter strips are connected to the ground plane thru vias and the power supplies return is soldered to the emitter strip next to the caps. Seperate supplies are used for base and collector.

Hello,

The emitter side looks like it's just two strips. Do you have a ground plane? This stuff really needs a full ground plane.

If there is no ground plane it is possible that the circuit breaks into RF oscillations, upon which the current depletes the 22uF cap on the collector side because the 1k may not allow enough current to sustain sufficient voltage. Then the voltage drop causes the oscillation to stop, the cap charges back up and the whole cycle begins anew. Kind of a "pumping" oscillation. On the scope you may only see the low frequency pumping.

Regards, Joerg

They're awfully close to the emitter leads, without getting in the way of the leads. The holes are spaces like 1/4 in apart, 4 - 5 holes up each strip. then foil bend over the edge, connecting each strip to the ground plane.

I have beads I'll try in the AM.

If nothing else, I can do another board with the shortest emitter strips possible. Oh, I'm trying to get S params at 1700 MHz.

[snip]

Who

frequencies.

Can

John's right. Unless these are really special parts the likes of which I've not seen, the 47uH bias inductors will self-resonate at frequencies much lower than your 1,700 MHz test frequency.

That's bad.

Ditto for those 22uF and even the 100nF capacitors, which are resonators here, not capacitors at all. The inductors, meanwhile, will be *highly* capacitive at 1,700 MHz.

Cures? One might add resistors in series with the inductors, the larger the better, use a skinny trace instead of or in conjunction with the inductor in the bias lines, use much smaller inductor values, and use real RF bypass caps, e.g. 22pF & 100pF. (100pF = 1 ohm at 1,700 GHz)

Were it me, I'd ditch the inductor feeding the transistor's base entirely & feed said transistor directly from the 50k-50k divider, eliminating undesired resonances and reducing the opportunity for unwanted coupling/feedback in the bargain.

Best luck! James Arthur

P.S. John's also right about putting the inductors at right angles to minimize coupling, and shielding and separating help too, but not before taking the measures above.

This will undoubtedly oscillate at uhf+ frequency unless you have some sort of rf load at the collector. With no load the inductor privodes 90' phase lead at the collector, the collector base capacitance provides an aditional 90' phase lead back to the base and you have a nice oscillator.

Colin =^.^=

No oscillations!!! Thanks to all! I removed the bypass caps and went with 22pF and 100 pF, as per James Arthur, and all is well now! My mistake, the coils are 47nH, not 47 uH.. No I can fool around some more and try stub tuning for a good 50 ohm match. I'ld like to try a noise figure measurement. I have an HP noise source and NF meter. I assume I need to tune to a good match for the noise test?

more

I

Good work on the oscillations.

The 47nH inductors still make me uneasy. A quick scan through the DigiKey catalog shows self-resonant frequencies as low as 260 MHz (muRata LQM18NN47) to 1,800 MHz (Panasonic ELJ-RE47NGF2) to 3,000 MHz (Panasonic ELJ-NK47NAF). Be sure to check the SRF of the units in your circuit. Non-shielded inductors invite unwanted coupling, while magnetically-shielded units have low SRFs.

The collector inductor has the benefit of increasing the available gain, which might justify its use over a resistor of equivalent impedance (e.g. 500 ohms). A mischief-reducing compromise is to use the inductor and a small (22-100 ohm) resistor in series.

The base-feed inductor, however, offers no advantage, and several dangers that I can see over simply connecting the base to the 50k-50k divider, which offers higher impedance & isolation than the inductor anyhow.

As to matching, the matching needed for best noise figure is commonly not the same as the matching for best power transfer (i.e. highest gain), but matching for best gain is a passable starting point for experimentation. The optimum match varies with Vcc, f, and Icc. If you're lucky the vendor's datasheets will offer parameters close to your intended region of operation.

Way back when, I wrecked a small box of MRF571s one after another learning these same things ... darned if those amplifiers don't love to scream.

Best, James Arthur

went

more

noise

assume I

its

parallel.

Thanks for the link Joerg. The MLC caps are better than I remembered. Especially heartening was the apparent lack of parallel resonances in Fig. 5, which were my chief concern. I've encountered such resonances in days gone by and marvel at these plots.

Even so, the 0603 was measured to have an effective series inductance of 0.87nH, making a 100nF unit series-resonant at 17MHz and, more importantly, contributing 9.3 ohms of reactance at 1.7 GHz. That's not horrible, but it's not fantastic -- power supply feedthrough and/or unexpected feedback are possible.

One can parallel caps to reduce the inductance, as you say, or, in many instances, deliberately select a lower-valued cap, using series resonance to advantage. An example of this can be found here:

in the document titled "Circuit Designer's Notebook - Capacitors in Bypass Applications."

Figure 2 shows a net impedance of just about 2 ohms for a single 100pF bypass capacitor at 1.7 Ghz, a 4-fold improvement.

Best Regards, James Arthur

Hello,

Great!

Just one word about ceramic caps. Even a 0.1uF shouldn't have let you down. Check AVX corporation under Tech Papers for the article by Dr.Jeffrey Cain about parasitic inductance. Figure 5 is pretty interesting here:

Any cap of reasonable size, such as 0603 has an inductance just by its sheer size, forming a "loop". So it helps to have several in parallel. Also, it just has to be SMT. Through hole is going to be next to impossible at 1700MHz.

Regards, Joerg

Agreed: that the packaging itself limits ESL to finite values.

It looks to me like the physically smaller caps _are_ somewhat better. Table 1 in the AVX paper shows 0.87nH for the 0603, vs: 1.2nH for the 1206, while Figure 4 depicts the effects of this.

Even better than "small" is to change the aspect ratio of the device to "wide and short," which is effectively paralleling devices as you've already suggested. I poked briefly about for these--and they do exist--but I didn't find any right off.

Umm...there is/was a method to my registration, and I had to be _someone_, so I applied the principles which Genome has already spoken to rather elegantly in the thread "Nationals Webench:

Grins, James Arthur

Hello James,

Thanks. That is an interesting paper. What struck me in the AVX paper was that smaller SMT caps weren't really any better at high frequencies. Much is probably attributable to the packaging itself. So smaller should be better.

Just curious: Where does your email address come from? Dagmar is, at least in Europe, a female name but James certainly isn't. Or is it a company name?

Regards, Joerg

Syfer makes 0612 sized capacitors.

Jeroen Belleman

collector

could

anchovies

drink

Love it! Once upon a project I fashioned cavity filters from Clabber Girl Baking Powder cans--tin-plated light gauge steel--and copper refrigeration tubing. Q of 300 at 900MHz with insertion loss

Hello James,

At RF that would almost be like placing caps parallel.

The real trick with RF is to use the circuit board itself. Areas such as the supplies on the OP's circuit board should be large and form a substantial capacitor with the ground plane underneath. Above 1GHz this can then become the major decoupling contributor while the 0.1uF just takes care of other noise coming in from the outside. Cost would be next to nothing.

Then there is shielding. A nice metal strip between base and collector can be very useful, at a few pennies a pop. Once when I absolutely could not find any metal at a client's site we all shared a can of anchovies during our lunch and cut up the can later. Problem was, we had to drink lots of water that afternoon.

ROFL. I believe someone did the same on the Freescale site lately.

Regards, Joerg

6 meter beer can cavity resonators?