JFET gain control "resistor" in Weinbridge oscillator

What's wrong with the J201? Can be had for around 20 cents. Of course, oldtimers would ask "What's wrong with the light bulb all of a sudden?" :-)

Check out the 2SK3372. Digikey has them but these only come in teeny SMT packages. One sneeze and it's gone.

How many thousand of these circuits are you building?

John

Please, bore us! We aren't mind readers - without details, we're just stabbing in the dark.

Tell us EVERYTHING. (about the problem circuit.)

And please don't use foreign unicode characters, like that messed-up apostrophe - set your character encoding to just plain ASCII.

Good Luck! Rich

Thousands???? Fifty will be a good year.

"Tell us EVERYTHING." Hi Rich, The circuit worked fine for years, then there started to be some "motor-boating" (amplitude modultion at some sub-frequency) on some of the units. The motorboating only happens at the highest frequency ranges (High in this case being 3kHz.) The problems seems to have started after a PCB "rebuild" where there was only supposed to be some silk screen changes. But this could be a red herring. Oscillator circuits are "sitting on a knife edge" and the motorboating would sometime go away just by touching the front panel. (changing the capacitance somewhere.) So misbehaving circuits were "hacked" by adding a little capacitance to the feedback loop.. rolling off the gain at high frequency. So now I've got half a dozen bad boards sitting one my shelf for which the hack doesn't work. And then a message yesterday from a user in Europe with a motorboating problem.... Arghhh! So I'm ready for the next stage solution. The first time I read Jim Williams article on the Weinbridge oscillators (this is also in one of his books.) I didn't quite "get" the integrator FET solution to the gain control problem and the light bulb looked like such a simple idea. (I also loved that it was the begining of HP) But I've advanced since then and am now ready to try this other circuit

he

ge

with

m J.

ET

the

"And you can't afford $215 for the right parts?" It's not so much the money, but I'm afraid that a $5 FET is going to dissappear next year.. or at least turn into a $10 FET.. and so on.

In that case it's worth a an hour or two to get the cost down.

How about a small mosfet? Positive gate drive might be convenient.

John

Yep, or a dual-gate RF FET which is what I use in AGC apps a lot. Really cheap because of all the TV tuners.

But small JFETs abound, such as the Panasonic 2SK3372 I suggested in the other post. Should do the trick and 20 Cents isn't a whole lot of money, doesn't even buy a lollipop anymore these days. Heck, it could beat the light bulb in cost.

This would be a good time to enter the design into LTSpice. Looks like something in there has a phase margin really, really close to zero.

The app note is AN47 with diagrams on page50.

The fet part number used is 2N4393, a P-Jfet normally used as a switch with Rdsmax

You might want to restrict yourself to fets that are graded for Vp, as the 2N4393 (please note the correct part number) used by Williams only had to work the one time. Analog switches or VCRs with lower maximum Vp (ie less than 3V) would be more predictable in this circuit, when powered from +/-5V.

The 4393, by the way, retails for less than US$0.10 in small quantities.

RL

It's usually only avaliable as the MMBF4393:

Say, this reminds me. I was recently mucking around with a lightbulb- regulated oscillator, and under certain conditions it would modulate uncontrollably. Sometimes reducing the gain would help, but not always. It would often be a knife's edge between clipping (which usually regulates amplitude, but gives *visible* distortion, obviously no good) and nothing at all (too little gain), with only bouncing inbetween.

Particulars include a TL072 op-amp, a resonant tank at 68kHz and a

Tim

P.S. Using googlegroups for now... please add me to your whitelist!

the

dge

with

e a

om J.

FET

.30

the

Thanks Joerg, I've got a list of Fet's I'm ordering, but it'll be at least a week before the parts come in and find time to protype it. If you guys with small RF fets have some part numbers I'll look up them and order some too.

The series I used a lot for variable resistance is the BF9xx, for example the BF998. However, they aren't so good for the kohm range and also these are RF-hotrods. You might want them to run at 3kHz but they don't know that, they want to sing at a few hundred MHz. They are like a Porsche on an icy winter road :-)

So I'd stick with the Panasonic, J201 and stuff. They aren't that "high strung". Better for your situation. It would not be cool if your customers tell you that the motor boating is gone but their FM radio always quits when they turn on your circuit.

th

be

Joerg, It'll take me a lot less time to prototype the circuit. Concerning the phase margin. I'll admit to not knowing exactly what my phase margin is, but I thought the idea of an oscillator was to make the phase margin zero. Actually I think I need a wee bit of postive gain??

We had a touching-the-front-panel issue recently. Turned out to be a 1 GHz RF oscillation in some opamps, way beyond the official frequency range of the circuit.

If that's what's happening to you, all the obvious loop margin fiddling is a waste of time. A small pcb layout change could easily change the RF situation.

Get an ordinary pencil and start touching nodes. That won't affect the KHz-range stuff much, but will sure change an RF oscillation. Got any bipolar transistors in the circuit?

Post a schematic, or email me one, and the culprit may emerge.

John

True. However, then you won't know what the margins are. It is very easy with SPICE to change components in rapid-fire fashion. No soldering, just seconds and you'll see what it does.

I just did that with a high voltage generator for a client. Caught a spike that would have caused a FET in the control path to "lift off".

Yes, but it looks like you have a regulated oscillator. So if the regulating loop has no margin it begins to oscillate, or motorboat.