Would an "ideal" circuit even oscillate in that configuration? I think the only reason it's oscillating at all is because there's some parasitic capacitive feedback from the output to the non inverting input, and that will affect the oscillation frequency.
Cool. Sounds like capacitance between opamp pins is making a series-resonant circuit with the L. The opamp's phase lag and the new LC network's lead cancel out. You only need a small resonant boost to push the loop gain over the top.
In general, operational amplifiers are not perfect. They have phase lag, they have coupling between pins, they have non-zero open-loop output impedance, they have finite open-loop gain, etc.
Often this does not matter. But any time you hang a highly reactive load off the end -- it does.
I rather suspect that if you put a healthy amount of resistance in series with your inductor -- "healthy" meaning 10k or maybe even 1k -- that you'll kill the oscillation, without killing the Q of your circuit as much as you're doing when you load things down with the 100k-ohm shunt. Further, it shouldn't have a whole lot of bad effects on the circuit as a whole (with the possible exception of lowering your bandwidth).
If you really want to isolate the inductor consider using a source follower. Assuming that you can find a small-enough signal FET with low enough capacitance, it should do a pretty good job of isolating things without killing the Q.
--
My liberal friends think I'm a conservative kook.
My conservative friends think I'm a liberal kook.
Why am I not happy that they have found common ground?
Tim Wescott, Communications, Control, Circuits & Software
http://www.wescottdesign.com
" Would an "ideal" circuit even oscillate in that configuration? "
That's what I don't understand... where's the feed back?
I think
OK I can try tightening up my layout a little bit.
Or how about if I add some capacitance from output to non-inverting input. From the... "If you can't make it better, try making it worse." .. school of thought.
Use an opamp with less capacitance or less phase shift
Use a jfet follower or some such ahead of the opamp
Neutralize it with a "negative capacitor"
Possibly improve the physical layout.
Try a different inductor, or just swap the leads. I tested some transformers recently for adjacent magnetic and electrostatic coupling, and the electrostatic coupling was much less if I grounded the "outside" of the winding.
I added 1 k ohm in series.. no change. (I can't afford 10 k.)
Well I tried a few different amps. Only the AD827 'video' amp did not show the oscillation. (But it's only got 300k ohm of input input impedance.)
Yeah I thought something like that might work... but it complicates the front end. I could try and see if it stops it.
ughh.
I tightened up the output. Less than 1" of buss wire from the output to the BNC input. This actually seemed to make it worse. a bit higher oscillation freq and I had to load it down with more than
100k ohm to get the oscillations to stop.
My inductor selection is a bit limited. The others I've tried all have the same pathology.
I also tried adding a bit of gain to the buffer.. but still the same with a gain of two.
I'm not really convinced it's capacitive feedback from the output.
On the 'scope traces the output is always negative. Almost bumping into the negative rail.
George H.
Whoa, look at this,
formatting link
I put a x10 scope probe on the input. (chan. 2 upper about 30 Vp-p)
Chan. 1 in the middle is the output (15Vp-p and always negative)
The circuit loop from output to (-) input can inductively couple to the field around the inductor, of course. More to the point, the two op amp input pins are sourced from very different impedances; try putting another inductor in the negative-feedback link; you likely have to account for the input capacitance of the op amp to get the phase shifts calculated correctly. Like as not, the 'high input impedance' assumption for this amplifier is ... not operative at this= time.
I've have that very pb when I designed my 200pV/rtHz preamplifier (several paralleled huge input jets which means lots of parasitic capacitance) and I think I reported it here in its time.
That pb comes from when you use a closed loop amplifier with a gain-bandwidth product GBW, which means that, at a high enough frequency above the low frequency pole, the amplifier differential input voltage has 90degree phase shift WRT to the input (or output) voltage. This phase shifted voltage then injects some current into the opamp positive input node, thanks to the opamp differential input capacitance. You can do the math for a perfect single pole opamp and a cd capacitor shunting the inputs.
You then obtain a synthesized input impedance which is:
Zin= - (2 pi GBW + j w)/(cd w^2)
When you decompose this in parallel real and imaginary impedances, that gives you:
Rp = -(4 pi^2 GBW^2 + w^2)/(2 pi cd GWB w)
Cp = cd w^2 /(4 pi^2 GBW^2 + w^2)
As you can see your stage input impedance has an always negative parallel admittance which depends only on your opamp GBW and capacitance.
Now for the oscillating frequency, apart from the obvious additional parasitic capacitance you also have to add the opamp CM input capacitance.
And now again, when you take the limit for w -> infinity, the parallel resistance limit is:
RpLimit = -1/(2 pi cd GBW)
which is roughly -10K for your example OPA2134 and have no high frequency limit! That is, no frequency limit for a perfect one pole opamp model plus differential input capacitance. The additional poles above GBW for real world opamps change that somewhere above the GBW frequency, but it's no surprising at all that you see some oscillation above the 8MHz GBW limit of your 2134.
Hello Gentlemen (and any Ladies), Just want a =91reality=92 check.
+---cf---+ | | + |\ | opa2134 +---+--+---|+\ | | | | | >-+out C L R +|-/ | C L R ||/ | C L R +-----+ | | | +---+--+ GND
So here=92s my current model of the oscillations and feedback. (thanks for the help in letting me see it.) CCC is the capacitance in the coil, opamp and any strays to ground. cf is some feed back capacitance from output to input.
RRR is a parallel resistance that I add to kill the oscillations. In a hand-wavy way the added resistance should about equal the cf impedance.
As a test I tightened up the layout and then added a copper shield between the output and input. (here=92s a pic, you can see a corner of the dip sticking out from the copper tape.)
formatting link
So I measure the resistance needed to kill the oscillations with the shield, without the shield and then with an added 2.2 pF between the output and input. Oscillation frequency were all near 1.5 MHz. (a tad lower with 2.2 pF C)
Test RRR shield 45.7k ohm no shield 38.9k
2.2pF 27k
The numbers don=92t quite add up.. but the trend is pretty clear.
So I=92ll search for a 1mH inductor with a high SRF, and also a bit better opamp.
Hi George, Looks like my hypothesis missed the mark, (I didn't note it wasn't supposed to be an osc.) However, does the frequency get lower with additional capacitance across the inductor. Mikek
Excellent! I'll have to read that a bit more carefully. I think I just measured that differential capacitance. (~50 kohm at 1.5 MHz about 2 pF which is exactly what it says on the spec sheet.)
The payoff, for me, is that I'm finally getting a 'real world' handle on the differential and common mode capacitance.
If you were to mention what you were actually _doing_ with that coil it might help us help you.
--
My liberal friends think I'm a conservative kook.
My conservative friends think I'm a liberal kook.
Why am I not happy that they have found common ground?
Tim Wescott, Communications, Control, Circuits & Software
http://www.wescottdesign.com
Hi Tim, Well let me first say that the opamp is the input to a wide band amplifier stage. (DC? to 2-3 MHz.... exact HF spec once the pcb is stuffed.) It's sort of a general purpose amp. One of the applications will be the front end of an AM radio. (You guys already helped me with that!) So the antenna is just a piece of wire... but if I hang an inductor on the input (where the wire enters the amp) I can get about 10dB more signal.
But having the inductor turn the amp into an oscillator is not so good. :^)
I=92ve been trolling digikey for other opamps without much luck. I think I=92ll have to stick with =91good enough=92.
As others have pointed out, you have pin-to-pin feedback. You might try some shielding along with guarding the +Input. ...Jim Thompson
--
| James E.Thompson, CTO | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| Phoenix, Arizona 85048 Skype: Contacts Only | |
| Voice:(480)460-2350 Fax: Available upon request | Brass Rat |
| E-mail Icon at http://www.analog-innovations.com | 1962 |
I love to cook with wine. Sometimes I even put it in the food.
They can be made, but I don't think that's the real problem here.
Why don't you make a real AM radio front end, with a real tank, and follow that with a real RF amplifier that works in real applications?
Just because an op-amp is a versatile thing, doesn't mean that it is _every_ thing.
--
My liberal friends think I'm a conservative kook.
My conservative friends think I'm a liberal kook.
Why am I not happy that they have found common ground?
Tim Wescott, Communications, Control, Circuits & Software
http://www.wescottdesign.com
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.