Jim hit the nail right on the head, make it linear as possible with local feedback first. Add global later if it applies - I put that page together as a quick reponse so that you might see for yourself what I mean by feedback being used to stabilize things. It is far from an in depth analysis and there very well could be something desperately wrong with the circuit, but it does reduce the drift to 1/4 of the circuit without feedback and that is the point of my post. Again , try to find the book, it is the best advice I can give.
Good application of local feedback has other advantages... it generally makes it easier to apply global feedback without getting into stability and ringing nightmares. ...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.
Thanks. Now why do I instead find myself sometimes having to actually _argue_ about understanding building blocks well? Why is the answer so often, "throw gobs of global NFB at it?"
I wonder if the availability of all-too-perfect A_ol = 1E15 opamps (not really, but what's the difference?) is part of the problem, here. Maybe it's making things too easy.
Ah. That was you? Thanks for the effort, then!!
I think I already understood the majestic power of global NFB. It's so important, I am sure, that if NFB didn't _also_ wrap the output stage itself, the results would be indeed very lousy no matter how good the earlier part of it turned out to be. So it is not an option.
I did learne the basic gain/feedback equation years ago: Vout/Vin=A/(1+A*B), with B being the feedback and A the open loop gain. With gobs of A available in these all-too-perfect opamps these days, the whole thing drops back to 1/B 'real fast.' Which is nice because then just set B and get handed a fixed gain on the so-called silver platter.
I have to still believe, as broadly ignorant as I indeed am about these things, that crafted design with localized NFB remains useful even in the case of audio amplifiers. That doesn't mean the power of global NFB isn't of the overarching importance that it is. There is no option there. But there remains more to life than merely that, too. Local NFB seems to remain important to me. And it was nice that Jim took a moment to confirm that impression, when so few had done so beforehand.
... You also noticed that I took the web page in a totally different direction? ;)
Speaking of which, what spice model did you use for that
2N3904? Can you post it? I'd like to stick it into LTspice and see why I got different results. Since my calculations didn't depend too highly on wrong estimates of beta and since kT/q doesn't care about the BJT, I'm curious about exploring it a little more.
Excellent! Thanks Phil. Do you just use a current source as bias as in the Amps from Pass labs? (Do I need to repost the previous link?) I don't quite see how this works for push-pull. Do you have any schematics? (Hmm thinking that a few resistors should take care of things...)
Say I thought I read that tubes can't do push-pull because "it's hard to get positrons from the filament" to parapharse what I read.
Hi Jon, I'm loving your threads... (did any one ever tell you your kinda long winded?) I wanted to take exception to your opamp statement,
"I wonder if the availability of all-too-perfect A_ol =3D 1E15 opamps (not really, but what's the difference?) is part of the problem, here. Maybe it's making things too easy. "
As the frequency of interest approachs the GBP the gain goes to 1 and things get intresting...
Well, I was thinking exactly this when I was earlier writing... imagining instead of writing this way, "... of all too perfect A_ol*BW = 1E15" rather than as I did write. Just to get that in. But it wasn't important to the point at hand and I edited it back out as I figured "it goes without saying."
And you didn't even complain that I spelled "you're" worng.
"> and I edited it back out as I figured "it goes without
Well sorry then, I'm perhaps more of a novice than you... but I find opamp circuits complicated enough.... And tend to stick transistors only on the edges of things. (Mostly on the output side... on the input you have to 'know more' than the guys who designed the opamp.. hard to do for a novice.)
I guess if I was designing an audio amp I'd figure on an opamp driving some sort of FET output stage. The question of how to bias the output stage is interesting. And also of how all the NFB works.
Pass designs are interesting reading. Nobody says you have to build them. ;-) Nelson is the opposite of Randy Slone, who is Mr. Practical. Doug Self is somewhere in the middle.
I don't use LT so I'm not sure if the commented (*) text needs to be removed.
These spice model files might be of interest but possibly obsolete - google for them OnSemiconductorAllModels.zip OnSemiconductorDiscreteModels.zip OnSemiconductorIntegratedFunctionsModels.zip
Don't sweat it. While some opamps leave _some_ issues nearly ignorable, there is always some tough problem at that scale that makes it non-trivial and interesting to work on, I imagine. Each macroscale view has it's own complexity. Telescoping levels, where the complexity at one stage doesn't take away from interesting complexity at another level.
(I would say more but I'm keeping in mind your warning about long-windedness and will now muzzle myself.)
I do the same things except that I enjoy math and BJTs give me an excuse, perhaps. Maybe that's the only difference.
Might as well just get a power opamp like the OPA502 and be done with it. Give it two rails, feed the input, and just drive the hell out of a speaker. Or get two of them and do a bridge amplifier. But where is the enjoyment in that? Or the learning? Someone else already did most of the fun stuff and there's nothing really left to do except some hook up and heat sinking. It's not at all satisfying to me, anyway.
An audio amplifier is basically a power opamp. Using an opamp to make one feels to me like building a car by first buying a car without the tires, selecting and installing some tires, and then saying you designed and built yourself a car.
Note that it does NOT include a figure for ISE, which your model does. LTspice is setting it to zero, instead. Your model may very well be right. If so, it affects the beta value (degrades it due to some recombination cause) and that affects my calculations. In fact, that value of ISE cuts it down to around 100, or so. Which about doubles the contribution to the final term in the equation I gave where a term is divided by beta, so that adds back a few tenths of a volt to the prediction! Great!
Now, with your model, LTspice shows the same results you got.
No, that is pretty much the universal spice comment character. It's fine.
I already had the first one laying about. I downloaded the second one just to have it and extracted 2N3904.LIB from it, using that model, as well. It works much more like the LTspice one and perhaps even more different than the one you were using. However, it does produce about the same _beta_ as your model does.
Very interesting. I need to consider the reasons carefully.
Thanks, Jon
P.S. Just for completeness, the OnSemi model is this:
--
"it's the network..." "The Journey is the reward"
speff@interlog.com Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog Info for designers: http://www.speff.com
Thanks for the hints, I do need more practice in nodal analysis and I'll make a try at solving it myself. If I understand correctly the purpose of the circuit is to minimize the voltage difference between the two inputs.
It's funny that it uses the LM324, as I was playing around with one the other day and for some reason attempted to make a 10Khz square wave oscillator with it, not remembering that its slew rate is pretty limited.
And a 2.5V "dead-band", but it _is_ precisely known, and temperature stable. Interesting thought if you have high enough power supplies. ...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.
--
| 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.
No. The object of the _overall_ circuit is to amplify the input signal (bottom graph of Page 2).
The differential inputs _are_ servo'd to zero when the amplifier is in its linear region... that may be what you are thinking of.
The second OpAmp and resistors are to null out at least a portion of the bias current needs (top graph) at the _circuit_ input... result is middle graph.
In the '70's LM324 and LM339 were my jelly-bean parts, for general purpose stuff, but TL084 OpAmp for fancy work ;-) ...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.
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.