Prev: Cushman CE-24A
Next: TIA Photodiode Bootstrap at 10MHz
From: bg on 10 Feb 2010 16:20 Jon Kirwan wrote in message <9kk3n5tjcntp1b1itrlbldjgta7st0oai8(a)4ax.com>... >On Tue, 9 Feb 2010 14:05:32 -0700, "bg" <bg(a)nospam.com> >wrote: > >><snip> Have a look - http://www.xmission.com/~bmge/Feedback.htm
From: Jon Kirwan on 10 Feb 2010 18:37 On Wed, 10 Feb 2010 14:20:43 -0700, "bg" <bg(a)nospam.com> wrote: > >Jon Kirwan wrote in message <9kk3n5tjcntp1b1itrlbldjgta7st0oai8(a)4ax.com>... >>On Tue, 9 Feb 2010 14:05:32 -0700, "bg" <bg(a)nospam.com> >>wrote: >> >>><snip> >Have a look - >http://www.xmission.com/~bmge/Feedback.htm I noted the author's comment, "If I knew what I was doing, I could probably optimize this circuit to produce even better results with feedback..." The author the also talks about "bias stability", which I might have taken to mean keeping the bias point at the same voltage as Vs1 varies, but could instead mean making it linear, without any "bow" to it. (As I understand it, for simple amplifier stages the collector is usually set somewhere near the midpoint and thus one might actually _want_ the change but want it to be linear in some known fashion and certainly not bending over in a bow as in the second circuit's case!) So the 'goal' isn't clearly stated to me. By the way, that first circuit _is_ a Vbe multiplier with a resistor limiting current from Vs1. So computing point A on it isn't much different than what I did, earlier, when I first posted. The first circuit shown there on that web site (and I don't know if you intended this, or not) is basically a Vbe multiplier sourced by a 10k resistor with point A being the Vbe multiplier output. The second circuit shown there is, once again, basically a Vbe multiplier with the collector resistor in place, if you think of point A in that case as being the output, but now with the topside tied directly to a voltage source instead of a resistor or current source -- which obviously isn't the way a Vbe multiplier would operate. However, the curve he shows for it remains interesting. Although the author is talking about something else, the importance of NFB instead, the first case he makes actually presents the _problem_ I was talking. In his first circuit case, the variation in the Vbe multiplier's output vs sourcing current through R1 is shown clearly. I pointed this out in the first post in this thread (with terms now changed to match his first circuit's usage): V_bias = Vbe*(1+(R2/R3)) + R2*Ic/beta (Which ignores the tiny kT/q 26mV always present in the emitter.) He shows a Vbe of 0.633V, R2=98k, R3=5.6k and I think the beta of his Q1 is over 200. Ic is about 770uA from his values. From these, I compute V_bias = 12.0878V. His circuit shows 12.3V there. Perhaps close enough, but I was interested in seeing what LTspice would show. After duplicating his schematic and running it, I see 12.081V at point A. Much closer to my computed value. I'd also gone to the trouble, that the author does not, of computing R_ac for the system. In his case, the value works out to around 1200 Ohms. Roughly speaking then, we have a 10k/1.2k divider for small __changes__ in voltage. This suggests about .11V/V while his graph shows something more like .75/5 or .15V/V. However, once again my schematic in LTspice shows instead .61/5 or .122V/V, which is closer to the value I calculated using R_ac as an approximation. I'd already done some useful analysis for his circuits and I'd not even read his web site, yet. Another interesting point. In the second circuit's case, although it uses a voltage source at the top -- which is decidedly NOT what I'm considering, obviously -- the _shape_ of his curve is exactly what I _want_ to have. Obviously driven differently than shown, I would set the collector resistor value to be approximately the R_ac computed without it and that nice curve should appear -- just not the very large magnitude excursions since the drive is different and the collector resistor is smaller in magnitude (it would be set to around 1.1k, not 10k, other things being similar.) In fact, I think I mentioned this either in this thread or the one over in .basics, last week. That curve helps to allow me to tweak for an optimal spot and then minimize voltage output variation over current that is sourcing through it. So I again modified the schematic to vary a current source instead of a voltage source, from 500uA to 1000uA (roughly centered over the estimated 770uA drive from before), and plotted the voltage curve. Using that 1.1k collector resistor in place, it is a very nice bow centered very sweetly around the target of about 750uA, drooping by only 38mV out at the skirts. Exactly as I predicted the shape should be with that value. Interesting page, sadly lacking in equation development. What I took away from it may have been different from what the author (or you) perhaps intended. But there it is. I still _get_ the idea of NFB!! So I don't mean to argue against that! I just went somewhere else with that page. Jon
From: Jim Thompson on 10 Feb 2010 18:42 On Wed, 10 Feb 2010 15:37:00 -0800, Jon Kirwan <jonk(a)infinitefactors.org> wrote: [snip] > >I still _get_ the idea of NFB!! So I don't mean to argue >against that! I just went somewhere else with that page. > >Jon First rule of "NFB": Make it as good as you possibly can without NFB, _then_ apply NFB ;-) But it's sort of a trick and a lie... you use _local_ feedback to make the individual pieces as linear as you can, then add overall _global_ feedback. ...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.
From: Jim Thompson on 10 Feb 2010 18:52 On Wed, 10 Feb 2010 16:42:47 -0700, Jim Thompson <To-Email-Use-The-Envelope-Icon(a)My-Web-Site.com> wrote: >On Wed, 10 Feb 2010 15:37:00 -0800, Jon Kirwan ><jonk(a)infinitefactors.org> wrote: > >[snip] >> >>I still _get_ the idea of NFB!! So I don't mean to argue >>against that! I just went somewhere else with that page. >> >>Jon > >First rule of "NFB": Make it as good as you possibly can without NFB, >_then_ apply NFB ;-) > >But it's sort of a trick and a lie... you use _local_ feedback to make >the individual pieces as linear as you can, then add overall _global_ >feedback. > Like this _very_linear_ differential pair.... http://analog-innovations.com/SED/TL431DiffPair.pdf :-P ...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.
From: Jon Kirwan on 10 Feb 2010 18:57
On Wed, 10 Feb 2010 16:42:47 -0700, Jim Thompson <To-Email-Use-The-Envelope-Icon(a)My-Web-Site.com> wrote: >On Wed, 10 Feb 2010 15:37:00 -0800, Jon Kirwan ><jonk(a)infinitefactors.org> wrote: > >[snip] >> >>I still _get_ the idea of NFB!! So I don't mean to argue >>against that! I just went somewhere else with that page. >> >>Jon > >First rule of "NFB": Make it as good as you possibly can without NFB, >_then_ apply NFB ;-) > >But it's sort of a trick and a lie... you use _local_ feedback to make >the individual pieces as linear as you can, then add overall _global_ >feedback. > > ...Jim Thompson Now _this_ is what I wanted to hear. Many seem to just tell me "use global NFB to fix things" almost, it seems, to simply stop me from bothering to struggle at all or even care about understanding things. Maybe it is just because it _takes work_ to actually engage a quantitative discussion and the lazy way out is to just hand wave and tell me to "move on by." But it was my sense at the outset, and it is my motivation for starting this thread as well, to do exactly what you are talking about here. I'm so glad to see it said. "Make it as good as you can without NFB, then apply NFB." Yes! For example, the Sziklai pair is really a BJT wrapped with a local NFB using the other BJT for that purpose. Nice. I couldn't state it this clearly because I'm just learning things. But what you said is what my instincts tell me, despite attempts to say "move on, there's nothing to see here." Jon |