It's was Bigun's "if you don't believe me" that sucked me into this thread, because I've always been agitated by the fact that it takes power to drive a BJT when people like to discuss whether it is voltage or current controlled. It definitely doesn't seem practical to think of the BJT as current controlled, or voltage controlled either (even though it is at the atomic level. It seems like the Gummel-Poon model is used a lot in software analysis because how the transistor will respond for some input is actually that complex in most instances.
I was tricked into oversimplified misconceptions at an early age by an instructor who suggested it would be for the sake of "convenience"! It seems like the intuitive "calculus" of which terms matter at different levels might really be the discussion. I think possibly the analysis softwares can be forced to somewhat automatically make simplified calculations without too much data loss, or maybe a competent programmer is actually required at each pass to do that. Either way, too much intuition and not enough model turns into design by belief, which leads to funny design and lots of discussion for the sake of it.
Member
Joined 2009
Paid Member
Did you say something, Gareth?
Lumba, I think you have been a bit naughty these past days with some controversial viewpoints
Last edited:
Member
Joined 2009
Paid Member
It does take power to drive all active devices because there is resistance and therefore dissipation. The BJT requires charge carriers in the base, you have to be able to supply a current or sink a current to the base in order to hold the desired voltage. This current provides the charge carriers needed to change the minority carrier charge distribution in the base region. And so we can equate a base current limitation to current flow through the device and express this as a current gain. Then we end up with the idea that the BJT is a current operated device.
I agree, all this is an oversimplification, a convenience. To understand how the device works in detail is quite a challenge and we see that even more complex models start to break down so we invent names for all these deviations of reality from our models (Early Effect, Kirk Effect, Auger Recombination, Shockley-Hall-Read bandgap narrowing, etc.)
The summary as I understand it:
a) the control voltages applied to the BJT determine the charge carrier densities
b) the emitter and collector currents are determined by the minority charge carrier distribution.
In other words, for a given device, Ie and Ic are functions of the applied voltages - both Vbe and Vce but for my simple convenience I think of it first and foremost as Vbe being the main control voltage.
One question I have about feedback is the following:
It can be shown that loop gain reduces non linear coefficients if feedback is applied around a non linearity.
But , if there is positive feedback the distortion coefficients are rising sharply. Positive feedback with gain less than one exists beyond the unity gain frequency . Could it happen that this rise in distortion beyond unity gain couples back via intermodulation in the audio band explaining why wider open loop gains could ( ???) sound better.
JPV
It can be shown that loop gain reduces non linear coefficients if feedback is applied around a non linearity.
But , if there is positive feedback the distortion coefficients are rising sharply. Positive feedback with gain less than one exists beyond the unity gain frequency . Could it happen that this rise in distortion beyond unity gain couples back via intermodulation in the audio band explaining why wider open loop gains could ( ???) sound better.
JPV
Oh please, not the chicken-egg vicious circle again, On Semi, Fairchild, TI, Natl Semicon, etc. all profess that at the macro scale, the bjt is current controlled, and at the micro scale it is charge controlled. Charge is more fundamental than either I or V. I & V are defined in terms of charge, time, & energy.
The original Ebers-Moll paper models the bjt as 2 current controlled current sources. The collector current Ic is given as alpha_n*Ie. The 1954 paper depicts Ic as a current source controlled by Ie & alpha_n. A 2nd bjt is shown upside down operating in reverse mode, where collector & emitter are swapped. Here Ie = alpha_I*Ic.
The alpha_n factor is for normal mode, & alpha_i is for inverse mode. I'll post the paper after I get home tonight.
I used the Sze text in grad school 2 years ago (Ph.D.), and nowhere is a bjt described as voltage-controlled. In fact, the text covers detailed quantum mechanics, and space charge distribution. In semi physics, it's all about charge control, as that is the most accurate model at the micro level. At the interatomic level, QM is employed, i.e. Schroedinger's wave equation, & the Kronig-Penny model.
As far as Ic goes, is a change in collector current, ic, due to a change in base current, ib, or base-emitter voltage, vbe? Neither one can ever exist w/o the other. My 1964 GE ref manual for bjt says that the change in collector current is due to both ib & vbe. In order to change the E field value, current & voltage are both needed. An E field carries energy. Changing E value requires changing energy/doing work. This takes time. For a limited time, energy changing means that power is non-zero. A non-zero power can happen only when ib & vbe are both non-zero.
To change the E field in the junction mandates non-zero ib & vbe. They are both necessary. Referring to bjt as current controlled is perfectly valid at the mAcro level, where we treat the device as a black box, focusing on its external I-V properties, neglecting internal physics.
At the mIcro scale, ib & vbe are both important, and the device is modeled as charge controlled, not voltage controlled, not current controlled. Again, changing collector current cannot happen unless ib & vbe both change. The change in ib takes place ahead of vbe change. Neither is the cause, or seat. Neither is more basic than the other. They both participate in unison.
Is this clear?
It could be measured by changing the rolloff after crossover without to much interaction below, impossible
Another way would be to use a swept sine wave impulse, this is a linear sweep between two frequencies during a short time, a chirp. The beauty of this impulse is that the fourier transform is more or less like a square in frequency concentrating the energy between these frequencies.
By applying the pulse with a lower frequeny at the crossover frequency of the amplifier and then changing the amplitude of the pulse, we could trigger enough intermodulation products in the audio band, the original signal having very low spectral content in the audio band. We have to be carrefull to avoid triggering other mecanism of distortion
To be investigated.
Member
Joined 2009
Paid Member
its more correct to refer to loop gain intercept frequency as the transition between negative and postive feedback - higher intecept frequency is enabled by faster (output) devices
don't confuse it with the open loop gain corner frequency which is often talked about as high vs low bandwidth in a dominant pole compensated amplifier - "low" open loop gain bandwidth is a consequence of high (DC, LF) loop gain which reduces IMD compared to the same GBW product and a "high" open loop gain bandwidth
don't confuse it with the open loop gain corner frequency which is often talked about as high vs low bandwidth in a dominant pole compensated amplifier - "low" open loop gain bandwidth is a consequence of high (DC, LF) loop gain which reduces IMD compared to the same GBW product and a "high" open loop gain bandwidth
Last edited:
I don't know for IMD but it can be shown that in a miller opamp with a single non linearity at the second stage, second and third harmonic distortion ( increasing with frequency) are inversely proportional to the gain bandwith product.
This means that if you decrease the LF gain and increase the corner frequency keeping GBW product constant, the distortion do not change. This is calculated using Volterra kernels in an accurate way for this simple case.
So what matters is the GBW product and not the LF gain or wide open loop gain.
JPV
it appears to be true if you load the VAS with a R to gnd to change the open loop corner frequency
but practical measures that increase the LF gain of VAS such as buffering or cascoding can improve linearity of the VAS itself
however even in the 1st case, the low bandwidth, high gain does reduce the distortion contributions of the other stages even at constant GBW - for the lower frequencies at which the gain is increased
added gain in one stage reduces the previous stage's distortion by reducing signal levels in their nonlinearities and the following stages by the increased loop gain
but practical measures that increase the LF gain of VAS such as buffering or cascoding can improve linearity of the VAS itself
however even in the 1st case, the low bandwidth, high gain does reduce the distortion contributions of the other stages even at constant GBW - for the lower frequencies at which the gain is increased
added gain in one stage reduces the previous stage's distortion by reducing signal levels in their nonlinearities and the following stages by the increased loop gain
Last edited:
Very many years with this issue. I long ago rejected the local feedback. I use overall feedback, very high gain and bandwidth in an open loop. Sufficient reserve gain. A lot of the written here: http://www.diyaudio.com/forums/soli...y-amplifiers-listening-characteristics-9.html
Well look at the impact of feedback on gain and bandwidth.
Well look at the impact of feedback on gain and bandwidth.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.