Actually, it is both local and global; 2 feedbacks, strictly speaking. One is formed by a resistor in emitter, and second is formed by a voltage divider that is at the same time load of an output transistor.
It is slightly more complicated than that. The local feedback comes from the parallel arrangement of 'emitter resistor' with the 'feedback resistor' in series with whatever the load is. This lets you calculate the open loop gain (i.e. just with local feedback), then you can add the global feedback from the second transistor. I suppose what is 'local' and 'global' depends on what you mean by 'stage'.
It is slightly even more complicated, if to consider internal properties of transistors.
Local means within a single active device; global means around a whole amplifier that contains more than one active device.
Adding extra coloured lines to a circuit could create confusion. In this case the red and green lines would have to be in the same place! Both 'local' and 'global' feedback involve both resistors. Strictly speaking, as Wavebourn has said, you have to include the internal impedances of the transistors too.
If you find the lines helpful, fine, but there is a danger that you could end up misleading yourself.
If you find the lines helpful, fine, but there is a danger that you could end up misleading yourself.
Member
Joined 2009
Paid Member
Wavebourn et al,
I find it confusing with all this talk about current feedback. For me, it makes things more confusing than I think they should be. Perhaps it would be nice if beginners could avoid learning this way ??
Active devices (transistors, tubes) are Voltage controlled devices. You apply a voltage between their control terminals (base-emitter, grid-cathode) in order to modulate the current flowing through them (collector-emitter, plate-cathode). Feedback and the Signal are both voltages. You can use a resistor to convert current flow (including current flowing through the amplifier load) into a voltage and in this manner I'd agree that the feedback signal is a measure of the current, but what is being fed-back is a voltage signal. When you think this way, you are careful to ensure you know what you are feeding back and why.
I find it confusing with all this talk about current feedback. For me, it makes things more confusing than I think they should be. Perhaps it would be nice if beginners could avoid learning this way ??
Active devices (transistors, tubes) are Voltage controlled devices. You apply a voltage between their control terminals (base-emitter, grid-cathode) in order to modulate the current flowing through them (collector-emitter, plate-cathode). Feedback and the Signal are both voltages. You can use a resistor to convert current flow (including current flowing through the amplifier load) into a voltage and in this manner I'd agree that the feedback signal is a measure of the current, but what is being fed-back is a voltage signal. When you think this way, you are careful to ensure you know what you are feeding back and why.
Last edited:
Bigun et al;
it is not about what controls device (however I feel better when assume that I use current to control BJTs). It is about what feedback senses: output current, or output voltage. If it senses output current, feedback signal will be proportional to output current. If it senses output voltage, it will be proportional to output voltage.
PS: also, speaking of audio I always assume that I amplify POWER. Such a way I am getting lower distortions and wider dynamic range.
it is not about what controls device (however I feel better when assume that I use current to control BJTs). It is about what feedback senses: output current, or output voltage. If it senses output current, feedback signal will be proportional to output current. If it senses output voltage, it will be proportional to output voltage.
PS: also, speaking of audio I always assume that I amplify POWER. Such a way I am getting lower distortions and wider dynamic range.
Member
Joined 2009
Paid Member
There are no voltage or current devices, only impedance devices. The former approach works for block diagrams, but for actual circuits high performance depends on the details. All real amplifiers are impedance drive, just more voltage than current or vice versa.
Transistors can be controlled by voltage or current, it just depends on the impedance of your source. If high impedance, it's more current control. Low impedance, as in most cases is voltage control. We prefer low impedance sources only because beta/Hfe is unpredictable. IE, which do you want to affect the signal or operating point more, Ib or Vbe?
- keantoken
Transistors can be controlled by voltage or current, it just depends on the impedance of your source. If high impedance, it's more current control. Low impedance, as in most cases is voltage control. We prefer low impedance sources only because beta/Hfe is unpredictable. IE, which do you want to affect the signal or operating point more, Ib or Vbe?
- keantoken
All depends on convenience. Sometimes it is better to view some devices as voltage driven, sometimes as current driven, the difference is different equations to omit unnecessary variables and simplify calculations. But thinking about transistors I always keep in mind base currents and betas. Thinking of their combinations I can think in terms of voltages, of course. But thinking about audio amplifications it is always better to think about power amplifications. That of course assumes impedances that are always complex and non-linear!
You can always use voltage or current, but usually one is a better match to reality in that the complications of real impedances appear as extra terms in an otherwise simple equation. At the grid of a valve it is generally best to think of voltage, unless working at VHF when the valve has a significant input conductance. At the base of a transistor either may be appropriate.
When series feedback is used it is generally best to think of voltage; for shunt feedback current is usually better. So we keep all the tools in our toolbox shiny, and use the appropriate one. If you only know how to use a hammer, all your problems have to look like nails!
When series feedback is used it is generally best to think of voltage; for shunt feedback current is usually better. So we keep all the tools in our toolbox shiny, and use the appropriate one. If you only know how to use a hammer, all your problems have to look like nails!
Member
Joined 2009
Paid Member
This is OT for this thread but I'll finish my thought here: when you think of a transistor as current controlled from a high source impedance you are simply bringing into the equation the resistance (external to the device) through which the current must flow in order to generate the control voltage. It's still a voltage that is controlling the active device (heck, just look at the math)
So back to global negative feedback. In my logic, the signal controls the current flow through the output devices of the amplifier and since it must flow in loops we notice that it flows through the output devices (including any cathode resistor bypass caps on tubes), the psu (mostly the caps) and the load. With a low impedance psu there should hopefully be negligible voltage drop across the psu and so it is the voltage drop across the load that determines the voltage signal at the output of the amplifier. This is the signal that is fed back to the input to achieve global negative feedback. If the load is non-linear (e.g. reactive voice-coil) then the feedback will include the effects of the load and this is a whole 'nother topic. You can put a resistor into the circuit to sense the output current, but this still results in a voltage being fed back. Signals are always a voltage, currents always flow in loops.
Last edited:
No, yes. The second part of your statement is guaranteed by conservation of charge, which is fundamental physics. The first part of your statement is simply untrue. Signals are sometimes voltages, sometimes currents, sometimes power, sometimes fields. Don't forget that potential (i.e. voltage) is strictly only defined for situations in which there are no changing magnetic fields, so it is less fundamental than current. Engineers, and audiophiles, ought to learn more about Maxwell's equations.
Whether the signal controls the current or voltage at the output depends primarily on the type of output stage. The feedback signal which is sent back is often a voltage, but it could be a current. The inverting opamp circuit senses the output voltage, and then sends a feedback current. Internally, a triode senses the anode voltage but feeds this back as an electric field - neither current or voltage! The Miller effect is caused by current fed back through the grid-anode capacitance.
An interesting experiment is to use a 'scope to measure the AC voltage needed (before the OPT in a tube amp) to drive a just audible signal. Now play some louder music and meaure the AC voltage it impresses upon the PSU rail. If it's greater than the first you'll hear your PSU
An amplifier is in fact a voltage controlled power supply, you cannot separate the amp and PSU in any meaningful way, except mechanically.
As for feedback, there is always non-linear stuff going on at the speaker terminals, the only decision is whether to leave it alone or not. The screechy transistor amp approach is to wrap loads of feedback round to 'correct it'. Except by the time it's happened it's too late, and the huffing and puffing to correct the error is heard long after the error has escaped, ruining the rest of the sound. The tubey approach is to let it be and not make the situation worse, and in fact to design to lessen the initial errors.
Another name for GNFB is 'feedback that's too late'. Especially in discrete transistor amps like the dreadful stuff on sale today.
Global feedback has it's uses in servo systems but not in amplifiers, like energy distortion cannot be destroyed, GNFB simply smears it all over the transients and upper harmonics where is measures less but sounds flat, lifeless and tiring. many tube amps sound better than transistors in my view simply because the amount of GNFB is less because you can't whack it up too high without increasing tube count and crossing the phase margin with the OPT.
The whole point of competent audio design is to separate the amp from its power supply. The fact that this cannot be done perfectly is not a reason to deny it. However, if no global NFB is used (due to prejudice?) then it is much more difficult to separate them.
There are good transistor amps, and bad valve amps. Bad transistor amps can be built by people who overdo GNFB; bad valve amps can be built by people who underdo GNFB. In both cases the basic problem is ignorance, often masquerading as superiority.
Personally, I prefer valve amps, but I certainly don't want any "tube sound" so I use global NFB to improve an amp which sounds quite good without it.
There are good transistor amps, and bad valve amps. Bad transistor amps can be built by people who overdo GNFB; bad valve amps can be built by people who underdo GNFB. In both cases the basic problem is ignorance, often masquerading as superiority.
Personally, I prefer valve amps, but I certainly don't want any "tube sound" so I use global NFB to improve an amp which sounds quite good without it.
I'm going to throw out an idea wrt feedback that may sound counterintuitive, but after considering it you may decide that some popular thinking about gNFB is in fact counterintuitive.
It's not the gNFB itself that's too slow being evil, but rather it's when gnfb is used to try to "speed up" an amp circuit that's already slow.
Another way to say may sound more familiar. If the open loop gain*bandwidth product of the amplifier is too small, or the unity gain BW limit is too low, no amoung of feedback can keep it linear in it's upper frequency range.
In tube amps, it's often using NFB around the OPT trying to get away with a cheap OPT that doesn't really reproduce the HF due to interwinding capacitance.
With SS circuits, it's often the difficulty in driving high intrinsic capacitances at high slew rates.
Trying to swing a gain stage harder to speed up a following driver or power stage may in fact work with servos but in audio it sounds horrible. The cure is to design circuits with adequate "slew margin", i.e. adequate current wrt the load capacitance.
IMO if it was simply a matter of linearity, as in local feedback around an outstage, there is no issue at all of the feedback being "too late". Rather, the feedback is only too late if it's picked off after an amplifier stage that is too slow, and feed back in attempt to speed it up.
Cheers,
Michael
It's not the gNFB itself that's too slow being evil, but rather it's when gnfb is used to try to "speed up" an amp circuit that's already slow.
Another way to say may sound more familiar. If the open loop gain*bandwidth product of the amplifier is too small, or the unity gain BW limit is too low, no amoung of feedback can keep it linear in it's upper frequency range.
In tube amps, it's often using NFB around the OPT trying to get away with a cheap OPT that doesn't really reproduce the HF due to interwinding capacitance.
With SS circuits, it's often the difficulty in driving high intrinsic capacitances at high slew rates.
Trying to swing a gain stage harder to speed up a following driver or power stage may in fact work with servos but in audio it sounds horrible. The cure is to design circuits with adequate "slew margin", i.e. adequate current wrt the load capacitance.
IMO if it was simply a matter of linearity, as in local feedback around an outstage, there is no issue at all of the feedback being "too late". Rather, the feedback is only too late if it's picked off after an amplifier stage that is too slow, and feed back in attempt to speed it up.
Cheers,
Michael
Yes, if you overdo NFB in an attempt to force an amp to do that which it cannot do then you hit trouble. You can make an amp move faster than its open loop bandwidth would suggest, as servo theory says, but in real life even if you don't get slew-rate limiting you may get extra nonlinearity. The NFB then has extra work to do, undoing its own effects.
The other problem which arises with many transistor amps is that because of the low dominant pole the remaining distortion residuals are, in effect, differentiated so they become more spiky. To put it another way, you get less NFB at high frequencies which is exactly where you would like more NFB. Most valve amps don't have this problem, but they have a double-pole from the OPT resonance just above the audio range and often a decidedly flaky LF response due to the many LF poles within the feedback loop.
The other problem which arises with many transistor amps is that because of the low dominant pole the remaining distortion residuals are, in effect, differentiated so they become more spiky. To put it another way, you get less NFB at high frequencies which is exactly where you would like more NFB. Most valve amps don't have this problem, but they have a double-pole from the OPT resonance just above the audio range and often a decidedly flaky LF response due to the many LF poles within the feedback loop.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.