In the ideal situation, with as much distortionless forward gain as you like, it is the feedback network which sets the amplifier response. If the feedback network has a loss of 10:1 then the amp has a gain of 10:1 (or maybe 11:1). If the feedback network has a distortion of 0.5% then the amp has 0.5% distortion. That is why we use feedback: typically resistors have much lower distortion than transistors.gabdx said:
The capacitor in question is part of the feedback network, so any distortion here is distortion in the feedback network and so becomes distortion in the amplifier as a whole. You need to be clear: feedback reduces distortion produced in the forward path; it does not reduce distortion produced in the feedback path.
In the ideal case it is the difference between the signals which is amplified (by a lot). Hence you really do not want anything added in here.Tibouchina said:
The feedback signal is about the same amplitude as the input signal.Tibouchina said:
Only when he is talking sense. When he shows that he didn't understand the course he attended many years ago then he needs to be corrected.
Not at all! With an amp that has infinite gain, the feedback signal is exactly the same as Vin. In a real-world amp, there is a small difference between Vin and Vfb.
This difference is exactly the output voltage divided by the amp open loop gain: Vin-Vfb=Vout/Aol.
BTW This also clearly shows that any imperfection in Vfb is amplified tens or hundreds of thousands of times!
Jan
Hello,
About using tantalum;
1) let's have a look at a data-sheet, for ex CTS20 capacitor: Condensateur tantale a electrolyte solide - CTS 20 - CTS 20E - Sonectrad-BRN,
So, you will discover that tg δmax = 6,
Or 108 Ω at 100 Hz.
It's not really a low ESR !
2) And, now this AN from Maxim: Do Passive Components Degrade Audio Quality in Your Portable Device? - Application Note - Maxim
Particularly in fig 8.
So, now, what do think of tantalum ??
HTML:
Always use a tantalum cap in the fblAbout using tantalum;
1) let's have a look at a data-sheet, for ex CTS20 capacitor: Condensateur tantale a electrolyte solide - CTS 20 - CTS 20E - Sonectrad-BRN,
So, you will discover that tg δmax = 6,
Or 108 Ω at 100 Hz.
It's not really a low ESR !
2) And, now this AN from Maxim: Do Passive Components Degrade Audio Quality in Your Portable Device? - Application Note - Maxim
Particularly in fig 8.
So, now, what do think of tantalum ??
Tants have their place, but it is seldom audio and modern electrolytics have stolen much of their market.
Basically back in the day tants were effective capacitors to about a decade or two higher in frequency then the electrolytic caps of the time were really good for, this is largely no longer the case, a modern reasonably small electrolytic cap designed for switched mode supply output service is as good in that respect. They are still (sometimes) a goto option in high temperature applications.
The other major virtue was low leakage, which is still a reason to use them, useful in things like PLL loop filters where you need the value but don't want the microphonics of a class II dielectric.
Apart from that, mostly meh.
Basically back in the day tants were effective capacitors to about a decade or two higher in frequency then the electrolytic caps of the time were really good for, this is largely no longer the case, a modern reasonably small electrolytic cap designed for switched mode supply output service is as good in that respect. They are still (sometimes) a goto option in high temperature applications.
The other major virtue was low leakage, which is still a reason to use them, useful in things like PLL loop filters where you need the value but don't want the microphonics of a class II dielectric.
Apart from that, mostly meh.
If input voltage and feedback signal voltage were exactly equal, the amplifier would have only a null voltage to amplify.
|Vin|-|Vfb| > 0
It is a question of limits (in the calc. sense), as open loop gain approaches infinity, Vin - Vfb tends towards zero.
Of course the thing that seldom seems to be discussed is that while feedback reduces the effects of non linearity (A LOT, if you have a lot of it) it also makes the products very much more complex.
If you are using feedback (and you really should be) then you really, really want to be using a lot of it.
Consider a fairly mundane open loop stage having a non linear transfer function say Vo=1000(Vi + 0.01Vi^2), so maybe 1% of second harmonic and intermod, and 60dB of open loop gain.
Now lets wrap it is say 10dB of feedback, so the gain is now 50dB, our 1% second harmonic distortion has dropped 10dB and is now 0.3%, but now there is some third harmonic, that second harmonic coming back thru the feedback path mixing (due to the non linearity of that V^2 term) with the wanted signal, and that comb of harmonics now extends all the way out! It is small but it was not there before. Worse the intermod spectrum when not amplifying sinewaves just got a LOT more complex.
The answer of course is that you should be using WAY more then 10dB of feedback, if the open loop gain was say 50+dB above the closed loop gain then the mess of multiplication products would be pretty much negligible.
This is a nice trap with opamps as they are basically integrators and open loop gain typically falls at 20dB/decade from somewhere pretty close to DC, the intermod concern means that it is NOT enough to have open loop gain at 10KHz, you really want to have available loop gain a decade above that.
Of course the thing that seldom seems to be discussed is that while feedback reduces the effects of non linearity (A LOT, if you have a lot of it) it also makes the products very much more complex.
If you are using feedback (and you really should be) then you really, really want to be using a lot of it.
Consider a fairly mundane open loop stage having a non linear transfer function say Vo=1000(Vi + 0.01Vi^2), so maybe 1% of second harmonic and intermod, and 60dB of open loop gain.
Now lets wrap it is say 10dB of feedback, so the gain is now 50dB, our 1% second harmonic distortion has dropped 10dB and is now 0.3%, but now there is some third harmonic, that second harmonic coming back thru the feedback path mixing (due to the non linearity of that V^2 term) with the wanted signal, and that comb of harmonics now extends all the way out! It is small but it was not there before. Worse the intermod spectrum when not amplifying sinewaves just got a LOT more complex.
The answer of course is that you should be using WAY more then 10dB of feedback, if the open loop gain was say 50+dB above the closed loop gain then the mess of multiplication products would be pretty much negligible.
This is a nice trap with opamps as they are basically integrators and open loop gain typically falls at 20dB/decade from somewhere pretty close to DC, the intermod concern means that it is NOT enough to have open loop gain at 10KHz, you really want to have available loop gain a decade above that.
No, the power supply is not in the signal path. The PSRR tells you how much it is seperated. And the DC servos are not signal path, the signal is higher freq.
Yes this is a trap many tube designers fall into. They say, 'oh, I only use moderate feedback, just 10dB', not realizing it makes matters worse than without fb.
Bob Cordell has some nice graphs in his book (1st Ed, don't know if they are still in the 2nd Ed.) that nicely illustrate this. I posted them once here, many moons ago ...
Jan
Clamp an AC current probe on the wire between the output device and the power supply
filter capacitor. What do you see? Compared to the input signal, it looks pretty similar.
Of course, one half may be chopped off, depending on the class of operation.
The servo is just a first order LPF, set to around 1Hz or less. Some signal does get through.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.