Two important points. Don't forget that, in a CFB amp, changing the serial feedback resistance value will change the slew rate. Changing the emitter resistance value will change the current in the input transistors.
So if you want to add some gain, the best is to replace the serial resistances by an attenuator, keeping the impedance seen by the input transistor near the same.
By example, if your original serial resistance is 1K and the emitter 50 Ohms, from the output, do 1Kserial +1k to gound and 500 Ohm to the 50Ohm.
So if you want to add some gain, the best is to replace the serial resistances by an attenuator, keeping the impedance seen by the input transistor near the same.
By example, if your original serial resistance is 1K and the emitter 50 Ohms, from the output, do 1Kserial +1k to gound and 500 Ohm to the 50Ohm.
Last edited:
Hi Esperado.
With a reasonable bandwidth, reducing the emitter resistor will not cause any high currents in the audio range. But i have and will recommend any one to have a lowpass input filter.
For two reasons:
- Reduce emi from the line signal - cellphones in particular
- Reduce slewrate allowed even if the amp is capable of more. It stresses the transistor and zobel network a lot with "extreme" slewrates.
By leaving the Rf value and reducing the emitter resistor you will raise the loop gain... So by adding +6dB system gain, this way the loopgain stays the same.
I will maybe change the DOC to say that they should start out without the bypass caps in the feedback network.
They can always add them later.
I too, of course. But keeping the ratio signal speed/amp speed as little as possible is the target. If no stability issue, i would reduce the serial feedback resistance instead. Keeping a low-pass filter at 200Khz before the input.
Last edited:
May-be i need to precise my point. Decreasing the emitter resistance of the first stage will increase the gain of the input stage, as its collector one will stay the same. If the amp has been carefully designed for each stage produce the same distortion in open loop, you will increase-it. And the quiescent current will be increased as well. Right ?
That's why i recommended to not touch this resistance.
On the contrary, reducing the serial feed-back resistance will stay the input stage unchanged, and reduce the influence of parasitic influence near the emitter of the input transistor because impedance will be reduced.
That's why i recommended to not touch this resistance.
On the contrary, reducing the serial feed-back resistance will stay the input stage unchanged, and reduce the influence of parasitic influence near the emitter of the input transistor because impedance will be reduced.
No the quiescent current will not increase, It is controlled by the current source and Rf. The emitter resistor is AC coupled to GND.
That is correct but you will raise the bandwidth. And my test showed that this is NOGO . 33pF and Rf=1K is the limit. Reducing Rf means larger Millercaps ....
Can't you add a // cap with the feedback resistance to keep the closed loop bandwitch the same ??
For the quiescent, my amp use a ccs referenced to rail, so we are not in the same situation, you are right.
Last edited:
If it is okay, you can show me the circuit. Just so that i can follow you.
No as with rising frequence the cap will start act as a smaller Rf. And it currentfeedback this oscillation will start
Well, i understand better your choice, now. I would try a 2K for R11 - R12 and a little paralleled cap with it for stability and no overshoot on square waves, if possible.
The thing i worry, looking at your schematic is the one octave upper high pass filter frequency if we decrease the R13 -R14 from 100 to 50, right ?
You have better experience on it, anyway... I suppose you have tried near everything ?
The thing i worry, looking at your schematic is the one octave upper high pass filter frequency if we decrease the R13 -R14 from 100 to 50, right ?
You have better experience on it, anyway... I suppose you have tried near everything ?
You are right. I remembered wrong. It is 47pF millercaps. Withe 33pF we starts to have small signs of oscillation driving capacitive loads.
There is always something new to try.
My problem is that my time flyes away. Right now i am doing a stepup LED driver for a new display for a customer as the OLD display was EOL!!! It is 23:37 .. stupid stupid .. When you can make audio instead....
You have to focus on the RF (R11 and R12) combined with the millercaps (C12 and C13).
They set the bandwidth. Do not place a cap across R11 and R12 . It will oscillate right away with a capacitive load.