I've not decided yet if the C3-R18 (~18kHz) combo will act as a zero or a pole together with Q10. Reminds me of the Colpitts oscillator.
Thanks guys, I'm reading and now I know something more than before.
About R18... I have read about it in a book (maybe by Douglas Self) and I found interesting. It's not a big problem I can simply not solder it if necessary.
I can also change capacitors to get good corner frequency.
I will use post 14 current sources, with 1/1.25 mA in the tail pair, and well balanced input impedance like suggested in post 4.
Feedback CAP will be increased starting from 220 uF, here I have to chose a tradeoff between corner frequency and PCB space.
About R18... I have read about it in a book (maybe by Douglas Self) and I found interesting. It's not a big problem I can simply not solder it if necessary.
I can also change capacitors to get good corner frequency.
I will use post 14 current sources, with 1/1.25 mA in the tail pair, and well balanced input impedance like suggested in post 4.
Feedback CAP will be increased starting from 220 uF, here I have to chose a tradeoff between corner frequency and PCB space.
Some more thoughts -
regarding the current sources, your more formal approach should and do work. But in practice, ripple on the supply lines will get into the amp unless the feed resistors are decoupled with either a capacitor or Zener diode. In fact they are better than the usual two diode approach as in principle the use of one junction to compensate another should give better thermal stability. This is a trade off because the VAS CCS transistor dissipates more than a diode-connected transistor, so either approach needs the devices to be thermally coupled. Often it is not done and the thermal drifts assumed to be small enough using AC feedback (using a decoupling capacitor as you have). The voltage drop across the emitter resistors in your circuit need not be particularly large. 600mV drops across them would be suitable.
I have simulated a version of your circuit and it seems it should work. However, 464 ohms is rather low. In itself that is not a problem, but the 22 ohm degen resistor in the VAS means that the current gain in the stage will be limited to about 22.
Resistor R18 should push the onset of the roll-off of the Miller capacitor upwards, but otherwise leave the high frequency roll-off largely as it was. It is not a bad idea, but once again the biggest problem is that it could introduce ripple from the supply, and in this case decoupling is not a solution, only a filtered and stabilised rail.
Using R18 gives local feedback but burdens the input stage further. You may find a better performance by connecting the 464 ohms (470 in the E24 range) between the emitter and base of the VAS rather than to Vcc which will increase gain, but make the 22 ohm less effective. You could keep a resistor there and add two diodes as a current limit for protection.
regarding the current sources, your more formal approach should and do work. But in practice, ripple on the supply lines will get into the amp unless the feed resistors are decoupled with either a capacitor or Zener diode. In fact they are better than the usual two diode approach as in principle the use of one junction to compensate another should give better thermal stability. This is a trade off because the VAS CCS transistor dissipates more than a diode-connected transistor, so either approach needs the devices to be thermally coupled. Often it is not done and the thermal drifts assumed to be small enough using AC feedback (using a decoupling capacitor as you have). The voltage drop across the emitter resistors in your circuit need not be particularly large. 600mV drops across them would be suitable.
I have simulated a version of your circuit and it seems it should work. However, 464 ohms is rather low. In itself that is not a problem, but the 22 ohm degen resistor in the VAS means that the current gain in the stage will be limited to about 22.
Resistor R18 should push the onset of the roll-off of the Miller capacitor upwards, but otherwise leave the high frequency roll-off largely as it was. It is not a bad idea, but once again the biggest problem is that it could introduce ripple from the supply, and in this case decoupling is not a solution, only a filtered and stabilised rail.
Using R18 gives local feedback but burdens the input stage further. You may find a better performance by connecting the 464 ohms (470 in the E24 range) between the emitter and base of the VAS rather than to Vcc which will increase gain, but make the 22 ohm less effective. You could keep a resistor there and add two diodes as a current limit for protection.
I think mtx's circuit would suffer pretty bad offset and drift, on the order of 10's of mV, but I can't see it drifting to the rails.
Ronaldo's fixes in post 16 are good. but I think he should have omitted R1 and R4 from his drawing, because those are potentially confusing to mtx. I especially like how he replaced the CFP output smoke bombs with a dual EF with much more stable bias current.
I personally would delete the 330K resistor across the VAS, because I don't like limiting VAS gain. But it is a good experiment to see what effect it has. I see lots of VAS with a resistor there.
Ronaldo's fixes in post 16 are good. but I think he should have omitted R1 and R4 from his drawing, because those are potentially confusing to mtx. I especially like how he replaced the CFP output smoke bombs with a dual EF with much more stable bias current.
I personally would delete the 330K resistor across the VAS, because I don't like limiting VAS gain. But it is a good experiment to see what effect it has. I see lots of VAS with a resistor there.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.