What defines the bias current through Q16 and Q4? Doesn't it spread all over the place when there is some mismatch in the input differential pairs?
And a capacitor. It works, the resistor might need to be quite large for the dissipation, and there aren't many linear regulators that can handle that voltage, and this is probably the cheapest approach and current limits nicely.Regulating 60VDC down to 15VDC using one resistor and one Zener diode? Really?
Works greatRegulating 60VDC down to 15VDC using one resistor and one Zener diode? Really?
Feedback works greatWhat defines the bias current through Q16 and Q4? Doesn't it spread all over the place when there is some mismatch in the input differential pairs?
OK, then where is the feedback loop that prevents the currents through Q16 and Q4 from increasing or decreasing together? I don't see a common-mode loop that does that.
looks like normal emitter followers to me.
with emitter resistors and sharing same heatsink.
MJE340/350 is questionable choice.
as with any design everyone will nit pick it
with emitter resistors and sharing same heatsink.
MJE340/350 is questionable choice.
as with any design everyone will nit pick it
Emitter resistors don't help much when the base voltages are undefined, do they?
I think that there is a good chance that the amplifier will either create cross-over distortion or go up in smoke, depending on the signs of the offsets of the input differential pairs. You can call that nitpicking if you like, but I don't think that's appropriate. I also think the issue can be solved by adding a common-mode loop that senses the sum of the currents through Q16 and Q4 and regulates that to some target value.
I think that there is a good chance that the amplifier will either create cross-over distortion or go up in smoke, depending on the signs of the offsets of the input differential pairs. You can call that nitpicking if you like, but I don't think that's appropriate. I also think the issue can be solved by adding a common-mode loop that senses the sum of the currents through Q16 and Q4 and regulates that to some target value.
Looks like there are two feedback loops in parallel; two different error amplifiers but each one trying to control the same output variable in the same approximate time frame. Might behave something like a coupled pendulum problem due to mismatches in components and or their operating points/conditions, and or any other disturbance that would get it going.
Last edited:
Yes at the VAS a smaller die is preferred, or several in parallel, however MJE340/350, it should also work, but smaller die preferablelooks like normal emitter followers to me.
with emitter resistors and sharing same heatsink.
MJE340/350 is questionable choice.
as with any design everyone will nit pick it
Interestingly the pendulum will work just fineEmitter resistors don't help much when the base voltages are undefined, do they?
I think that there is a good chance that the amplifier will either create cross-over distortion or go up in smoke, depending on the signs of the offsets of the input differential pairs. You can call that nitpicking if you like, but I don't think that's appropriate. I also think the issue can be solved by adding a common-mode loop that senses the sum of the currents through Q16 and Q4 and regulates that to some target value.
Let me elaborate, the used solution is the cleanest solution as a buck converter or a linear step down supply attached would be too much hereRegulating 60VDC down to 15VDC using one resistor and one Zener diode? Really?
Attachments
2R+1C+1Zener would increase PSRR by 20X. Cost increase: 1 new resistor and 1 new capacitor per rail.
2R + 2Zener would increase PSRR by 100X. Cost increase: 1 new resistor and 1 new Zener per rail.
1R + 1 JFET + 2Zener would increase PSRR by 300X. Cost increase: 1 new resistor and 1 new JFET per rail.
"Works great" with 1R and 1Zener ... passes up some very cheap opportunities for much better performance.
2R + 2Zener would increase PSRR by 100X. Cost increase: 1 new resistor and 1 new Zener per rail.
1R + 1 JFET + 2Zener would increase PSRR by 300X. Cost increase: 1 new resistor and 1 new JFET per rail.
"Works great" with 1R and 1Zener ... passes up some very cheap opportunities for much better performance.
Seen that in some of your other posts with parallel small dies or T0-92Yes at the VAS a smaller die is preferred, or several in parallel, however MJE340/350, it should also work, but smaller die preferable
Something I thought of trying as well.
sure many would overreact to it as well.
But was interesting to see.
Far as this design
common 3503/1381 could be used for Q4,Q16, Q11, Q13, Q17, Q18
Think it would really wake up the performance.
Be interesting to see the numbers in same sim
Interestingly the pendulum will work just fine
If you already built a dozen or more of them, checked the bias currents of Q16 and Q4 for each of them and found that the spread is perfectly acceptable, then I agree. In that case, apparently the offset of the input stage and the current gain of the second stage is small enough for the conceptual issue I found not to be a big deal in practice.
If this design only exists as an LTSpice simulation, then I would advice you to put a DC voltage source in series with the base of one and only one of the four input transistors, do a DC sweep where this voltage source is swept from -50 mV to +50 mV and see what happens with the DC currents through Q16 and Q4. It's a simulation that can be run in a few minutes and it will give you an impression whether the issue I mentioned is a big deal or not.
I just now saw this thread. A mismatch in the differential pairs will cause Q4 and Q16 to have zero current or to burn out.
The differential pairs loaded by current mirrors have too much gain. A "feature" of the fully complementary amplifier is that the gain must be relatively low in order for the two halves to track.
Ed
The differential pairs loaded by current mirrors have too much gain. A "feature" of the fully complementary amplifier is that the gain must be relatively low in order for the two halves to track.
Ed
MarcelvdG - You are right. Leach knew that and everyone who follows Leach's approach also knows.
Ed
Ed
I think it could be fixed with a common-mode loop, by the way. Spoiling the gain need not be the only solution.
A common-mode loop could stabilize the currents in Q4 and Q16, but the gain will still be poorly-controlled. The gain needs to have tight tolerance as well. This is a consequence of having two amplifiers connected at their outputs.
ETA: Relatively low gain is ~500.
Ed
ETA: Relatively low gain is ~500.
Ed
Last edited:
I agree with the guys expressing concerns. Was just thinking about it in a different way, which would be more along the lines of what might be its dynamical response be in the presence of noise? Say, for example, if one kicked one of the control loops to upset the output, then what would the dynamical response of the other control loop be, and how would the two loops settle (assuming the excited mode is sufficiently damped). Even if stable, what might the effect on sound be, depending on particular mismatches?
- Home
- Amplifiers
- Solid State
- NOT Blameless Amplifier (Red)