R29 is a rudimentary current source for the voltage reference (D3). For a simple resistor current source ideally the voltage across it to be constant, but that cannot be the case, because the rail voltage fluctuates. If you refer the midpoint to the output node, then the voltage will change with the signal - not what we want.
A better idea would be to split R29, take a small electro, say 10 to 47uF, connect the +ve terminal to the +ve rail, and the -ve terminal to the midpoint of the series resistors. The stored charge in the cap works helps stabilise the voltage (and as a consequence the current flow) across the resistors due to rail fluctuations.
Edit; I think improving the LTP CCS is the wrong direction to take. The problem with the basic design is that it doesn't have a lot of loop gain to start with and with the compensation as shown the GBW is closed well below 1 MHz. I would increase the gain of the LTP by reducing those degen resistors to about 10R and then adjusting the compensation to suit. Alternative you might buffer the VAS to increase its gain or perhaps try a CFP VAS if you'd prefer to linearise that stage without increasing the overall OLG. Either of those last two changes would make a very significant difference for the cost of one TO92 and one resistor.
A better idea would be to split R29, take a small electro, say 10 to 47uF, connect the +ve terminal to the +ve rail, and the -ve terminal to the midpoint of the series resistors. The stored charge in the cap works helps stabilise the voltage (and as a consequence the current flow) across the resistors due to rail fluctuations.
Edit; I think improving the LTP CCS is the wrong direction to take. The problem with the basic design is that it doesn't have a lot of loop gain to start with and with the compensation as shown the GBW is closed well below 1 MHz. I would increase the gain of the LTP by reducing those degen resistors to about 10R and then adjusting the compensation to suit. Alternative you might buffer the VAS to increase its gain or perhaps try a CFP VAS if you'd prefer to linearise that stage without increasing the overall OLG. Either of those last two changes would make a very significant difference for the cost of one TO92 and one resistor.
Last edited:
Yeah, that's exactly what e.g. German manufaturer Klein & Hummel did in the late series of their E120N amplifiers for public adress purposes. They provided a discrete true darlington for the VAS.
I suspect that even this idea could be improved by going into a quasi complementary VAS stage (npn small signal and PNP power devices, respectively). Thus the diff input stage's design won't need no adaptation at all.
Btw, what's the purpose of trimpot R10 in the tail CCS? Which current (in mA) will be adjusted with it? Why no fixed resistor instead?
Best regards!
Last edited:
The trimpot at R10 allows the tail current to be adjusted in order to null any DC offset. I'm not a huge fan of this technique because it alters the compensation, and I just generally dislike trimmers :-D
Better to use a trimmer at R8 and better still replace it with the closest E24 or E96 series resistor once the optimal value is found. Or do I as do, calculate the value, choose the nearest E12 and just live with a little DC offset. It won't damage the speakers.
Better to use a trimmer at R8 and better still replace it with the closest E24 or E96 series resistor once the optimal value is found. Or do I as do, calculate the value, choose the nearest E12 and just live with a little DC offset. It won't damage the speakers.
Thanks! But how does this work? As I see it, the common tail current divides into two, thus the same influence on both transistors should have to be expected.
Best regards!
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.