Considering its age, I bet the 741 existed as a breadboard of discrete components and transistor arrays before the first monolithic prototype was made.
I am not trying to replicate exactly the features of the LM741 but instead i am trying to find out ways of redesigning specific parts in order to improve things and especially noise in high gain applications.
Would simulation help with this?
There is a 741 model in the Educational files included with LT and we can configure it to show noise just by adding this line and labelling the output node:
It is what it is but you get some ideas and can see how noise changes with gain and so on.
Just a thought.
There is a 741 model in the Educational files included with LT and we can configure it to show noise just by adding this line and labelling the output node:
Code:
.noise V(out) V3 dec 100 1k 100kIt is what it is but you get some ideas and can see how noise changes with gain and so on.
Just a thought.
There's an internal LM741 example schematic in Ltspice under educational. If you look at the PNP's as in #6 have hfe of only 25. So your not going to get the same result if you build discrete. You could google discrete op-amp circuits and find something more up to date. Sorry just repeated Mooly.
Q10 is a current source. The common-mode loop Q3&Q4, Q1&Q2, Q8, Q9 adjusts the base voltage of Q3&Q4 until the collector current of Q9 matches that of Q10 (except for the base currents of Q3 and Q4). The current through Q8 is then determined by Q10's current and the ratio between Q8 and Q9.
The differential signal path doesn't involve the collectors of Q1 and Q2, but their emitters. Their emitter currents pass through Q3 and Q4, get subtracted by the current mirror Q5, Q6 with helper Q7, and drive the next stage.
It's a quite unusual input stage. One advantage is that the 741 has twice the slew-rate-to-gain-bandwidth-product ratio that a bipolar op-amp usually has. Another one is that the awfully slow lateral PNPs are only used as common-base stages. Q20 is used as an emitter follower, but because its collector is grounded, that can be a slightly better substrate PNP.
The differential signal path doesn't involve the collectors of Q1 and Q2, but their emitters. Their emitter currents pass through Q3 and Q4, get subtracted by the current mirror Q5, Q6 with helper Q7, and drive the next stage.
It's a quite unusual input stage. One advantage is that the 741 has twice the slew-rate-to-gain-bandwidth-product ratio that a bipolar op-amp usually has. Another one is that the awfully slow lateral PNPs are only used as common-base stages. Q20 is used as an emitter follower, but because its collector is grounded, that can be a slightly better substrate PNP.
I would say the input pair is an emitter follower, hence a darlington. Emitter Q2 driving Base of Q15
Emitter of Q2 driving emitter of Q4, like in a differential pair driven single-endedly, but with opposite polarities, and the same on the other side.
I considered Q4 a cascode, but you are right. Q3, Q4 are the input differential amp, while Q1,2 are simple emitter followers. Anyway, i would not expect any good noise figure from that configuration.
There are four transistors directly 1:1 contributing to the equivalent input noise voltage, versus two in a normal differential pair and one in a single-ended input stage, so in that sense it is not nice if you want low voltage noise. The collector currents are also rather low because of the poor current handling of the lateral PNPs, leading to low current and high voltage noise.