With a differential pair with folded cascode, the relation between input voltage and output current is still tanh(vin/(2kT/q))-shaped, rather than something a bit straighter.
Id argue that the feedback transistor in the differential has to be faster and have higher transconductance than the input and rely more on the feedback to fasten the seat belts .I dont see the argument for a linear approach in the differential when using feedback...I believe theres a prevalence of lowering distortions over the objectively perceived noise in the apt-holman aproach, although an argument for the cheapness of the circuit can be readily made too....
Why do you need complicated solution if the distortion can be corrected using negative feedback? In folded cascode, we can say in the LTP almost have no gain, so it is depend on the VAS. If LTP loaded by current mirror then the gain can be very high and need resistor emitter to reduce the gain and make LTP more linier and slew rate can increase.
Yes, that'd work, a configuration to keep in mind. However, you'd loose the matching of the transistors being on the same die and therefore at the same temperature.
I note that the symasym design has no emitter degeneration. So I'd be interested to know how the symasym handles large differential input signals? Does the distortion go through the roof in this case?
Hi Marcel
I think I would go for Fig. 9 without the connection from the Bases of T1A and T1B in the patent that you are linking to.
I have only skimmed through the patent, but its a bit strange that "multitanh" isnt mentioned.
Stein
Nonsense
What you are posting has absolutely nothing to about linearization of the LTP.
Depend on the load impedance of the LTP. I think you do not understand about folded cascode topology that VAS is common base.
Of course it is nothing to with linearization of the LTP, but it can use LTP without degeneration with simple solution. And if design properly, folded cascode topology can have low enough distortion and high slew rate.
I think it's time for you to stop saying "I think you do not understand"
I think that he knows very well what a folded cascode is.
In my opinion it seems like you have problems undrestanding a thing or two.
In any case, it has absolutely nothing to do with the subject of this thread: linearizing differential pairs.
Do we not understand why we degenerate the LTP? It is not about distortion in general or linearizing differential pairs. It is actually a compromise at lower frequencies. The problem comes from slew induced distortion where the LTP clips before the VAS can respond to a fast edge. Reducing the gain by reducing the LTP current just aggravates the problem because the slew rate is a function of the LTP current. If we could put the frequency compensation ahead of the LTP then there would be no point in degenerating the LTP.
A degenerated LTP is the probably the best solution we have, so if the emitters are internally connected then the best solution is not available, and we must accept a compromise. If you limit the scope of acceptable solutions, then you forgo finding any solution.
The expression "the tail wagging the dog" comes to mind but we see a lot of that in DIYA.
The value of a perfectly matched LTP is not worth compromising the circuit topology. If you are obsessed with DC offset, use a servo. The CFA folks are having a chuckle.
A degenerated LTP is the probably the best solution we have, so if the emitters are internally connected then the best solution is not available, and we must accept a compromise. If you limit the scope of acceptable solutions, then you forgo finding any solution.
The expression "the tail wagging the dog" comes to mind but we see a lot of that in DIYA.
The value of a perfectly matched LTP is not worth compromising the circuit topology. If you are obsessed with DC offset, use a servo. The CFA folks are having a chuckle.
I agree. but they said "it has absolutely nothing to do with the subject of this thread: linearizing differential pairs".