Hello again,
Ok, when you increase the gain past a certain point the output of the op amps (one or more of them) may saturate, which could cause problems. This means when you change the two 10k resistors on the output of the two input buffers to 1k then you should use an input test voltage of 0.1v instead of 1.0v. That means we should still get 10v peak on the output even with resistors 10k and 100k for the output stage resistor pairs.
BTW i did verify that a lone LM358 with the two current mirrors does in fact get it's apparent slew rate increased as per our previous discussion, where the voltage gain is increased because of the current mirrors and choice of output resistor and current mirror output resistor.
This means that i was able to pass a 50kHz signal through the amplifier with a voltage gain of 10. This would be unheard of for just an LM358 with no additional parts (other than gain resistors) as that would only go up to about 8kHz with 10v peak output, and anything higher in frequency would cause severe distortion.
So we see one benefit from using the current mirrors, the slew rate has increased 10 fold. I did not try any higher but i have a feeling we could go up more yet and still get acceptable operation.
Of course there is still the DC input offset but that's another issue which may or may not affect the end application.
Using values of 1k and 9k to get a non inverting gain of 10 with a lone LM358 (no additional parts) and a 10v peak output, the max frequency is close to 8kHz before distortion sets in. With the current mirrors and the LM358 used as a voltage follower, the max frequency goes up to about 80kHz, which is a 10 fold increase in usable frequency and that's pretty significant.
This is still in theory, so there may be practical issues such as current mirror transistor matching and stuff like that, but it looks very promising.
I could show a simulation picture but it just looks like two sine waves, one is the real output of 10v peak at 50kHz and the other is the 1v peak input voltage synthetically amplified to 10v peak just for direct comparison of the input vs the output waveforms. They look identical except for the small output DC offset caused by what i think is the input offset of the LM358 itself.
I might try to take this further up in gain using the current mirrors.
One side issue is that the last stage of your current circuit is using an op amp at the output too, with a gain other than one. If we can change that gain to 1 we can see a higher bandwidth for this circuit provided we can make the gain of 10 elsewhere in the circuit.
The theoretical maximum frequency would be the unity gain frequency of the op amps, but at whatever gain we can get it up to before secondary factors set in, which means we would have effectively increased the gain bandwidth product.