Have you ever succeed in buiding a input stage with current mirror?

[thanh]

D.Self said that current mirror will increase slew rate for 2 times, I use some discrete transistors but isn't successful for 2 times
So how can I do?
If I use array transistors ,will I be successful?

[robert_chien]

I am not sure how it is not successful. Could you give more detail?

Current mirror is seldom, as I know, used in an audio amplifier.
It is usually used in situations that need very high loop gain, such as comparator. This does not inhibit using it in an audio amp. Whereas care has to be taken in the whole amp design such that it is stable.

[Ouroboros]

Current mirrors are nearly always used these days as the load for the collectors of the input long-tailed pair. The current mirror effectively gives a push-pull drive to the next stage. This approximately doubles the current swing available, thus increasing the slew rate.

How did you implement the current mirror? If you are using discrete transistors, what value of emitter resistors did you use?

[MikeB]

hi thanh !

I tried currentmirror in my actual design, and had no big problems.
I matched 2 bc556b, added 100ohm emitterresistors.
The only problem was to get the amp stable again. As predicted
from sims, currentmirror did not really add stability.
I got it stable with a RC, 100pf+330ohm, directly connected
between the collectors of the diffamp, which also did a great job
in reducing noise.

What exactly was your problem with the currentmirros ?

Maybe you used currentmirrors in a full complementary design ?
This works in sims, but i can't imagine how they should work
in real world, as you have no "reference"-current. So the output
from the diffamps should be unpredictable.
Is this your 20bjts-design ?

My actual design is a simple asymetrical single diffamp. (+cfp)

Mike

[jam]

Question is, why use a current mirror if you don't really need it. If not used carefully they can screw up the sound. Case in point, try a current mirror on the Aleph circuit. Distortion goes down and so does the quality of sound.

Then how about the stability problems you might have to correct.........and the sonic penalty of the compensation of the compensation components that you have to use.

This is not to say current mirrors are all bad.......then can work well in a folded cascode. Slew rate is an overated spec. anyway ( not to mention D.Self )

[Bruno Putzeys]

There's nothing wrong with current mirrors per se. The problem is that they usually figure in Miller amps. A miller amp explicitly uses the top rail as a reference for the voltage amplifier (the emitter of this transistor is the positive input, y'know). This means that if it weren't for the loop gain of the closed loop system, PSRR on that rail is literally zero. A PSRR vs frequency plot on an op amp (for the rail the VAS sits on) usually resembles the loop gain plot to a remarkable extent. Not just the shape of course, but also the absolute value).

A miller amp is a good way of insuring the power supply puts its signature on the sound. This is not something to be blamed on the current mirror.

The better way of designing an amp is building a transconductance stage (the input pair followed by an arrangement of current mirrors and/or folded cascodes to get push-pull operation), followed by a compensation network that -crucially- refers to ground, followed by a buffer stage (the power stage) with the highest possible input impedance (AC-wise).

PSRR of such amplifiers is easily 20dB better than of Miller amps, meaning 20dB less of power supply colouration. The reason why this arrangement is so rarely (if at all) used in IC op amps is simply because it would require an extra pin to be connected to ground.

The loop gain becomes transconductance times compensation network impedance (in parallel with the buffer's input impedance). Don't overdo on the transconductance, because you'll have to compensate most of it away if you don't watch out. That would cost unnecessarily in slew rate.

It's simple really

[cunningham]

Works great...if your rails are regulated to a steady DC. They are however acurate.

[thanh]

Is this your 20bjts-design ?

yes . Thank everyone!
I have not added any emitterresistors yet. Vce of current mirror goes round .I have just added a 12 ohm emitterresistor. The result is quite good . How is the best value of it ? My current source is 2mA ,

[thanh]

I mod Slone's design which posted by Lumanuaw at "dual differential or single differential"

[andy_c]

Originally posted by Bruno Putzeys
The better way of designing an amp is building a transconductance stage (the input pair followed by an arrangement of current mirrors and/or folded cascodes to get push-pull operation), followed by a compensation network that -crucially- refers to ground[...]

As pointed out by Self (who references Baxandall) the ground-referenced compensation cap gives poor distortion performance. Due to the decreasing capacitor impedance with increasing frequency, the open-loop distortion of the VAS will increase with increasing frequency. This, combined with the decreasing global feedback as frequency increases gives a "double whammy" to the distortion performance. OTOH, a Miller-compensated VAS has decreasing open-loop distortion as frequency increases, due to the increasing local feedback through the compensation cap as frequency increases. This nearly cancels with the decreasing global feedback as frequency goes up, giving a total VAS distortion contribution that's nearly independent of frequency. Improved PSRR can be achieved using a circuit similar to figure 8.6 in Self's book. I've shown a picture of a similar circuit below. Connecting C9 to the collector of Q1 instead of the base of Q13 gives a large improvement in PSRR. See Self for more details. I believe he got this idea from "A General Relationship Between Amplifier Parameters, And Its Application to PSRR Improvement" by Sackinger, Goette and Guggenbuhl, IEEE Transations on Circuits and Systems Vol 38, #10, Oct 1991 (thanks jcx for originally pointing out this paper to me). Press email button for more details. (Not shown below is a DC voltage source biasing the bases of Q3 and Q6).