Input stage idea

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hello,
recently I've been thinking about best possible input stage for my amplifier and I've got an idea. I don't know if it's something new, or already used before, but I haven't seen anything similar yet.
Principle is in differential input stage fed from current source, but output is attached to 2 current mirrors and then to another current mirror. I've made this with my old mosfets (IRF540) and everything seems that it works, but I don't have anything to measure performance. So I'm trying to ask you (proffesionals (some of you)).
Schematic is attached (i hope)
 

Attachments

  • amp.png
    amp.png
    69.1 KB · Views: 849
That topology is very much like an OTA, operational transconductance amplifier. It's a voltage to current amplifier. If you observe, with zero differential input, you'll have zero current going through the load.

If you want to use it for your input stage, you'll have to size a load resistor to set your voltage gain.

One thing you should notice is that the output resistance this looks like the parallel combination of two current mirrors which is fairly high. Good for current source, bad for voltage source.
--
Danny
 
Thanks for your replies.

If this is used in operational amplifiers, I think that overall performance is good.
High output impedance is not a problem, I'll be driving power mosfets directly from this stage and I've set circuit currents quite high (200mA for current source, and 40mA AC for driving output mosfets). Output stage is ciclotron (so I'll need this circuit twice, inverting and non inverting).
 
Of course I'm not going to use IRF540 for an input stage. I just had a dozen of them, so I've been testing. I'm going to use IRF610 & IRF9610, maybe in combination with BC550 or some similar small-signal bjt (don't know how you call that circuit when small bjt controls big BJT or FET)
 
blu_line said:
I am thinking of using the same ckind of circuit.
My main problem is how to match the upper current mirrors and the current source in the tail of the diff-stage !


grtz

Simon


Current mirrors must consist of matched pairs of transistors and they must be thermaly coupled, because one side has much more power dissipiation. If they were not matched and thermaly coupled distortion will occur and there will be more DC component in the output.

I use the same source resistors in current mirrors as in differential stage, so every part of circuit works with the same current and the same Vgd - distortion will cancel.
 
Beware that the capacitance (possibly Miller capacitance) of the output stages will come in parallel with the output resistor. Make sure that pole does not dampen your high-frequency signal or give you oscillations if you choose a wide feedback.

The output voltage as a function of input voltage of this circuit can be tanh-shaped. That means flat-linear-flat. Choose your resistors wisely to get both gain and linear range.

I usually visualize linearity simulations by plotting d(Vout)/d(Vin) in PSpice.


--
Børge
 
Christer said:


Noise is random and doesn't cancel.


If you mix 2 random noise signals of the same level, noise will cancel each other (of course no completely, but lower, I think tat it's about -6dB). It's like 2 non linear transistors in differential stage where distortion cancels (is lower). (Electronical basics)

It's similar to dithering in DAC, you add some noise of exactly specified level to lower quantization noise.
 
No, two uncorrelated noise sources add in rms fashion. Assume
two noise voltages x and y. The resulting noise will be
sqrt(x^2 + y^2). The noise genereated in two transistors are
completely independent of each other and thus uncorrelated.

Edit: I have looked around for good references on noise, but
I have a tendency to download documents to my computer and
then not know where they came from. However, one good
paper is App. note AN-104 from National Semi. which is about
noise in op amps.
 
BTW:why are we talking about noise, this input stage uses the same amount of transistors or less than any other amp, only one stage amplify signal, current mirrors are set to gain=1. I use source resistors in current mirrors for better linearity. Nelson Pass uses the same kind of transistors and claims 155dB dinamic range, but I've read a review in Stereophile, where they measured s/n about 80dB and nobody cares about it (and JA wrote that noise figure is good compared to others). Others claims s/n 130dB, but if you measure it, you'll se 80bB.
You can use bjts for this amp and have "no problems" with noise.

So I don't know why are we talking obout noise.
 
This design has no input stage gain, so the dioded mirror fet and the output fet on each side will add their noise essentially equally to the input fet. This can be prevented by using low value input source resistors and high value load resistors, but the circuit performance will then be compromised.
I normally get paid to do a complete analysis of something like this, so I can't give details, but you could do a Spice simulation and find the exact answers.
This is a subtle design detail and is usually missed by beginners. I missed it back in 1969, when I added a fet current source as the first stage load for a tape recorder. It got VERY noisy. This was because the fet current source amplifies its own noise as if it were an independent source. A similar problem happens with the 741 op amp. The load transistors add significant noise to the overall input noise.
 
john curl said:

This is a subtle design detail and is usually missed by beginners. I missed it back in 1969, when I added a fet current source as the first stage load for a tape recorder. It got VERY noisy. This was because the fet current source amplifies its own noise as if it were an independent source. A similar problem happens with the 741 op amp. The load transistors add significant noise to the overall input noise.

And here we get another gem for free to jot down in our
notebooks. Thanks John.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.