This blog grew out of my getting interested in wideband voltage amplifiers as a result of

doing the transistor load. Something that's fascinating with all this blogging is that,

whatever subject I choose to cover, it naturally seems to lead on to other related material

to explore and experiment with.


With the transistor load, a voltage amplifier formed part of the circuit. That controlled a

MOSFET to adjust the output current, with overall feedback to keep the current accurate. The

overall feedback was essential because the characteristic of the MOSFET at low currents is

quite non-linear and needs to be linearised by the feedback and also, given the kinds of

currents being controlled, temperature changes would also affect the response if the

feedback weren't there to remove them [even with the feedback, the current moved around

quite a lot due to heating in the current sense resistor which it couldn't account for]. An

evident disadvantage of having a multi-stage amplifier with very high open-loop gain and

overall feedback, for all its advantages in terms of precision and accuracy, is that it has

an impact on the closed-loop bandwidth that can be achieved because of the difficulties of

maintaining stability without rolling off the response early.


An alternative approach is to move to having [reasonably] precise feedback locally for each

gain stage, with each having fairly modest gain.


The circuit I'm going to look at here comes from a book by T O'Dell. He published this pair

of books as texts on circuit design about 30 years ago.



They are an interesting try at teaching actual design - how a designer pieces-together a

circuit to do a job. (Note that they aren't 'cookbooks', so if you want complete projects to

build they aren't for you, and neither are they text books, so you won't learn a lot of

basic theory.) Each chapter introduces a topic, discusses possible approaches, and then

leads on to an experimental circuit that could be built and experimented with. Sometimes

there are a couple of related topics in a chapter. The first book [1] (on the left in the photo

- mine is an ex-library copy which is why it is a bit scruffy) is a scatter of different

subjects, but for the second [2] he based it on the theme of test instrumentation and a fair

proportion is inevitably based on how the various sub-systems of a traditional analogue

oscilloscope would have been designed. Because he was targeting education, where the test

equipment then might not have extended to much more than a fairly modest 'scope, his

circuits were often slow versions of the real thing. One problem, though, I've found with

trying to work with the books now is that some of the integrated circuits he uses are no

longer available. This circuit is from the chapter where he discusses the kind of vertical

pre-amplifier that might have been used in an oscilloscope after the FET input stage but

before the high voltage plate-drive section. It's simplified and has limitations but it

allows some basic circuit ideas to be explored, which is what I'm going to do here.


How it Works


Here's the circuit as I've drawn it in the simulator



Although it's his circuit (more or less - I left out the zero adjustment and just tied the

other input to ground through a 47R resistor), they are my component values and aren't

necessarily very good, though it works in the simulator.


Although I'm learning and still not very good with some of this, let's see if I can explain

what is going on (feel free to point out anything that's wrong in the comments below).

There are two ways we can get local feedback with a transistor in order to stabilise the

gain and move it away from the very variable transistor properties: series feedback and

shunt feedback.



The series feedback stage is basically a transconductance amplifier - voltage in, current

out - and the shunt feedback stage is a transresistive stage - current in, voltage out.

[Trans-resistor is where the name 'transistor' comes from, though in some ways it might have

been more appropriate to call it a 'transductor'.] I haven't drawn the second one very well

as I've labelled the base current rather than the current at the input which would be in the

other direction.


An obvious thing to do with this pair is to join them together, so in this circuit both are

used in two successive stages. The first differential stage (T1 and T2) uses series

feedback, developed over R1 and R2 for each respective input transistor. The second stage

uses shunt feedback, utilising R9 and R10 for each of T3 and T4. The amplifier output is

differential, between the two collectors where I've placed a voltmeter on the schematic (but

for the traces below I looked at the output single-ended just by probing the collector of

T3). The differential output amplitude is twice that of the single-ended output. As an

amplifier for an oscilloscope, working with a differential output was natural because,

ultimately, after further amplification, this would be driving a pair of plates within the

tube in order to steer the electron beam.


T1 and T2 both have an input signal applied to the base and the output is taken from the

collector, so they are each in a common-emitter configuration (the emitter is common to both

input and output). [Although, for the purposes of the experimental circuit, T2 has no actual

input connected and is just tied to 0V, pretend that it's an input for this explanation.]

By swapping to NPN transistors for T3 and T4, the dc level can be folded back towards

ground, rather than wandering off and requiring higher and higher supply voltages for

multiple stages.


The small-signal performance looks fairly good in the simulator. Here's a frequency response




It's flat up to about 1MHz and the -3dB bandwidth is about 50MHz. The real circuit won't

manage that because of parasitic capacitances, but it's an indication of how it's making

reasonably good use of the raw 300MHz fT of the transistors.


I don't have a signal generator that I can look at that with, so instead I'm going to focus

on the large-signal response and how it gets on amplifying a square wave at a relatively low

frequency (1MHz).


Performance on a Circuit Board


To see how the circuit behaves, I've built it on a scrap of single-sided PCB material.



The centre strip is 0V, and the top and bottom strips are +15V and -15V respectively.

Normally I use a craft knife to cut the V-shaped groove to separate the different areas, but

this time I stupidly decided to have a go with a hacksaw blade and you can see the mess I

made of the upper cut before I reverted to the knife to finish the cut and do the other one.

The layout is more or less the same as the schematic. The large metal component that stands

above the first pair of transistors, to the left of the board, is the capacitor between the

emitters. This is a trimmer removed from some old radio chassis and has a range of about 10

-35pF, which is ideal for what I want to do here, and means I can see the effect of varying

the capacitance - more on that in a moment. The gungy brown material on the top is the

remains of a hard, waxy substance that was used to hold the position of the trimmer cap

after it had been set in the original circuit.


The input signal is connected to a BNC socket just off the left of the photo - the twisted

pair then brings it to the 47R resistor and the base of the transistor. Other than that,

there is just some decoupling for each rail (a 10uF electrolytic and a 100nF ceramic). For

the output I can just probe straight on to the resistors.


So how does it perform with a signal going in?


Here it is with a 1V pk/pk square-wave input [yellow trace] at approximately 1MHz. The

output [blue trace] is ac coupled and the capacitor has been trimmed for what I considered

the best wave shape. This is the waveform at one of the output collectors viewed singled-

ended with respect to ground [I chose the one that is in phase with the input].



Here is the positive edge in more detail



The gain is slightly over five. It follows the input quite nicely, with a delay of about

15nS, but can't quite keep up. The wave shape at the top where it comes in to land is

reasonable - it's just a fraction underdamped.


Here it is on the falling edge.



The reason it's a bit of a mess at the bottom is because it's faithfully following the input

from my cheap function generator which is also a bit of a mess.


What's the effect of changing the capacitor value?


Here it is with the capacitor set to a capacitance less than what I considered the optimum




and here it is with the capacitance larger



For me, this was a bit counter-intuitive; initially, I expected that more capacitance would

damp the response but it's the other way around. That's because the capacitor is on the

feedback. I'm not exactly sure about this, but I think that, with the fast edge, it

momentarily delays the feedback and allows the transistor to operate more as though it were

a simple common-emitter stage with a higher gain. That then gives more current to charge the

base-collector capacitance and improve the slew rate for large signals. So the capacitor

isn't there for stability, instead it's improving the way the circuit behaves with large

signals. (Unfortunately, having spent too much time looking at stability of circuits in the

simulator, I tend to see everything in those terms.) With too little capacitance we get the

basic response of the transistor, with too much it overshoots and is slow to return to the

feedback gain, and somewhere between is an optimum value where the two effects balance and

we get the best waveform shape. [Having thought about this a bit more, I think it must be having

to compensate for the base-collector capacitance of both transistors, the series feedback

one and the shunt feedback one.] 




The biasing, as I've done it, is far from ideal. For the above traces, I showed the output

ac-coupled. Here it is dc-coupled and you can see it isn't centred very well and extends

four-and-a-half volts below ground. That takes it down to the point where it's only a few

volts above the collector of the current source and almost straying into the area where the

transistor will no longer be able to hold the current steady.



Offsetting the input signal away from ground also has a bad effect. I imagine that that is

partly down to using a resistor for the tail current rather than a source.


It wouldn't be too difficult to correct the poor biasing - it is set by the tail currents

and the resistor values chosen for the collector load resistors - but I'm not going to do

that at the moment or the blog will never end.


Just be aware that, if you wanted to use it for real, it would need some further elaboration

and development.


If this blog interested you, a fairly comprehensive list of others can be found here: jc2048 Blog Index


[1] Electronic Circuit Design: Art and Practice. T. H. O'Dell. Cambridge University Press, 1988.
[2] Circuits for Electronic Instrumentation. T. H. O'Dell. Cambridge University Press, 1991.