Transistors: Saturation

This is another exploratory blog relating to bipolar junction transistors (BJTs). This time I'm going to look at saturation.

Be aware that I'm not an expert in all this: I'm studying it and you're looking over my shoulder as I experiment and explore the subject.

So this is not a tutorial and some of my explanations may, at times, be wrong.

What is saturation and why does it occur?

If you take an NPN transistor, like the 2N3904 devices I've been using in most of my experiments, and measure the base-emitter

junction using a diode test function on a multimeter, it behaves like a silicon diode (that's because it is one). We're used to that diode

because we are conscious of it when we design. If we were to probe a bit further, we'd discover that there's also a diode from the base

to the collector. That diode is a little less visible to us because when the transistor is operating with a reasonable collector voltage it is

reverse-biased and doesn't affect the operation too much (technically, it certainly does affect the ac operation, because the capacitance

across that junction is one of the things that limits the high frequency performance, but that's not where I'm going with this blog).

The forward voltages of those two diodes are slightly different (for a 2N3904, I see a difference of about 20mV at whatever the test

current is). That's because the properties of the collector and the emitter regions are different - the emitter is more heavily doped than

the collector.

So, is this just of academic interest? Well, no. Let's think what happens as the collector voltage falls and comes closer to the emitter

voltage. At some point the base-collector diode is going to cease to be reverse biased and at that point the operation of the transistor

starts to change. The neat self-regulation that the transistor does, where the collector current is proportional to the base current starts

to fall down. The base current will no longer simply relate to the collector current and we can keep increasing the base current without

it further affecting the collector current. This is about as far as I can go with an explanation at the moment - I'm not good enough with

the solid-state physics that underpins transistor operation to be sure of the mechanism and describe it properly.

A side-effect of that whole process is that charge starts to accumulate in the transistor. It's that accumulation that the term 'saturation'

comes from. It doesn't do any harm but there is an important consequence if you are a designer - the device will remain on for some

time, even if you remove the base current (ie if you are using the transistor as a switch it will be slow to turn off - 'slow' here means

hundreds of nanoseconds for the kind of transistors I've been experimenting with, so no problem at all if you are switching a relay

but much more of a problem if it's supposed to be a logic switch). The way to get around that, if you did want fast switching, is to drag

the charge back out again rather than wait for the normal transistor action to use it up.

So, when we talk of a transistor 'saturating' or 'going into saturation', we are referring to a consequence of what happens, rather than

the direct cause. [I guess people saying 'the transistor has saturated' was more natural than saying 'the base-collector diode of the

transistor has ceased to be reverse biased'.]

Because the base current can now continue increasing without a change in the collector current, the manufacturer has to be careful

to specify the relationship between the two when giving saturation voltages on a datasheet. The standard that manufacturers seem

to have settled on is to have the base current a tenth of the collector current (I imagine that was chosen to be safely underneath the

minimum gain that a large power transistor might have at high current, a figure of perhaps 40 or 50). Here's an example from a

2N3904 datasheet where they give indicative voltages for two different collector currents

They also give this handy graph and it's that that I'm going to have a go at replicating in the next section.

That graph is very useful because, as well as showing the Vbe(sat), they also give a second curve for the Vbe with the collector

held at 1V where the transistor is out of saturation. That shows us the difference the saturation makes. It also emphasises something

that I'm going to skate around here and that is that, at high currents, there is also an ohmic component from the bulk resistivity of the

semiconductor [at low currents, the Vbe voltage increases at about 60mV per decade of current (it should be a straight line on the

graph with its log scale for current), but at high currents it increases faster because of that additional factor]. The collector current

will also be affected by the same kind of ohmic effect.

Measuring Saturation Voltages

I'm now going to try measuring Vce(sat) and Vbe(sat) over a range of collector currents. Here's the simple circuit I've come up with.

There's a quad, precision op amp to run a constant current into each of the base and the collector. The respective currents will be in

the ratio 1:10 [the two current sources are identical except for the current-sense resistor which sets the relationship between the

voltage at the input and the current generated].

I'm going to have a 0V to 2V ramp from a function generator going in and adjust the oscilloscope so that it displays over the width of

the screen (I'm using the sync output from the generator to trigger the 'scope). The 2V input corresponds to 200mA in the collector

and 20mA in the base. That way the 'scope will draw the graph for me with the horizontal scale being equivalent to 0-200mA at the collector.

Here it is wired up on a breadboard (almost as messy as the circuit!):

Here are the resulting curves for a 2N3904 transistor (yellow base voltage, blue collector voltage):

It wasn't all that stable. My function generator is noisy, which doesn't help. Including a simple RC filter at the input helped cut down the noise

going in. I also added capacitors across the feedback resistors to the differential amplifier outputs to compensate the op amps more. Something

else that helped was giving the amplifier a negative rail. In spite of that, the transistor hoots a bit under some circumstances but, if you ignore

the area up to about 20mA collector current (the first horizontal division), the rest looks reasonable. In the simulator it's stable, so I'm probably

looking at the way three rf transistors interact on a real, messy, plug-in breadboard.

The Vce(sat) extends from just below 100mV at 20mA up to just below 300mV at 200mA collector current. The Vbe(sat) is higher than you

would see for operation out of saturation and goes from about 800mV at 20mA collector current to just over a volt at 200mA.

My curve looks different to theirs, but that's largely due to my having a linear scale for the current - if you pick a few spot values and compare,

there's not too much difference. One important difference between the way I'm doing this test and the way a manufacturer would it is that they

use a pulse test where the transistor is only on for 2% of the time. That removes the temperature effects from the otherwise high dissipation

at the higher end of the scale.

Here are the results for a 2N30532N3053 part. This is a somewhat slower device (fT = 100MHz) and is much more stable on my test setup.

In this case, the Vce(sat) at 200mA collector current is much less, being slightly over 100mV. Vbe(sat) is also lower, being about 850mV.

If you found this interesting and would like to see more blogs I've written, a list can be found here: jc2048 Blog Index