This is a short follow-up to A Transistor Load
which I blogged about as part of the Project14 Test and Instrumentation competition.
In the last blog A Transistor Load: Stability
I got as far as compensating the amplifier section with a single 'dominant pole' capacitor and looking
at the way the dynamic response was affected by different capacitor values. I also looked at the issue of lead inductance
in the output current loop. I've done a little bit more work on it since then but it's all rather piecemeal and bitty and
that makes it difficult to blog about, so this is an update rather than a full blog.
Is 'Dominant Pole' Compensation the only Way to Ensure Stability?
In the last blog, I used a capacitor around the amplifier for the compensation. That capacitor turns the amplifier
section of the circuit into an integrator over a wide range of the bandwidth and the result is that the phase of the
open-loop amplifier is forced to a well-defined 90 degrees giving a good phase margin as the gain comes down to unity.
I've been experimenting with the simulator, but my optimism about being able to find an alternative hasn't come to much.
The fundamental problem is that to ensure stability we have to get the gain down well below unity by the time we get up
to the kind of frequencies where all the awkward poles will give us trouble. But at the same time, the rate at which the
gain lowers as we approach that unity point can't be too steep or the phase will get to 0 degrees whilst there is some
gain. That seems to be a circle that is difficult to square.
More On Lead Inductance
When I did the last blog, I looked at how lead inductance affected the stability. One thing that was evident was that the
inductance can see some capacitance that it will resonate with. Unfortunately for me, that resonance comes just where I
don't want it and with only fairly modest values of lead inductance (frequency-wise it pushes the gain around just above
the area where the open-loop gain gets down to unity). I wasn't clear what that capacitance was though I speculated that
it was one of the intrinsic capacitances in the MOSFET.
Since then I've discovered something that gives me some more clues. The behaviour with a drain voltage of 12V is very
different to how it behaves with a drain voltage of 3V. These next two traces show the two conditions - the top one is
12V and the bottom one 3V. I've set the input so that it moves the output between 1A and 3A to get away from my other
major problem that the load behaves much differently down at low output currents.
All conditions are the same for both traces except for the drain voltage.
At 12V, where the drain voltage is above the gate voltage, there is a small amount of residual ringing that takes a long
time to die away (not much damping). A simulation doesn't show that ripple, so its coming from something that doesn't get
modelled. At 3V, where the drain is below the gate voltage, there's very definite ringing which dies away more quickly.
That does show up in a simulation.
Here's what a simulation shows for the two cases, again 12V top and 3V beneath. These are for a lead inductance of 75nH.
These are the curves that the datasheet gives, but these are for the condition where Vgs is 0V, not with current flowing
through the channel. I thought that I might have picked a particularly bad MOSFET for this, but comparison with
datasheets for similar devices shows that it is actually quite reasonable.
That shows the way the output capacitance of the MOSFET increases as the drain voltage comes below the threshold voltage and which is
presumably the root of most of my problems.
I've come up with a new version of the design, which seems to deal better with the lead inductance, so
the next thing on the agenda is to build it and see how far short of the promise it falls.