Lab Supply Project 1: Concept

A couple days ago I started watching the series of videos that Dave Jones (of the EEVBLOG) made about the design of a little lab power supply. Really great stuff!

He says that building a little power supply like this is a great project for budding electronics designers, and I think I agree... so now I got the bug.

The specs for the project as he gave them:

A couple things struck me as I watched the first few videos and thought about it a little...

• A lot of effort has to be spent dealing with the peculiarities of components.
• If I need to include a microcontroller anyway (for current/voltage control, reading buttons and knobs, and display), it might be interesting to push even more functionality into the MCU...
• I think Dave gave up too easily on the zero volt output capability... at least I personally want a supply that can go down (very close) to zero volts.  I already have a supply that can go down to a volt or so...
• One of my obsessions is motor control for CNC/robotic applications, including "servos", which are feedback controllers that accurately control the position or speed of devices.  It struck me that op amps are really like little servo controllers.  Hmm.... 
• I am the Circuit Jerk, so one awful thing to do would be to skip all the elegant analog crap and try to use a microcontroller in place of op amps (or similar control loops inside regulator chips)... turn the whole thing into a software problem!

I think it could be difficult to get that to work properly, but Dave always says that the way you learn is to fail... so here's a sketch of the basic design I'd like to try to build:

Easy in concept!  Use an ADC to fetch the output voltage and current (and a voltage reference to use for calculations); use a DAC to control the pass transistor, code up a servo loop, and "Bob's your uncle!"

I'm pretty sure that the biggest problems will come from the latencies in the VOUT -> ADC -> ALGORITHM -> DAC -> BASE pathway.  One of the greatest things about op amps (besides their coolness and elegance) is their speed.  A relatively slow digital feedback cycle means that the always tough tradeoff between stability and response time will be an even bigger problem than in an analog circuit (and even there it can be bad, or so I have read, haha).  The biggest area of concern in this regard will be the transient response -- will this solution be too slow to be useful?  If a demand for more current leads to a drastic and long-lasting voltage drop, the circuit on the other side of VOUT could easily malfunction.  Yuk!  And even worse, a demand for less current could lead to a voltage gain large enough to fry sensitive circuitry.  Desperate efforts by the control loop to compensate can easily lead to overshoots in the transient response, which is a problem even for analog hardware.

Adding some buffering capacitance will certainly help, but as Dave pointed out in his videos, too much of that causes other problems...

Anyway, it should be interesting to try!

Now... in order to test this, I really wish I had a flexible current sink I could use for testing... maybe I will add yet another project to the list and build one of those first!