Bench power supply with switchmode pre-regulator

[Leonelf]

Hi!
I'm currently trying to develop a 0-30V 0-3A bench PSU. I'm using a TI simpleswitcher as a preregulator to limit the amount of heat the PSU will produce. I want the PSU to be controlled by an MCU, since I'm lazy, I'm using an Arduino Pro Mini. I plan to use a 12-bit DAC to control the comparator voltages for voltage and current sensing and a 12-bit ADC to read back the regulated voltage/current so I can trim in software if needed.
I'm pretty new to analog circuit design (I know all my transistor stuff, but have no idea about opamp-selection etc.), so I don't know which opamps I should choose. For transistors I'm curently using 2n2222 and its pnp counterpart, the 2n2907, both in a SOT-23 package. I only chose those trannies since they seem to be pretty standard, so suggest better ones if you have suggestions^^

Are there any caveats to this design? Should I use seperate regulators for the DAC/ADC supply and the arduino or would a ferrite bead or capacitors suffice?

Greetings,
Leon

[Paul Price]

The common-mode input range of the TL071 will not allow you to adjust the output voltage to less than a few volts.

Your circuit doesn't measure the output current and voltage and give feedback to the MCU, you have a wild-horse pulling the baby carriage. Your open-loop design is under utilizing the ability of a MCU to be a major part of your power supply control.

Use fast op-amps, at least something like the CA3240 or other rail to rail input/output op-amps.

Consider the slewrate /settling time/overshoot/undershoot of your op-amps to fast changes.

You could eliminate the DACS completely and cut cost and circuit complexity by using comparators in the V and C feedback loops driven by MCU 12-bit PWM r-c filtered control for setting output voltage and current.

Q1 is likely to fail with a short C-E,  and this will result in a power supply output of a uncontrolled 30V+ push into the load with any fast short-circuit condition at more than a few volts output or less. Q2 will quickly likely fail as well once Q1 shorts.

With a faster MCU(64MHz clock PICC MCU) with a 12-bit PWM you can also improve power supply response time to changes in load or overload/short-circuit conditions.

You must consider using the MCU better for your design, but do not put the MCU directly in the control loop, but instead use it for monitoring the hardware control loop. In this way the fast hardware can optimally manage overload conditions while the MCU can measure output, set output V and C and correct and respond correctly to out of regulation conditions.

In this case, you must consider the time for your MCU to A2D to measure output voltage and current and to evaluate out-of-regulation/short-circuit/setting load power supply states, the time to decide and calculate a tactic to correct the problem and the response time problem of the power supply due to the  time needed to serially transmit the new DAC output settings and also the maybe slower than optimum 1v/uS slew rate of the DACs you've chosen.

You can also achieve 12-bit resolution with 10-bit PWM using software PWM dithering with the same comparator control idea.

It any case, it would be beneficial to create a low-voltage, low-current regulated  V-- rail for any op-amps controlling the power supply output to achieve precise control down to 0-volts.

[Leonelf]

Couldn't I add diodes between the opamps and the transistor base to increase the voltage drop for more precise control? Then they wouldn't need to go completely down to 0, right?

[Paul Price]

No.

[Leonelf]

What about loop stability? Can such a simple loop (direct feedback into opamp) be stable? Or would I have to add capacitors at the bases etc?

[fcb]

I think your current design will struggle to hit 3A (that poor SOT23 driver transistor), very little chance of it being stable and I don't think your tracking pre-regulator FB loop can be frigged like you think.
 
I'd split the design into two major parts, develop and prove them seperately:
1. Output stage /w current limit.
2. Tracking switched pre-regulator.

The first part (the output stage), i'd probably look very carefully at what HP have done in the 80's - these supplies were excellent. and the schematics are readily available.

Would probably tackle the switched pre-regulator using a microcontroller (dsPIC33 would be my first choice), as this is really only to reduce the heat generated the absolute performance is not *that* critical. Probably build a control loop at 10KHz with perhaps 100KHz PWM, would be converting absolute current, output voltage, input voltage and requested voltage, probably quite a straightforward algorithm. As Paul Price says, use PWM's to generate the target voltages, these don't need to move quickly.

One problem I've had with PWM generated voltage rails is that they are only as accurate as the processor voltage regulator.  I recently got round this using a REF5025 (this was elsewhere in the design) and an ADG749 (being driven by the PWM), super stable.

[Kleinstein]

It depends on the OPs  if they are slow enough (e,g. like OPA251 at 35 kHz GBW), and the load is well behaved it might be stable. Usually one needs extra provisions to make the loops stable - this is the main difficulty with a lab supply. As there is no resistor at the emitter of the power transistor, it might be rather difficult to make it stable with capacitive load. The way current and voltage regulation is combined is also unusual, so it could be tricky.
One usually should use a simulation to test such a circuit and get the compensation right.
The fist thing is to test just the transistor output stage and see if it is well behaved and easy to control.

For the set voltages, there are 3 options:
1) Use a high quality DAC (e.g 12 Bit and more with low INL)
2) use a lower quality but higher resolution DAC (maybe even 2 combined) and use an ADC channel for corrections
3) use a kind of PWM from the µC and a higher order filter as a slow, but accurate DAC
    Software modulation can be used to extend something like an 8 bit Hardware PWM of the µC.