Laboratory Power Supply June 2020

One can get infinite options of laboratory power supplies on the market, either a cheap Chinese one for few bucks or expensive high quality one. Why should I even bother making my own that will now be as cheap or good?

challenge and customisation

Analog design and regulation is not as straightforward as writing few lines of code. We were thought how to calculate and simulate stability, but never executed it on the PCB where many problems arise: PCB parasites, non-ideal components, unexpected blackbox component behaviour because of complexity...

Also, many solutions are made without practicality in mind, made with "alarm-clock" interfaces, clumsy touch screen only control, huge and inefficient UI design...

The presented design is a result of many iterations even after the article release and is frequently used when working with electronics.

Design specifications

In order to make my dream laboratory supply, I have prepared a list of features I want:

voltage/current regulation
differential regulation for low error
differential voltage/current measurement that makes analysis easier
simple UI with big characters and voltage, current, power, resistance, efficiency data
multiple reliable load connection points
floating output with earth plug
switch between transformer windings 1x12V or 2x12V
compact size

A huge research was done to see what kind of solutions already exist on the market and to customize them to my needs:

Basic transistor theory
Elektor's laboratory supply - is simple and popular design without differential regulation
Old HP laboratory supplies service manuals - are a great source for simple topologies made from discrete transistors for better understanding of the backbone operation.

The beginnings

1 - Simple non-differential regulator consists of an operational amplifier (opamp) and power transistor in either common source or common drain configuration. Common drain is preferred to maximize the bandwidth, but limits the output range without the use of additional power source to drive output transistor. Common base configuration is useless since it limits output range and requires the same current flow in output and input signal. Single transistor output stage is capable only of current sourcing to load, the load needs to take care of current sinking. In case we need quicker and better regulation or current sinking we need to use full output stages such as Class B or AB with additional transistor which sinks current.

2 - Differential regulator adds a differential amplification stage before error amplification stage. Differential stage makes the regulation more accurate since it senses positive and negative output voltage and amplifier/attenuates the difference against reference voltage common node. The disadvantage of additional amplifier stage is lowering frequency response, because of additional delay introduction into the feedback loop.

3 - "Analog gates" are needed to regulate voltage or current depending on which quantity exceeds the set value. This is easily accomplished by using diodes on the output of voltage and current amplifier connected together. The diodes are connected in a way that amplifiers can only sink current. In no limit situation only pull up resistor regulates output by completely opening the output transistor. Also in this situation both quantities (voltage and current) are lower than set, amplifier go maximum positive voltage. If voltage exceeds the set value, voltage amplifier pulls down the pull-up resistor. Similarly when current exceeds set value, current amplifier pulls down the pull-up resistor.

4 - Complex operational amplifier issues! In design a very good performing operational amplifier OPA196 was intended to be used. It has low noise, offset voltage and high output current capability. BUT! Even after topology change and huge feedback loop compensation (slow operation) the regulation was unreliable and sporadic. One time it was working and no oscillation occurred, other time in the same situation oscillations were happening and slow voltage to current regulation happened which resulted in overdriving the load with voltage or current. After many hours lost with debugging, simulations and datasheet reading I decided to use more simple operational amplifier with only three passive amplification stages and no active circuit to lower offset and noise. I have used a standard LM2904 amplifier which also comes with schematic. All the problems with instability immediately went away.

5 - Output capacitance improves stability. Because we are dealing with static voltage and current references and loads that are expected to work at constant voltage or current, output capacitor does not worsen but it improves the stability. A load that pulsates the current needs a really fast regulation. Since regulation feedback loop cannot be infinitely fast because of components and circuit layout parasites, a small output capacitor acts as buffer for high frequency current spikes.

Schematics and layout

A quick block diagram (Figure 1) was drawn before drawing the schematics in Altium Designer. Block diagram is very useful to evaluate if all necessary functions will be implemented. A regulation part of schematic is shown in Figure 2 and layout in Figure 3. On layout level typical rules were followed:

Separation of power, regulation, sensing and MCU part
Usage of differential routing to lower noise influence improve accuracy
MCU with continuous power plane
Thick power connections

Figure 1: Block diagram of laboratory power supply

Figure 2: Schematic of regulation part of laboratory power supply

Figure 3: Layout design of laboratory power supply

MCU programming

STM32L552 was used as a MCU for controlling user interface. A I2C protocol was used to communicate with Microchip MCP3422 18bit sigma-delta ADC. Current, voltage, resistance and power were all clearly displayed on the display. In order to switch between tasks a simple scheduling was used which checked in a loop if current task needs to be executed in case if enough time has elapsed from the last run. Main code can be found HERE. The following tasks were running (with interval in ms):

LCD (300)
MCP_ADC (333)
BUTTONS (30)

Conclusion

The project of making my own power supply that is reliable was very challenging. Not a lot is written about using operational amplifiers to regulate power. Older amplifiers that use simple topology was found appropriate for regulation, where as newer operational amplifiers that utilize many additional circuits to compensate for non-ideal characteristics introduce issues when trying to regulate voltage or current on a high power load. This resulted in many circuit redesigns and bodges on the PCB as can be seen on Figure 4. Nevertheless, the project gave me confidence to start thinking of my own design of operational amplifier and how to reliably make feedback loop. A back side of power supply is shown on Figure 5.

Figure 4: Many bodge wires on circuit that is successfully working

Figure 5: Back side of laboratory power supply