In part 1 of this series of blog postings, I showed my preliminary design of a dual tracking linear power supply. In part 2, I showed the completed design and also the finished power supply circuits. And as promised last time, I have made a couple of videos showing the measurements of a few key parameters and will share the results here.
Load regulation is one of the key parameters of any regulated power supply. It tells you how well a power supply is able to maintain its output voltage when the load changes between 0% to 100%.
In the video below, I measured the load regulation with the output set at approximately 15V. Using the electronic load I built earlier, I changed the load current from 0 to 10A and observed the changes in the output voltage (less than 0.01V). So a very rough estimate of the load regulation over the full current range is better than 0.01/15*100=0.067% at 15V. And more accurate measurements showed that the load regulation is actually just 0.03% over the full current range, and far less than 0.01% from 0A to 80% of the current capacity (8A). This kind of load regulation is far superior than what can be achieved using a standard three terminal regulator (e.g. LM317/LM337) as the typical load regulation among these linear regulators is at around 0.1%. The video also showed the current setting capability.
View on YouTube in a new window
Tracking accuracy defines how well the output from the slave rail tracks the output from the master rail. For this power supply, the negative rail is designed to track the positive rail. The video below shows the tracking accuracy measurements when the positive output is adjusted from 0V to 30V. At each measurement point, the negative voltage pretty much matches the positive voltage spot on except for a couple of mVs offsets. The following video shows the measurements:
View on YouTube in a new window
The voltage/current setting resolution and accuracy are largely determined by the quality of the voltage reference and DAC used. Given the 12-bit DAC I used, the voltage adjustment resolution is roughly 7.3mV and the current adjustment resolution is roughly 2.4mA. Of course, you can always use higher resolution DACs (e.g. 16 bit) to obtain even finer adjustment resolutions. But for typical use, 12-bit resolution is more than enough and it gives you the best bang for the buck. The INL/DNL error from DAC can be calibrated and corrected in software, which makes the voltage/current setting even more accurate.
Ripple noise is usually not an issue for linear power supplies. It can actually be quite tricky to measure it correctly as other electronic devices connected to the same power line can introduce significant noise and interference into the measurement. I had attempted to measure the ripple noise, but soon realized that all those CCFLs and computer switching supplies in my lab had introduced so much noise making an accurate measurement almost impossible. But by comparing the output waveforms to the output waveforms of my Topward 6603A (spec’d for <0.5Vrms), I can say that the outputs waveforms (use AC coupled, 2mV/div) of these two power supplies are quite comparable. One thing we need to pay special attention to is the decoupling of the digital circuitry power supply. Because the power supply is essentially driven by the DAC, inadequate decoupling could introduce unwanted ripple noise. But in general, because this is a linear power supply, as long as the input capacitance is large enough (I used two 22,000 uF filter caps) and as long as there is no oscillation in the control loop, ripple noise should be minimal. Perhaps one of the biggest advantages of this digitally controlled power supply is the stability of the output voltages. When a power supply uses potentiometer to adjust its output voltage, the voltage tends to drift over time as the wiper of the potentiometer settles, especially when the quality of the potentiometer is poor. Since the DAC output is very stable, the long term output voltage drift is largely determined by the temperature coefficients of the sensing resistors (R10 and R11). In my limited testing, I left the power supply on over night and the output voltage drifted only around 1 mV when the output voltage was set at 20V.
Overall, I am pretty satisfied with the performance of this power supply. Because the set output voltage is controlled by an ATmega328p, you can easily reprogram the MCU to add in more sophisticated programmability. And as I explained along the way there is a lot of flexibiliy when building this power supply. You can build just the positive portion for instance (without tracking for instance), and you can adapt the number of pass transistors to handle your maximum current requirement.
OpAmp oscillation could become an issue depending on the type of OpAmps used and the actual circuit layout. A clear indication of such oscillation is the drift of the output voltage with load. The best way to verify is using an oscilloscope to observe the AC component of the output signal. The snubber network and the RC filter within the feedback loop I mentioned previously can be adjusted accordingly to restore the correct phase margin and eliminate the oscillation. My recommendation is to build the circuit without the snubber network or the RC filter and add them back in only when stability becomes an issue.
Of course, there are a couple of areas you can still improve upon. While this power supply offers over current protection, it does not protect the output from over voltage. An over-voltage situation typically occurs when one of the pass transistors fails (e.g. a shorted collector emitter) and in that case the input voltage is applied directly to the output. When the power supply is correctly designed and the power dissipation among the pass transistors is adequately provisioned, this situation should not occur even during a prolonged dead short at the maximum allowed current for prolonged period of time. Thus, most of the lower end commercial lab power supplies do not offer over current protection. But you can always add a crowbar configuration to protect the output against this unlikely scenario. Another area some readers suggested is the over temperature protection. Again, this can be easily added if it is needed. But for the massive heatsink I used and the designed operating conditions, this is not an issue.