Do you sometimes need a cheap precision power supply? This design scheme shows a precision power supply with thermal overload protection and short circuit protection functions, which can provide 100mA of current. I needed such a power supply, so I started using Google to find a suitable circuit that might be available on the Web. I did not find an acceptable solution, but I did get some ideas that led me to start this design.

This circuit is designed around the LM317L regulator. Yes, this is an old type of regulator, but it is still in use, it is cheap and easy to buy. This circuit (Figure 1) uses a rail-to-rail (input and output) operational amplifier in the feedback loop from the output of the LM317L to its regulation pin. The inexpensive LM4040BIZ precision voltage reference is used to establish a 0.2% voltage reference for the operational amplifier.

Figure 1 The power supply circuit is designed around the LM317L regulator.

I found the SPICE model of LM317 on the LTspice groups.io website (LTspice@groups.io) and performed some simulations to see if a stable configuration can be achieved. According to LTspice simulation and SPICE model, the obtained configuration is conservative and stable, and the phase margin is about 70 degrees. LM317L output series resistance of 3.3 ohms may be counterintuitive, but it can very well increase the stability of the circuit, and when the output current is 100mA, the power loss in the resistance is only 33mW. 

Wow the engineering world with your unique design: a guide for design concept submission

The feedback loop keeps the output impedance of the circuit at a very low level (Figure 2), so 3.3 ohm resistors are not a problem. The output filter capacitor C4 contributes to loop stability. The loop is stable when the C4 value is equal to 10uF, but the larger value of 22uF provides more margin for a conservative design. The low values ​​of R7 and R8 ensure that the required minimum output current is achieved, and they also make the feedback loop more stable.

Figure 2 The feedback loop keeps the output impedance of the circuit at a very low level.

The schematic diagram of LM317L varies from manufacturer to manufacturer, and the LTspice model of LM317 may not fully simulate the device, so I choose to be very cautious when stabilizing the loop.

Note that the R9 – C5 filter on the reference output may not be needed. It is used to eliminate most of the noise generated by LM4040BIZ. The refiltering provided by the op amp circuit and output capacitor C4 may be sufficient.

I built and tested this circuit, and it worked as expected. I used two matching resistors with a nominal value of 243 ohms for R7 and R8, the circuit is accurate to 0.2%, and the output voltage is 4.99V. Moreover, as predicted by the LTspice simulation, the feedback loop is stable when the load is less than 10mA and the maximum load is 100mA.

When 1% resistors are used for R7 and R8, the worst-case accuracy of the output voltage is 1.2%. When R7 and R8 use a matched pair, the voltage accuracy is close to 0.2%. In order to obtain higher accuracy and higher cost, you can use the LM4040 specified as 0.1% voltage accuracy. My results show that the circuit works well as a low-cost precision power supply with up to 100mA output current. 

Jim McLucas retired from Hewlett-Packard after 30 years in production engineering and design and testing of analog and digital circuits.

I've seen this situation many times when employees in the electronics industry for more than 40 years don't read data sheets and "recommended" application notes. 3 The terminal regulator is undoubtedly one of the easiest things to use, but it is also one of the most error-prone things. Unfortunately again, the spice model did not pay attention to the application design recommendations. I use the suggestion, but in fact, if you want a reliable circuit, it can work for decades without causing the pain of 3 terminal devices. It is very stupid to read and understand them as free application circuits and ignore them. . I now use the National Semi data sheet Ti of course, so you can find out how many violations are on the recommended track. I calculated 8 diverse areas from the recommended exercises in the data sheet.

Constructive criticism is a good thing. Please provide detailed information to facilitate useful discussions.

Jim, thank you so much for sharing the interesting circuit! I found that there is no problem with your circuit, and the fact that you have tested proves this. I did not see any content in the circuit contradicting any application notes. I hope to see more of your designs!

Hi, JIm Since low cost and high precision are your obvious design goals, why not add 9.1KR between pins 1 and 2 of U1 to increase the minimum Vout to U2's 1.8V minimum recommended V+? Then both U2 and U3 can be powered by U1. The former will eliminate D1 (reduce the cost of smidgen), and the latter will provide a constant bias current for U3 (improve accuracy-also through smidgen).

Oops! It is best to set it to 4.7K!

If there are no startup issues, this may work. It should be tested with LTspice for verification. The original circuit avoids potential startup problems. The stability of the loop should also be checked.

The interesting thing is that you mentioned the startup problem.

Considering that the time constant of R1C3 (560us) that provides V+ to U2 is nearly 10 times longer than R3C4 (66us), which limits the rate of rise of power output. Therefore, if U1 takes off before U2 wakes up to control it, then it seems that there is indeed the possibility of destructive output transients-not the power supply itself, but any load connected to it.

The probability of this failure will depend on the behavior of U2 when pin 5 is below the minimum 1.8V specification specified in its data sheet to ensure normal operation-this is not Spice (or any other simulator) known for accuracy The kind of thing predicted.

Thanks for your insightful comments. Your point of view is very good. Maybe C3 should be 0.1uF. The Zener diode provides a low impedance, constant voltage to pin 5 of U2, so it may not need much bypass anyway.

You must log in or register to post comments.