I'm working on PWM power supply to supply a load which needs a constant current of 1 mA and 2.5V of operating voltage. The Pulse width Ton is 1 to 10 µs Maximum and Toff is 10 ms which results in a duty cycle of 0.1% Duty Cycle. The main purpose is to reduce the average current consumption of the whole device since it's needed only to operate during that short pulse. Now this comes to several issues. One is delivering a constant current using something other than a current limiting resistor whilst having a fixed supply voltage of 2.5 V. All that has to be provided during that Pulse. The rest of The time the power supply shouldn't consume more than 1 µA. A buck step-down converter is a good way to start for providing the right voltage and efficiency at a certain load, a device like TPS61098. One question about PWM switching buck converters. During the Toff time the current is supplied from inductor which means it needs to drain more current from the battery to maintain that constant and average load current. Am I correct here ? So in other words, the average output and the input current will be the same ideally. There are plenty of DC-DC converters, do you have any suggestion of one which can provide the behaviour I described? The circuit might be similar to LED driver where you can dim the LEDs using PWM input.

To conclude, there are two possible options here. One can control the buck converter using external PWM input and is able to operate with the timing i mentioned (1µs min enable time). The second option would be to supply my PWM timing generator from the buck converter and drive the load from the PWM generator. In both cases, the buck converter should be able to limit the current and voltage and consuming below 1µA averagely.

Your support is much appreciated.

Edit: The PWM pulses are generated from A MCU. I need a fixed voltage with a current limit. The device is considered linear. The voltage is 3V and is from a battery.

  • \$\begingroup\$ You cannot regulate both voltage and current simultaneously. So you need to decide which is more important: A constant current with a voltage limit, or a constant voltage with a current limit. You also need to supply a lot of other details, such as the voltage of your primary supply, the nature of the load (linear/nonlinear, capacitive/inductive/resistive) and where the control pulse comes from. How precise does the control need to be? \$\endgroup\$
    – Dave Tweed
    Commented Dec 8, 2015 at 15:29
  • \$\begingroup\$ With such low current requirement, I would suggest using a linear circuit as a current source or sink. Since your question suggests a lack of knowledge and/or experience, I suggest you provide much more detail about what you are trying to do. What is the load? Why do you think it needs 1 mA? What is the overall system? Is it battery powered or??? \$\endgroup\$
    – user57037
    Commented Dec 8, 2015 at 17:37
  • \$\begingroup\$ The lack of knowledge or experience is ultimately not the issue. It's the existence of multiple choices at once where a small detail makes a difference. I took the liberty to ask for support. All the specified requirements are based on calculations and simulations. \$\endgroup\$
    – Sadok
    Commented Dec 9, 2015 at 15:44

2 Answers 2


If you have a device which is outputting pulses, its output can't be constant current because it's , well, pulsing...

However, you could easily push 1 milliampere pulses through a 2500 ohm load with 2.5 volt pulses from a voltage source if the load resistance and the pulse amplitudes didn't change.

  • \$\begingroup\$ What i'm trying to say here is the current during the pulse is limited to 1mA and ~0 mA otherwise. The pulse control should act like a fast switch of power supply. \$\endgroup\$
    – Sadok
    Commented Dec 9, 2015 at 14:51
  • \$\begingroup\$ Thank you for you answer. A resistor limiting current would dissipate some power. That was my initial setup, I just wanted more efficiency. It seems it's a tradeoff of complexity/efficiency here. I'd go with the current limiting resistor as a last resort. \$\endgroup\$
    – Sadok
    Commented Dec 9, 2015 at 15:32

If you use a simple buck converter (probably one that regulates output current for your application), the bandwidth of the control loop will generally be no more than a few hundred kHz. There is no way to get the output to pulse at microsecond rates with this approach.

These days, there are special LED drivers with built in PWM functionality that can freeze their internal regulation state and open-circuit the LED during a Toff period, resuming control during the next PWM on-cycle. I believe Linear Technology calls this feature True Color PWM (trademark). This type of thing MIGHT barely work for you, but even so you will be limited by the RC time constant of the LED. It might get you down to a few microseconds, but probably not 1usec.

What you CAN do, however, is divert the current from the LEDs during the off time to a current source, such that the PWM pulses are invisible to the regulator. Unless you design another switcher for this current source, this method will be dissipative: if your only goal was to save power, it won't help. But if you really need fast and tightly controlled pulses, this will allow you to get down to hundreds of nanoseconds. Check out the LT3909 for an IC that does it all in one package.

Also, note that the TPS61098 is a boost converter not a buck converter

  • \$\begingroup\$ Well, LED driver was an example. I'm not driving LEDs and you are right it depends on the RC constant for the load to react. In my load i can operate on the RC, thus the load can be active in tens of nanoseconds. Let's assume i have a pulse generator now and i can get 10µs pulse width. The question remains on how to limt the current to a 1mA maximum without using a resistor. \$\endgroup\$
    – Sadok
    Commented Dec 9, 2015 at 15:59
  • \$\begingroup\$ All of the same principles apply for any constant-current load. Current regulating converters are often marketed as LED drivers, but work as well for any load that wants a constant current to bias. I believe the solutions I mentioned are still the solutions you should consider. I am having trouble understanding your requirements from your post and comments - is it true that the load will be different from pulse to pulse? I.e. on some pulses it will draw no current and on some pulses it will look like a short, in which case you want to limit the current to 1mA? \$\endgroup\$
    – user49628
    Commented Dec 10, 2015 at 20:27
  • \$\begingroup\$ Exactly, During pulses it will consume 1 mA of current. When no pulse is generated (during Toff) the load won't receive any current as the voltage is 0 V. \$\endgroup\$
    – Sadok
    Commented Dec 12, 2015 at 2:08

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.