AMC7135 PWM LED control – works?

The AMC7135 is a constant current driver (sink) meant for 1W power LEDs. It sinks 350 mA into its OUT terminal, thud driving the load (power LED) with 350 milliamps of current. Assuming that the supply voltage is under 6V, the setup can drive a single 1W white power LED easily with minimal component count and no current sense resistor. But what happens when you use PWM to drive the AMC7135? The AMC7135 PWM circuit must operate in a particular frequency and duty limit to work smoothly!

Can AMC7135 handle PWM?

Yes, it can.
Is it recommended to do this? Probably not! There is nothing in the datasheet to hint that the current sink can switch in kHz range (usually required in lighting control). However, when I tested the AMC7135 out, it seemed as if it can handle fairly high PWM frequency and thus control the brightness of the connected LED flexibly.

However, assuming that the AMC7135 is a feedback based current control system, you must consider the time that the device takes to settle to a stable constant current output. Not caring about the settling time of the current sink will cause loss of total effective power delivered to the LED.

How fast can you run the AMC7135 PWM?

The Vdd pin is used to enable/disable the output of the LED driver. The input is very high impedance and so any microcontroller that outputs a 3.3V or 5V logic level should be able to directly switch the AMC7135 current sink. How fast can you run the PWM? Here is the test circuit I used to check how long it takes for the circuit to draw a stable 350mA through the LED:


Apologies for my terrible drawing skills! But this should be sufficient to get the point across. In the scope screenshots shown below, the nice rectangular wave is the signal generator output (3.3V CMOS level) and the other waveform that is struggling to follow the PWM input is the AMC7135 current output.

AMC7135 PWM drive response speed


At about 2kHz PWM frequency, things seem okay! The time required by the current sink to become stable is about 8us.
That is,
Turn-on time: 8 us
Turn-off time: 4 us

This should tell you that you do not want the frequency anywhere close to these time periods. If you make the frequency so high that the current sink cannot turn on properly, you will lose on your full potential power that can be delivered to the LED, i.e. PWM control will not be effective over all duty cycles.
This can be seen here:


As in this screenshot, the duty cycle is at its minimum possible here. If you go to a lower duty cycle, things will not stabilize enough before the PWM turns the LED off again. That is not a good way to drive the AMC7135 as the brightness will suddenly decrease in the lower ranges of duty cycle.

So if you want control over the lower ranges of duty cycle, you need to reduce the PWM frequency. The best approach to drive the AMC7135 PWM based LED systems is:

Frequency of PWM: 6 kHz
Minimum duty: 5%
Maximum duty: 100%

This will ensure smooth brightness control over all duty ranges, as well as good enough precision of PWM control even on slow microcontrollers. Note that reducing the PWM frequency to under 5 kHz may cause visible flicker in the LED!

9 Responses

    • pratik

      It did not seem to make any huge difference. The time to get stable actually depends on the load capacitance and how fast the source can ramp up the current (ESR of source, etc matter too). But keeping all the things modest… I think this is the right frequency. You can of course go higher, but it will reduce smoothness in brightness control.

  1. ivan

    This very informative. Would it be acceptable to run a 10kHz frequency with duty cycle of 30% or greater?

  2. ivan

    I guess the better question would be what max frequency could I use if I keep the duty cycle to 30% minimum?

  3. Emil

    Why do you say that lower than 5KHz will produce visible flicker? Does a 120Hz monitor produce any flicker – no.
    The human eye persistence is about 1/16 seconds so a frequency 10 times bigger will largely suffice for flicker free appearance.

    • pratik

      120Hz monitor is being constantly looked at. If you move your phone around fast in dark, you will see a trail of flicker at 120Hz… running LED lights at 120Hz at a low duty cycle of say 10-20% will cause considerable flicker because the off time is long enough to be noticeable. Even give you a headache. 500 Hz is the absolute minimum for PWM controlled LED lighting that doesn’t cause moving objects to flicker.

  4. Lagom

    This is very good information. Research shows that even fairly high frequencies, not perceptible, cause flicker that the human vision system in the brain still recognises, leading to discomfort.

    What I would find also interesting would be to see which components are actually needed at Vdd to PWM the voltage from a single Li-Ion battery, the most popular way to use this IC with. Here, we see a bench supply for a good reason, but then, one has to PWM with something else, while the Li-Ion battery depletes from 4.2V to 3V and then shuts down by way of its protection circuit.

    • Martin

      I have just finished designing my own flashlight, including fabrication of the housing in alu om a lathe.
      I used the Cree xm-l2 LED, driven by 6x acm7135 (2.1A), controlled by a PIC microcontroller and powered by 2×18650 in parallel.
      I ran into the problem that you where queries about, where I couldn’t supply the amc7135 with high enough voltage from the controller, as the i/o pin only supplies vdd-0.7v
      I solved this by adding i FET to switch vbat directly to vdd on amc7135. I have no problem driving the amc7135 with a battery voltage of 3V.
      BTW my PWM frequency is 5KHz and duty in these steps 10, 25, 50,100%.

Leave a Reply

Your email address will not be published. Required fields are marked *