An Engineer in Wonderland – Op amp needed for a current source

I was pondering a simple way with a low drop-out voltage to power a grounded LED from a grounded supercapacitor. Something that would be tolerant of the LED being powered from elsewhere when the capacitor is empty.

constant-current-122.jpgI was pondering a simple way with a low drop-out voltage to power a grounded LED from a grounded supercapacitor. Something that would be tolerant of the LED being powered from elsewhere when the capacitor is empty.

The easiest option is to use a resistor, but this will mean brightness is dependent on capacitor voltage.

A complicated way, which could deliver maximum efficiency and maximum power extraction, is to implement a buck/boost converter.

The LTC3454 1A LED Driver from Linear Tech looks pretty suitable .

More investigation would be needed to see if it can take volts on its output when there is a huge capacitor on the input.

The following would be one way to implement a linear solution


However, the op amp choice is a little tricky – the inputs must include the positive rail in their common mode range, and the output needs to get near the positive rail too.

In ancient times, there were op amps with p-fet inputs that were completely happy working up to 0.1V above the positive rail.

However, these were +/-15V parts.

So my question is, is there an op amp that includes the positive rail in its input common mode range, that will run from 3V to 5.5V single rail?

By the way, I claim it is cheating to just specify a rail-to-rail part.

An alternative to this would be the rather neat two transistor circuit that uses different current through two transistor emitters to sense a small voltage across a low value resistor to do the current measurement, with another transistor to modulate the current accordingly.

Sadly, right now I cannot remember how the circuit goes. In this case AFAIK it would need two pnps to measure, a larger pnp to modulate, and an npn to link the two. 

And also by the way, the Fairchild FPF2165 load switch seems to completely fit the bill from a different direction.

From the data sheet:

When the switch current reaches the current limit, the parts operate in a constant-current mode to prohibit excessive currents from causing damage….
For the FPF2163 and FPF2164, if the constant current condition still persists after 30ms, these parts will shut off the switch….
The FPF2165 will not turn off after a current limit fault, but will rather remain in the constant current mode indefinitely. The minimum current limit is 150mA.

And, explicitly stated:

“Internally, current is prevented from flowing when the MOSFET is off and the output voltage is higher than the input voltage.”


Should you feel the need, respond to with ‘Constant current’ in the title.

Please don’t respond below as our spam blocking system doesn’t work and the inbox is overwhelmed by all kinds of generous offers from all kinds of rather annoying people.

No email addresses are collected for marketing (or any other) purposes from responses to this blog. I will keep it that way for as long as possible.



  1. Thanks Geoff L
    That Microchip rail-to-rail types would certainly do the job.
    Maybe I was being a bit pedantic, trying to find one which only inluded the positive rail.
    But it was an interesting challenge, to which there seems to be no answer.

  2. Hi ‘Alice’
    You might find the Microchip MCP6021 series of op-amps interesting.
    The common-mode range exceeds the supply rails by 300mV and the output can go to within a few mV of the rails.
    The MCP6023 even has a /CS function to put it into a low power mode.
    Geoff L

  3. Thank you gentlemen.
    I take your point Mk Kurt, that 2.7V is very near the 3.0V forward voltage of good leds so I would not gain much (0.85J from a 1F cap if my calculations are correct).
    But the LTC3454 would make better use of the 5.5-3V range where most of the energy is in the capacitor (10J from 1F).
    I just did a quick calculation and came up with 4J available from a 1F capacitor between 3.0-1.0V, and 0.5J below 1V.
    A 250mA linear current source would deliver 4.5J to a 3.0V led from the same capacitor initially charged to 5.5V.
    So the 3454 would give me 10J (making a linear reg look pretty lame, now I have done the calculation :-), and a better boost converter could add 40% to that.
    And if it could be regulated, Mr Fred’s flashlight boost, how ever frugal the circuit, would probably deliver a couple of extra Joules.
    I remember using a ZXSC310 in boost-buck configuration (impossible in this case) to drain a 5.5V cap to 0.8V, but the output is nothing like constant current.
    I believe Mr Kurt also has experience with this chip?
    As for a fully-linear configuration, the FPF2165 fit the bill completely.
    Were you proposing the LTC6101 as a sensor for a switching supply Mr Kurt.
    I remain pondering ‘is there a low-voltage non-R-R op-amp that includes Vcc in its common-mode?’.

  4. Some of the cheap l.e.d. torches using a 1.5volt single battery (AA or AAA) are using something similar to power the emitter. So cheap in fact it might be more convenient to strip one down and use it? Just a thought!

  5. The LTC3454 is rated down to 2.7v, so it may or may not be suitable, depending on the supercap and charge level. Haven’t tried searching for low voltage boost converters myself, so I can’t say if anything better exists.
    For the analog current limit circuit, aren’t there high side current sensors that would be suitable? The LTC6101 is such a device, although it only operates down to 4v, which is probably not adequate for the application.
    Could you sense current on the low side of the load, even though you are controlling the high side?
    Steve Kurt

Leave a Reply

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