Switching Constant-Current Source?

[BrianG]

I am working on a project (NiMH battery charger) which requires a constant current of 1.4A. But no matter what design I look at, there seems to be a lot of heat generated. This project is a portable device so size (and therefore heatsink area) is at a premium.

An LM317T configured as a CC source requires at least 2v dropout on the regulator itself, with another 1.25v on the program-current resistor. 3.25V @ 1.4A is over 4.5 watts.

A PNP configured as a CC source can be made to have a bit lower dissipation, but there is still quite a bit of heat.

The circuit needs to be powered from a 5-6v source. The current source has to be on the rail side of the circuit (not the ground side - hence the PNP transistor requirement).

And even if I can overcome the high dropout issue, the load resistance can fluctuate enough to where heat is still a problem.

When I looked at the LM317T datasheet, I saw some application hints for low-cost (and size) switching voltage regulators. Can a similar scheme be used to create a switching current source?

 

[crutschow]

Yes, you can use a switching regulator to be a constant current source. If you Google "constant current switching regulator" you'll come up with numerous hits.

 

[MikeMl]

Previous threads:

Linear

Linear and SwitchMode

 

[Hero999]

2-transistor Black Regulator

 

[BrianG]

Thanks for answering guys!

I read through those links and not sure if I understand them fully or can use them. Some requirement bullet points:

- The current source circuit has to be "on top" of the load, so hooked directly to the supply. Page 7 of the DS2714 datasheet should show why.

- With a 5v supply, the highest battery (being charged) voltage of 1.65v, and the Vce_sat of the switching transistors (again on page 7 of the link, but not using those exact PNPs) of ~0.1v, I have a minimum voltage differential of 3.25v to work with. When a weak cell (~1.2v) is inserted, the differential goes up to 3.7v.

Another issue: From the little I know about SM regulators, parts placement is VERY important to avoid induced voltages in the wrong places. If I use any of these projects, I would be etching my own PCB, which I have done exactly once about 15 years ago using RadioShack stuff (masks and pen to make the traces).

As I was doing some more searching, I came across this (scroll to the end). I happen to have a few completed PCBs using a version of the main IC, but set up to be a voltage regulator. Adding the additional sense resistor, two Opamps, and the other support components to the feeback loop would be fairly easy (I think).

I was hoping to use an LM317T somehow and just add the coil, diode, and a few other support components. I guess this is what I get for jumping into something a little more than I can handle.

 

[MikeMl]

Did you notice the need for a -15V supply in the EDN circuit?

So you need use high-side current sensing because one end of the load is grounded?

 

[BrianG]

Did you notice the need for a -15V supply in the EDN circuit?

Yeah, I saw that. Current requirements are very low for that so an itty bitty transformer could be used. I'm not even saying I would use this scheme, it was just something I came across and looked do-able.

But to be honest, I don't know why simply hooking the feedback line directly to the load-side of the sense resistor won't work. The output voltage will rise as needed to develop Vref across the sense resistor. With a Vref of 1.237v, a 0.88Ω resistor would be needed for my required 1.4A. Yeah, 1.7w will be generated on the resistor, but that's a lot better than over 4.5w total or more with a linear design. So, if I want less voltage on the resistor (and less heat), I could use the addition of the op-amps, but if I value simplicity, I can still go with what I described at the cost of heat (just not nearly as much). I have a few similar ICs at home I can play with to see if it will work.

So you need use high-side current sensing because one end of the load is grounded?

Pretty much (I think). The way I see it, the DS2714 reads the cell voltages with reference to ground. If I use a current sink (instead of a CC source), it will see the cells as way more than the max of 1.65v. To get around this using a CC sink, I'd have to reference the IC's and cell ground at the same point "above" the CC sink. And that would mess with the 5v line coming into the chip.