Issue with MC34063 current limit simulation

[optotester]

Hi all and best wishes for this new year,

I am trying  to drive a 1W LED (350mA / 3.3V forward voltage) using MC34063. Device is powered with 12 to 24V DC adapter so linear regulation is not an option. I do realize it is much better to use current control mode and that MC34063 is not really efficient. I have other switchers available but I prefer to understand why simulation (or circuit ?) does not work before giving up with that option. I am currently evaluating Proteus for circuit simulation.

When I try to simulate the step-down circuit from the datasheet in Proteus, voltage regulation does work, but not efficiently at all (less than 30%, I would expect around 70%) and current limit does not work. With Rsc = 0.47ohm, I would expect the current in R3 to be limited to 350mW so the wattmeter value should be around 0.36W (R3 * 0.35^2). However, the current limit does not seems to work.

Any ideas ?

PS: R3 is just a random load to test the circuit, in real life it will be the LED. VCC is 24V in the example below.

Kind regards

[T3sl4co1l]

Well, it's not a current limit -- it looks like that under certain conditions, but really it's just skewing the PWM, and frequency I think?  It won't go to zero under like, fault conditions.  All the more reason why it's such a dumb design.

For that, you need the updated model, NCV3063 or 4 for example.  This is actually peak current mode.

Or just a regular, proper current mode controller like UC3842, but not that one exactly because it's boost/flyback, and needs some work to be used as a buck (although maybe not that much, a bootstrap supply would probably work? -- fly the whole chip).

Tim

[optotester]

Thanks for the quick reply, indeed I was bit misguided by the datasheet.

With MC34063, what would be the disadvantages of the "Feedback Control" design as shown here ?
http://www.jiggerjuice.info/electronics/projects/power/LED-drivers.html
Is there any risk of overcurrent condition when the device starts ?

I have some MAX5033 which both includes thermal and short-circuit protection, so I can use one of them with feedback loop if needed. I do not really want to lose an eye.
https://datasheets.maximintegrated.com/en/ds/MAX5033.pdf

What would be the advantages of a true current mode controller compared to step-down DC/DC controller with feedback loop ? Would
https://datasheets.maximintegrated.com/en/ds/MAX1953-MAX1957.pdf work for this application ?

Final question: is that design https://www.instructables.com/Poormans-Buck/ reliable ?
It provides PWM dimming and current regulation, and does not really use much parts compared to an IC + feedback loop. However, the reference voltage does not seem really accurate.

Valentin

[T3sl4co1l]

Thanks for the quick reply, indeed I was bit misguided by the datasheet.

With MC34063, what would be the disadvantages of the "Feedback Control" design as shown here ?
http://www.jiggerjuice.info/electronics/projects/power/LED-drivers.html
Is there any risk of overcurrent condition when the device starts ?

No more than usual, I suppose.

It simply outputs pulses while the input is below threshold, and stops when above.  By setting feedback to a voltage drop across a series resistor, output current is controlled.  The amp is simply to reduce the voltage drop on the resistor.

I have some MAX5033 which both includes thermal and short-circuit protection, so I can use one of them with feedback loop if needed. I do not really want to lose an eye.
https://datasheets.maximintegrated.com/en/ds/MAX5033.pdf

Hmm, the current limit may be right, such that you can just run it full bore into the LED.  It does claim a "hiccup" mode which would be undesirable, but it doesn't say how it works.  (Often, FB voltage is sensed, and when it's well below the nominal threshold, the oscillator frequency is stepped back, or the soft-start circuit is triggered, etc.  Neither of those possibilities is shown on the block diagram, however.)  It would need to be tested.  Otherwise, sure, with the current sense amp, that method will work too.

FYI, I generally avoid Maxim parts, as they tend to be special purpose and suffer from poor availability.  For one-offs, meh, if it works it works.

What would be the advantages of a true current mode controller compared to step-down DC/DC controller with feedback loop ? Would
https://datasheets.maximintegrated.com/en/ds/MAX1953-MAX1957.pdf work for this application ?

A whole-ass controller seems overkill for a 1W LED, but yeah, that'll work.  Same thing, not sure if the limiting behavior will be well enough behaved to use directly or if it needs a different feedback path.  Neat part here is, by sensing inductor current via Rds(on), you can trim that with a series drain resistor to reasonable accuracy (note Rds(on) varies widely with manufacture and temperature).

If nothing else, you can use voltage regulators like these, in the usual way, setting the voltage just a bit over what the LED needs.  Simply use a series resistor to set LED current.  The efficiency is somewhat worse, but still a whole heck of a lot better than with no switching reg at all!

Consider shopping for purpose-made devices.  There are oodles of LED drivers out there, should be something to do exactly what you want, just add inductor and capacitors, maybe a resistor or two.  (I don't know any offhand; the last few LED drivers I used were boost type, for series-connected display backlights.)

Final question: is that design https://www.instructables.com/Poormans-Buck/ reliable ?

The schematic is a... video?  It's cropped, I can't even see the whole thing?  WTF?

Instructables is, in general, a poor resource...

Tim

[magic]

MC34063 can only output about VCC-2V in buck configuration, so do the math if power dissipation is OK for you, considering voltage loss in the chip, buck output current and expected duty cycle.

The current limit works on real world samples. I have seen this chip used as a CC/CV regulator in a cheap Li-Ion charger. Prolly not the best thing in the world, but gets the job done for some value of "done".

[optotester]

Thanks for these answers. For the last link there is actually a PDF on the page :
https://content.instructables.com/ORIG/FO2/8MUZ/H2CE78WC/FO28MUZH2CE78WC.pdf
I usually do not really trust instructable but the designer of this circuit seems quite specialized in analog circuits. It does seems to be a two transistor oscillator controlling a FET with a feedback control loop.

I would be more than happy to go for an IC fitted for this purpose but I have been looking at Farnell and Mouser and finding a driver able to drive a single LED at 350mA with a « high » voltage drop (and ideally with dimming control)  is much more difficult than I thought. In fact... I could not found any except XL4001 which is quite difficult to source.

magic, I am not sure I understand your answer. You would mean that it can only output 24V-2V = 22V ? It would have to dissipate 18V@350mA = 6W ?

[T3sl4co1l]

Ah, hmm, then depending on what the PWM node is supposed to do, it's probably just a hysteretic controller.  You need a relatively large inductor, but the LED current is literally all you're doing, no voltage feedback to worry about at all.  I suppose the PWM node is supposed to be pulled to GND to disable it, which will work (and explains D2), and left to float (at ~ +V/2) otherwise.

Yeah, that's fine, it's just kinda shite on propagation delay (some 100s ns due to the sluggish comparator and drive circuit, and the PMOS being poorer performance than NMOS).  Frequency isn't well controlled, it will rise significantly at low throttle -- PWM is better than continuous dimming here.

Looks like some promising options here:
https://www.digikey.com/short/4cn478
The AL8860 looks to be roughly the same thing, at heart, but with lots of nice features added onto it, all ready to go in a SOT-23.  (Haven't looked if they're made for single LEDs or what if any restrictions there are on pulse width or min/max voltage, etc.)

Tim

[magic]

magic, I am not sure I understand your answer. You would mean that it can only output 24V-2V = 22V ? It would have to dissipate 18V@350mA = 6W ?

Yes, on 24V it will output about 22V when it conducts. This means 2V dropped in the chip, not sure where you got 18V from. 2V·350mA = 700mW, but output current will be pulsed at 20% or so because it's buck, so maybe 150mW. Well, not as bad as I though, but not a start of efficiency either.

This 2V input-outptu difference is specified somewhere in the datasheet, by the way.

[T3sl4co1l]

Note that the voltage drop is only while the switch is on.  Basically, it's like you start at a 2V handicap in supply voltage, and that drop is all loss.  So the efficiency is less than (22) / (24) x 100% to begin with, and continues to drop due to switching loss, rectifier, etc.

This isn't so important at 24V (it's about 10%) but is rather bothersome at lower voltages.

Tim

[optotester]

Thanks for all these very insightful answers. I now remember that I filtered on DIP packaging when ordering LED drivers, which explain why so little options were available (I wanted to prototype on breadboard first). With the high switching frequency of the well-performing LED drivers, I may have to build the LED supply on a small PCB instead. I will definitely go for that option.

Additional question (mainly to continue learning). I tried to implement PWM control for dimming with DC/DC controller (with Schottky diodes to make sure the current varies between the amplified feedback voltage and the PWM high (>1.25V)).
It works, but the the output ripple is really high with PWM frequency @ 1Khz, and going above that is not possible with my PWM generator.

Is there a way to reduce the ripple and to improve this design ?

Regards