Self oscillating buck LED driver has too many components

[Flyback]

Hello,
I have here a simulation (LTspice) of a self oscillating Buck LED driver. It just regulates the LED current by switching ON/OFF its FET when the inductor current hits the requisite peak and trough levels. One comparator acts on the peak of the inductor current, the other, on the trough. (there is a resistive inductor current sensor)

The 'glue' logic that I have is a bit messy, (basically it just allows one comparator to disable the other comparator when it trips, to allow proper operation).

Do you know how I can use say NAND latchs in order to tidy up this circuit?

LTspice simulation and schematic attached..

 

[alec_t]

The 'glue' logic that I have is a bit messy, (basically it just allows one comparator to disable the other comparator when it trips, to allow proper operation). Do you know how I can use say NAND latchs in order to tidy up this circuit?

Couldn't you just replace the two comparators by a single comparator with hysteresis?

 

[Flyback]

yes but the led current regulation would then not be so accurate over line and load voltage changes.
The system in use above is the same as that used in the ZXLD1366 LED Driver...

**broken link removed**

 

[alec_t]

the led current regulation would then not be so accurate over line and load voltage changes

I don't see why not, if the hysteretic threshold levels are defined by a reference voltage source.

The system in use above is the same as that used in the ZXLD1366 LED Driver

Then I hope you won't be infringing anyone's IP rights .

 

[ronsimpson]

If you buy a reel of ZXLD1366s the price is just over $1 us each.
Is this an experiment? or are you serious about building your own part?

 

[Flyback]

its for a smps course, we want the hardware, and the simulation.
Hysteretic converters are not as accurate over line and load changes....that's why zxld1366 contains the current from 100mA +/- 15%.... its the most accurate way.

 

[crutschow]

What load current versus input voltage coefficient and load current versus output load do you need? Or what is the input voltage range and and output load change allowed for the 100mA +/- 15% change in load current?

 

[Flyback]

The simple hysteretic converter is more than 15% in accurate when compared to the two comparator one.

We need the two comparators...of that there is no doubt...clearly, to get accuracy in current setting, one needs to fix the peak AND trough of the inductor current.

I think having a toggle flip flop connected to the output of each comparator could do the trick...then or'ing their outputs together?

 

[ronsimpson]

the led current regulation would then not be so accurate over line and load voltage changes.

For 1 LED; 7 volts to 30 volts the regulation is 2%. To 50 volts 4%. How good do you need?

68uH

 

[ronsimpson]

With 220uH the regulation is much better. 1% from 8 to 50 volts.

 

[Flyback]

yes that's right, but that's the zxld1366, and that uses the "two comparator " method, rather than the "single comparator with hysteresis" method

 

[crutschow]

The simple hysteretic converter is more than 15% in accurate when compared to the two comparator one.

We need the two comparators...of that there is no doubt...clearly, to get accuracy in current setting, one needs to fix the peak AND trough of the inductor current.
....................................

It still is not clear what accuracy in current setting you want/need.(?) You can't do a design without clear specifications.

 

[Flyback]

the accuracy needed is within 10% , and the single hysteretic comparator doesn't cut it over line and load changes.

We are down to fracking now, we are running out of fuel, we need lighting to be as efficient as possible, surely you agree that the single comparator is too inaccurate and thereby inefficient.....

Basically there are two comparators, one trips on the peak of the inductor current, one trips on the trough of theinductor current....how do you make it so that that turns the fet on/off correctly, doweneed a state machine, karnaugh maps, and nand/d-type logic?...or can it be done with just a toggle flipflop

 

[alec_t]

We need the two comparators...of that there is no doubt...clearly, to get accuracy in current setting, one needs to fix the peak AND trough of the inductor current.

When running your sim I see both comparators tripping at the same point, corresponding to the current trough?? Peak current is being limited by the inductor rather than being set by either comparator.
Here's a sim of a hysteretic single-comparator simplification of your circuit. LED current variation is ~4% for a 3:1 variation in supply voltage, and less than that for a load change from 1 to 2 LEDs. Rd and Cd allow the frequency to be reduced (if required).

 

[ronsimpson]

surely you agree that the single comparator is too inaccurate and thereby inefficient.....

I don't see that inaccurate has much to do with inefficient. accurate <> efficient (does not equal)
You could be 95% efficient at -10% or 95% efficient at +10%. Who cares.

When the price of the bulbs is right they will have a market. They do have a marked based on money. Only part of the market is driven by "lets save the earth".

Two comparators verses one comparator and a moving reference. No differance.

 

[Flyback]

OK but why does diodes.com bother having two comparators if only one will do?

Alec_t......thanks for the sim, i enjoyed that, I had not previously known how to do stepped comparisons like that...im storing that one away where I can find it again, very useful.

Though Alec_t,i used your sim sometimes with just one led,and sometimes with three, and with the 10 to 30v input range, and got an overall 10% max difference. Remember that these buck converter led driversa re likely to appear in multiples in a singleproduct, andbe supplied by an upstream offline power supply.....if every buck is supplying 10% more than it should, then the offline psu is going to overheat, or need its components to be too high rated, ie, higher rated thannecessary.

Diodes.com have done the zxld1366 like it is for a reason.

Ive forgotten how I did the logic in my sim, it was more try and see, keep changing it till it worked,and eventually it did, but I feel there must be a cleaner way to do it with the two comparators?

 

[alec_t]

that's why zxld1366 contains the current from 100mA +/- 15%.... its the most accurate way.

Isn't that worse than the 10% variation you found with my sim??

sometimes with three

I didn't try that.

it was more try and see, keep changing it till it worked

Ah, the beauty of simulation .

Diodes.com have done the zxld1366 like it is for a reason.

I'm sure they have, but it's not unknown for manufacturers/designers to omit vital details, or add spurious details, in published schematics to obfuscate the real circuit as an antitheft/anti-piracy measure. Of course, I'm not saying Diodes.com would do such a thing . Since both comparators trip at virtually the same point in your sim, perhaps one is a belt-and-braces backup to ensure the FET switches off at some safe point? I don't see why you/Diodes.com have R12 and R13 shorted together? That places R2 and R3 in parallel.

 

[Flyback]

I see what you mean about those resistors, I think I just got it working and that was it, i left it, I have been trying to do it properly with memory logic bat haven't succeeded yet.

 

[alec_t]

I think I just got it working

Did you get the comparators to trip at current peak and trough respectively? Personally, I don't see the trough trip level being as critical as the peak trip level ....providing the FET switches on and off. The less the gap between the two levels, the higher the switching frequency and the lower the LED current ripple amplitude.
You've 'cheated' by using a behavioural voltage source to get the current-sense voltage V(A)-V(B) . Since V(A) and V(B) are both at/near the 20V supply rail, how will you actually get them into comparators running from the 5V rail?

 

[Flyback]

Did you get the comparators to trip at current peak and trough respectively?

I tried to do that, that's what I was trying to do.
Yes I used behavioural volt source, but to be honest, its just a representative simulation so that the students can see how its working...I mean, we have the zxld1366 hardware working, its just that the students can only be in the lab when supervised, other-times they only have the simulator -hence we need something to represent the zxld1366 in the simulator.