HV9910 and inductor size

[justDIY]

I would like to build some hv9910 based constant-current regulators for using in my experiments with LEDs. I would like to build one converter, that can handle several different output options, by changing the current programming resistor(s).

I have a basic knowledge of the forumla involved with sizing the components, and have built an excel spreadsheet to help keep it all straight. I have had prototypes working on the breadboard of both the buck and buck-boost configurations.

it would seem from my experience, the less output current required of the converter, the more inductance is required, in an almost linear relationship.

for example:
given an input voltage of 24v and an estimated output voltage of 36v, I come up with these inductace values:

0.550 A requires roughly 880 uH
1.100 A requires roughly 440 uH

now my question is, if I build the converter to handle 0.55 A, but later change the programming resistor to regulate for 1.1 A, what effect will the larger than required inductor have? Will its extra inductance simply be wasted capacity, or will it have some detremental effect on the circuit?

 

[Oznog]

Did you mean 3.6v instead of 36v? 36v would be quite an LED string- and I'd suggest another part than the HV9910 for boost conversion.

Larger inductors in a current mode buck converter simply reduce the ripple further. There's no inherent problem. But be sure to still get enough average current and saturation current capacity; for a given inductor package, one wound for twice the inductance generally has half the current capacity. Check the specs.

The HV9910 has a prob though. The current required to turn on the gate capacitance goes through the shunt resistor. In cases where the ripple is low, the cycle begins with inductor current only slightly below the threshold value. The gate surge added on top of this will cause the compartor to reset the output latch. The answer is an RC filter between the shunt resistor and the input pin of the HV9910. I believe an RC time constant of 300nS was the minimum to make it ignore the gate surge.

Setting different currents might be better accomplished with the LD terminal. Use the smallest resistance for the highest current case you want to design for, then use a resistive voltage divider between VddOut and Gnd with the center to LD to make a new threshold voltage below 250mV that will override the built-in 250mV threshold.

 

[justDIY]

Did you mean 3.6v instead of 36v? 36v would be quite an LED string- and I'd suggest another part than the HV9910 for boost conversion.

what do you suggest? I've looked into the product offerings of TI, Maxim and some others ... Linear has a few options, but the limited (expensive) availability of Linear makes me stay away from them. Ideally I'd rather go with a straight step-down, from say 48v ... however, most of the high-efficiency step-down converters out there are designed to step-down 48v to 3 - 12v (obvisouly for POE / telcom applications)

I did not mistype, 36 volts is correct, perhaps a bit low even ... at 1a. I have two strings of ten 2-watt Nicha Jupiters, with a forward voltage of 3.6 v per diode. Due to the physical installation of the diodes, wiring fewer than ten in a larger parallel array is not an option (like 5 x 4 for example)

The reason I've chosen the buck-boost mode, is I cannot find a DC supply which would provide (at least) 72 volts the hv9910's buck mode requires. And I'm not willing to direct-drive the hv9910 from the mains (rectified of course)

Larger inductors in a current mode buck converter simply reduce the ripple further. There's no inherent problem. But be sure to still get enough average current and saturation current capacity; for a given inductor package, one wound for twice the inductance generally has half the current capacity. Check the specs.

Thanks for the info - I will be watchful of the current rating.

The HV9910 has a prob though. The current required to turn on the gate capacitance goes through the shunt resistor. In cases where the ripple is low, the cycle begins with inductor current only slightly below the threshold value. The gate surge added on top of this will cause the compartor to reset the output latch. The answer is an RC filter between the shunt resistor and the input pin of the HV9910. I believe an RC time constant of 300nS was the minimum to make it ignore the gate surge.

I'm afraid you've lost me here ... I am following Figure 3 in the hv9910 datasheet, I have no resistor between the gate pin and the mosfet gate. Is the resistor internal to the chip?

Setting different currents might be better accomplished with the LD terminal. Use the smallest resistance for the highest current case you want to design for, then use a resistive voltage divider between VddOut and Gnd with the center to LD to make a new threshold voltage below 250mV that will override the built-in 250mV threshold.

that's my plan b!

 

[Oznog]

This type of reg does not work all that well. There are a number of issues:
1. The reg cannot regulate very effectively. The inductor current may be set at 1 amp, but the average current is still proportional to duty cycle and duty cycle depends on input voltage. So if the DC input voltage isn't regulated the LED intensity will change with input voltage.
2. The LEDs depend on a cap to keep from turning off while the inductor is charging. The isn't an ideal situation because an LED has a sharp IV curve, the cap need only discharge slightly before the LED current decreases substantially. And it's going to be discharging at 1 amp. So there are 3 requirements for effective filtering- high capacitance (and/or high freq), very low ESR, and a high ripple rating (generally only big quality caps are rated for 1 amp). I thought another inductor and schottky would do a better job but I haven't seen this design suggested.
3. There's a flaw if the power supply is having trouble producing the specified current. The latch will stay on so the MOSFET just becomes a dead short, as does the inductor.

It would be much, much easier if you could do 4 strings of 5. You really, really sure you can't? Just some sharing ballast resistors.

The HV9930 is better designed for boost conversion from what I can tell, but I don't know much about it. I think Zetex may have made a good one too. Ideally I'd look for one whose average output current doesn't depend on duty cycle. My statements about the stability of larger inductors may or may not stay true with a different circuit.

I'm afraid you've lost me here ... I am following Figure 3 in the hv9910 datasheet, I have no resistor between the gate pin and the mosfet gate. Is the resistor internal to the chip?

No. Generallly a gate resistor isn't necessary (it'll slow the transitions and increase the heat) and it wasn't what I was describing. I'm taking about adding an R and C on the feedback voltage from the shunt.

 

[justDIY]

thanks for your input!

I think part of my problem is trying to cover too many needs with a single solution.

for experimenting on the workbench, I'll build a step-down HV9910 with 1.5A capacity, and use a pot to change the 250mV sense voltage (using the LD pin), so I can dial down the current as needed. With a 48 volt input, I should be able to run a decent number of diodes off the max of 24 v output

for my Nicha Jupiter array; I think I will build a step-down model, and feed it 48v... but I will build two converters, one for each half of the string. instead of one array of 20 leds, wired 10x2, I'll wire it as two arrays of 10 leds, wired 5x2. This will reduce the power needed to 18v @ 1a, instead of the 36v @ 1a. supplying the converters with 48 volts will be a lot easier than supplying it with 70+ volts. To keep the wiring clean, I'll mount the converters inside the led assemblies, instead of having it live with the lighting controller.

Is there anything to watch out for feeding one smps with another smps? I mean, the inexpensive 48vdc supplies I have are certified up the wazoo for compliance, so they should be pretty 'clean'. I think they're intended for telcom applications.

 

[Oznog]

You don't need 2 converters, the LEDs can go in parallel with appropriate ballast resistance. Since you're not trying to ballast for variations in source voltage, a relatively lower resistance is used.

Actually feeding one SMPS with another may be a big problem!
SMPS always have stability concerns, particularly with low loads or quickly varying loads. Capacitance may or may not stabilize it, and it would be tough to predict how stable this will be without knowing more about the SMPS design.

Fortunately, the HV9910 is quite tolerant of terrible input noise. It will simply turn on the transistor for as long as it takes for the inductor current to reach the specified level, so it won't matter much if the input is 38V or 60V or oscillating between at 1khz. As long as it stays below the max voltage of both the HV9910 and the transistor. The HV9910 logic is driven off its internal 7.5v reg and the reg has a (large) external storage cap so it can function with even a rectified yet unfiltered AC input.