Solution Finder

Adding more leds to existing design

Posted by: Richard Wright on Mon, 04/20/2020


I changed my PI password, just in case it was something on my end that was causing the spam.

RE previous messages, Am I just expecting too much from the driver? Have I reached the limit of what it can put out? All I want now, is 2 strings of 9 leds, 330mA total current, approx. 24-27V, but I need constant current drive. Do I need to go up to a 3.3mH inductor possibly? Is the noise on the FB pin causing the issue? I put a cap with a higher voltage rating on the Cfb, and it got rid of the flashing with 16 leds, but now I need to go up to 18 to get the lumen output we need. As soon as I add them, the flashing starts again. I can't change the driver circuit alot, as our customer doesn't want a whole new design. I figure the design is around 8.3W, is this too high? I'm still thinking it is the FB that is doing it, but I need a way to get rid of that noise, I'm thinking it is from the switching diode. Any kind of RC snubber I could put on it?
Any other info I can get from the circuit that will help you? Please let me know. Any more help would be greatly appreciated. Thanks.


Submitted by Richard Wright on Tue, 04/21/2020

Well, I changed my password, and it still appears to be spam. However, I still need help with this.

Submitted by PI-Martok on Tue, 04/21/2020

Hi Richard,
Kindly double check the probe connection? The FB pin voltage looks different from the first one the you sent to me .The FB pin voltage should be DC voltage with respect to source pin due to C3. But it looks like the waveform corresponds to the device switching behavior and I observed the non-operational waveform is a switching without pulse skip and still its not providing full power to the load. This will lead to auto-restart mode condition. On the other hand the normal operation waveform is cycle skipping which means it were able to deliver the required output power.
For LNK306, the current limit and Fsw is fix so the only way is to increase the inductance to be able to deliver higher output power and generate more pulse skipping which is required for normal full load operation.
8.3W is still ok as long as the device has enough heat-sinking. Device temperature at room temperature should not exceed 100C.
With 2X9V LED string can you please list down what's the difference on the operation with the below changes?
1. Increase L1 inductance - 1.5mH to 3mH- You can add 2X1.5mH in series connection.Please capture the VDS waveform.
2. Increase C1 Capacitance from 10uH to 22uH
3. Removed R4

Submitted by Richard Wright on Wed, 04/22/2020

We've decided on a different led with 2 strings of 10, but only 290mA total. I have changed a few things, including a 4.7uF 50V on the output, a 10uF 25v for the Csense, and changed Rsense to 6.8-ohms. I also increased the input cap to 33uf. It seems to be working continuously down to 100Vac input now. I think I will still increase the inductor so that the chip isn't working as hard. As you mentioned, there is very little pulse-skipping going on. I mainly wanted to be sure I wasn't missing something in the fb circuit, as far as values for the two resistors, and what role they play. Everytime I start to read about the 49uA current source in the chip, I'm trying to figure out if the current has to mirror that at the fb pin, or what exacly is going on. I keep thinking the fb input would look like a 33k ohm resistor in parallel with the 2k-ohm external resistor, when it meets that requirement. Thanks for the advice. Much appreciated.

Submitted by PI-Martok on Sun, 04/26/2020

Hi Richard,
Thanks for your inputs. 1.65V external feedback voltage corresponds to an internal 49uA feedback pin current. Through 1.65V (FB)feedback threshold, the IC monitors the voltage in the sense resistor to regulate the output current. The IC samples the FB Pin current only at the start of the switching cycle(rising edge of ton) so the noise you are seeing on FB pin probably disragarded since these are during turn off period of the internal mosfet.

Since you are using ceramic capacitor for Csense, Increasing the voltage rating will help lower the ripple. Using electrolytic capacitor is much better in terms of ripple reduction and audible noise performance.

Submitted by Richard Wright on Mon, 05/04/2020

Believe it or not, that is with a Tantalum cap on Csense. I've had the circuit running for days now at different input voltages, and it seems to be doing good down to 95VAC input. We're getting prototype boards in later this week, so hope it works as good as the mock-up on our old board. Thanks again for the FB explanation.

Submitted by Richard Wright on Thu, 05/21/2020

Hello again PI-Martok,
Hope you are doing well. As mentioned before, we are using a tantalum cap for Csense. However, I am getting minimum operating input voltage that are varying from unit to unit, plus it rises as the unit warms up. I tried using an electrolytic type, which does seem to be better, but not as good as I would like to see. Is there anything I can do with the two resistors in the feedback circuit to make this more consistent from unit to unit? Also we are having a heat problem when the unit is put into it's housing, as there is no ventilation, unfortunately. I noticed that the input capacitor which is a 27uF, 400v device, that has a ripple current rating of 720mA at 120khz, is getting quite hot, too, as well as the inductor, which only has a 2-ohm DC resistance. I am only running at 290mA now, and still seem to be having issues. Any more ideas? Thanks in advance!

Submitted by Richard Wright on Fri, 05/22/2020

What happened with the Lytswitch-1 series that it is being phased out? That seems like it would have been a perfect solution to my problem! Still seeking info from my last post, if any more info might help. I don't really want to get into a design with a transformer involved. Thanks!

Submitted by PI-Martok on Sat, 05/23/2020

Hi Richard,

1. For unit to unit variation. Kindly increase the output capacitor a little bit more to reduce the ripple current? The ripple current will vary with respect to input voltage and LED voltage variation and it will reflect to the Rsense.

2. For the thermal Issue. I suspect the drain copper pad where the input cap sit is too hot. You can try to add solder mask around the drain copper pad to reinforce the heatsink. Have you already removed 100k across the input cap? this resistor dissipates 1.5W at 277V. Can you also verify the drain current waveform if the power inductor is saturating? Saturation will increase rms current and Power dissipation on the device mosfet.

Submitted by Richard Wright on Tue, 05/26/2020

Hello PI-Martok,
I am changing the 100k to 470k. I'm not following you on the solder mask around the Drain copper pad to reinforce heatsink...more space between copper? To measure the drain current, do I have to lift the pin and put a current loop in place with a current probe, or is there another means that would be easier? I can try to increase the output cap, but I think the restart kicked in with a higher value. Thanks.

Submitted by Richard Wright on Tue, 05/26/2020

Hello again,
I have attached a screen capture of 1 pulse of the drain current on the IC. it works out to be 592mA peak, not including the initial spike on the waveform. My probe was 10mA/mV, and the peak was 59.2mV.

Attachment Size
Drain current of single pulse. 33.36 KB
Submitted by PI-Martok on Tue, 05/26/2020

There was no sign of saturation on your drain current waveform which is good however the leading edge spike is so high and would cause high RMS current through the MOSFET and Inductor. This might be one of the reason of MOSFET overheating due to high conduction loss. Can You please try to use ultrafast diode( trr ≤35 ns) for the free wheeling diode (D4). This should minimize the leading edge spike and consequently reduce the MOSFET and Inductor temperature.

Submitted by Richard Wright on Wed, 05/27/2020

Hello Pi-Martok,
Thanks for the quick reply The diode we are using has a trr of 25ns. It is a ES1G-13-F. I reduced our output current to 280mA, to see how much things changed, too. It works initially down to 85vac in, however, I noticed on the current waveform I am getting double spikes, one larger than the other. Could the issue be associated with the open load circuit partially conducting? Please see scope capture. Thanks.

Attachment Size
TEK0001.jpg 36.1 KB
Submitted by Richard Wright on Wed, 05/27/2020

Why do these links keep appearing? Am I being spammed on my end? I changed my PI password not too long ago.

Submitted by PI-Martok on Thu, 05/28/2020

Hi Richard,
The first spike is usually called leading edge spike common to MOSFET high voltage power switch. This is due to parasitic capacitance between the drain to source called the Coss. During off time, the input voltage will charge the Coss and when the MOSFET turns on, it will discharged the Coss that will result to a high leading edge spike. The leading edge spike is worst during continues mode operation due to free wheeling diode reverse recovery current is higher and this energy will be reflected to the input that will worsen the leading edge current spike.
Can you please try to use schottky diode and see if it will help lower the leading edge current spike?