# LED Power Supply Design Recommendation

I have a power supply based on VIPer22a. It is an isolated supply with output 17V, 400mA. The input voltage ranges from 100-240V, 47-63Hz. It drives the LED Array module perfectly. I have attached the circuit for that supply as well. But I am designing other product which requires Dual outputs. Here are my specifications :-

• Output 1 :- 17v, 400mA
• Output 2:- 3.3v, 1A
• Constant current and voltage

The circuit I have attached can help identify what exactly I need because it is working fine. I just want to improve on that or may be make it better. I don't know whether it is constant voltage & constant current or not, please let me know that after going through the circuit.

Please recommend me some design for my application or any improvisation on this.

Update :-

I just found ST's eDesign suite which helps in designing power supply. I filled up my requirements and it displayed all the information I require for my supply.

I just want to confirm the completeness/accurateness of this design and if it fits my purpose.

Here it is :-

Schematic :-

Summary :-

Parameters :-

• You have not defined the Vac input specs. Your power specs exceed the universal range input range and also the EU specs for SO-8 chip of 12W, so does that imply you plan on using the DIP-8 with adequate large heatsink? – Sunnyskyguy EE75 Nov 23 '16 at 19:12
• I have updated question with Input voltage specs. Yes, it uses DIP-8 with no heatsink at all and does not heat very much. The circuit works fine – Embedded Geek Nov 23 '16 at 19:15
• Double check the TL431 pinout on your schematic .It does not look right to me . – Autistic Nov 23 '16 at 19:22
• @AnujMattóõ OK was that for design spec 17V*0.4V+3.3V*1A= 10.1W at nominal Vac? ST indicates max WITH a large heatsink Typical power capability is 12W. , not knowing efficiency at 10W while Tjc=15'C/W – Sunnyskyguy EE75 Nov 23 '16 at 19:24
• @SpehroPefhany I guess you are right. – Embedded Geek Nov 23 '16 at 19:32

You are producing a regulated 17V voltage. You could use an inexpensive non-isolated buck regulator to produce 3.3V from that. If it was 100% efficient, 3.3V @1A would draw 200mA from the 17V supply.

It's possible, but not guaranteed, that you might do better by picking a somewhat lower voltage than 17V to regulate down, in which case you could add a second winding to the transformer, wound so as to minimize leakage inductance, and use that semi-regulated supply to run the buck regulator. That probably won't change the efficiency by much, but it might reduce the cost if you can use a lower voltage chip and a smaller inductor.

• I can't use non-isolated supply. Moreover 17V is fixed. The Led array design uses 17V to work. Also constant current is need to glow 20 White Leds with 60mA each. – Embedded Geek Nov 23 '16 at 18:43
• I didn't get any useful information from your answer. Any other suggestion would be helpful. – Embedded Geek Nov 23 '16 at 18:44
• @AnjuMattóõ You're missing the point. The 3.3 and 17 V does not have to isolated from each other since they share ground. Hence go for a small buck regulator to get from 17 to 3.3V. I would recommend TPS62175. – winny Nov 23 '16 at 19:17
• Maybe it can. You may have to sharpen your pencil and do thermal calculations in detail since you will be about at the limit with losses etc. – Spehro Pefhany Nov 23 '16 at 19:47
• You didn't mention the RGB LEDs earlier. 1W seems a bit high for 3 x 0.06 * 3V- more like 0.6W. – Spehro Pefhany Nov 23 '16 at 19:55

I am glad you are happy with the design, I have a few comments not answers. The design might be great for linear loads, but poorly regulated for LEDs which prefer RMS current feedback.

If it is designed for 6.8W @17V or 400mA operating at 14%d.f for example , it is expecting to see a 17 / 0.4A = 42.5 Ohm linear load.

Unfortunately your load is non-linear and at with typical RGB LEDs operating at 17V with 400mA Max that won't be RMS . I estimate the LED array to have an ESR of about 2-5 Ohms at 400mA rms power or less , thus your current ripple is going to be 100% in Grn,Blue , but not R and is very senistive to the ESR in both 220uF caps, DCR of transformer, RdsOn of switcher and diode power ESR=1/Pd rating.

Is this acceptable to you? Also is your duty factor is a lot higher than design calc?

The problems with regulation and stability using a perfectly good regulator for a nonlinear load do not follow the rule of BOde plot phase margin , nor precise regulation and hence much higher gain error sensitivity to all parameters even with infinite gain in the 2.5V comparator with internal Vref. The feedback is mainly linear average voltage, rather than peak current or even RMS current.

As you know, the transfer function is nonlinear for any diode and with and array of 3P5S is offset by that Zener like threshold and amplified by the incremental slope of diode ESR such that for each colour, the string Voltage , Vs, is the same , but each string will equate a different current transfer function as follows;

$Vs_{LED}=5*(V_f+I_f*ESR)$. +Rs for 5*1V delta for Red with 5S string.

If your goal was to achieve the max. rated Lumen per LED with 400mA peak, equally shared by RGB and thus 133mA peak per colour with an RMS value of 65mA , each string if balanced would have a ratio of RMS/peak of 65/150= 1.3/3=0.43.

However given your feedforward conversion is half-wave sine the RMS/peak is closer $1/\pi=0.32$ this error is amplified by the voltage variance of $If*ESR$ which is controlled not by the ripple voltage but the difference between (17V-5*Vth+Rs) where Rs = 0 for Bl,Grn.

The two 220uF=440uF caps are probable chosen for the expected ripple with a linear load of 42.5 ohms giving a design parameter of a linear RC = 18.7 ms. With an optimistic ESR of 1 ohm then T=ESRC = 1*440uF= 440us. With pulse current at 60kHz, 1/f=16.7us However, the power ripple in the LEDs is not linear like a rectangular switched PWM and very sensitive to LED ESR variation, Cap ESR variation, Line input Vac range and variable load current if modulated by PWM.

You haven't reported power efficiency vs design goal, but I suspect losses are higher than estimates and sensitivity to Pout with each component loss parameter variations.

• Thanks for the elaborate explanation and this has made me more confused then ever. I am stuck now. Because the schematic I posted earlier to the update is working fine from past 3 years, how can it can't work?? I am just adding another voltage and adding more amps. Kindly shed light on that as well – Embedded Geek Nov 25 '16 at 18:46
• Was anything verified to specs given? Vac min/max . Is load constant? is there any overshoot? or load regulation issues? Works more than just shedding light. it means tolerance and sensitivity to variables, so please illuminate us. there are many unstated assumptions like BOM parts, ambient and modulation of load, history , test methods etc – Sunnyskyguy EE75 Nov 25 '16 at 19:14
• it also begs the question, have you done power cycling tests with dropouts < thermal time constant of PTC which protects the primary circuit diodes and caps from stress. 10k random dropout cycles and durations? equiv to turning on/off with switch bounce and power interruptions . and how will that affect over voltage given there is no OTP, OCP, OVP. Does it pass CE conducted noise? there are no y caps. does it pass radiated CE noise, given high pulse currents at 60kHz with ~ 100kA/s slew rate with unknown loop area? – Sunnyskyguy EE75 Nov 25 '16 at 19:22
• No I haven't done these many test, nor I have gone into much detail. – Embedded Geek Nov 25 '16 at 20:51
• not possible to drive 4S Whites with 17V while 5S for Blue – Sunnyskyguy EE75 Nov 26 '16 at 0:26