EEVblog Electronics Community Forum
Electronics => Beginners => Topic started by: LooseJunkHater on June 15, 2024, 01:15:38 am
-
Background: Trying to fix a 30W LED passively-cooled spotlight with adjustable light output controled via encoder [image 3 of PCB]
How the spotlight works: 12v power input is boosted using a typical boost converter circuit, to something around 30v-ish. The MCU sends a 30Khz signal to a SOP8 I.C (labelled "U1") which has MOST of its numbers rubbed off, with only "394" being visible. I could not locate the I.C with a matching footprint + schematic. 100% duty cycle (from MCU) = full LED power, while 0% duty cycle = no light output from LED. [image 6 of the signals]
The Problem: Lots. Essentially the MOSFET was initially found shorted. It was replaced with an (EBay special) "identical" FET. I had originally replaced the 30N06L FET with a known-good FET with similar specs, but the same problems persisted so I don't think my EBay FET is the problem.
- U1 gets an input voltage of about 5v (steady). When signaled by the MCU, U1 forward the same signal from the MCU to the FET gate (ONLY AT 5V)
- The output voltage is a horrible 25v with ripple from 20-30v.
- As the duty-cycle is increased to U1, the FET becomes hotter and hotter, absorbing in excess of 15w+ (NOT BY DESIGN), with the output voltage not changing (the ripple gets worse) and the light output actually getting dimmer
- ~640KHz oscillations are seen on the output AS WELL AS the gate of the FET
?Ideas on the problem?: The 10uF tantalum cap tested fine and the voltage to U1 is steady. The input/output electrolytic capacitors also tested fine (accurate capacitance reading + Megaohm resistances), no obvious burns/buldging on any components. Is U1 faulty? The output electrolytic capacitors? The inductor? Something else?
-
Image 1: My guess of the I.C pinout and what each pin does
Image 2: My drawn board-view + schematic
Image 3: Image of the schematic (U1 removed to view traces under). No traces on rear, just ground.
Image 4: Close-up of U1 and the scratched-off writing
Image 5 + Image 6: As indicated in images
-
Also note-worthy is that there are essentially two grounds for an unknown (to me) reason, seperated by the 0.36ohm resistor. There's also a 0.05ohm resistor on the drain of the FET. Both are maybe to monitor the power output of both input and output?
-
Yeah the shunt resistor in the led return ground is to measure the current.
Probably that circuit is designed to work at 12V, not 5.
Most fets will conduct pretty bad at 5VGS, and do exactly that, dissipate a lot of heat.
You might be lucky if it works by replacing the fet with a low gate threshold one.
-
Based on the information I provided, do you think anything is wrong with the circuit? Do you think I should feed more voltage into the unknown I.C (and risk blowing it up) and see if it increases the gate voltage? Do you think it may be something else in the circuit leading to the fault exhibited?
Edit: Per the original FET's datasheet, it looks like the FET is near-entirely conducting at 5v.
https://pdf1.alldatasheet.com/datasheet-pdf/view/604829/FAIRCHILD/FQB30N06L.html
What I guess I can do is hookup the gate of the FET to my signal generator and see how hot the FET gets at a 5v vs 12v signal when switching the output. I guess this could tell me whether the EBay-FET is the problem. Should I do this???
-
Maybe, yes.
-
IC looks a lot like the ZSLS7025. If it is the same IC, your lamp circuit has implemented the PWM dimming incorrectly. (Or maybe there is an error in your schematic)
-
IC looks a lot like the ZSLS7025. If it is the same IC, your lamp circuit has implemented the PWM dimming incorrectly. (Or maybe there is an error in your schematic)
Looks like the pinout matches, including the VDD pin being 5V. Based on that information, I'm lead to believe that the I.C is fine and it's the EBay-FET.
My plan is to desolder the I.C, feed 12v @ 30khz to the FET gate and see if it gets warm or not.
-
IC looks a lot like the ZSLS7025. If it is the same IC, your lamp circuit has implemented the PWM dimming incorrectly. (Or maybe there is an error in your schematic)
Looks like the pinout matches, including the VDD pin being 5V. Based on that information, I'm lead to believe that the I.C is fine and it's the EBay-FET.
My plan is to desolder the I.C, feed 12v @ 30khz to the FET gate and see if it gets warm or not.
Update:
Desoldered the "ZSLS7025" IC, fed 5v 30khz to the gate of the FET with a varying duty cycle and indeed it becomes VERY hot (100C+) at a duty cycle of less than 5%. I then fed 12v 30khz to the gate of the FET and while the EBay-FET did become warm (around 60C), the 0.05ohm current-sense resistors actually became very hot (120C+) at a duty cycle of around 5%. I have no understanding as to why the 0.05ohm current sense resistors became hot (someone know why?) but none-the-less, to my understanding it does seem like the EBay-FET is a fake as it does not entirely conduct at a gate voltage of 5v.
My next step is to remove the EBay-FET and look in my spare parts bins and see if I can find a FET which fully conducts at a 5v gate voltage.
-
Where's that 50mOhm resistor?
You pictures are all incomplete, making harder to understand the thing.
Make sure the diode is working!
Capture the pwm input and the gate waveform to make sure it's switching properly.
50V+ logic level fets are rarely seem, you'll probably need to order few.
You can test RDS on with a small pwm duty (So it doesn't
overheat) while watching VGS, which should be very low.
Probably at 5VGS it's dropping a lot of voltage, hence the heat, but there can be other causes like poor switching, ringing..
-
The fault has been found! Whoever was thinking that it was a faulty inductor (alongside the FET), you're the winner.
I had replaced the EBay-FET with a Vanguard 013n10m (https://alltransistors.com/adv/pdfview.php?doc=vsf013n10ms.pdf&dire=_cn_vanguard) which was the only FET that I had stocked that closely-matched the original FET's VGS @ 5v, Continuous drain current, and Drain-Source breakdown voltage, but the exact same problem persisted.
On the left of the attached image [attach=1]
is the original (faulty) inductor, acting entirely as a short over a span of 50KHz-1MHz, showing less than 1uH over the entire span. On right is an inductor essentially measuring as ~48uH over the entire frequency range.
For funsies, I destroyed the original inductor to see if there was any obvious burnt windings inside but there was not.
What was interesting was that the output of the circuit was indeed reaching around 25v and just barely turning on the LED, so the inductor was partially working, but nowhere near its original inductance. What led me to think about the inductor being faulty was that it would get very hot, as well as the FET. I knew that the inductor shouldn't have gotten hot (~40C) at low loads, hence why I turned my attention to it.
-
Congrats on the fix. All it takes is for the thin insulation on the magnet wire to break down or be damaged during manufacturing. Where the last turn crosses the first turn on the coil, on it's way to the component leg, is particularly prone to that. That's where the highest voltage across the coil is.
Yea, the clue was in the ZSLS7025 datasheet. But now it's more obvious:
The inductance value of the inductor (L1) directly determines the switching frequency of the converter. Under fixed
conditions, the inductance is inversely proportional to the switching frequency; i.e., the larger the inductance, the
lower the switching frequency. A higher switching frequency will reduce the value required for the inductor but will
increase the switching loss in the external MOSFET, Q1
The recommended switching frequency is 20kHz < f < 200kHz. Lower than 20KHz will cause audio noise of the
inductor, and a frequency that is too high will increase the switching loss in Q1.
Combined with what you measured:
- ~640KHz oscillations are seen on the output AS WELL AS the gate of the FET
-
Alright now that I've found the fault, I'm trying to grow my knowledge further in terms of inductors. I'm trying to use a NanoVNA (with the NanoVNA software, calibrated) to determine the best inductor but I'm not 100% sure what to look for. I've found 3 inductors with similar inductances at ~400KHz-1MHz, but at around 30KHz (the operating frequency of the boost converter circuit), their inductances are quite different.
[attach=1]
What exact value should I be looking for? The Q-factor at 30KHz? The inductance at 30KHz? VSWR at 30KHz? Impedence at 30KHz? I honestly have no idea.
I know that this is starting to go deep-end of inductors and I have essentially no knowledge in this area, hence I'm reaching out again.
___
The "yellow-shrink wrapped inductor" got to ~120C when ran in the circuit (and caused the FET to run at 65C), while the "60c round inductor" only actually got to 55C with the FET running at 50C. This obviously means that the "60c round inductor" is more efficient in this circuit, BUT WHY???
-
I would say, "The Q-factor at 30KHz and the inductance at 30KHz". You also have to take into account the saturation characteristics of the magnetic core.
You can ignore VSWR completely. Impedance (reactance) is the function of the inductance and frequency anyway.
-
I would say, "The Q-factor at 30KHz and the inductance at 30KHz". You also have to take into account the saturation characteristics of the magnetic core.
You can ignore VSWR completely. Impedance (reactance) is the function of the inductance and frequency anyway.
If inductance and Q-factor are the most important variables, do you have any idea why the "60c round inductor" ran cooler compared to the "yellow-shrink wrapped inductor" (per my previous image)?
-
Not sure. But then I'm not sure how the NanoVNA measures Q or what Q in S11 mode means. S11 is usually for measuring return loss so I'm not sure how this would relate to inductor Q.
Maybe the core on the "yellow-shrink wrapped inductor" was saturating, when too much current flowed through it, but the 60c didn't? Just a guess.
Hopefully someone else knows the answer to that.