Solution Finder Get Tech Support

LNK306DG-TL heat problem

Posted by: Qawasmi on

Dear PI Expert,

We are designing a solution that use LNK306DG-TL , we experince a very high temperature in a short period of time ( 30min Appr.) , attached is the schematic, we noticed that D4 is DIODE U.FAST Single 1A 1000V SMT DO214AC (SMA) is Trr is < 100ns in the schematic , so we changed to STTH1R02A DIODE U.FAST Single 1.5A 200V SMT DO214AC (SMA), Trr 15ns.

Also we tried to change coils L1-L2 ,they were FPI0503-102K INDUCTOR FIXED 1mH 310MA 8 OHM SMD so we changed to RC0810U-102K INDUCTOR FIXED 1MH 750mA 10% RADIAL,

after using the new Diode and new fixed inductor , temperature are good so far in the first few hours , but diode reach 75-80 C degree. please advice as we were been testing for weeks.

Please let us know if you still need any other information,

Best Regards
H.Qawasmi

Files

Attachment Size
schematic 217.97 KB

Comments

Submitted by Qawasmi on 01/14/2020

Hello,

We are still looking for help urgently please

Best regards

Submitted by PI-Yoda on 01/14/2020

Hi Qawasmi,

It will be helpful to know some generic parameters of your design, like output power and input voltage.
When you say 'very high temperature' how high? On which component?
What is the actual measured RMS voltage Vin over C1 of the time of the experiments?

Best Regards,

Submitted by Qawasmi on 01/16/2020

Hi PI-Yoda,

Here are the parameters:

Vin: 85-265VAC (we are testing with 220VAC)
Vout: 12VDC
Iout: Load requires 200mA on most of times. 300mA for a short period (5 seconds)

Voltage at C1: 293VDC (tested with a 22uF 400V capacitor)

We measure up to 86 C on both LNK306 & D4. And as the circuit is too small; we are measuring around 60-70 C on some other power parts.
I wonder if 85-90 C is high or normal. If high, we need to learn about how to decrease it.

We use PIExpertOnline tool and with these values; it should provide up to 330mA without problem.

Thanks for your help.

Submitted by PI-Yoda on 01/17/2020

Hi Qawasmi,
60 DegC temperature rise (on 25DegC ambient) sounds high.
1. Is it possible the surface heat-sinking area not be provided as required in page 5 of NK302/304-306 LinkSwitch-TN Family document?
2. Or some additional thermal resistance to present between this area and the pins 'S' of the controller?
3. Or this heat-sinking area to get heated from other 'lossy' components like the output inductor?
4. What is your total efficiency in comparison to 75%. If it is lower you need to work out where the additional losses are. Switching shapes OK? Inductor saturation?

From the design checklist in the LNK302/304-306 document:
Thermal check – at maximum output power, minimum input voltage and maximum ambient temperature, verify that the LinkSwitch-TN SOURCE pin temperature is 100 °C or below.
Does this check pass?

Best Regards,

Submitted by Qawasmi on 01/17/2020

Hi,
Please see pcb layouts attached. For power part we just have 10 x 27 mm area as application requires this. 2nd revision of pcb has 4 layers, two of them are GND plane but result is same
Previously we made 3 different products with same components and we didn't have heat problem. But these boards were bigger sized (almost double).
For the heat-sinking area there is a drawing in datasheet but no dimensions were indicated. Our heatsink area is 9x11 mm.
As we used PIExpertOnline while making calculations and component selections; we would reach up to 350mA without problem but even at 200mA; temperature rises into 85-90 C.
I will check switching shapes but how to measure inductor saturation? In fact we even used a 1mH & 750mA inductor, result was same.
Regarding thermal check; we even have thermal problem at 25 C ambient temperature, 70% output power and max input voltage. So, it will not pass.

Attachment Size
top.JPG 116.03 KB
bottom.JPG 120.37 KB
Submitted by PI-Yoda on 01/18/2020

Hi Qawasmi

Please have a look at the picture attached.
Rough estimation shows that on the bottom side you have 40-45mm^2 copper area. Some chunk of it is covered by D4 which is already hot and area around D4 is not actually sinking heat.
So, more real estimation – bottom side is close to 35mm^2 actual heat dissipating copper area.
Let’s assume that on the top side is 10x10 = 100mm^2, but this heat-sinking part could be only used when the heat passes through single connection – C4 +pin. This bad thermal connection between both layers makes the effectiveness of the top layer cooling very low.
It is not far of to approximate your actual effectively cooling copper area as total of 100mm^2.
PXLS Designer shows that for 220mA load the power in the controller is 360mW.
1. Datasheet case (Page9/package D) - Copper 232mm^2 – gives Rth = 100DegC/W
0.36W*Rth = 61Deg C on the pin on the controller for 25Deg C ambient.
2. Solving back for you design case (80DegC on the pin) gives your actual thermal resistance as 152DegC/W.
In summary - even you circuit works ideally - as per the examples, the cooling copper area of the controller is by far too small from the one needed to guarantee 100DegC on the pin at min. mains and maximum power (Datasheet, p8, Quick design check list, point 4).
I will recommend to redesign the PCB to provide more cooling copper area. If this area is on two different layers to provide PROPER thermal paths between the two layers. Maybe using a lot of metalized vias.
Best Regards,

Attachment Size
PCBLINK_TN_heating_problem.png 248.95 KB
Submitted by Qawasmi on 01/19/2020

Hi,

Thank you very much for this very helpful information.

We will increase heatsink area and tie them better between top and bottom layers.

1) Our middle (2nd and 3rd) layers are GND planes. Do you recommend to convert them as heatsink plane? Thus we will have two full layers (PCB total size is 29x27mm) with this plane

2) We are using 12V/16A relays and 5V & 3.3V ldo regulators supplied from 12V (LNK)306. Today I tested to reduce LNK output into 5V and I have seen temperature is now at 43 C max. Do you recommend to us apply this as well? With a simple calculation; power consumption of LNK reduced more than 50%. Thus heating is reduced.
Disadvantage for us is 5V/16A relays are more expensive for bulk production but for making safer product this sounds like a better choice. At least we will be able to remove 5V LDO!
Do you think 5V will be more safe for us as if we think on temperature side?

3) After reducing from 12V to 5V; I can now hear vibration sound from inductors. I didn't understand why it changed with voltage? Does this look like a problem?

Best Regards.

Submitted by PI-Yoda on 01/19/2020

Hi Qawasmi,
1)
Sinking heat means providing a low therm.resistance path from a hot spot to a low temperature area. This low resistance path will cause the heat to flow (get sunk) from the hot spot to the low temperature point (ambient).
In your case of internal copper layers with big areas when connected thermally to the controller are good path for the heat. BUT they are not connected to the cool point - ambient . It means that the heat flow process will not exist because the sinking point is missing.
2)
The information I have about your project's specification is tiny and does not allow me to give advice.
For sure, in a tough thermal design it is worth every effort to optimize efficiency and power consumption. You may even be able to get rid of the expensive 4 layer PCB. Also reliability is directly connected to the components operational temperature.
3)
It looks like there is some interruption in the controller operation. It maybe starts and stops.
You need to look at the switching wave-forms, feedback signal etc.. to find the reason for this operation.
Best Regards,

Submitted by magicmicros on 03/11/2020

Hello PI Expert.
I've now been working with this for a while, for Mr. Qawasmi.
I've made a complete re-layout for more cooling, but still see about the same results.

Please refer to schematic below for the exact components used.

The area of the Source pad (cooling), is now 268 mm2, 100 mm2 on the top side and 168 mm2 on the bottom, with several vias.

Board layouts are attached, some with the Source pad highlighted.

Also, a chart of efficiency and temperature vs current output.
Input was 225V AC.
As you can see, efficiency is terrible, tops 47% at the most.
Temperature is a little better than before, but still way too high (which is, of course related to efficiency.

Are we missing something here?
I've tried several inductors, as listed on the schematic, which made very little difference.
Component selections should be good according to the DS.

Why is the efficiency so low? We would expect >70%, but get <47%
What can we do/try to improve it?

A second problem is that the Source pad needs a lot of copper to dissipate heat.
On the other hand, as it is the switching node, it needs to be as small as possible for EMI reasons.
I haven't measured noise, but there was enough EMI to upset my handheld scope that was lying 10 cm from the board, and make it flip around in the menus by itself.
Not a good sign.

I hope to hear from you soon, this is quite urgent.

Best regards,
/Jesper

Submitted by PI-Yoda on 03/11/2020

hI Jesper,
yes, the PCB looks much better, but 47% efficiency is ridiculously low. Could you please send me a screen shots of the following waveforms a. Voltage over the switch, b. Inductor current. Both in 1, 5 and 50 pulses time scale. Copy of the calculation PIXLS spreadsheet as well please.
Best Regards,

Submitted by magicmicros on 05/17/2020

Finally, here's the measurements as requested.
Switch voltage (CH1) over D5. 100V/div
Inductor current (CH2) over L2. (Current is inverted, sorry). 200mA/div
There are two screenshots for each timesetting, 0mA out and 200mA out.

Looking forward to hear your comments.

Attachment Size
400ns_0mA.jpeg 378.41 KB
400ns_200mA.jpeg 404.27 KB
10us_0mA.jpeg 397.7 KB
10us_200mA.jpeg 411.87 KB
100us_0mA.jpeg 401.45 KB
100us_200mA.jpeg 507.59 KB
Submitted by PI-Yoda on 05/19/2020

Hi, magicmicros

Thank you for the information supplied.
From the graph 400ns_200mA,jpeg is clear that the inductor gets into saturation after 1.2us from the beginning of the pulse. (The slope of the current di/dt changes dramatically).
Practically after 1.2us from the beginning of the pulse the inductance is not anymore in the circuit to limit the current rise because the inductor lost it’s the magnetics property due to saturation.
That is completely unacceptable operation and is most likely to be the biggest contributor to the low efficiency problem.
Please change the inductor to the one which can handle the peak required from PIExpert calculations +20% margin.

Best regards