LNK625 design not working properly

It is part of the usual design process to check and optimize components of the design.

Can you send me your PI Expert files (with transformer diagram), your actual circuit schematic and PCB layout (PDF is fine)?

-The Traveler

Thanks TT

Files attached - the PCB is a view from the bottom - little hard to see as Copper pour  has been reduced to outline.

Thanks C.

Hi TT

Is there anyway I can reply directly to an email as I need to move on to a plan B - the transformer was wound by one of your design cosultants and the design is directly from PI software. 

Contacted the transformer mnf. and they said it was wound directly according to the PI directions.

Changed the phasing of theFB winding - did not work - proves FB correctly phased on PCB.

Connected  6.3V to the  BP winding - made no difference.

Carl -

There are definitely some things in your PCB layout that are less than optimal but it sounds like the design is, at least partly, regulating.  

Here are some of the steps I go through when bringing up a supply for the first time and need to debug issues:

-With light load and an input voltage high enough to *just* start the device switching, look at the voltage across the output rectifiers (positive lead to the TXFMR side of the diode) with a 10x or 100x probe and verify that there isn't any excessive ringing when the diode turns off.  Changing RC snubber components may be necessary.

-Using a current probe and  100x scope probe, look at the drain current and voltage waveforms.  Look for excessive ringing in both waveforms.  Of extra importance is to ensure that any ringing in the drain voltage waveform has died out before 2us after the mosfet turns off.  Adjusting clamp components may be necessary.

-With those steps out of the way, you can start looking at adjusting the upper and lower feedback resisters.  If you put your design parameters into PIXLs, there is a section at the bottom for fine-tuning Rupper and Rlower.  As the 4V output is the highest output power, I would suggest using this one as the basis for the fine-tuning adjustments.

The high frequency ringing is important to look at in these types of designs.  With primary side regulation, the IC samples the feedback winding voltage ~2us after the MOSFET turns off.  If any ringing exists at the time of sampling, it can create an error in the output voltage set point.  This error will then off change as a function of input voltage and load due to changes in the amount of leakage energy being disipated in the clamp. 

A couple of other thoughts:

You PCB layout could use some improvement.   

-Rupper and Rlower need to be located much closer to the LinkSwitch IC.

-Your routing of the drain and B+ traces along with the clamp components might cause some EMI issues.

-Changing some of the pin allocations on your transformer might make the layout easier to route.

-Ground planes are useful but simply dropping a large copper pour into a PSU layout often won't work as you intend.

-Giving the FB winding a dedicated SRC connection back at the bulk cap (along with the bias supply if you choose to include one) will will help with surge and ESD immunity.

-Since your two output voltages are even multiples of each other, you might experiment with some novel transformer construction.  One idea might be to use some added parallel 4V windings (if there is room on the bobbin) and then AC stack those windings to get your 12V output.  

If you can, please follow up with some scope plots of your drain voltage and current waveforms, highline/lowline and lightload/fullload would be helpful.   

-The Traveler

Carl -

I also just noticed that you don't have any pre-loads on your 4V or 12V outputs.  You will need these in your final design. 

-The Traveler

Thanks TT

I have a 16V zener on the 12V line and forgot the 4V line in rushing.

There is some sort of regulation but not great - with a 5E load on the 3V, I get about 2.6V out. I did fiddle the resistor to get 4V out and regulation was about the same - 5E resistor got 3.6V out

I will follow your guidelines as above.

 Thx C.

The zener will help clamp the output voltage but you're going to get better results with a small pre-load resister on both the outputs.

-The Traveler

Hi TT

After scoping the diodes I discovered the 4V diode was not a Schottky so replaced it and got 4V out BUT regulation is extremely bad i.e. with a 10E load the output drops from 4V down to 2.9V.

I have taken a few measurements and attached the pictures.

Thx C.

Carl -

What kind of sampling rate are you using on your scope?  You need to be sampling > 20MS/s (preferable 40MS/s or more) to see the high frequency noise we're looking for.

For the diode waveform, we're interested in looking at the part of the waveform when diode turns off (the large falling edge). For this capture, you can take off the RC snubbers across your output diodes.  Power up the supply with a very low input voltage (just enough to get it switching) and only load the supply down lightly.

On the drain voltage waveform, it looks like you might have some bad ringing going on but I can't tell with the large time base you're using.  Change the scope to 1-2us/div and use as small of a volts/div that keeps the waveform on the scope screen. 

When you say a 10E load, do you mean a 10 ohm load?   

Here is a two output reference design using LinkSwitch-CV:

http://www.powerint.com/sites/default/files/PDFFiles/der213.pdf

I'm attaching this reference design because it highlights some good PCB layout practices for this type of design. 

-The Traveler

Hi TT

Sampling rate was 5uS on all except File1 which was 20uS on to show more of the waveform.

Attached are the new waveforms with the relevant data inserted.

Thx C

The drain waveform looks ok.

The output diode looks like it will need to have the RC snubber fine tuned.

Have you gone through and fine tuned the upper and lower feedback resisters to adjust the output voltage yet?

-The Traveler

Hi TT

I have the output sitting at the correct voltage (4V).

I have very poor regulation. 4V output - place a 10 ohm  load and I get 3V out.

Thx C

Hi TT

You asked me to remove the snubber from the diode and send you the waveform as well as the DRAIN waveform.

I did that  and your response was:-

"The output diode looks like it will need to have the RC snubber fine tuned. "

It does not have the snubber - it was removed as you asked. 

Thx C

Rupper and Rlower need to be fine tuned at full-load.  Fine tuning at light/no-load will give you regulation issues. 

-The Traveler

The purpose of looking at the output diode ringing without the snubber is so that you can optimize the snubber component values.  PI Expert will often get you pretty close for the snubber design but depending on the amplitude and length of ringing it is sometimes necessary to change these values.

The time length of the ringing does not appear to be so long as to cause regulation issues but it is something I would definitely fine tune before you finalize your design. 

-The Traveler

Hi TT

I thought the measurement was made 2uS after the switch off of the LNK625 in which case the ringing is reduced below the required amount.

To fine tune resistors at no-load or at full load will yield the same poor regulation.

Is there any way to exclude the transformer design from this poor regulation or at least prove its not the transformer ?

LinkSwitch-II samples the output voltage approximately 2us after the MOSFET turns off.  From the scope plots, it looks like the amplitude and duration of the ringing isn't  large or long enough to cause problems with regulation.  For non-regulation issues (such as EMI), you may want to go in and fine tune the snubber component values.

For regulation, you do need to fine tune Rupper and Rlower with your power supply at or near full load.  Adjusting Rupper and Rlower at no-load will cause a ton of problems, especially since you have no pre-loads in your design.  

-The Traveler