Ripple on TOP247YN power supply

The seen ripple is line frequency ripple,

The main problem seems the response of the controller is too slow, decrease R9 or increase C12, you can also try a series capacitor to C12..

additionally, 

You can increase the value and current rating of C2, or use Super Low ESR cap. for C9 and C10.. 

Madhuri - 

Can you please attach your PI Expert design files as well as your actual schematic (if different), along with your PCB layout and transformer documentation?  

Have you run any control loop tests on your power supply yet?

This problem is definitely something we can help with.  We'll just need a little bit more information.

-The Traveler

Many thanks for the support. I have attached the design file and a PCB layout for your reference. There is no change to the design, I used as such. I could also make the transformer successfully according to the requirements. As mentioned by Kemred in the previous comment, I tried all changes and the reduction of R9 to 1K from 4k7 reduced the ripple drastically. Is that alright? Another problem I realised that when I connected 1A resistor load the diode UF5408 failed within a minute. Does it need a heat sink?

Regards

Madhuri 

Sorry I attached wrong file. The design file now attached.

Regards,

Madhuri

I am not an expert in design, but when I read your problem since I see such a problem in one of my design I comment on probable solutions, and the comments I wrote in the previous message are my experiences to solve the problem, and they work fine,but PI Enginners must know the details as I said I am not expert,

For UF5408 is not its reverse voltage too high for the design? Is there any heatsink? did you measure the peak voltage of the diode? from my experience when the Vrrm increases forward voltage also increases due to the semiconductor design of the diode, which appears in high loss and temperature.. Did you measure the temperature of the diode? if it is not too hot before the explosion may be there is an error in another part.. If I were you I would choose more current capable low reverse voltage diode for this application. like 5-6A 400-500V diode if the voltage enough. since V forward will be less and this will heat less..

galaxyelect - 

Did you send the entire layout?  I looks like it might be a two layer board but I'm only seeing one of the layers.  Could you generate a PDF with each layer a different color but stacked on top of each other?

-The Traveler

The layout attached covers the entire PCB. The PCB has only one layer. I am attaching herewith another copy of the same design with a low file size. A provision for using MBR20200 diode is there on the PCB for high output current supply which is not used for the current design.

Regards,

Madhuri

Regards,

Madhuri - 

Looking at your design,  I suspect that the majority of your problems are PCB layout related.  Both your primary and secondary side layout and routing of components needs to be completely redone. 

On the secondary side, remove the continuous ground plane.  It is a misconception that a large continuous ground plane is better than actual traces.  What you want to do, is route the current through the components in a controlled manner.  The large ground plane allows high-frequency ripple currents to "go around" your filter components and is likely causing radiated EMI problems. 

On the primary side,  you need to move the controller closer to the transformer and re-layout the clamp circuitry.  Also, the control loop circuitry needs to have it's own SRC connection back to the input capacitor.  

I would *highly* recommend that you go through some of our TopSwitch RDK reference designs and use the PCB layouts in their reports as a starting point for your PCB layout.  We provide our reference designs specifically for this reason...we want you to be succesfful in building power supplies with our products.  We won't be even slightly offended if you leverage or use one of our existing designs.  I think with a much better PCB layout, the vast majority of your problems will go away.

-The Traveler

Many thanks for the tips. Attached please find a latest PCB design according to the datasheet. Please let me know your comments.

Regards,

Madhuri. 

Regards,

That layout looks much better!

My only suggestions would be:

-Move your Vpin resisters closer to the IC and use a wire jumper.  The node between the two resisters is a high-impedance node and susceptable to noise pickup.  You want to keep high-impedance traces/nodes short and small and make use of low-impedance paths when possible.

-I would reccomend flipping your clamp circuitry around so the B+ trace can come down on the PCB. This will help reduce your drain current loop area. Placing R6 horizontally, you could jump it over the B+ trace and locate your clamp circuitry in almost the exact same location.

-See if you can get C14, C13 and R16 closer to the IC.

-You might be able to tighten up the layout on the primary side if you rotate the IC 90deg CW.

-Adding a couple through holes in the drain trace can be helpful for debugging purposes but itsn't totally neccessary.

-Likewise, adding some easy spots to break the control loop so you can analyze the control loop gain/phase is also useful.

-Another thought for the clamp circuitry, you might add a few "dummy" components to the layout so you can experiment with different clamp circuitry components.  This can be really helpful on an initial prototype.

Overall though, this layout is MUCH better. 

Just out of curiousity, is D3 mounted on a HS?  I can't quite tell from the PCB layout.

Now while you're cutting the PCB, do you have the ability to measure your transformer parameters?

-The Traveler

Many thanks for your encouraging support. Please forgive me for the question but I did not understand the Vpin resistor. I feel those are the resistors connected to the "L" terminal of the TOP247. Is that right?

I have attached a brochure of heat sinks. Could you please let me know if I can reduce the heatsink size from PI51 to PI 49 or PI48? If I can use smaller size I can do better PCB layout.

What is B+ trace? Is that a bridge +ve?

I understood all other guidelines fully.

Yes, D3 has a heat sink. The diode was getting overheated on the first trial PCB. So I added a heat sink. Is that necessary?

Regards,

Madhuri.

My apologies.  Yes I should have said the L pin.  For this series of parts, it's called the Line-Sense pin (L pin for short) and on some other series of parts it's called the Voltage-Sense pin (V pin for short).

It's really difficult to make a recommandation on heatsinks for you.  Power levels, air flow, operating temperature ranges, dust conamination, etc all become factors.  What I would recommend is seeing you can find a series of heatsinks that encompass a handfull of different sizes and organize your layout so that you have the option of experimenting with different heatsink setups. You should be able to layout your components to give you this option.  Also, if you can maximize your SRC plane area, this will act as an additional heatsink.

The B+ trace refers to the bulk capacitor positive trace.  This also often gets referred to ask the HVDC+ trace.

If you're going to use an actual heatsink, I'd recommend switching over to a diode in a TO-220 package or something similar.  If you would like to use a through hold diode, i would recommend adding additional copper area around the diode but only on the side connected to the output capacitor.  This will help full heat out of the diode.  You can also add two diodes in parallel to help with the heat.

-The Traveler

Now I fully understand. Many thanks. 

The PI Experts says for 40V output the diode should have rating above 200V. The easily available diodes in TO220 package have voltage rating up to 200V only. That's why we have not used it.

Regards,

Madhuri

galaxyelect - 

In that case, I would suggest maximizing the copper area associated with the side of the diode connected to C9.  This will give you a larger heatsink created by the copper area of the PCB.  I would also suggest adding the option of a second output rectifer in your PCB layout.  This can help reduce the power loss associated with the output rectifiers and might also give you an increase in efficiency without the cost of using a heatsink.

I've attached a PDF of your layout with some sketches and notes on it.

-The Traveler

Beautiful! Many many thanks once again. I was almost through with the modifications you suggested in your last post. However the scanned image with your graphical comments made my all doubts very clear. I shall follow all what you mentioned. 

Regards,

Madhuri

/p

We are still struggling to make the power supply working properly. I am once again attaching the schematics and the PCB layout as suggested by you. The difficulties are as follows:

1. The Power supply only starts if we change the R9 to 1K and C12 to 0.1uF. With the suggested R of 4k7 and C of 0.047uF the power supply does not start.

2. The Output diodes UF5404 are used in parallel as you suggested still they heat and fail within 10 to 15 minutes of operation with a resistive load of 1A.

3. Y capacitor also heats. The temperature measure was above 95 deg.C. within ten minutes.

4. Transformer is made using easily available ferrites in India by Cosmo, the core is ETD29.

Please guide at the earliest.

With best regards,

Madhuri 

1 - Yes I agree that value of R9 seems way to high.  A value in the 1k-1.5k seems a bit more reasonable.  The value of R9 is going to set the overall DC loop gain of your feedback circuit.  With the 4k resister, the loop is probably not going to function as the resistance is too high to drive any current through the optocoupler.  I would also suggest adding a soft-finish capacitor across the TL431 to help minimize overshoot during startup.  Please make sure you verify your control loop design to make sure that it is stable under all operating conditions.

2- Have you done any transformer measurements yet?  The suggested values of R8 and C7 on the PI Expert schematic are based on a calculated value of leakage inductance based on the transformer layout used in PI Expert.  The primary leakage inductance, when reflected to the secondary side, rings with the diode junction capacitance when the output diodes turn off.  The values of R8 and C7 are chosen to snub the ringing of the output diodes.  If the transformer construction is different and the resulting leakage inductance value changes much, the values of the snubber components will also change.  Excessive ringing on the output diodes can easily cause them to fail.   I would check the waveforms on your output diodes with a very light load and load voltage input to see how bad the ringing is and redesign the snubbers/transformer if neccessary.

3 - I would suggest connecting the Y-cap directly to pin 1 of the transformer or directly to the bulk cap through it's own dedicated trace.  While technically the same node, where it is connected now it's probably seeing a decent amount of clamp noise.

4 - Did you use an off-the-shelf transformer or have a magnetics vendor build the design from PI Expert?  Either way, I would highly recommend measuring all important transformer values with an impedance analyzer to verify the design of the  transformer.

Regards,

The Traveler

I forgot to mention that most TL431's are rated for only about 35V.  R9 and your opto will dissipate some additional voltage but I would check and see what kind of voltage is present across the TL431 to make sure it won't fail during normal use or transient conditions.  A 30V zener across the TL431 might be a wise safety item to make sure the TL431 doesn't get damanged.

Regards,

The Traveler

Thanks for the important tips. My report back to you is as follows:

As you mentioned, TL431 was rated 30V so I replaced it by a proper voltage zener which worked well.

1. R9 was reduced to 1K when there was TL431 and now it is 750E for zener. I did not find any difficulty in the loop circuit.

2. The leakage inductance of the transformer used is little on higher side than suggested so I ensured that the clamp circuitry is sufficient to keep the drain voltage within the limits and I found it much below the max limit.

To save the output diodes I checked the ringing on the same and I found it. I then increased the capacitor value and decreased the resistor which removed the ringing almost. However that could not save the diodes. The supply is rated for 40V 1.8A. I am trying it for just 1A load still the two parallel diodes fail within 10 minutes of operation. The diodes are UF5408 and genuine.

3. The dedicated track to the Y capacitor helped keeping the temperature low. This worked very well.

4. We would the transformer according to the instructions in-house. We have LCR meter and nothing else to test the transformer.

Regards,

Madhuri

Simply increasing Csnub and decreasing Rsnub isn't a good design approach for snubbers.  Since you don't have a low-frequency impedance analyzer, you're going to have to do this the old fashioned way.

1:

-Remove the snubber from across your diodes (There should be a snubber on each diode.)

-With a VERY low input voltage and a VERY light load, measure the unsnubbed ringing frequency of the turn-off waveform.

2:

-Put a capacitor across the output rectifier that is a bit larger than the diode junction capacitance (look it up in your datasheet).  If the datasheet says 20-100pf, try a 1nF or so.

- With a very low input voltage and a very light load, measure the NEW ringing frequency at turn-off.

3:

-Repeat with a slightly high value capacitor added accros the diode.

4:

You secondary leakage inductance at the turn off ringing frequency is this:

Llkg = [1/(4*pi*pi(Cx1 - Cx2)]*[(1/fring1*fring1)-(1/fring2*fring2)]

fring1 = the ringing frequency with the smaller added capacitor Cx1

fring2 = the  ringing frequency with the smaller added capacitor Cx2

You actual diode junction capacitance is this:

C = 1/(4*pi*pi*funsnubbed*funsnubbed*Llkg)

where funsnubbed is the ringing frequency observed with the output diode just by itself.

5:

You snubber resister value will now be:

R = Z = 2*pi*funsnubbed*Llkg

where Z is the characteristic impedance of the ringing leakage inductance

Choose C such that:

C = 1/(2*pi*funsnubbed*R)

6:

These values will be a starting point but should get you really close to their final values.

Regards, 

The Traveler

Many thanks. Please suggest us an economical Impedance analyzer and also let us know the minimum requirements of the same. We shall try and buy one. Also please let us know the procedure to test the parts and to decide the snubber parts values using the impedance analyzer.

Regards,

Madhuri

Impedance analyzer will not help since capcitance of diode is nonlinear. I would start with 50 to 100 ohm series resistor and use perhaps 220pf capcitor. I would not go beyone 470pf as that will be too dissipative. You can also try high perm ferite beads in series with winding to damp spike.  There are also amorphous beads that are more expensive

but work quite well

http://www.toshiba.com/taec/Catalog/Line.do?lineid=6004&familyid=12