Output Voltage dropping and supply not delivering full output power with TOP259EN

26 posts / 0 new

I have designed a 12.2V, 9.4A fly-back power supply using the PI Expert Suite.
I have attached the design file and circuit diagram below.
Device: TOP259EN
Transformer: ETD34/17/11

1. Output voltage drops to 7.9-8V when drawing more than 3A
2. It is unable to supply more than 4.4A of current

1. No load voltage is 12.2V as designed
2. Output voltage drops to 11.8V under 1A load
3. Output voltage drops to 11V under 2A load
4. Heat-sink attached to TOP259EN is warm to touch
5. Heat-sink attached to Output Rectifier is very hot

Both heat-sinks are of the same size: 20mmX60mm of 2mm thick Aluminum strip
The transformer was not made by us but one sample was opened and verified to be matching with the specifications provided by the suite, with one exception. The terminations for the foil winding were done using a single wire as opposed to 2 wires suggested by the designer.
Output capacitor bank is made up of 4X 1000uF Low ESR capacitors instead of 4X 47uF as suggested by the designer.

What would be the best method to use the remote ON/OFF function? Currently we have connected an opto-coupler output between the X pin and High Voltage Negative and controlling it via a daughter board. The supply was also tested by connecting the X pin to High Voltage Negative to rule out malfunction because of remote ON/OFF, but it still functioned as described above.

Design File2.48 MB
Circuit Diagram438.69 KB

First off, the input bulk capacitor (C2), is too small for the input voltage range and power level. This capacitor should be sized somewhere between 2-3 microfarads per watt for a universal input supply, so, at the very minimum, 220uF, and 330uF on the generous side. Running with 3 uF/w (~330uF) will result in a higher average bus voltage and better efficiency, important for operation at low line. The capacitor size shown in the attached schematic (68 uF) would be more appropriate for high line (180-264 VAC) operation.
If you look closely in the attached design file in the input capacitor section a capacitor value of at least 270uF is suggested.

The supply is indeed targetted for 180-264VAC operation. The space is limited so we had to forgo having a larger capacitor.
What would be the appropriate uF/W for high line like you stated for Universal?

Also, on further testing it was found out that the output inductor was preventing the supply from operating properly. Another issue was with heat dissipation. With the present heatsink we are able to obtain 10 minutes of continuous operation at 5A, 12.11V and then it goes out of regulation to 8.5V. We will be affixing a longer heatsink.

Is there any tool or application note to calculate the dimensions of flat panel aluminium heatsink like the ones used in your reference designs?
The finned heatsink suggested by PI Expert is 60mm in height, so it wouldn't fit in our enclosure of 25mm height.
We were inclined to using the flat panel/strip of aluminium which is both cost effective and easy to make on custom order.

Around 1uF/W should be appropriate for high line operation, so 100-120uF in your case.. Again, sizing the capacitor on the generous side will bring up the average bus voltage and result in more efficient operation. You will probably need to lay down the capacitor in order to meet your 25 mm height requirement.
As for the output inductor, for high current outputs I generally use a second stage inductive filter consisting of a small powdered iron toroid (~T-50 size from the Micrometals catalog) with 18 AWG winding to make a low loss second stage filter.

I will run a PIXLS spreadsheet to take a look at the transformer design - the number of turns are suspiciously low for a transformer with that output power on an ETD34 core. I got high flux warnings when I tried a 2-turn secondary while checking the design previously. If you are truly designing for a high-line only input voltage, the optimum primary inductance will be different.

You might want to consider an InnoSwitch-EP design rather than TOPSwitch, as then you have the advantage of a built-in synchronous rectifer controller, lower output loss, and the distinct possibility of eliminating the secondary diode heat sink, and maybe using a minimal heat sink on the primary side. Eliminating large heat sinks (and the TL431 and optocoupler) may more than make up for the price difference between the InnoSwitch and TopSwitch designs.

Here is a PIXLS spreadsheet for your design configured for high line operation. It looks like an ETD34 core is possible for the design. You can use the parameters in the spreadsheet as a guide for the next design iteration.

I will try to get a 100uF or 120uF capacitor of appropriate size then.
But, is the supply unable to power the load for over 10 minutes because of the input bulk capacitor?
Thanks a lot for that spreadsheet. It seems to be similar to the design I am currently testing.
The uds file I uploaded was edited afterwards to check if the core can be used to obtain even more current.

Isn't Innoswitch3-EP for low power designs? Only INN3670C seems to be suitable for our requirement that too at peak power. By the way which family would you suggest for 100W+ designs

We are a bit short on time and would like to get TOP259EN working as designed. Will we able to get away with the current design if we increase the heatsink dimensions? We will be fitting a 120mmX24mm of 2.0mm thick Aluminum strip.

The suite suggests 1130ohms as the loop gain resistor but I have been using 1000ohms. Is it causing the supply to lose regulation at higher currents?

I would strongly suggest at this time to use a current probe to monitor the TOPSwitch drain current in order to detect whether the TOPSwitch is hitting its peak drain current, or if the transformer is saturating. Both of these conditions would cause the output voltage to droop or worse, for the supply to go into autorestart mode. A minor change in the loop gain resistors would not cause the issues you are seeing. It is also a good idea to monitor the input DC voltage at the bulk capacitor to see the voltage droop of the B+ at higher power levels.
I also suspect that your output rectifier is thermally running away due to inadequate heat sinking. It may be necessary to use two output windings and two output diodes to spread out the secondary power dissipation. I would strongly recommend using Schottky diodes on the secondary side rather than ultrafast to reduce the switching losses. These should be oversized, as the secondary peak current will be on the order of at least 20A peak. Also using a larger size TOPSwitch and dialing down the current limit will help reduce dissipation on the primary side. This may be more cost effective than s small TOPSwitch and a huge heat sink. Here is a link to a web page with basic information on heat sinks. It shows the heat sinking capability of a basic sheet metal heat sink. The best heat sink for your design might be an extrusion with the fins pointed up, and an l-bracket on the reverse side to connect to the devices in question., as it may not be possible to get sufficient area with a sheet metal heat sink.


Suppose the transformer is saturating. Does that point to the cores being not to spec? Will gaping the same cores have any effect?
Due to the peculiar design and height limitation, we are stuck with sheet metal heat sinks. A new design with double the length/volume is on the way. Will test and get back to you.

We are already using schottky at the output stage: MBR20100CT, 20A(Dual 10A), 100V with the same heatsink as on PI device. The PI device was also upgraded to TOP259 from TOP258(according to design) because of the reasons you mentioned. Would we need even higher device?

We have also designed a buck convertor using LNK3206 on the same high voltage bus to power a 5V micro-controller. Can we use a transistor between X pin and High Voltage Negative to act as remote ON/OFF switch? Presently we are using an optocopupler with on state resistance of around 600ohms.

I hope you are running your transformer with an appropriate gap, as it will not be able to store enough energy to process the required power without a gap. The inductance in the spreadsheet for the specified number of turns is achieved with a gap. This gap, as well as the number of primary turns, is adjusted in the spreadsheet to achieve the proper operating characteristics without saturating at worst-case peak current.

Having said that, ferrite materials are rather optimistically specified. The saturation flux density shown in the ferrite data sheets is obtained with the material in deep saturation. However, power supply designers should be far more interested in the flux density at the knee of the B-H curve where the permeability starts to rapidly decline as a function of primary current. This is not specified, but in our spreadsheets we generally set a warning at a flux density of 3900-4000 Gauss. This number will be highest for room temperature and declines as the core temperature rises. A very hot transformer will have a lower saturation flux density than at room temperature.. A cheap ferrite may have a lower saturation flux density, both at the knee and at deep saturation

Depending on the amount of heat sink you can use, you do have the option of using an even larger size TOPSwitch to reduce the conduction losses, making sure to dial down the peak current limit via the X pin to avoid saturating your transformer.

The TOP-HX datasheet specifies using an optocoupler/transistor in series with the X pin and current limit setting resistor for on-off control, so you should be OK with that approach.. A well-saturated transistor should work just as well, and a small signal mosfet like the 2N7000 or 2N7002 would be even better.

There is a GaN-based InnoSwitch that would be appropriate for your power level. It has a current limit sufficiently high so that you can run your flyback in discontinuous mode, allowing the supply to run in quasi-resonant mode, greatly reducing primary switching losses and offering the possibility of using no or a greatly reduced primary heat sink. The integrated synchronous rectifier controller gives you the option of secondary synchronous rectification, again greatly reducing or even eliminating the secondary heat sink. I am including a spreadsheet using that device for your reference. A reference design is also available (DER-751). It currently uses a PFC input, but could be adapted for high line operation. This may be an option if you run into trouble with the TOPSwitch-based design.

I received the specification sheet from the manufacturer and it is confirmed that the core is gaped and the primary inductance is matching with the spec provided by the design suite.
We were running the circuit with 2X 47kohms MOF(older design had this) as opposed to 2X 36Kohm specified by the design file uploaded. The clamp diode, ES1J(600V, 1A, 35ns) was failing after 30 minutes of operation. We even used ES2J(600V, 2A, 35ns) to rule out failure due to current and it too failed after 30 minutes. On putting 3X 56Kohms this issue was resolved and now the circuit shuts down due to thermal overload after 2.5-3 hours after continuously operating at 12.2V, 5.5-6A output.
The input capacitor was also doubled to 120uF.
The designer suggested TOP258 and we are already using TOP259 even though maximum output of TOP258(141W UNIV) is well above the 70Watts we are getting right now.

Which TOPSwitch-HX device would you suggest be used?
Thanks for confirming the remote ON/OFF method. We will go ahead with optocoupler based design since the daughter board is at a different location.

Does the overheating problem lie elsewhere? Like the clamp for example?

Where can I get the details for SC1957C used in the spreadsheet you provided. I am unable to find them for purchase also.

Another approach that might be fruitful for your application would be to use a forward converter instead of a flyback. This would result in much lower peak currents in both the primary and the secondary, reducing heat sinking requirements. Stress on the output filter capacitors would also be greatly reduced,. and the transformer is also much smaller.
Attached is a photo of one of our eval boards, the DER-368. This is a 150W, 12V 2-switch forward DC-DC converter using the TFS-2 series of devices from PI, running at 132 kHz. We use a proprietary scheme to allow duty cycle of >50% for a 2-switch forward converter. The TFS-2 also incorporates a standby controller w/mosfet that could be used to generate your supervisory supply. The eval board is currently configured to accept a 380V DC supply, but it could be modified for use with rectified and filtered 180-264VAC. This would probably require the next size larger TFS device to reduce heat loading. The eval circuit currently uses schottky rectifiers rectifiers at the output, but the forward rectifier could be replaced with a self-driven synchronous rectifier for greater efficiency and less power dissipation. This is described in the report for eval board DER-428.

IMG_2919.JPG456.05 KB

Well that is a great solution.
But unfortunately due to time, money and material constraints we are stuck with TOPSwitch for this 100W design.
We are considering changing the entire layout to accommodate the supply inside a metal enclosure so that heat-sinking is improved.

My question is what can I do to reduce the power dissipation at the clamp diode? The PCB is turning brown around the Ultra-Fast ES1J diode possibly because of its high forward voltage(1.7V).
What can I do to mitigate this?
Will increasing capacitance(Currently:3.3nF) in RCD-Z clamp do any good?
Will increasing current rating of diode help? If yes, then should I go for ES2J or ES3J?
Will reducing Resistance of RCD clamp help?
Repeating my previous question: Where can I get the details for SC1957C used in the spreadsheet you provided? I am unable to find them for purchase also.
We will use this in future designs.

It would be a good idea to post your circuit (at least the high power portion) so I can see what is going on. It would also be a good idea to add a damping resistor in series with the primary snubber diode to help reduce the stress on the snubber diode, which may be forced to recover several times in succession as the leakage inductance discharges and the snubber circuit rings, causing excess power dissipation in the snubber diode. I would try a starting value of ~10 ohms to start for the damping resistor. It may be also necessary to go to a 3A snubber diode at your output power level.
For a high power application such as yours it is necessary to minimize the loop area in both your primary ans secondary circuits, as a loose secondary layout acts the same as high transformer leakage and will result in more stress on your snubber. It is also important to minimize the transformer leakage inductance for the same reason. A picture of the Topswitch drain voltage and current would help provide guidance.

The SC1957C is a custom part, but available to the general public. Please contact your local PI representative/office for details.

I have attached the circuit diagram and the board layout.
Please suggest any other changes too if necessary.
We are already using a 5.1ohms 0.25W chip resistor as damping resistor.
As suggested we will change the snubber diode to ES3J(600V, 1.7V, 3A). On a side note, is 600V diode sufficient? We were unable to source the 1000V Ultra Fast diodes like US series, locally.
We will be modifying it to fit in an all metal enclosure while keeping the rest of the circuit similar.
I will post the readings later.

Regarding SC1957C, which InnoSwitch3-EP part is it equivalent to? We would like to make an adapter for another prject and it would be better if we can evaluate the design on the PI suite and then order the part since the local PI office don't provide samples for these parts.
We are using an auxiliary supply using LNK3206 for generating 5V buck supply. They had suggested SC1099 to be used in place of LNK3206. But I couldn't find any information regarding Buck configuration in it's datasheet. Can you suggest which IC can be used in place of LNK3206?

Circuit_Diagram.pdf438.69 KB
Front.jpg284.47 KB
Back.jpg173.57 KB

Both the primary and secondary sections of the board have large loop areas, adding to the effective leakage inductance. If you cannot move the primary bulk capacitor closer to the snubber and TOPSwitch, I suggest adding a 100nF capacitor as a bypass to provide a shorter path for high frequency currents. The layout on the secondary side needs to be considerably tightened up to reduce the loop area and stray inductance, which adds to the effective leakage inductance that must be handled by the primary snubber. One way of doing this would be to use organic polymer capacitors instead of standard electrolytic capacitors. They have higher ripple current ratings and lower ESR for a smaller size, allowing a more compact layout. A tight secondary layout is always important for flyback converters, but even more so for higher power designs that experience higher dI/dt on the secondary side.
The schematic still shows universal input and only a 68 uF bulk capacitor.

We observed a strange behavior where in the output voltage drops to 9V, while drawing 5-6Amps when the heat-sinks are cold. While it warms up, the voltage gradually climbs to 12.2V(as designed) under 5 minutes. After running for 30 minutes, the supply was again run at similar load after 1-5 minutes of rest and the output was always 12.2V at the same load. If given enough time to cool down, it again behaves similar to cold start.

We haven't been able to do much to reduce primary loop areas due to tight dimensions and board layout constrictions.
Regarding secondary section, is there any other way to reduce the loop area with 4 capacitors at the output side. Will removing solder resist from the tracks help in some manner?
I have uploaded new PCB with ES3J, removal of solder resist at some points(with the aim to reduce resistance and act as heat sinks) and an even larger heat sink. Please suggest any changes. Is the 100nF Capacitor (C19) to be added parallel to bulk capacitor C2?
On reading DER-218(https://ac-dc.power.com/sites/default/files/PDFFiles/der218.pdf) which was somewhat similar to our requirements, we observed that the flat panel heat sink used was much smaller than the one we are using now and the device used is also closer to our TOP259EN. The clamp diode, MUR160 is also similar in spec to ES1J. I wonder how they got away with smaller diode and heat sink.
120uF capacitor is being used now.

On a different note, is it advisable to connect GND(DC) and EARTH together? If yes, can it it be done via chassis instead of connecting them on the PCB. We will be placing thermal pads between the heat-sinks and the device/schottky in case they need to be connected to the chasis.

Front.jpg284.47 KB
Back.jpg190.35 KB

On loading the supply for 8A, the supply begins at 11.8V and then loses regulation after 5 seconds, falling to 8V and 4.5A. It continues to run like that and would occasionally jump back to 11.8V, 8A falling back again after 5 seconds or so. If I stop and restart the supply, it will again do the same thing. It is behaving the same regardless of the device/heat-sink temperature.
Is this also linked to previous problems?

I don't understand why you are giving the low power supervisory supply a preferred location next to the bulk capacitor while the main supply is far away - the positions should be reversed.. The standby supply could be located at a considerable distance from the bulk capacitor with a 10-20 nF bypass capacitor located next to it.
Do you have a current probe to monitor the drain current? You should have that in place to look at the drain current during high power loading. Something very obvious is going wrong, and a look at the drain current would tell you a lot, including whether the transformer is saturating, and whether the supply is oscillating violently. Posting some screen shots of your drain current would be very helpful.

Something very obvious is going wrong indeed. I made an entirely new board and its behaving even worse. Swapped out TOP259EN from two different vendors. Used ES3J and ES2J for clamp diodes and it just won't work. Now the supply would drop to 8V at just 1.2A.
Unfortunately we only have 300Vpp probes for voltage measurements and no current probes. We have been trying to get hold of test equipment and will post as soon as we get them. Will it be any good if I use a divider circuit using 10M resistors so that I can measure using 300Vpp probes?

I am doing a redesign of PCB and change transformer pin outs to minimize primary loop area.

I have noticed that most reference designs use US1, RS1, BAV series diodes for bias winding rectifier and not 1N914/1N4148 as suggested by the PI Expert Suite. Is there any specific reason? 1N4148 diodes are much cheaper and faster too.

Adding a divider network to your low voltage probes will most likely mess up the frequency response, as the divider will not be compensated to properly work with the probe. 100X probes are relatively inexpensive from Chinese vendors or manufacturers like Probe Master.
For the types of problems you are experiencing, a current probe is essential to determine if your transformer is saturating, among other things. Trying to suss out an SMPS design without one (especially a high power flyback) is like working with a blindfold on. It may be possible to rent a DC current probe from a local vendor.

Okay, understandable.

The problem was in fact where I assumed in post #19.
The supply is now operating as designed after I replaced both 1N4148 with ES1J. Now I don't understand why it was not functioning with 1N4148. Please do look into this since the PI Expert design suite mostly recommends 1N914 by default. Significant time was spent trying to hunt down the problem.

I have a 180W flyback supply running on my bench using a BAV21 (similar in current rating to the 1n914/4148) for the bias supply rectifier, with no issues. I suspect something is still wrong in the circuit, perhaps the transformer construction. What does the voltage waveform look like at the anode of your bias rectifier? You should be able to capture this waveform using a normal 10X oscilloscope probe.

We have updated the design and tested the prototype.
I mirrored the transformer pin-outs and reorganized the components to reduce the distance.
The supply is now operating at 8A, 12.02V.
Yet to test at 9A after we receive the case which doubles up as the heat sink.
I had a few questions regarding the operation:
1. Why was it operating earlier at 8V if the voltage reference circuit was designed to operate at 12.2V?
2. How can we size the clamp circuit appropriately according to our design? We have ordered for 1.5KV Vpp probes. Will that be enough for it?
Please have a look at the new loop area and suggest improvements if any.

Front.jpg228.44 KB
Back.jpg183.91 KB

The new layout shown looks precisely the same as the original shown in post #14, at least from the top side view.

I still suspect that the polarity is wrong on the bias winding. The bias rectifier anode should go negative during the TOPSwitch "on" time. If the polarity is wrong, the bias rectifier will experience high peak current when the TOPSwitch turns on, as it will be trying to directly charge the bias filter cap. This can cause the bias diode to thermally run away. Sometimes this will also cause the supply to prematurely shut down.You should be able to see the bias winding polarity with an oscilloscope, even with the probes you have. Also, a tell-tale sign that the bias winding polarity is wrong is that the bias voltage will track the input voltage rather than remaining relatively constant. This should be examined at a relatively low and constant output current, as the bias voltage can also vary with output current.

The clamp design is complicated by a poor layout, as the excessive stray inductance can cause the clamp diode to recover several times in rapid succession when the clamp engages, causing excessive dissipation in the clamp diode. Improving the primary snubber layout and adding a series damping resistor to the clamp diode will both help address this issue. A 3A rated clamp diode also wouldn't hurt.

I have attached the uds file for the new design
I have also attached the PCB comparison highlighting the change in loop area and Transformer Pin-out comparison of both old and new design. Please give your opinion.
The supply is operating as intended but at higher loads the clamping resistor R5(series damping resistor) fails open, shutting down the supply.
I am using a carbon film 0.250W 5.1ohm SMD resistor as specified by the PIExpert suite.

It works fine at a load of 4-5Amps at 12.2V over extended periods of time. Anything above 6Amps and the resistor fails after a few minutes. The voltage regulation is well within 12.15 to 12.3V at all loads. The clamp diode no longer fails and I have tested with both ES2J and ES3J.

The bias winding is connected as given in PIExpert suite. Pin1 to Source and Pin2 to Rectifier(D2) in new design.

The TVS diode has been marked on Comparison1.png and it was getting really hot during operation. But since it has been placed away from clamp diode in new design the clamp diode is comparatively cooler.

New Design2.47 MB
PCB Layout578.44 KB
Transformer pinout189.8 KB

The primary layout is improved, but the secondary layout is still fairly loose, looking the same as your previous layout. Again, I suggest using aluminum polymer capacitors in you secondary filter, as they can be more compact than conventional electrolytics, with enhanced ripple current rating, and will result in a more compact secondary layout.

Stray inductance on the secondary side acts just like excessive leakage inductance, putting an extra burden on the primary clamp, and also causing spikes and ringing on the secondary rectifier that may give rise to HF EMI problems.
Regarding the damping resistor for the primary clamp, with a supply of this power level, I use two 1206 resistors (in series or parallel) for extra power dissipation capability. If you now have your 100X probes, you can monitor the drain voltage and optimize the value of damping resistor to reduce ringing on the drain waveform after the leakage inductance discharges. The goal is to keep the clamp diode from repeatedly recovering each cycle, which will result in extra power dissipation in the clamp diode. You may be using more damping resistance than is optimal.

Reducing the value of your primary clamp resistor will help reduce the power dissipation in your clamp TVS, though the resistor will get hotter as a result. The other factor that can cause the clamp TVS to overheat is using a value that's too close to your transformer's reflected voltage. What value of VOR did you use in your design? It may be possible to increase the TVS clamping voltage to reduce heating, while still protecting the primary switch against excessive voltage.