First off, thank you for your interest in our products and for using them in your new design!

For the higher than expected power dissipation you're seeing, how did you verify this?  What was your measurement method?

Can you attach your actual PI Expert Design file (should have a  UDS file extension).

-The Traveler

Hi and thanks for your reply. We've been using your products for years (LNK-TNY) but only now we embarked on designs >20W, hence problems with optimization. I have uploaded the uds file for your attention.

With regards to dissipation, TOP246YN runs up to Δ60°C in the open air with 15K/W heatsink. System runs in continuous mode, with Duty 0.27 at full load. Both primary and secondary currents seem as expected, apart from nasty transients on the primary current before turn on. These transients extend up to 1uS and reach peaks of ±9A. Leakage inductance measured at 0.1mH on primary and 0.01mH on secondary. 

Measured we have Ip(peak)=1.2A ton=2.5uS which rumps up from almost 0 as secondary current decays linearly from 9.4A peak over 5uS. 

Here we have primary 27T+27T and 5T 4 filar secondary. Could you please advice as to how do equations 2-3-5-6 in your AN16 change in this case? Or in other words how are Lp and Ls related and what is the turns ratio?  Measured we have Vor+Vin=478V where Vin=300V (AN16 figure 8). As we have stable 12VDC on the secondary, this would indicate turn ratio of 14.8? Or would this indicate that some secondary turns have not been collapsed properly?

Also would you suggest we try fast rectifier such as MURB820-1 instead of 8TQ100? Do you think this schottky could be too slow for our purpose?

Your help is much appreciated.

Best regards

Rallie

Hi

in the mean time we tried UG4D ultra fast recovery diode and there was no improvement.

Regards

Rallie

Can you get me a waveform capture of your MOSFET current and voltage waveforms?   The 0.1mH leakage inductance you're giving is REALLY high. 

Did you build the transformer in-house? 

How did you measure the leakage inductance?

I'm not seeing anything that is too suspect in your PI Expert design.  Could you attach a PDF of your PCB layout?  Sometimes there can be some quirks in a PCB layout that can cause some nasty problems in power supply designs.

-The Traveler

Hi the leakage inductance looks huge indeed. The measurement method: with shorted secondaries we measured primary inductance with 100kHz 150uA sampling signal. We have built one transformer in house and also sample quantity at the authorised winding contractor. All units yield similar leakage inductance. Could you please check our transformer specification? We made some changes to the original PIExper design in order to simplify manufacturability. Also, i have attached the PCB layout, maybe you can see something wrong there. I am also preparing drain current and voltage waveforms. 

Thanks again.

Also do you know when will the "Techniques for Estimating Transformer Parasitics" course become avaiable? Thanks.

Rallie - 

I'm not seeing anything in the PCB layout that looks like a deal breaker.  I'm curious to see your drain voltage and current waveforms though.  When you grab your drain current waveforms, make sure to insert your current probe loop between the drain pin of the IC and D6. Placing it between D6 and the transformer will add in the clamp circuitry current which we don't want to look at.  If you could also get a waveform of the output rectifer voltage at the time when the IC MOSFET turns on, that would be helpful as well. This will let us see if the output rectifier snubber components have been chosen correctly.

For the transformer leakage measurments, I'm really surprised you're getting a leakage figure of around 100uH.  Is this measurement with the bias winding shorted as well?  I'm also surprised your magnetics vendor supplied you with your prototype and didn't say anything about the leakage inductance being well over the 20uH figured spec'd in your document.  If it was me, I would double check the transformer measurements for both inductance and leakage inductance seen from both the primary and secondary windings.  With the secondary side measurements, we can calculate the output rectifier snubber component values pretty easily.

I'm not sure when the transformer parasitics information will be available but when I see the guy who works on these projects I'll ask and see what he says. 

-The Traveler

Hi

we checked the leakage inductance on the primary with shorted both the secondary and bias windings, and the value is slightly better = 50uH.

We looked closely into primary current waveforms and with Lp=0.7mH the primary DC current component was almost non-existent. We then calculated that for peak primary current of 3x IpDC we needed Lp=1.4mH. Hence I enclose measurements for Lp=1.4-1.2-1.0-0.9-0.7mH. The lowest dissipation on TOP246YN we achieve with Lp=0.9mH of Δ45°C but with 17K/W heatsink, which yields 2.6W. This is still twice as much as per PIExpert calculation. Do you have any suggestion as to how we could reduce this dissipation? Same problem we have with 8TQ100 where temp rise is  Δ65°C with 17K/W heatsing, i.e. 3.8W, but PIExpert suggests 2.3W dissipation?

I would have to get rid of at least 2W for this design.  Please take a look at th waveforms, any suggestions are welcome. Thanks.

Comments regading primary current versus Lp:

• * with Lp 1.4-1.2 we are getting saturation issues.
• * all primary RMS currents for Lp 1.0-0.75 are ranging between 550mA and 250mA depending on the Lp. With Rdson of <2Ω, TOP246YN should not be dissipating more then 1W, unless the FET has not opened fully and Rdson is much larger?
• *************************************************************************************************************************************
• Where do we go from here? Please let me know if you need any other measurements. Thanks very much.
  

Rallie - 

A couple things:

 - You won't have any primary DC components.  You're probably thinking there is because your design is running continuous as opposed to continuous.  This isn't the same as having a DC primary current component.

 - If your transformer design is calculated for 630uH...why are you regapping it for 1.2-1.4mH?  This is going to significantly (probably 2x-3x) raise your peak flux density in the core by a huge amount leading to saturation issues.  Add some high ambient temperatures and you're to start blowing up power supplies left and right.

- For the waveforms with lower primary inductances, they look pretty good.

- When it comes to calculating the losses in your IC, you can't simple take your primary RMS currents and then multiply it by the RDS(ON) value of the MOSFET.  This isn't how MOSFET switching losses are calculated.  Conduction losses are *sort of* calculated this way...but it's a bit more complicated than you think.  If you want to get a better idea of your MOSFET losses, you will want to take a scope capture of your drain voltage and current waveforms and multiply them in a math function and then integrate it over multiple switching cycles.  If you're using a hall-effect current probe you need to take into account the time delay that your probe is introducing and time shift the current waveform to account for this.  The time delay for the probe can change the power calculations a decent amount.  

Some things you can look into to track down power losses in your design: 

- Get a thermal camera or a thermocouple datalogger and track component temperatures to give you an idea of where the mojority of your power losses are coming from.  

 - For excessive IC losses, you can move up to a larger device that has a lower RDS(on) value.

 - For excessive clamp circuitry losses you can look into raising the clamp voltage level (provided you're leaving enough safety margin on the MOSFET BVDSS), change the clamp components, try a different clamp style, etc.

 - For excessive output rectifier losses, you can try a output diode with a lower forward voltage like a Shottkey, make sure your output diode snubber circuitry is setup correctly, parallel two output diodes (will lower the power dissipation for each diode by reducing the forward voltage), etc.

- For excessive transformer losses you have a few options.  This is an easily overlooked area of power supply design.  If you're pushing the core really hard (as you are with regapping it for 1.2-1.4mH) your core losses will rise substantially.  In these situations, it's advantageous to move to a larger core size.  If you have an impedance analyzer, looking at the real component of each winding's impedance (|Z|) can give you a good idea of the AC losses in your transformer design.  Reconfiguring your transformer design is worth the time in these situations.

-The Traveler

Rallie - 

• * with Lp 1.4-1.2 we are getting saturation issues.

Of course you are.  Your transformer was design to gapped for 630uH.  Raising the inductance will increase your peak flux density in the core causing saturation.

• * all primary RMS currents for Lp 1.0-0.75 are ranging between 550mA and 250mA depending on the Lp. With Rdson of <2Ω, TOP246YN should not be dissipating more then 1W, unless the FET has not opened fully and Rdson is much larger?
• You can't calculate your IC losses this way.  Please see my response to your previous forum comment.
• *************************************************************************************************************************************
• Where do we go from here? Please let me know if you need any other measurements. Thanks very much.
• I think you need to do a bit more troubleshooting and investigating.  Go back to using the inductance calculated in your initial design and don't saturate your transformer. 
  

-The Traveler

Hi again sorry for not replying earlier, I was looking into this dissipation issue and even though I did not make any progress with my main aim of reducing the dissipation, i've learned a lot about PIExpert. In any case, we have moved to TOP249 then optomized the project with this device. Included two output diodes, experimeted with different types of rectifiers. We also looked into the clamp design and tried to optimise this element of the circuit, but as our primary leakage is several times higher then suggested, there was nothing to be done, while keeping the save BVDSS margin.. Notning really produced goods so we have not decided to migrate to EE28 core, optimise the design for efficiency and keep TOP249 and look a winding house that can meet the design parameters. I will inform you of the progress as soon as i have EE28 proto up and running. Thanks for your help.

If you have the ability to wind transformers in-house, I would highly recommend taking the same basic transformer design and experiment with reconfiguring the windings.  When it comes to transformer winding layout, there is often a fair amount of trade-off between leakage inductance and pri/sec winding capacitance.  A tighter coupling between your pri and sec windings will give you a lower leakage inductance but will increase your pri/sec capacitance (in general).  Optimizing a design for low pri/sec capacitance will often lead to higher leakage inductances.  High leakage inductances place more of a burden on your clamp circuitry and will make you take a ding in efficiency due to the extra losses.  Increased pri/sec capacitance will tend to cause problems with conducted EMI (radiated as well).

If you want some ideas on winding configurations to experiment with, let me know.

Regards,

The Traveler

Hi and thanks for your prompt reply. We have tried to experiment with the winding topology. From the PI expert all windings are run clockwise and, in general, from left to right. We tried this approach but with 4 filar secondary, this approach, we think, leads to half turns. Hence we tried winding anti-clockwise right-to-left, in order to fully colapse all windings. This is how the final prototype was made with minor improvements. If you have further suggestions, we would love to try them out. Thanks very much.

 Reorganizing the left-to-right and right-to-left winding direction probably won't make too much difference with leakage inducatance.  It might be able to help with stray capacitances though and help EMI.

For leakage inductance, you want to focus on topologies that will help you get better coupling between your pri and sec windings. 

Options you can experiment with:

-Interleaved pri and sec (split primary)

-Multiple levels of interleaving

-Move to a larger core and reduce your secondary windings.  This would let you experiment with using a foil for your secondary.

-Experiment with using single layer windings

Sometimes it can be helpful to just come up with a handful of different winding arrangements and measure their device parameters to see how they perform.

If you have a low-frequency impedance analyzer, this can be really helpful in analyzing your different transformer designs.

If you try experimenting with different cores/bobbins, you might give an EER28 a try.  You should be able to get down to 3 or 4 secondary turns with this core (would let you experiment with a foil secondary).  The physical size of the transformer pins isn't that different than an EF25 so you could probably test it in your existing PCB with only a little rework.

Regards,

The Traveler