3
\$\begingroup\$

I built a flyback SMPS intended to produce ~800 V (variable) from a 12 V input (basic schematic below). I'm using a TL5001 PWM controller with an inverting gate driver, to drive a SiC MOSFET (GeneSiC 750 V). The PWM controller and FET seem to be doing their job. I get a PWM signal at the expected 100 kHz, with a max duty cycle of ~80%.

The problem is that I am getting poor efficiency (10-15%) and my transformer core is getting very hot (see thermal image below).

Basic Schematic

enter image description here

Looking at the thermal image, it's clear the core is getting extremely hot. Note that it is definitely the core and not the windings that are hot (24 AWG on the primary).

I'm using a Proterial MP2310 distributed gap core (good for 100+ kHz operation, 1.56 T sat. flux density. Datasheet specifically says its great for flyback SMPSs https://www.mouser.com/datasheet/2/957/Micorlite_technical_bulletin_opt-1509702.pdf). My first thought was that the core was saturating. I have B_induced = V_int_max / (A_coreN_primary), with V_in = 12 V, t_max = 8 us (100 kHz and an 80% maximum duty cycle), A_core = 0.43 cm^2 = 4.3E-5 m^2, and N_primary = 10. With these values, you get ~ 12V*8us/(0.43cm^2 * 10) = 0.22 T, so saturation should not be the issue (unless I'm doing something wrong, which is always possible). I've increased the primary windings to N=20 and operated at other frequencies (55 kHz and 200 kHz), and in all cases, the results are more or less the same (low efficiency, hot core. Sometimes not able to get up to the desired ~800 V).

My questions: Did I select a bad core? Am I just using it incorrectly? What kind of efficiencies would be considered reasonable? How hot should the core be in these types of applications? What improvements can be made?

Editing to add additional info: I am not using a snubber (can add later). The 750 V FET was selected knowing that it was overkill. I'm winding these by hand and was not sure what kind of turns ratio I would be able to achieve, so I selected a device with an absurd amount of voltage overhead.

R_sense is the 1 MOhm in the primitive schematic. I changed it to 500k for the below waveforms.

I am showing gate-to-source voltage and drain-to-source voltage in the waveforms below. I do not have a current diagnostic, sorry!

KiCad Schematic (note that I'm actually using the UCCC27523 gate driver and not UCC27524 shown in the schematic, only difference is inverting or non-inverting). The diode is the S1V-13-F, which has a 2 kV blocking voltage (https://www.mouser.com/datasheet/2/115/DIOD_S_A0004884737_1-2542723.pdf). KiCad Schematic of the flyback SMPS

Waveforms for 100 V output Waveforms for 200 V output Waveforms for 400 V output Waveforms for 600 V output

\$\endgroup\$
20
  • 1
    \$\begingroup\$ Welcome! ”1.56 T sat. flux density” This is of no use in your application since you are loss limited, not saturation limited. You mention 0.22 T peak, but what peak flux density swing (deltaB) have you designed for? DCM or CCM operation? What made you not choose bog standard N87 or PC90 material? 750 V MOSFET for 12 V rail seems extreme. What does your snubber situation look like? \$\endgroup\$
    – winny
    Commented Jul 27 at 17:32
  • 1
    \$\begingroup\$ Yes, powder cores (and this, which has roughly similar specs) aren't very useful in DCM, you need deep CCM to lean into Bsat. The winding distribution (or what it looks like through the thermal cam's vis detail feature) is quite poor as well. \$\endgroup\$
    – Tim Williams
    Commented Jul 27 at 17:45
  • 1
    \$\begingroup\$ Yes, if you’re in DCM, then you have 0.22 T flux swing and you’ll heat your core accordingly. Since you have a very cool MOSFET, you’d be better off operating in CCM as core losses are lower. Distributed gap is good, but expensive and a regular E-derived or PQ if you are low on height, with a simple single air gap will suffice. Compare different core materials and their loss characteristics and report back. \$\endgroup\$
    – winny
    Commented Jul 27 at 17:46
  • 1
    \$\begingroup\$ I didn't dig too deeply, but the Proterial MP2310 I found seems to be a ferrite toroid with a discrete gap, not a distributed gap. Looks like you have excessive core losses, so maybe see what you should expect based on frequency and flux swing and the core loss curves for your toroid. You'll probably see why it's getting hot pretty quickly. \$\endgroup\$
    – John D
    Commented Jul 27 at 17:48
  • 1
    \$\begingroup\$ As far as I can see, you are designing an HV-output flyback, nothing much special. With a 100 kHz switching frequency, a soft ferrite with a carefully selected core shape/size and bobbin (due to the HV output i.e. isolation is a safety must here), you still can get the desired performance. Soft ferrite is widely available in different sizes/shapes, so it's easier to find in the market. \$\endgroup\$
    – Rohat Kılıç
    Commented Jul 27 at 18:32

1 Answer 1

Reset to default
1
\$\begingroup\$

I'm going to run through how I calculate peak flux density with a little more care than I did in comments. Then I'm going to suggest that you might be running in CCM and this could certainly raise peak flux density to daft levels if you don't properly control duty cycle on light loads.

  • Your primary inductance is 23.37 μH
  • This comes from the number of turns (10) and \$A_L\$ (233.7 nH/turn²)
  • Maximum duty is 80% at 100 kHz. This means Δt is 8 μs
  • Therefore Δi is 4.1 amps (rearrangement of the inductor formula)
  • 4.1 amps is also the peak current in DCM
  • MMF is amperes × turns = 41
  • H is MMF divided by core length (5.6 cm or 0.056 metres) = 732 At/m
  • From the data sheet, permeability (relative) is 245
  • Therefore, peak flux density is 732 x \$4\pi \times 10^{-7}\$ × 245 = 0.225 teslas

Sorry for being out by a factor of ten in comments. My value above now agrees with your value so, either your core has a lot more permeability than what is stated in the data sheet or, the circuit is launching into CCM and creating an overload of flux density.

But, 100 °C isn't an unexpected temperature for this type of converter. Maybe you are worrying about something that is unimportant?

Here's what your current waveforms should look like: -

enter image description here

Calculator image from my basic website.

\$\endgroup\$
8
  • \$\begingroup\$ I'd say it is rather unexpected, given the temp rise of everything else is almost nil in comparison, and the quoted efficiency. But we seem to be lacking adequate information to properly answer this question as such. Which, actually, someone should probably get a close vote started, huh? Guess I'd better get on that... \$\endgroup\$
    – Tim Williams
    Commented Jul 28 at 13:34
  • \$\begingroup\$ @TimWilliams I'll work on adding some more details. I was trying to avoid overloading a reader with unnecessary information. R_sense is the 1 MOhm in the primitive schematic, it just was not labeled as such. By the way, I'm new here. What does it mean to "close" this? \$\endgroup\$
    – Jacob Stephens
    Commented Jul 28 at 14:14
  • \$\begingroup\$ @Andyaka Thanks for the second look. I'm surprised that 100 C is not considered too high. I'm okay to run things a bit hot, but thought that was a bit much. I'm still not happy with the 10% efficiency. I'm working on collecting some waveforms, but I believe you are correct in that the device will transition into CCM at higher duty cycle. Will collect data and post soon. \$\endgroup\$
    – Jacob Stephens
    Commented Jul 28 at 14:31
  • \$\begingroup\$ @JacobStephens it might, or it might not. It all depends on the output voltage (if not 800 volts) and current/load resistance. You need to state these things. \$\endgroup\$
    – Andy aka
    Commented Jul 28 at 14:51
  • 1
    \$\begingroup\$ @JacobStephens if your load is 500 kohm and the output voltage is 800 volts then something is going badly wrong somewhere. I'll attach a picture into my answer of a calculator result that shows but, for now, it looks like that the diode you are using has a breakdown reverse voltage of 56 volts and this is a significant problem. The diode needs to be rated at least 900 volts for this application. \$\endgroup\$
    – Andy aka
    Commented Jul 28 at 15:58

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.