3
\$\begingroup\$

I have a SMPS acting weirdly and I believe to be caused by its transformer. I need to find a replacement but I'm having trouble identifying it.

SMPS transformer: pri: 240V, sec1: 12V, sec2: 24V, aux:10-27V

The marking has too lines: 1st containing the VDE logo followed by 30090314, 2nd containing REG.-Nr. E895 CH 20 18. I know "VDE" stands for VDE Association, most likely indicating the standards by which it was built/tested. The rest of the info reveals nothing.

Its external dimensions are 34mm x 23mm x 21mm. It is a through-hole component, soldered on a PCB labeled VESTEL 17IPS12 R3, for which I found a generic schematic. In an effort to identify the cause of the problem, I replaced and changed parts to reflect the actual circuit.

VESTEL 17IPS12 R3 - adapted to the circuit

The primary can take up to 240V or more. It has 3 secondaries: one for 12V, another for 24V and an auxiliary with no voltage mention in the schematic but from the controller's specs (FAN6300A) I can infer it can be anything from 10V to 27V. Also, the controller can work with frequencies up to 100kHz.

Searching on mouser and digikey gave one candidate: ECO2425SEO-D01V019. However, this doesn't seem to have a suitable AUX voltage.

Also, since the circuit seems to be popular I can assume there's a at least small marked for parts but I could not find results in conjunction with the circuit info. Other similar circuits seem to have transformers with other markings, suggesting brands (LiShin) or containing Thermal class B.

Can anyone suggest what else to look for?

Update

I made 4 measurement, all having on channel 1 the drain-GND signal for trigger and timing reference. On the 2nd channel I captured CS-GND, then DET-GND both before and after modifying the voltage dividers. For obvious safety reasons, all measurements were made with the DUT powered through an isolation transformer.

Before anything I wanted to catch the moment the switching stops when I found the controller's Vcc pin to have a stable 16V. I then probed the DET pin and, after the blanking interval (4us), there's clearly more than 2.5V. All these check against the R210, (R211 + R212) divider ratio (6.4). According the the FAN6300A datasheet, a DET voltage above 2.5V is sufficient to latch it off.

FAN6300A stopping with 16V on Vcc

FAN6300A stopping with 16V on Vcc

Based on the above measurements I assumed the feedback output voltage doesn't reach 12V in time and causes the DET to trigger OVP latch-off. The timeout for lack of FB signal (55ms) seems generous enough not to cause problems. With the 24V rail disconnected and all 12V output rectifier diodes, capacitors (well within tolerance and nearly zero ESR) and the entire feedback network (including optocupler) checked, the high-voltage rail also having a stable 340V (with capacitors and rectifier bridge checked as well), I further assumed something else causing the slow to catch-up output. The only culprit, in my opinion, was the transformer.

The following captures are made for the circuit in its original form.

FAN6300A - ch1: MOSFET's drain-GND voltage, ch2: CS-GND voltage, 2us time base

Channel 1: MOSFET's drain-GND voltage, Channel 2: CS-GND voltage

FAN6300A - ch1: MOSFET's drain-GND voltage, ch2: DET-GND voltage, 2us time base

Channel 1: MOSFET's drain-GND voltage, Channel 2: DET-GND voltage

I then got the idea of altering the divider network for the DET voltage, such that it doesn't trigger OVP anymore yet keeping Vcc below 27V (the other OVP threshold). I placed 2 resistors (51k and 47k) in parallel with the R210. This caused the 12V rail to raise higher but still not enough to generate feedback. I then altered the divider (one 51k resistor in parallel with R122) feeding the TL431 on the feedback such that it lowered the output voltage. This proved successful in the sense that there was feedback and a rather constant output voltage. Nothing overheated (or worse) for the approx. 5 minutes I kept the circuit on.

The following was taken for the modified circuit.

Circuit with modified dividers - ch1: MOSFET's drain-GND voltage, ch2: DET-GND voltage, 2us time base

Circuit with modified dividers - Channel 1: MOSFET's drain-GND voltage, Channel 2: DET-GND voltage

I was glad to see the controller entering extended-valley mode to cope with the low current drawn by a LED + its limiting resistor. That said, the circuit still seems to have long pauses (~1s) between trains of switching pulses. Probably, I could further adjust divider networks such that the Vcc reaches and remains between around 17-24V (assuming that's the cause of the pauses).

All those, of course, are made for testing purposes and to confirm/infirm suppositions but I would like to understand what's happening and eventually fix the unit.

\$\endgroup\$
18
  • 4
    \$\begingroup\$ SMPS transformers are usually custom-made. It's rare to find an off-line SMPS with an off-the-shelf transformer. What makes you suspect that the transformer is bad? They're just wire wound on a core and are therefore rather robust, so before they die, everything else in the circuit is already charcoal. \$\endgroup\$ Nov 19, 2023 at 14:22
  • 1
    \$\begingroup\$ I feared this might be the case but hoped somehow standard transformers may exist. The 12V and 24V are common and the SMPS in question has been used in many consumer electronic units. Answering the second question, all parts surrounding the transformer are good. I was able to adjust voltage dividers around the transformer to bring the SMPS back to live (reduced the feedback output to from 12V to 7V and the DET voltage from AUX output). My only guess is that neighboring transformer turns have somehow shorted, creating an imbalance that threw the controller in a failure latching mode. \$\endgroup\$ Nov 19, 2023 at 14:40
  • 3
    \$\begingroup\$ That's possible, but unlikely. A shorted turn causes extreme power dissipation in the transformer, which usually leads to melting / burning. It's more likely that one of the surrounding components is degraded. Likely candidates are the bootstrap rectifying diode (these love to become weak / open), bootstrap Zener, bootstrap capacitor, and current shunts. In the circuit you posted, this would be D128, D105, C115, and R239..R244. If you have spares in your parts bins, try replacing these diodes and the capacitor to see if that changes anything. \$\endgroup\$ Nov 19, 2023 at 14:47
  • 1
    \$\begingroup\$ Thanks for the suggestion. The current sensing resistors are sparkly clean, in fact everything seems to be. I checked every resistor and every junction, some out-of-circuit as well, including those you mentioned. I do find weird that the Vcc voltage is continuously varying between 8V and 16V once the circuit was revived but I assumed this is because it's still functioning out of specs. That said, according to electronics.stackexchange.com/questions/58702 high potential difference turns (between layers) are separated by insulation. Nearby shorted turns may be less obvious to detect. \$\endgroup\$ Nov 19, 2023 at 15:07
  • 1
    \$\begingroup\$ Additionally, I disconnected the transformer and placed a 100kHz sine wave on the primary. I then measured both the feedback and auxiliary outputs in parallel with the primary. The scope indicated 720mVpp on the feedback output and 5.92Vpp on the primary, giving a roughly 8:1 ration. Similarly, I got 1.15mVpp on the aux output for a 5.76Vpp on primary, giving exactly 5:1. This may be withing specs, but I cannot tell for sure. \$\endgroup\$ Nov 19, 2023 at 15:30

1 Answer 1

3
\$\begingroup\$

The inductance and resistance of a transformer can be measured with simple tools, for example a signal generator (sine preferred) and oscilloscope. At least an AC voltmeter, or preferably vector voltmeter, is required, but the oscilloscope being so indispensable, it's commonly available.

I have several measurement methods on my website: Calculators: LC Resonance | Seven Transistor Labs, in particular consider the vector method. Set up a voltage divider between a signal generator, resistor, and DUT (device under test), and measure the voltage on both sides of the resistor.

(A simultaneous measurement across a resistor is required to get a phase-accurate reading without other signals. We could also use a generator's trigger output and phase reference to that, making sure to subtract unloaded and loaded angles, to measure with minimum resistance i.e. the generator's (typically 50 ohm) output itself, which will extend sensitivity to somewhat lower |Z|.)

This appears to be a flyback supply. Presumably, the transformer in question is a gapped ferrite design, and should have a Q factor on the primary in the ballpark of 100 or so. Losses this low are difficult to measure accurately with a vector method, but we can rule out very low Q, and unexpectedly low L, as would be the case for a shorted turn.

(The resonant (RLC) method further down the page, is recommended for measuring higher Q accurately.)

Probably, unloaded primary inductance should be in the ballpark of upper 100s µH to low mH.

\$\endgroup\$
2
  • \$\begingroup\$ Thanks for the suggestion. I will try that next as I have no other leads at this point. \$\endgroup\$ Nov 21, 2023 at 0:08
  • \$\begingroup\$ The added waveforms look like the transformer and controller are both behaving more or less normally. Your problem seems to lie elsewhere. \$\endgroup\$ Nov 21, 2023 at 1:37

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.