TIDA-01605: Negative bias supply not regulating

Tool/software:

Hi all,

We took the TIDA-01605 reference design as an example for the design of our SiC gate driver, but unfortunately during functional testing we have hit a bump where the regulated negative supply is not doing its job. The push pull transformer is working and we measure a neat 23V output. The negative supply is not providing -5V, we only see -0.2V. We are aware of stability boundaries for the TL431 as mentioned in application report SLVA482A. The strange thing is, we do not really see any oscillation or whatsoever. It's just very stable at -0.2V. Still we went ahead assuming it must be somehow related to stability, so we did further testing (see below).

The schematic is included for the isolated push pull supply with regulated negative bias. I've added the picture for the decoupling capacitors as well, because these are connected accross the gate driver chip and are therefore adding to the capacitive load of the TL431 (so in parallel to the 2x100nF because v_gdd == v_gdda, v_ks == v_ksa, v_gss == v_gssa). As you can see, the design is very similar to the reference design, we choose a different transformer ratio to match our design and the MPNs for the passives are slightly different, although specifications are again similar.

The following tests have been carried out success to determine a cause, but so far no luck:

• Reducing cathode bias resistor to increase bias current (from 10k to 1.1k), no visible difference regulation voltage;
• Adding 'ESR' resistor between cathode and kelvin source output (see slva482a.pdf). First trying with 180Ohm resulted in approximately -3.6V negative bias supply. However, this does not match the graph in the document, so it was decided to lower it to 30Ohms. Then its back to a bias of only -0.2V.
• Further increasing the capacitive load to increase stabilty. Adding one or two extra 4.7uF did no change significantly change the behavior.
• Assuming the assembly service placed the wrong part. Under a microscope, the device shows a clear marking . T3QG is indeed mentioned in the TL431 datasheet tl431.pdf. Still assuming the pinout is wrong, two further tests were conducted. First, a simple shunt regulator with only a bias resistor (180Ohm) was created. This shows the device would indeed correctly produce a 2.495V cathode output for 5V input. Going to 10V would keep the cathode voltage at 2.489V. The second test assumed pins 1 and 2 are reversed. The first test could not verify this because these were shorted, so the second test was on the PCB. The SOT23-3 was flipped upside down. No output voltage at all. Hence conclusion of these tests, correct part and pinout correct.

We've now ordered a TL431LI, maybe this will help, but we are a little lost. It seems as we are overlooking something obvious. Does anyone have new ideas?

Thanks a lot!