TPS40304 not working

Hi Christine,

Have you checked what part on the power circuit is died? From the schematic, I would suspect that the spike and ringing on the switching node will be quite high for 16V input and that might cause FETs breakdown. CSD16321Q5 is 25V device. I would suggest use at least 30V device and add R-C snubber to dampen the switching node spike. The board layout should be done with care for miniziing parasitics.

Regards,

Na

Hi Na

We probed the signals on every pin of the power regulator and all outputs 0V except for the Vin and the enable pin. Does this indicate that the power regulator has died?

Actually all the parts that we used on the circuit are the recommended parts in the BOM list provided in Switcher Pro. CSD16321Q5 was one of the suggested parts.

Another thing is that if the 16V input is producing high spike or ringing on the switching node, wouldn't the 1.2V power circuit die as well? However, the 1.2V power circuit is still able to output 1.2V when it is plug-in to our test board.

Regards,

Christine

Hi Christine,

BP is also 0V? If EN/SS is not 0V, what does its waveform look like?

Have you measured whether there is any suspicious short in the power circuit? FET, IC or other components? If the IC is broken, which pin is short? If FET is dead, what is the impedance between each two of the terminals?

The switching node spike is the first thing I would suspect from the schematic. Since the 1.2V is working well, you may examine what does the switching node spike and ringring look like. In terms of the differences between the two power regulators, does the 1.2V share the same layout as 5.3V? How about the load on the two test boards?

Regards,

Na

Hi Na

Sorry, my mistake...not all the pins are 0V. Below are the pins that have non-zero voltages (taken using DMM). All others shows 0V

1. VDD = 16V
2. EN/SS = 0.947V
6. BOOT = 0.625V
10. BP = 6.47V

Attached is the screenshot of the waveform on EN pin.

I have measured and there isnt any short between the pins on the IC and these pins are also not shorted to GND or Vin or Vout.

the Vout is also not shorted to GND.

For the FET connected to the HDRV, the measured resistance between terminals are:

R(G-S) = 104 kohm
R(G-D) = ~280 kohm (fluctuating)
R(D-S) = ~350 kohm (fluctuating)

For the FET connected to the LDRV, the measured resistance between terminals are:

R(G-S) = 3.2 kohm
R(G-D) = 14.5 kohm
R(D-S) = 11.2 kohm

When you mention switching node spike, do u mean the SW pin on the power IC?

Even though the 1.2V power board is able to output 1.2V, our device on the test board cannot be detected when it is powered up by this 1.2V power card.

We designed a common board that can support 1.2V and 5.3V as the parts footprints are the same except for the inductors.  

Attached is the schematic for the 5.3V. There are some mistakes in the circuit that we have manually fixed; we have wired the net between R6 and R4 to Vout instead of GND and also changed the potentiometer to a fixed 1.3kohm resistor.3858.standalone_5v.pdf

Regards,

Christine

Hi Christine,

From the EN/SS waveform, the TPS40304 gets into overcurrent protection falsely. The two power MOSFETs seems not broken. Then it might be caused by high-side driver not working properly becaue BOOT voltage is not right.

Can you measure the diode voltage drop from BP to BOOT when the part is not powered up?

Can you also capture the BOOT, HDRV, LDRV and SW waveforms when the part is powered up?

The switching node, it is electrially the SW pin of the controller IC, also the drain of the low-side FET. There might be difference between SW pin voltage and the drain of low-side FET depending on where and how it is measured. I was worried the D-S voltage of low-side FET is too high, which needs to be measured close to the Drain and Source of the FET with short probe leads to minimize noise coupling. It seems not the cause here, but I am still worried that there might be reliability issue even if the circuit is working now. You'd better check.

Another thing I would like to know is what is connected on EN/SS. From the waveform, it looks like 0.2uF is on EN/SS. But I cannot figure out how VCC_ENABLE connected to other components in the schematic. Since the internal current source connected on EN/SS is used to determine the soft-start time and hiccup time in an overcurrent fault, you may not want to connect other soure to this pin. Open-collector or open-drain transistor to pull it down to disable the device is okay.

Regards,

Na

Hi Chirstine,

If the power board cannot be accessed on the test board, you may test the power board alone with resistor load instead of your test board based on your estimate of the load current and power.

Regards,

Na

Hi Na

The diode voltage drop between BP and BOOT when not powered up is 1.75V.

Below are the screenshot of the waveforms probed at the BOOT, HDRV, LDRV and SW pins.

 1. BOOT

2. HDRV

3. LDRV
    a) overview of the high spike

    b) zoom in on the high spike 

4. SW

There isnt any difference when measuring the SW pin from the IC pin and from the low-side FET pin. Below is the SW pin measured from the low-side FET.

The EN/SS pin is connected to a 0.1uF capacitor to GND.

All these signals are captured from the 5.3V power card which is externally powered up by the Agilent DC power module instead of from the test board.

Regards,

Christine

Hi Christine,

I think the high-side MOSFET driver inside TPS40304 is broken. You need to replace the IC. It is possibly due to the overvoltage stress on BOOT pin and still might be caused by the switching node spike and ringring. I suggest

(1) increase the boot resistor R5 from 2ohm to 10ohm to slow down the turn-on of high side FET and limit the BOOT charging/discharging current.

(2) add series connected R-C snubber to parallel the D-S of the low-side FET Q3. This can smooth out the switching node spike and ringing. You may start with 1ohm and 1nF. These two components need to be placed as close to the FET as possible to reduce parasitic inductance and better to be surface mount parts. The resistor may get hot and is better to be 0805 or 1206 size.

(3) Measure the switching node spike when the power regulator works properly, at different input voltage and load current. For test purpose, it's better to use the DC power supply and pure load, either resistor or electronic load in constant resistive load  mode, not the test board.

If possible, can you also share the layout? This may give an idea what other improvements can be done.

Regards,

Na

Hi Na

How do we determine that the high-side MOSFET is broken? And should i replace it with the same part or do you recommend using another higher voltage MOSFET?

Below is an image of the schematic with the modifications done in green. Is the R-C snubber circuit correct? How do you come out with the 1ohm and 1nF value?

Attached is the layout of the power card.

1185.cm_dut_power_card.zip

We have scoped the 16V from the test board and it is stable without any ringing. I have also attached the schematic of the 16V power circuit that is supplied to the test board.

5543.Vin_16V.doc

Thanks.

Regards,

Christine

Hi Christine,

The high-side driver inside TPS40304 looks broken, not the high-side FET. The BP to BOOT diode voltage drop should be measured around 0.8V for normal part when the TPS40304 is not powered up. But since it is 1.75V now and the BOOT voltage is quite low when TPS40304 is powered up, so it is determined that it is broken. The TPS40304 needs to be replaced.

From the impedance measurement results, the high-side FET should be fine. However, I would still suggest you replace both Q1 and Q3 with 30V or higher voltage rating FETs for more margin.

The changes in the updated schematic are correct. The following is a useful reference for snubber circuits design. The suggested snubber capacitor is 2~4 times the output capacitance of the low-side FET, which is around 1.5nF@Vds=16V. But the higher snubber capacitance, more power loss. From the reference, the power consumption on the snubber resistor can be estimated as FswxVin^2xCsnubber.  Using 1nF, the loss caused by the snubber is about 0.2W, 0.4% of the total output power at the maximum Vin. From my experience, it should be effective for the damping. Regarding the snubber resistor, it is suggested to be the characteristic impedance of the parasitis inductance and capacitance. The smaller resistance, lower the first spike, but takes longer time for the ringing to die out, probably the second spike of the ringing is even higher than the first spike. From my experience, 1~4 ohm works well.

http://www.ti.com/lit/an/slup100/slup100.pdf

I could not open the .brd file. Can you convert it to pdf?

I suggested use DC supply and resistive load for debug, instead of the test board, not because the 16V is not stable, but the ideal supply and load give you the access to probing the power board and rule out other factors caused by cascading switching power converters. You need to make sure the single power converter works well under possible Vin and load range, before connecting it to other system. In fact, even two switching power converters work fine alone, to cascade them, the output impedance of the first one needs to be smaller than the input impedance of the second one to avoid instability issue.

Regards,

Na

Hi Na.

I found this thread very interesting as I'm currently experiencing a similar problem.

I have designed a TPS40304 regulator with 15V -> 1.0V / 8A characteristics using online Webench tool.

I'm using a CSD86330Q3D as switcher. I've been using this power circuit in many designs without problem with various controllers (it sometimes requires a high side gate series resistor, depending how fast the driver switches but I did not find it necessary with TPS40304)

This regulator is usually working just fine with clean switching (no ringing on the SW node) and excellent efficiency but it failed after a while on many boards. After reading this thread, I have made measurements and found out the the CSD86330 power driver is OK on the failing boards (I checked that both FETs can switch on and off properly). Apparently, the high side driver in the TPS40304 is dead (too high forward voltage on the boot diode and too low HDRV voltage to turn the FET on).

I also have a TPS40304EVM-353. It is not using the CSD86330 as power switch but I used it to compare the TPS40304 regulator behaviour. I measured very high ringing noise on the EVM compared to what I have on my system but still, the TPS40304 is not dying (with the same input voltage and 2 ohms resistor in series with the boot capacitor). This refers to your suggestions above that ringing may cause the driver to die...

Any other idea what could cause this driver to die? Does CSD86330 switch too fast (it is switching really fast!) and kill the driver through the boot capacitor?

Thanks.

Arnaud Boursier.

BTW, what's the difference between TPS40304 and TPS40304A? I could not really tell from the datasheets and did not find any comparison...

Hi Arnaud,

Yes, I think it is quite possible for the switching spike to kill the BOOT pin of TPS40304. Increase BOOT resistor, high-side gate resistor and RC snubber between SW and GND are effective ways to lower down the spike.

The only difference between TPS40304 and TPS40304A is the reference voltage, 0.6V vs 0.591V.

Regards,

Na

Hi Na.

Following this suggestion you made earlier to Christine, I did try to raise the series resistor with boot capacitor from 2 ohms initial recommended value to 10 ohms. It actually seems to fix the problem as the regulator did not die on any of the fixed boards since then.

I have chosen CSD86330Q3D dual FET as power stage associated with TPS40304 for best compactness, efficiency and low parasitic inductance leading to no switching ringing. I had inserted a small resistor in series with high side transistor gate because I experienced ringing with another controller associated with this transistor. However, it was counter-productive with TPS40304 because the dead time between high side FET turn off and low side turn on is very short. The additional resistor in the high side gate delayed high side turn off and caused trans conduction  through both FETs at each cycle. This caused a lot of excess dissipation. Without the high side FET gate series resistor, the transition is very fast, yet, without ringing; ideal case! The drawback of this very fast transition on the switch node is that it apparently is too fast for the integrated boot diode and kills it while reverse-polarizing it. It looks like this diode is a bit weak, isn't it? The recommended application schematic already includes a series resistor which means that this diode is known to be sensitive to this transition... It may be interesting to add a note to users about this point: take care of the diode which is sensitive to large (if VCC is rather high) and fast transition on SW node (depending on high side FET choice) and potentially raise boot capacitor series resistor value.

After all this, the association of TPS40304 and CSD86330Q3D is an excellent compact, efficient, low noise solution (I'm doing 15V -> 1.0V / 10A with it)

Thanks for your support and hope this case can help other engineers.

Arnaud.

Hi Arnaud,

I appreciate you sharing the experience with our devices! Since the TPS40304 is released earlier, the diode does look a bit weak now with super fast dV/dt and high VIN when working together with NexFET (small Qg). We would take your suggetion add an note in the datasheet to highlight the importance of limiting the BOOT transient voltage. Thank you again!

Regards,

Na