TPS61022: Fails to short with 2A load

Using the TI calculation spreadsheet, I found an error in my unit conversion 9.7e-11 F = 0.097nF (I thought it was uF).

Tomorrow, I will try 97pF for the C3 value and will report back.

I'd welcome other ideas if someone happens to stumble upon this in the meantime.

Oksy, look forward to your update.

Hi Nicholas,

Do you mean all of the TPS61022 are damaged? If that, I think it may be caused by your layout, the area of loop SW-COUT-VOUT-GND is large (TPS61022 is very sensitive to this loop area), which may cause high voltage spike in SW node, then damage the device. 100nF cap is not helpful because the main power current will flow through the larger cap. At least 10uF may be helpful, and you need to place this smaller cap as close as VOUT pin.

You can take this app note as refence, it shows the recommended layout and lists some common mistakes we found from other customer before. Thanks.

www.ti.com/.../slvaes4.pdf

Regards,

Nathan

Thank you Nathan, this document is very helpful, and it seems the datasheet for this IC deserves a revision.

The document that you linked above shows a very different layout than the example in the datasheet (in gray below) which I followed...

It seems like fiddling with components to solve this is hopeless if the layout is incorrect, I will redesign the PCB and order a new one with a layout like the document you mentioned above (in color below) and will report back...

I've made an attempt at redesigning the layout (see image below).

I added the pre-bias diode from VIN to VOUT mentioned by the datasheet for applications where the input voltage is higher than 4.8V (USB). I also added the feedforward capacitor and series resistor for 2kHz zero frequency since I require output capacitance over 40uF.

I tried to keep the zone areas minimal, but in that document that you mentioned, it says to "use larger and thicker PCB copper for the power pads (GND, SW, and VOUT) to enhance the thermal performance". This makes sense from a thermal point of view, but I believe this goes against what you said with regard to loop area. Should I expand GND and add vias for thermal mass? Should I reduce the area of the SW zone?

Generally, might you have any feedback about this revised layout and schematic?

Thank you.

Hi Nicholas,

1.feed forward may not be necessary in your application, because the 40uF is effective value not nominal value, for ceramic cap, effective is smaller than nominal value. Of course we recommend you can leave a placement of feed forward cap here in case that there are some changes.

2.There is no conflict between that  larger and thicker PCB copper and keep the loop area as much as small. Because high frequency AC current will flow through the capacitor along the minimum impedance path (minimum impedance means minimum loop area), so it doesn't matter that you use large PCB copper as long as you place output cap as close as possible to VOUT pin.

3.One more thing is that you'd better use wider line between the cathode of diode and vout, because there may be large current here when startup.

Regards,

Nathan 

Thank you Nathan, I made some final modifications and ordered a new board to test.

From your feedback about the start-up current through the pre-bias diode, I made the trace to VOUT wider, and changed the component to a 2A Schottky with low voltage drop (280mV).

At the risk of growing the loop area, I added a few more vias for thermal mass in the GND zone, this better-matches the example layout in the application note that you sent.

On your recommendation, I kept the footprints for the feed-forward capacitor and its series resistor. The calculator suggests values of 100pf and 100K; I made a mistake with the previous schematic I posted with 1K for the series resistor, which is calculated to result in an unacceptable phase margin which may cause instability.

I will share the results here when the board arrives and I assemble and test it. I still have a few days to submit a revision to the fabricator it if someone spots a problem with this (below).

Thank you Nathan for your helpful feedback

Nicholas

Hi Nicholas,

You can try to remove feed farword cap in your calculation tool (be giving a very large zero like 1000kHz), if the phase margin > 45 and gain margin > 10, it means the system is stable, feed farword is not necessary unless you want to increase cross over frequency. (Generally feed farword cap has two functions, improve the loop stability or increase cross over frequency)

Of course, based on your calculation tool, the system is still stable with 100pF, depends on you.

Regards,

Nathan

Hello Nathan,

I've built a module using the revised layout as depicted earlier in this thread, I am using a WEL3005 electronic load for testing.

With the feed forward cap and series resistor (phase margin 48.5º), the TPS61022 failed to short with a 1.5A load.

I replaced the TPS61022 with a new one, removed the feed forward cap and series resistor (phase margin 64.5º), and my benchtop power supply reads 5V3A (VIN) with the electronic load reading 5V2.2A (VOUT), an effciency of 73%. Lowering the VIN to 3V2.87A, VOUT reads 5.43V1A, an effciency of 63%.

Unless I have made a mistake with calculations, the efficiency seems quite a bit lower than what I would have expected from the datasheet, but at least it is stable/working now. Thank you for your advice on bringing this IC up.

Might you have any ideas of what I can try to improve the efficiency, or some feedback about whether or not these results are reasonable?

Many thanks,

Nicholas

Hi Nicholas, 

I think it is abnormal because the efficiency is too low, and why you get 5V not 5.5V when you give VIN=5V?

Do you have a long wire at the input and output? If that, there will be line power loss and line drop voltage, you need to test the vin pin and vout pin voltage to calculate the efficiency, that would be a more accurate result.

Regards,

Nathan 

Hi Nathan, for this test, I am using flexible breadboard jumper wires 10cm long.

After reading your message I took some more organized readings with the electronic load, and recorded them in a spreadsheet...

Quite a low efficiency.

I tried to set the load to 2.5A, but there is a very noticeable coil whine and the voltage that the load sees is very low, around 3.4V. Perhaps this could be because of the 2A forward current rating of the pre-bias diode CUHS20S30?

I will try to change this diode and will report back.

*** UPDATE 1 ***

I changed the pre-bias diode to a B340 (3A forward current), I was able to adjust the load a bit higher. Seems like a slight regression with efficiency.

I'm now realizing that the coil noise and sudden low output voltage occurs when VOUT is loaded such that it becomes equal to or lower than VIN.

*** UPDATE 2 ***

Here are the results with no pre-bias diode...

10cm line is not short, there will be voltage drop and power loss caused by this, especially with larger current. So I recommend that you could test the Vin and Vout pin voltage in PCB board with multimeter, this is the real efficiency of TPS61022.

By the way, I think you set E-load at CC mode, and start-up with load on, right? If that, it means TPS61022 will start up with large load current, when you add pre-bias diode, when start-up, the load current will flow through this diode, so if you want to support 2,5A load current when startup, 2A diode actually has risk. When you don't add a pre-bias diode, it means the the load current will flow through Boost, the startup current limit is usually smaller than the current limit after startup, I am not sure if TPS61022 can support 5.5V/2.5A current when startup, I think it may be difficult... 

Maybe you can try to startup with no load and then load on after startup.

Thanks.

Nathan

Nathan Ding said:

if you want to support 2,5A load current when startup, 2A diode actually has risk

I used a B340 pre-bias diode which is rated for 3A, and it didn't change much (see table of measurements under "update 1" that I posted above)

Nathan Ding said:

10cm line is not short, there will be voltage drop and power loss caused by this, especially with larger current.

I measure 0.5ohm resistance on this 10cm jumper wire. I might be missing something, but I don't see 0.5ohm as causing significant power loss.

Regardless, I put the TPS61022 module in my project (plugging it on using 2.54mm headers, no wires), ensuring it is started with a low load ~50mA before I started the battery chargers to increase the load. It still has a 2A limit. I discovered this by reducing the charge-rate of the battery chargers until the boost converter was stable, and it totaled ~2A when all chargers are active.

Do you see any reasons in my layout or component choices (above) why this implementation would have a 2A limit? Or why the efficiency is lower than expected?

With 0.5ohm resistor, if you have 2A load current, then you get 1V voltage drop and 2W power loss. The total output power is ~10W, so you have 20% power loss on line.

I think your layout is ok.

Is it possible the 2A current limit cause by USB supply? Because you said the USB supply is 5V/3A, not sure if you mean the max current is 3A.

Or maybe you can test the waveform of VIN SW VOUT when 2A current limit condition and 2.5A coil whine condition to check if TPS61022 works normally.

Hi Nathan, for these tests with the results in spreadsheet-style format, I have been using a 30V5A adjustable linear bench-top power supply.

Although I am now realizing that, from the PSU to VIN, I should have been using leads rated for this power, an oversight on my-side that seems obvious now.

Regardless, performed another test with a shorter cable with a thicker gauge from VOUT to the load...

.

These results are better than before, but still worse than I expected from the datasheet (below).

the TPS61022 was stable at 3A, but at 3.5A the efficiency began dropping steadily, I believe that this was due to heat since the coil became noticeably hot. I tried adjusting the load to 5V4A, but it would not activate.

The fact that this test shows improvement over my previous test, and also when I installed this module in my project, leads me to believe that at least part of the problem is the resistance of the paths VIN or VOUT.

As you mentioned, perhaps the diode selection could be improved. I'm not sure if I mentioned, but the coil I am using is a Bourns SRP5030CA-1R0M.

I find vout=4.25V when iout=3.5A, I think the diode is conducted in this condition so vout is very low.

If you have a long wire in your input, you can also add a big input capacitor, like aluminum electrolytic capacitor or tantalum cap.