5V buck converter debugging

[davegravy]

I assembled a 12-to-5V buck converter from an LM64460-Q1 (datasheet) and I'm not getting any output. Schematics and board layout are attached. Here was my process and observations for testing the board:

* I slowly ramped the input voltage up with no load connected and noticed no current draw until 4V at which ~50mA started to flow. At around 6 or 7V the current spiked to 1A for a few seconds and then dropped to 6mA when I reached 12V. At this point I checked the output voltage which was only ~1.2V but was very slowly rising.

* I removed supply to Vin, briefly shorted the output terminals just to discharge the output caps, and reconnected 12V input. Now I have 0V output and that's all she wrote.

I checked all the other pins and other than the Vin1 and Vin2 pins and EN, all are 0V (including VCC which is nominally a 3.3V LDO output). I desoldered the chip and found the SW1, SW2, and SW3 pins shorted to GND, so the chip seems cooked. I guess I should have pulled the plug before I got to 1A draw.

Before I solder a new chip, is there something wrong with my design that one would expect to fail like this?

Before powering I did check for shorts but I'm not sure how to test for unconnected pins... doesn't seem likely that an unconnected pin would cause failure like this however.

[davegravy]

Ignore that the schematic says LM63460, this is the LM64460, specifically the non-adjustable 5V output model.

Well I decided to replace the chip anyways and this time current limit my bench supply to 100mA. My bench supply is outputting 3.8V @ 100mA, and the buck converter output is 1.6V.  I'm guessing I shouldn't expect that much current with no load, even though I'm not operating anywhere near my design input voltage (12V).

[Vovk_Z]

I guess it can't work as intended with low input voltage, and even can be damaged (becuse it can't operate). So you have to feed it with needed intended voltage or nothing.

[davegravy]

I guess it can't work as intended with low input voltage, and even can be damaged (becuse it can't operate). So you have to feed it with needed intended voltage or nothing.

The datasheet claims it can work as low as 3V input but do you mean perhaps with the external components I've connected that some instability can occur at low voltage which is damaging?

[__steven]

I don't see any issues with the circuit that would cause this failure. What nodes did you check for a short prior to powering on? Assuming you only checked the input and output rails, I will guess that there was a short from the switch node to ground under the inductor. You mention that there is still a short from the switch node to ground after removing the IC. Is this a short in the IC or on the board?

[davegravy]

I don't see any issues with the circuit that would cause this failure. What nodes did you check for a short prior to powering on? Assuming you only checked the input and output rails

Thanks for reviewing. I checked adjacent pins, at least those with connections (most of the functional pins have an NC pin between them), and I checked each pin to ground (in case of a short to the center pad).

I will guess that there was a short from the switch node to ground under the inductor. You mention that there is still a short from the switch node to ground after removing the IC. Is this a short in the IC or on the board?

A short in the IC because after removing the IC the board shows no short and the IC shows the short. I suspect the internal mosfet overheated and caused the short, the question is why.

I'm leaning towards the idea that the low voltage I was initially supplying caused a failure to start which (somehow) drew too much current and overheated. The IC has a "low voltage lockout" feature where the chip will remain disabled until a minimum voltage is reached. This is using a voltage divider on the EN pin. The Webench design I used didn't implement this (EN is tied to Vin) and I'm suspicious that this exists to prevent this very failure mode (e.g. if your voltage supply is a battery which might drift out of nominal voltage range).

[__steven]

You mention there was no load attached and no short, therefore the only current you would see is the inrush to charge the output capacitors. The IC has a soft-start feature that should handle this no problem. Starting up at 4V is what the datasheet describes during start up.

How are you measuring for a short?

The only other two things I can think of. Can you verify the power good resistor is 100k and not something much smaller? The other thing is perhaps the feedback voltage is disconnected. Can you check continuity/resistance between the feedback pin and the output voltage?

I do not see issues in layout nor schematic causing a failure assuming they are implemented as documented. I would think that replacing the IC and trying again is best, if you see the same behavior at 4V try to find what is heating up.

[davegravy]

How are you measuring for a short?

Using continuity (with audible beep) mode on my multimeter, probing at the nearest connected component (e.g. C4) on the same net so that I don't have to get the probe on the very fine pitch IC pad. I skipped checking for shorts between a pin and adjacent NC pin, maybe it's not a safe assumption that those aren't internally connected to something?

I would think that replacing the IC and trying again is best, if you see the same behavior at 4V try to find what is heating up.

I think I'm going to find the IC heating up, given that it developed an internal short. I could get an IR camera to pinpoint what part of the IC is hot but is this going to be valuable info? Or are you suggesting some external part of the circuit may also be heating up and that could be the root of the whole failure?

[__steven]

I think I'm going to find the IC heating up, given that it developed an internal short. I could get an IR camera to pinpoint what part of the IC is hot but is this going to be valuable info? Or are you suggesting some external part of the circuit may also be heating up and that could be the root of the whole failure?

I believe, based on your description, that there was an issue on the board and when increasing the power available to the IC it was able to destroy itself. I am suggesting that the failure was on the board external to the IC. The internal short was likely a secondary failure caused by the first.

Using continuity (with audible beep) mode on my multimeter, probing at the nearest connected component (e.g. C4) on the same net so that I don't have to get the probe on the very fine pitch IC pad. I skipped checking for shorts between a pin and adjacent NC pin, maybe it's not a safe assumption that those aren't internally connected to something?

I would suggest using resistance mode just in case you have a higher impedance short that would not give you an audible continuity beep. NC pins are defined as not carrying a bond wire to the die so I believe you made a good assumption not checking them.

I would suggest trying again. Bring it up to 4 volts and check some waveforms, I am hopeful with that information you will find the issue and be up and running 

[davegravy]

I would suggest trying again. Bring it up to 4 volts and check some waveforms, I am hopeful with that information you will find the issue and be up and running 

I don't have a great scope but these were all taken with my 10X probe. Does this suggest there's insufficient input filtering?

Vin at 3.6V RMS


Vin at 3.7V RMS (where the IC starts drawing current)


SW1/2/3 (Vin=3.7V RMS)


Vout (Vin=3.7V RMS)


Cboot (Vin=3.7V RMS)


SW4 (Vin=3.7V RMS)

[__steven]

I notice two things. One, Cboot/SW4 indicate a 50% duty cycle gate drive, however, we do not see the 0 to 3.7V pulse at 50% duty cycle that is expected at SW1/2/3.

The other is that indeed the input voltage rail is collapsing which likely leads to the IC turning on and off because of it's 3-4V under voltage lockout. What is your input current limit?

Can you confirm that the picture of SW1/2/3 is accurate? It looks identical to the "Vout (Vin=3.7V RMS)" picture. I recommend using one shot/single mode to remove the other waveforms from the screen. Easier to read that way.

I will say, assuming that the picture was indeed wrong, and you do see the 0-3.7V 50% duty cycle pulses at (SW1/2/3, pin 1 of L1), I think it is working fine. The output voltage is sitting at roughly half the input voltage which is what we expect with a 50% duty cycle. I cannot find information on the behavior when the input voltage is lower than the output voltage setpoint. With no load and no short, having an input voltage lower than the output target should not destroy the IC or cause an overcurrent condition.

If you believe it is safe to do so, I would suggest bringing the input voltage up and try and remove the undervoltage lockout from the behavior.

[davegravy]

I notice two things. One, Cboot/SW4 indicate a 50% duty cycle gate drive, however, we do not see the 0 to 3.7V pulse at 50% duty cycle that is expected at SW1/2/3.

The other is that indeed the input voltage rail is collapsing which likely leads to the IC turning on and off because of it's 3-4V under voltage lockout. What is your input current limit?

100mA. I think I saw 1A on the first one that I fried. If I need to go higher than 100mA, any suggestion of what's a safe limit? (I won't hold anyone responsible if I blow another  ).

Can you confirm that the picture of SW1/2/3 is accurate? It looks identical to the "Vout (Vin=3.7V RMS)" picture.

I reproduced the same waveform this morning, but read on for more context.

I recommend using one shot/single mode to remove the other waveforms from the screen. Easier to read that way.

I'm struggling getting 'single' to work on this scope. I see a blip of the trace when I hit the "reset" button on the trigger, but then the scope goes blank, ie the trace isn't persisted. The reference-manua isn't very beginner-friendly. I'm sticking with 'auto' for now.

I noticed my probe has a filter built in and looking up the model # (PM8918) it seems it's limited to 4kHz bandwidth. So I broke out a bnc cable to my own leads (6" length) confirmed I measured a clean 1MHz square wave from my function generator with it, then remeasured.

At the 0.1us/div setting it's all HF noise now. But something interesting is going on at ~1kHz fundamental:

Vin


SW1/2/3


Vout


SW4


Cboot


Vin at 0.1us timescale is just noise, the other nodes look fairly similar:


Safe to assume the noise is contained within that 1kHz "blip" that's occurring?

[davegravy]

I turned up the current limit to 0.5A, voltage to 12V, and turned it on for just a couple seconds.

The output voltage remained pegged at 4V, drawing the full 0.5A.

Returning to the previous operating condition of 0.1A current limit, I'm now unable to bring Vin above 3.5V before CC engages (previously this happened at 3.7 or 3.8V). The waveforms all look significantly different, Vout is now only 0.6V instead of formerly ~1.2V

I'm suspicious that damage occurred.

[__steven]

I was going to suggest bringing it up. Seems like the issue persists.

What gets hot by the way?

Would you be comfortable sending a picture of the board?

If you are absolutely sure you do not have any dead or higher impedance shorts on the board, the only item I can think of is the inductor saturating during startup. The one that is outlined in the schematic should be fine, 5.5A saturation current, where the high side switch current limit is 10.3A typical.

I am afraid this may be all I can offer.

[davegravy]

I was going to suggest bringing it up. Seems like the issue persists.

What gets hot by the way?

I figured I'd go for broke and removed the current limit and set the bench supply to 12V. Turning it on I was surprised to see only 12mA of current draw! For a few seconds, when suddenly it began drawing 3.2A (the max my supply can provide) and the chip started smoking.

Would you be comfortable sending a picture of the board?

Sure, will do as soon as I'm back from travels.

If you are absolutely sure you do not have any dead or higher impedance shorts on the board, the only item I can think of is the inductor saturating during startup. The one that is outlined in the schematic should be fine, 5.5A saturation current, where the high side switch current limit is 10.3A typical.

Is there a simple way to test inductor saturating?

I am afraid this may be all I can offer.

Thanks for your help.

I did find this in the datasheet which seems relevant (emphasis mine):

If the converter is connected to an input supply through long wires or PCB traces with a large impedance,
take special care to achieve stable performance. The parasitic inductance and resistance of the input cables
can have an adverse affect on converter operation. The parasitic inductance in combination with the low-ESR
ceramic input capacitors form an underdamped resonant circuit, possibly resulting in instability and voltage
transients each time the input supply is cycled ON and OFF. The parasitic resistance causes the input voltage to
dip during a load transient. If the converter is operating close to the minimum input voltage, this dip can cause
false UVLO triggering and a system reset.

The best way to solve such issues is to reduce the distance from the input supply to the converter and use an
electrolytic input capacitor in parallel with the ceramics. The moderate ESR of the electrolytic capacitor helps
damp the input resonant circuit and reduce any overshoot or undershoot at the input. A capacitance in the
range of 47 μF to 100 μF is usually sufficient to provide input parallel damping and helps hold the input voltage
steady during large load transients. An ESR of 0.1 Ω to 0.4 Ω provides enough damping for most input circuit
configurations.

The input voltage must not be allowed to suddenly fall below the output voltage. In this scenario, such as
a shorted input test, the output capacitors discharge through the body diode of the internal high-side power
MOSFET. The current is effectively uncontrolled during this condition, possibly causing damage to the device. If
this scenario is considered likely, then connect a Schottky bypass diode between the output and the input supply.

With no load I have no load transients but I have a couple chained sets of dupont wires from my bench supply to the input pads on the board. I think that's probably a few ohms of resistance. Is it worth putting an electrolytic across the input pads, and bodging in a schottky?

Another oddity I found. Highlighted is the device I'm using, which shows Fsw fixed at 2.1MhHz



The Webench design (link) shows a 412kHz Fsw

[DimitriP]

The Webench design (link) shows a 412kHz Fsw

The simulation link no-worky.

I tried WEBENCH using the highlighted part   LM64460CPPQRYFRQ1   
and it won't accept less than 6 V input.

WEBENCH also claims "simulation not enabled for this design"

So how did you get it to simulate to get the 412KHz Fsw ?

[Vovk_Z]

Replace the IC. Don't use too low current limit. Set at least 2-4 Amperes, so it can charge capacitors fast and start fast.

[davegravy]

The Webench design (link) shows a 412kHz Fsw

The simulation link no-worky.

I tried WEBENCH using the highlighted part   LM64460CPPQRYFRQ1   
and it won't accept less than 6 V input.

WEBENCH also claims "simulation not enabled for this design"

So how did you get it to simulate to get the 412KHz Fsw ?

Strange, I used the "share public link" feature in Webench. Try this link instead perhaps: https://webench.ti.com/appinfo/webench/scripts/SDP.cgi?ID=3856836FF479F532

My design also says "simulation not enabled for this design". I assumed that TI has not programmed simulation capability for this IC, similar to how they don't have SPICE models for all their ICs. Do you think instead this is a sign the design is bad?



This is where I got the Fsw = 412KHz from, under "operating values"

Indeed I'm not operating at 16V, it's designed for a minimum of 8V input, but I can't get to 8V due to the massive current draw (and resulting IC failure) at anything above ~4V.

[DimitriP]

Something has gone haywire with "something".
max current is too low (0.7A),
the Fsw is wrong, since the 5V fixed part doesn't have adjustable Fsw

Here is what mine looks like:
https://webench.ti.com/appinfo/webench/scripts/SDP.cgi?ID=5139B2DC7011C355
Fsw 2.1MHz, max current 6A @ 5V




 

[davegravy]

Something has gone haywire with "something".
max current is too low (0.7A),
the Fsw is wrong, since the 5V fixed part doesn't have adjustable Fsw

Here is what mine looks like:
https://webench.ti.com/appinfo/webench/scripts/SDP.cgi?ID=5139B2DC7011C355
Fsw 2.1MHz, max current 6A @ 5V

You can change the designed max current and vin, when you do that you should see it change the Fsw to a lower frequency.

So does this mean Webench is bugged because it's adjusting Fsw on an IC where you can't do this?

I downloaded the excel-based design tool for this IC and the efficiencies that are achievable with this design at low current at 2100khz are abysmal compared to the 97% that webench boasts.

But then that doesn't jive with the "high light load efficiency" claimed in the datasheet

[DimitriP]

So does this mean Webench is bugged because it's adjusting Fsw on an IC where you can't do this?

Yikes.
I tried lowering the max current and  Fsw does get reduced. 
I'd stick with a high enough current that keeps WEBENCH Fsw to 2.1MHz so it's running in "sane" mode  
I don't know how many good parts you have left, but it'll be intersting to see what it does if you change L1 to 800nH.
And keep Vin above 5 Volts

[davegravy]

I decided to assemble another board from scratch and this time "let her rip"...ie 12v input and no current limit. I also added a 2200uF bulk capacitor on the input and ran with a 7ohm power resistor load.

Same thing, basically - my power supply dropped to 4.5V and the IC started smoking immediately.

Here's some after photos, also attaching the before photo

EDIT: one difference is that the SW1/2/3 node isn't completely shorted to GND this time, although the resistance dropped considerably (from 150kOhm to 1.5kOhm)

[davegravy]

Is there some way I can bring this board up more incrementally?

E.g.

1) solder just the IC and C5, apply power, check for overcurrent/overheating, check for nominal VCC
2) solder C6, apply power, check for overcurrent, check for ? switching waveform on SW1/2/3
3) solder inductor and remaining components, check for overcurrent

Or is the issue almost certainly not going to present itself until the last step when there's inductor current?

[BennoG]

I had a similar problem with a buc converter that had only ceramic Capacitors on the input filteren.
It blew when connecting the power with the bench power supply already on.
And worked fine when first connect it and then turn the bench power supply on.

The problem was ringing of the wires and the ceramic capacitor because of the very low ESR and 50cm wires from the PSU to the board.
The ringing was double the allowed voltage of the buck IC.
Solved it by putting a 1 ohm resistor in series with the ceramic capacitor, and put a electrolytic capacitor in parallel with that.

Benno

[davegravy]

I had a similar problem with a buc converter that had only ceramic Capacitors on the input filteren.
It blew when connecting the power with the bench power supply already on.
And worked fine when first connect it and then turn the bench power supply on.

The problem was ringing of the wires and the ceramic capacitor because of the very low ESR and 50cm wires from the PSU to the board.
The ringing was double the allowed voltage of the buck IC.
Solved it by putting a 1 ohm resistor in series with the ceramic capacitor, and put a electrolytic capacitor in parallel with that.

Benno

Thanks Benno. You don't think the electrolytic would have sufficient ESR on its own, you needed to add series resistance to the ceramics?