Boost converter IC layout fails with increased load

Question

First post on here and not very experienced so let me know if there's anything wrong with my post as well as my design.

I am trying to layout a boost converter using the PAM2421AECADJR IC I bought as a prototype, but I'm having issues with the output dropping under any significant load.

Vin = 3.3 V, expected Vout = 12 V. I based the design on the recommended one in the datasheet

full size

It seems to work fine with a 100 kΩ load, but drops to 6 V with a 10 kΩ load and quickly drops to ~3 V with anything below 10 kΩ.

Parts List:

• boost IC (PAM2421)
• 10uH Inductor (B78108E1103K000)
  • Isat: 2A
  • Fres,min: 20MHz
  • Rdc: 136mΩ
• 470uF elec cap (KSY477M025S1A5H12K)
  • Vmax: 25V
  • ESR: 70mΩ
  • Iripple: 1050mA/rms (105celcius, 100KHz)
• 10uF elec cap (MCGPR50V106M5X11)
  • Vmax: 50V
  • ESR: 16.579Ω
  • Iripple: 70mA/rms (85celcius, 120Hz)
• 2.2nF (2200pF) ceramic cap (MCCC1KV222MZ5U)
  • Vmax: 1KV
• 1uF tantalum cap (MCTAR25V105KA)
  • Vmax: 25V
• 22pF ceramic cap (used as 2*22pF (parallel) = 44pF) (MCCC1500V220JNPO)
  • Vmax: 1.5KV
• 0.01uF (10nF) ceramic cap (MCCC100V103KY5P)
  • Vmax: 100V
• Schottky diode (SR1100)
  • If: 1A

This is the prototype layout:

And this is the switching node voltage with a 100 kΩ load:

And with a 100 Ω load:

The inductor is one of the B78108E1103K000s which on paper seems like it should be good enough for this circuit (saturation current 2 A, 136 mΩ DCR).

I've tried to follow as much advice as I could find in terms of layout.

• switch loop as small as possible
• separate the PGND and AGND as much as possible (but connect at a single point)

Not sure what to do from here to get it to work, so hopefully someone can give me some suggestions.

Just for clarification, in the reference schematic, the inductor is 6.8uH (not 68), it's a little hard to see. And the actual value used in the breadboard is 10uH (as that's the closest I had)

UPDATE

I managed to get the circuit working, no change to the schematic or components used, just by swapping to a protoboard and putting a massive blob of solder and bare wire together to make a large ground plane. After doing that, the output voltage was stable with both the 100 kΩ and 100 Ω loads as well as a 12V audio amplifier with good results

Answer 1

On top of what's been discussed in comments, here's my take on this.

First, based on the datasheet your 10uH inductor is a noise filter part, not a SMPS "power" inductor. Please find something that says "power" "DC-DC" or something to that effect. These are often SMPS and/or relative chunky through-hole parts Your SMPS is something like 500kHz so the inductor SFR must be much higher than that, usually it's like 10MHz and above.

Second, what are your electrolytics like? 470uF tells us nothing basically. For SMPS duty you need LowZ, Low ESR, Low Impedance type electrolytic, which all mean basically the same thing. Some kind of generic mystery Elcap is not going to get the job done. Usually for SMPS duty you choose elcap based on ripple current rating, not so much for ESR or capacitance per se, but this depends on power levels, for low voltages, high power density, high frequency you might want ceramics or polymerics. Don't confuse your design with this for now, file it for future reference. Elcap is fine for generic fiddling.

Third, what's your 3.3V supply like? Don't forget the input current is ideally the output current multiplied by the ratio of input and output and in practise there's a harsh current spike on boost circuits. Your 470uF input cap should take care of it if (a) you are not trying to be clever with grounds (b) no long wires (c) you have low impedance elcap.

Fourth, consider using a proper solder breadboard. I know it's more work than this kind of plug-in-wire board but you can actually create perfectly decent connections on a breadboard. And as a bonus, you learn some useful soldering skillz. Something along these lines:

The Vero board is one style with the entire row (or column) is shorted so you will need a scalpel (small box cutter works) to cut tracks manually but as an upside you get PCB-like copper connections. The other type is with every pin being isolated which requires you to create solder bridges everywhere.. Just look for protoboard on google.

Hint, get proper soldering iron that comes with a power supply unit that has a temperature control, even if it's a cheap one. If you're trying to make solder bridges, set the temperature too low, inadequate temperature tends to make solder follow the iron. Also, get a solder wick and a flux pen. Solder wick has bit of a gotcha for uninitiated, it works best where wetted with solder so don't try to get it to suck solder as-is, solder some on it as a starter and go from there. Ideally it'll look like a fat wire of solder afterwards.

Well as far practical soldering goes the kit still needs tweezers, wire cutters and one of those crappy PCB holder/cum rubbish magnifier glass things. Presumably you already have access to many of those items. Ideally a decent magnifying lamp thingy, in a pinch that fold-up magnifying glass thing (you want 5x or more, avoid plastic lenses and small lenses!) you'll use that fold up magnifying glass your entire career either way. Ebay is perfectly valid source for these and they cost little but remember caveats.

Answer 2

1. Solderless breadboard just won't work for switch-mode power supplies, there is too much parasitic inductance and capacitance. This absolutely must be laid out on a PCB. Check if the manufacturer has a reference design / evaluation kit / application notes. As an applications engineer (20+years at Maxim), this is the most common problem customers have trying to build boost or buck regulators.

2. That 10uH wire-wound air-core choke is not the right type of inductor. Its LI^2 energy storage is not enough, and its internal resistance is too high. You need a ferrite or iron core inductor made for switch-mode power supplies, such as TDK CD75 series. Usually these are surface-mount because of the high currents.

Answer 3

Designing this circuit on a breadboard takes way more engineering effort than laying it out on a PCB. This is a 500kHz switcher with switching currents on the order of amperes.

The primary problem I see is that “10uH” is a sufficient description for an inductor in this application. It’s not. 10uH by itself is maybe fine for high impedance signal chains like audio filters. Once you’re beyond that, you need a few more specs, and the most important one is the saturation current, or the maximum rated DC current, whichever is the lower of the two.

At the minimum, we’d need to see your actual parts list, and then we could comment on what’s wrong and why.

Answer 4

That's a tiny output current causing it to droop. I've built very similar circuits (using a different switcher chip) on a solderless breadboard without problems. You have 2 wires connecting the grounds, so that isn't the problem.

You aren't asking for an amp out at 12V, you are asking for 1.2mA and it's failing. Recheck your wiring.

Also, have you checked your input voltage? Is it stable? You don't have current limiting set on your 3.3V supply, right? You need more input current than the output current you are asking for. For a 10mA output at 12V, for example, you'll need something like 60mA input at 3.3V. Check the input with your oscilloscope to make sure it isn't drooping.

Answer 5

The #1 problem I see is the ratio of the Self Resonant Frequency, SRF, to switching frequency fs, is to low. Then when the FET output capacitance Coss is added, it reduces the ratio with a new SRF = 2MHz and fs appears to be 1/(4.6x400ns)=543 kHz or approx 4.3:1 instead of >20:1. You can see 4.3 cycles between pulses with No load (100k). This means the inductor cannot store the energy between pulses and just oscillates between pulses.

You have 2 considerations:

1: pick a better inductor with a 30 MHz SRF , instead of 6.5 but that would need to be a smaller L value than 68 uH as I only see 8MHz in 2A chokes in D-K's lineup, so it would have to be much larger than necessary (bigger winding gaps) and costly.

2: choose a much smaller L by a factor of 5 like 10 to 12 uH

So you have only the second choice as I see it.

This creates a new problem with faster risetimes and more crosstalk making parasitic jumper ESL and raises the Q if the DCR is also 5x lower. That makes the PCB layout more critical.

You know what happens when you measure a square wave on a 10:1 probe with a long ground lead. it rings at 20MHz due to the <10 nH ESL of a ground lead and the probe coax capacitance. So traces must be fat and short as this w/l ratio controls the inductance (logarithmically) from 1.5 to about 0.3 nH/mm (off the top of my head) for most PSU traces. Then you can compute the SRF from this and the Q = Z(f)/(DCR+RdsOn) for the reactance/resistance ratio.

This is why there is a tradeoff between ripple, resonance and efficiency.