D2 and C4 are much more important than you are giving credit.
In simplified terms (perfect components and no power loss) with 5V input and 12V 1A output. 500 KHz operation so 2 uSec period:
First:
The perfect inductor has 0V AVERAGE DC volts across it. (Most basic guiding principle).
Perfect capacitor has 0 A AVERAGE DC current through it (Another most basic guiding principle).
The IC internal switch transistor is switching on (0 V across it) and off (0 Amps through it.)
When the switch is ON, the inductor has 5V in one direction. When the switch is off, it has 12-5=7 volts across it. The switch ON duty cycle must be 7/12 = 58%. (I do not need the formulas but you should work this out. Let me know if this does not make sense).
The switch OFF duty cycle = 42%. D2 and C4 are being WHACKED (hit hard and fast) with very fast current pulses at a 42% duty cycle. If the DC load current is 1 Amp, then these pulses must each have an average value of 1/42%=2.4 Amps. This is ignoring inductor ripple current.
The high current switches in nanoseconds, and any layout inductance (mostly in the tracks) causes AC voltage spikes where you do not want them. That is almost definitely your biggest problem. Volts=Inductance*Amps/second
If we assume continuous conduction in a boost converter (look it up) then the inductor current is ramping up for 58% of 2 uSec = 1.16 uSec. Then we will have peak-peak inductor ripple current = 1.16 uSec*7V/6.8 uH =~ 1.2 Amps. If the inductor DC current = 2.4 Amps, that means that it has a peak current = 3.0 Amps and a valley current = 1.8 Amps.
C5 will have more internal inductance and resistance than C4 (if C4 is a multilayer X5R or X7R ceramic) so C5 function is more for added stablization and filtering.
Obviously ChatGPT fooled you:
The device model (to be attached to the symbol) is enclosed in a subcircuit
.SUBCKT PAM2423AECADJR IN OUT FB SS COMP EN GND
...
.ENDS PAM2423AECADJR
But this subcircuit contains an inductor L1, a rectifier diode D1 and a capacitor 100u. Inside of the IC, as element of the PAM2423!? This is of course not possible. You cannot integrate such devices in an IC.
The model given may be the model of a module, containing all circuit elements. But then your circuit schematic is wrong, as you add the cited devices externally (a second time when using the model attached the symbol).
Could you please put a note in the “PAM2423AECADJR.lib” file that is found in your github stating that this subcircuit is created using ChatGPT.
When someone tries to use this subcircuit or another is asked to solve the troubles it generates there is some hope that the note about the ChatGPT hallucination is found. Reread the remark above, made by Holger.
Prepend something like this on the first lines:
*
* ChatGPT generated subcircuit
*
* SPICE Model for PAM2423AECADJR
.… rest of the file …
Thanks for all the input, I’ll try and make a better circuit and I’m learning more each time I open KiCad.
As for the simulator, I think it makes more sense when manufacturers like Diodes Inc. who make the PAM2423 update their spice models to support these parts and not rely on novices like me trying to make it up.
The other take away from this thread is there are a lot of things the may affect the circuit outside of the ERC and DRC, and even the simulator. Not a problem for seasoned engineers, but still a challenging area for novices.
At the end of the day, I’m in half a mind about weather this circuit will work, I value the inputs from everyone in this thread, but I have no “evidence” to support it working, and there’s a high chance sending it off to PCBWay or similar will be a waste of time and money and end up with more e-waste.
My hope was the simulator would give me more confidence that the design would work, but I don’t think I’m going to get that.
SPICE model availability is really variable - and available models are of varying quality. Just another thing to contend with unfortunately.
On your circuit in particular, if you follow the datasheet (which is a skill in itself), and the advice here, it really should work. Looking at your schematic, you’re not doing anything crazy. It’s a shame the datasheet doesn’t show recommended PCB layouts, as many datasheets for sensitive parts do. But, with the advice in the thread, and the posted example layouts from similar circuits, there shouldn’t be any issues.
I’d chalk it up to a learning experience, which does sometimes cost time and/or money. I think everybody here has had failed boards, blown up components (including expensive ones, which is always annoying), and learned things the hard way. Embrace it as all part of the process!
I would like to add that SMPS are not beginners circuits and do need experience to get right. I had a few noisy and failure prone boards before designing quiet and reliable designs.
I also have the necessary test equipment.
The manufacturer (Diodes Inc) of this IC (PAM2423) does not publish a SPICE model, so when you find a model you should look carefully to see what characteristics the model’s publisher tried to model, and how comprehensive the modeling is. It is often, for example, a waste of time and effort to look for subtle operational anomalies if the device model is created from ideal components.
In addition to what “holger” said about modeling energy storage components inside the IC:
this model for the PAM2423 uses a P-channel MOSFET, while the datasheet’s block diagram shows an N-channel device.
In short, I have very low confidence in this model’s ability to return useful information from a simulation. My time would probably be more profitably spent by kludge-wiring a prototype on the bench and poking around with a multimeter and oscilloscope. As the late Bob Pease once remarked, “My favorite programming language is . . . . solder.”
To be fair, Diodes Inc do seem to have spice models for some things they make:
These may help folks trying to design circuits using their parts which is great I think. Just unfortunately, there was nothing for the part I had in mind.
I did also see the write up on Electronics-Lab there using the exact part, and I thought it was pretty nice they included digikey links to the specific components they used.
In my case, I needed a different format as my board needs to fit in a 3D printed case I made. So I’m learning as I go, the first 2 variants failed, but I think I’m going to have a shot with this latest revision and hopefully it will work as I hope.
Thanks to all here for the design help, an apologies for anyone looking specifically for the simulation model we started the topic with.
Do take a good look at the Evaluation Boards and User Guides document linked above - it’s got an example known-good PCB layout you can use as a baseline.
About operating from 0V9 input: Seriously…in the context of a boost converter, that almost definitely means to RUN from 0V9 input but not to START from 0V9 input. If you are boosting to 5V then you can probably power the chip from the boost output…remember that the converter does not need to run before you get current passing to the output via the diode. The chip might start at 3V or 4V; that is probably specified in the datasheet.
