https://community.element14.com/products/pcbprototyping/Join members as we design, prototype and fabricate printed circuit boards, for fun and for practicality!en-USTelligent Community 12https://community.element14.com/products/pcbprototyping/b/pcb-blogs/posts/reprogramming-an-air-fryer-for-reflow-solderingSun, 29 Mar 2026 16:35:00 GMT93d5dcb4-84c2-446f-b2cb-99731719e767:45d773ed-f91d-4f41-95b0-0b7ed9af3e52Fred27As was browsing EEVBlog a while ago... (I know. It's great here, but I sometimes like go look elsewhere too.) Anyway, I was browsing EEVBlog and came across an interesting post where someone had taken a Ninja SP101 domestic air fryer / pizza oven and replaced the PCB to use it for reflow soldering. https://www.eevblog.com/forum/projects/ninja-smt-reflow-oven/msg6103711/#msg6103711 I already have a small mini-oven that's been adapted with an external controller. However I liked the way he'd gone about this - especially the way the original display and controls had been retained. I put a search for the same model on eBay and thought that it might make a nice project if a cheap one came up. Well, what do you know - a brand new one that had been dropped and dented came up for only £25, so I went for it. I fully intended for it to ba a back burner project and for it to perhaps never get finished, but once it was in the gagade I decided to give it a poke. The original project was a total PCB replacement, but I was interested in a bit of reverse-engineering and to see if it was possible to re-use the existing PCB and replace the firmware. The existing board consists of an STM32F031 microcontroller. This controls some relays, triacs, fans, and an internal light. It connects to a front panel with some 7-segment displays, a rotary knob and a few buttons. I know from the other project that it uses an NTC thermistor for temperature measurement, but that this reacts a little too slowly. If this is going to work then I will need to interface with all these components and also see if I can add a thermocouple. The first thing I did was to hammer out some of the dents and get it looking respectable, if not exactly perfect. Apart from a few scars and a broken doo rhandled it did indeed work just fine. The next thing I did was to see if I can connect to the onboard microcontroller. I found an old STM32F0Discovery board from many years ago to use as a ST-LINK and decided to see if the usefully provided debug header would reveal anything. As the oven was running from mains power I ensured my laptop was running on battery to prevent any ground loops. STM32CubeProgrammer did indeed reveal an STM32f03x microcontroller but it had Read Out Protection (level 1) enabled. It looked like it was possible to reset the Read Out Protection bits using the programmer so I decided to give this a try. It worked! However I shouldn't have been surprised to see that it also erased the on-board firmware. Oh well - no going back now. My air fryer is getting a new purpose in life or it's being left for dead. My reverse engineering journey is about to begin...stm32reflow solderinghttps://community.element14.com/products/pcbprototyping/f/forum/56793/led_matrix_mosfet_driver_10x10/234625Thu, 26 Mar 2026 21:38:00 GMT93d5dcb4-84c2-446f-b2cb-99731719e767:0fc99a67-edbb-4eb1-92d9-6842efe945efbattlecoderNice. I love transistor-based project, so I like your approach. Having said that, and echoing what @vmate already said; for this amount of LEDs it's normally desirable to have a higher level control interface with more features. There's one controller I like a lot; the MAX7219 8x8 driver. It has a serial interface, so you don't need to worry about multiplexing or toggling LEDs individually, just feed the pixel data to the IC and will do the multiplexing for you. It also handles the LED current, which you can set externally, and it can also do BCD decoding if you use it to drive 7-segment displays instead of LEDs... And all of that with just 3 wires. Now, it's only 8x8, but it can be daisy chained to extend the display resolution without increasing the pin count. There are advantages to directly addressable LED matrices like yours though, like the immediate response time of each LED, the fact that you can refresh the LEDs non-sequentially, and the ability to control them with analog or mixed logic, but I think those are slightly more niche applications so I expect that most people would probably be looking for something that frees the microcontroller from toggling the LEDs one by one.https://community.element14.com/products/pcbprototyping/b/pcb-blogs/posts/kicad-10-saving-ram-pruning-librariesWed, 25 Mar 2026 18:29:00 GMT93d5dcb4-84c2-446f-b2cb-99731719e767:ebf47a12-736b-45f5-873d-8397b77c562bshabazTable of Contents Introduction How are Libraries Stored? Pruning Strategy Source Code Using the Code Summary Introduction KiCad 10, the PCB design software, has finally been released, and I've started getting up to speed with it. One thing I've noticed is that KiCad by default consumes just over 5 GB of RAM when running, and a part of that is due to the number of default libraries it loads. RAM is tight these days, due to increasing costs! I don't use all the default libraries, and have no foreseeable plans to use them in the coming year either. This blog post discusses how to prune things to reduce the RAM usage. www.youtube.com/watch How are Libraries Stored? KiCad creates a folder KiCad\10.0\share\kicad during installation. Within that, there are footprints and symbols folders, and within those are a load of yet more folders, one for each library. In other words, each library is a folder (this wasn't the case in older KiCad versions, but it is in the past few KiCad versions). For instance, here is some of the content in the symbols folder: Clicking on (say) the Comparator.kicad_symdir folder reveals files, one per symbol: There are hundreds of libraries, and KiCad maintains a list of them; two lists actually, one for symbols, and one for footprints, called sym-lib-table and fp-lib-table, respectively. They are stored in the template folder highlighted in green in an earlier screenshot. Here is what the two files look like (both are similar; this is just a screenshot from the fp-lib-table file): Pruning Strategy Ordinarily, you don't need to touch the files manually. You can create libraries, footprints, symbols, and so on using the KiCad graphical app. Using the Preferences selection in the top menu bar, you can click on Manage Symbol Libraries or Manage Footprint Libraries and work with the contents listed there. However, I wanted a programmatic way to 'remove' libraries. The method described in this blog post doesn't actually delete any library files as such; it just modifies the sym-lib-table and fp-lib-table files in the template folder, so that KiCad simply ignores libraries that are not present in those two files. With some AI help, I created a couple of Python programs. One program will read the two files and save a list of all libraries in a simple CSV format to a file called lib-list.txt , which looks like this: Then, using any text editor, I can delete line entries for libraries I do not care about and save the file with a name such as desired-lib-list.txt . Next, I run another program that will perform the following steps: (a) Backup the sym-lib-table and fp-lib-table files (b) Recreate the two files, but omitting any entries that are not present in the desired-lib-list.txt file (c) Save all the omitted entries in a file called lib-table-removed-lines.txt, just in case it's ever needed. Source Code The two Python files are on GitHub , called list-libs.py and write-table-files.py Using the Code First, I closed KiCad. I placed the Python files in a folder (in my case, I created one called C:\dev\projects\kicad_tools ) and then typed the following in a PowerShell window from that folder: python list-libs.py C:/DEV/vhd_mounts/kicad/10.0/share/kicad/template That resulted in a file called lib-list.txt Next, I opened up the text file, manually deleted all the uninteresting (to me) library entries, and saved the file as desired-lib-list.txt After that, I typed the following line (the two parameters are the desired list name, and the path to the folder containing the sym-lib-table and fp-lib-table files): python write-table-files.py desired-lib-list.txt C:/DEV/vhd_mounts/kicad/10.0/share/kicad/template That was all I needed to do. Here is a screenshot of the commands and their output: Next, I started up KiCad, opened an example project and schematic, and that consumes about 2.7 GB according to Windows Task Manager. Then I clicked Tools->Assign Footprints , which usually results in a short delay while the footprints are loaded into RAM. It was noticeably quicker. Also, Windows Task Manager revealed significantly reduced RAM usage by this stage (4 GB, i.e., a 20% saving of that resource). Summary KiCad consumes a lot of memory because it loads certain library-related content into RAM. Not all libraries are interesting, and it is possible to effectively remove unwanted libraries from KiCad visibility by editing a couple of files. Python code was written to perform the editing operations, and noticeable memory savings were achieved, as well as a noticeably quicker response when KiCad loads library content into memory. Please note that there may be bugs in the code; I don't believe there is any significant risk, but I cannot guarantee it. KiCad can be uninstalled and reinstalled if any wanted libraries are lost or the table files become corrupted. If you give it a shot, it would be useful to hear whether it worked or not. Thanks for reading!kicadkicad_10cadkicad 10https://community.element14.com/products/pcbprototyping/f/forum/56792/kicad-10-reviewTue, 24 Mar 2026 17:54:00 GMT93d5dcb4-84c2-446f-b2cb-99731719e767:5d975c07-6a17-4a63-85f9-b78a935a1142colporteurReview shows promise for kicad playing with the bigger boys. techexplorations.com/.../kicad-10-review-new-features-high-speed-tuning-variants-and-morereviewkicad 10https://community.element14.com/products/pcbprototyping/f/forum/56694/mp2172-step-down-switcher-anyone-have-experience-getting-one-going-please/233943Sun, 22 Feb 2026 23:26:00 GMT93d5dcb4-84c2-446f-b2cb-99731719e767:11687fec-2b66-44da-8616-7baa2114f78ejc2048Some other things to think about. Is the inductor shielded? A semi-shielded part may give you problems here because of the field extending to the sides and down into the tracks/plane areas, particularly with coupling to the feedback resistors. [The usual shielded part has a small airgap for the energy storage, but they put it as a ring on the top surface, so the external part of the small fringing field points up and away from the board planes and components.] Is the core material good for the frequency you're running it at? How much headroom do you have on the saturation current (remember that this is a soft ferrite, so the saturation current gets specified at 20% down on the nominal inductance value, so you really don't want to be anywhere near that at all)? How did you determine the coil value? Did you use their on-line tool or do it yourself from the datasheet? If you're running from a battery, is it single cell? If so, you'll have to be very careful to look at both ends of the voltage range since that's quite a dramatic change in input voltage. How noisy is your load? Is it something synchronous like a processor that could be sending noise back into your converter. The COT may be prone to resetting cycles early if there are noise spikes. The converter self oscillates using the voltage ripple, so there needs to actually be voltage ripple clear of noise for reliable operation, so the combined ESR of the output caps is going to be a factor. The datasheet should cover that in detail. Also, if you're calculating your own voltage divider for the feedback, try and keep the loading similar to their examples, as it probably needs a minimum load to work well - don't be tempted to increase the values dramatically [if you think noise is coupling to the feedback, you could try temporarily dramatically reducing the values of the resistors to make it harder for the coupling to occur]. I don't know that you've helped yourself with the second layout. You might have done better to experiment with the first one to determine where the problem lies - there's nothing to stop you adding wires, soldering on bits of copper foil, adding temporary shields between components, and so on. As to guidance for layout, look at the application notes that all the manufacturers of converter chips produce. They're as good as anything you'll find in books, and much more reliable than random people like me on the internet. But the basic principles are simple. Keep the input and output current loops as tight as possible, and keep the high dv/dt stuff away from the analogue feedback bits. Oh, and it's worth remembering that the designers of chips do think about layout issues and have in mind how the external layout is going to work, so look at the pinout and think your way back to what was going through the designer's head and why they arranged the pins in a particular way.https://community.element14.com/products/pcbprototyping/f/forum/56694/mp2172-step-down-switcher-anyone-have-experience-getting-one-going-please/233934Sun, 22 Feb 2026 16:49:00 GMT93d5dcb4-84c2-446f-b2cb-99731719e767:cd540e35-4982-4188-aeb1-e65187211e34davebullockmbeHI jc2048, I was unaware of the datasheet on the EV Board so thank you for this. AS you see above I reworked my terrible layout to mimic the 'suggested layout' in the IC's original datasheet, however the EV board layout is different again with components in a completely different placements. Well my new board is half way through production, so lets see if it works properly. If not you have now pointed me to what should be a 'sure fire' layout on the EV Board. Great stuff. Thanks for your help. Dave