 and welcome to Escalator. Hey everybody and welcome to my desk another Sunday night, another week of engineering and fun. We're not at CPAC, instead we're at my desk doing engineering, which I think is, that'll be a better use of my time. Now that stuff came in, first up, is there any like news or updates we wanna tell people about? No, regular show schedule this week. Yeah. So, you know, see you on Tuesdays, Wednesdays, Thursdays and Fridays for all of our shows. Yeah, I think Scott and you, Thursday or Friday, I remember, we'll find out later this week. But, He's Thursday this week. Thursday this week. Okay, well let's jump right into some of the stuff that I got. So, let's go to the overhead and I'll just do like samples and stuff, like things that we came in. Okay, first off, we actually got some compute boards in. This is the Raspberry Pi 4 compute module boards. It's pretty cool. They're in stock but I also grabbed them for myself because I thought this would be really neat for like, you know, if you wanna do like development with like PCIe slots or like, you know, multiple displays or cameras or like, it looks like there's a real-time clock module, chunky HDMI. This is a kind of nice little dev board that I might use for, you know, sometimes when I'm like doing Raspberry Pi projects and then you plug modules into them. So, this is a little module. We also got one of the ones with Wi-Fi. So, I'm gonna very carefully pry this up. So, this is the 2 gigabyte with Wi-Fi module. So, these are starting to come out. I think we only have the non-Wi-Fi in stock. I think we only got a couple of these. So, I snagged one. But it's got the onboard flash. It looks like, so the MMC Pi 4 chip, Wi-Fi Bluetooth and little antenna dongles. So, you can either use a built-in antenna or there's a, you can see the built-in antenna or the dynamic antenna switcher and use a UFL. So, good, good stuff there. Good stuff for the audience. Yeah. Did you do anything with that ESP32 from last week? No. I haven't sent out PCBs, but I wanna finish that ESP32 QT Pi and send it off. I've been actually focused on RP2040 and a couple other things that just business things. This part of the year when you run a business is quite busy because there's taxes and end of year stuff and reporting and it's just, it's a little busier. I usually, my time sort of starts freeing up a little bit in April and May. So, you'll see, not as fast on the hardware in the first few months of the year, but it's okay. Plenty of time in the summer, do more. So, this is the Compute IO board. So, cool stuff. Check these out if you're interested in doing compute module hacking. These will be available in the shop soon. Good, good, good. All right. Next up, I got some cellular modules. I thought I'd just show these off. So, these are BG95s. These are Quectel modules. So, I used to use like Simcom only modules and I've also used Ublocks. But Quectel offered me some samples and I'm not gonna say notice of samples. So, this is, you know, I got the BG95M3 and the 95M1 which is pin compatible, but I think one is probably Cat M1, one's Cat M3. We got the EG91, I don't know the difference. I'll check those out. And what's cool is that these come with a removable top. So, this is the module. And then the bottom has kind of, you know, your funky cellular module pinout. These aren't too bad actually. I mean, like the pitch is wide enough that I've never really had a lot of difficulty with these, you know, unusual module pitches like the pinout collection. You know, there's plenty of space here to, you know, have via, so you like these pads, you can via them down to another layer and these pads come out. So, it's actually a pretty easy module to use. And then the tin comes off. And you can see, I think it says like a Qualcomm chipset over here and then some supporting chipsets. It's a Qualcomm module, pretty cool. So, I've seen the BG95 used a lot. And I'll check it out. I don't think, yeah, it's a little bit too big to fit on a feather, it's a little sad, but it could make for a little breakout board or maybe like a shield or something because 2G is basically over. You know, we dragged out 2G as long as we could. Now we're gonna move on to LTE. 3G was always, you know, kind of expensive in a pain, but I'm thinking, you know, let's just move straight to LTE. Maybe, you know, CAT-1, machine to machine stuff. The only thing I hope is I think that these might have an audio interface. I really wanted to have a module that could do audio because people really love making DIY cell phones or DIY phones. And I noticed a lot of 3G modules started dropping the audio interfaces. So, cool stuff. Is it 3G, LTE? This is LTE, I think, but you could look up the datasheet, you know, if you wanna be sure. I believe it has 2G fallback and I think this is an LTE module. So, BG95, so cool. All right, so this is good stuff. I wanna get back into doing some cellular. I know I haven't done cellular in many, many years. It's always a little bit of a journey. Okay, so what else? Okay, next up, we got Feather Takes Flight with the Feather RP2040. Give this silk screen some. You can do it. Okay, give the silk screen some love because Philby did a beautiful job. These are the, this is the array panel of the RP2040 Feather. This is the logo that Mr. Lady Aida thoughtfully asked for and they're like, oh yeah, we should have a logo. And so they provided a nice silk screen friendly logo. You can see a little pie logo in the RP2040. So, we put that on the back. The back has the GPIO pins. So, you know, we wanted to make this feather compatible as much as possible with our existing feathers. So, on the front side, we use not GPIO pin numbering, but digital pin numbering. And this is sort of similar to how Arduino does, the original Arduino, the 32U4 and the 328, pin five was pin five. But then what's cool about when they moved on to the SAMD 21 and other chipsets is they started to have like a pin table. So, pin 13 wasn't necessarily GPIO 13. There was like a remapping table that would let you map a GPIO number to an underlying pin definition. And this was handy because if you wanna have your boards have pin 13 red LED, right? That's kind of a standard Arduino pin number for the red LED. You may not actually have a pin 13 on your chip. Like some chips don't even have 13 pins, right? Or that pin 13 might be used for something else. GPIO 13 might not be available. But with pin remapping, which is a wonderful invention, any pin can be numbered anything. So, what we did is in circuit pipeline, we definitely have pin them, remapping. Like GPIO 13 on this board happens to be 13, but like five is not. I'll explain in a moment exactly why. Five isn't GPIO five, but in circuit Python, we do that remapping for you. So, when you interface with D5 it's gonna be, I think it's GPIO seven or eight or something. But for those who are gonna be using this with Pico SDK and the Pico SDK, I think they haven't added pin remapping yet. You have the low level pin numbering. And so, the reason, if you're like, well, why aren't the pins in exactly the same order? The reason we did that is we wanted to have pin 13, these eight pins here, 13, 12, 11, 10, nine, eight, seven, and then over here, six. We wanted them to be in a row. And what I mean by that is the eight free GPIO pins, not I-squared C, not UR, not SPI. They are consecutive GPIO pins because the RP2040 PIO state machine system works on consecutive pins. And so, the most number pins you can work on is eight. And so, this means that there's, if you want to use the PIO for some funky like data pushing thing going on, you would be able to do it and have as many pins as possible in a row. So there's like, if you're trying to like say, drive eight neopixel strands at the same time in a row, you can do that. But I didn't, you know, I wanted to make sure that our existing feathers and feather wing examples worked well. And so, oops, that's why I am. We made the pin numbering up here, you know, five, six, nine, 10, 11, 12, 13, that's kind of the standard feather pin numbering order. So that explains that. So that's why the pins are different on the other sides. Otherwise, this is a nice board. You know, what's nifty is that even though the RP2040 is a 0.4 millimeter pitched part, really haven't had any issue with assembling it even by hand. So, you know, I put together, you know, you can see there's two missing. I put together one of them over here to test it out. And it's just running a neopixel, you know, swirling demo and blinking the red LED. So once we have the board designed, you know, as people who watch the show know, I have to make the tester. And my favorite testers are teensy-based testers. We do use Raspberry Pi computers for testers sometimes. Cool thing about Raspberry Pi is it's a computer. So you can plug a J-Link into it or run OpenOCD, you know, like run software, software. That was a question. That's what I was asking that all, Okay, why don't you ask the question? Yeah, will the feather be debuggable with OpenOCD and GDB like Pico or we need a different technique? You can. I didn't break out the SWD pins like on GPIO pads. They are available on like the bottom, you know, there's two pads down here that you can barely see it says clock and data. These are the SWD pads. But there's also, you know, if you are somebody who really wants to do debugging and if a lot of people demand it, I'll do it. There's an SWD connector that you can solder in place and that will use the standard SWD connector. You know, why not have it broken out into pads? There just wasn't room. Like I had to choose what I wanted to have. And so in the end, I was like, I think an SWD connector is more useful. Time will tell. Look, if it turns out absolutely nobody ever uses the SWD connector and they want break, you know, breakout pads, I'll re-spin the board, right? No big deal. There'll be more room even because the SWD connectors is quite large. And if people really demand it, I'll make a version that has that part placed. But for now, I'm leaving it blank because honestly, I don't, not a lot of people do SWD debugging, right? It's for some engineers and for those engineers, if you're doing that, you're also the kind of person who is comfortable soldering on a connector. But speaking of, you know, when we make boards with other Cortex chips like the NRF 52s or 51s or STMs or SAMD series, we program them over SWD. And I was kind of planning on doing that as well. I was like, okay, well, you know, at worst I can always connect them to a Raspberry Pi and then use open OCD on the Raspberry Pi to, you know, connect to these pads and then load firmware that way. But what's interesting about the RP2040 is it has a ROM bootloader and the ROM bootloader is UF2 and that means it's a disk drive. And I was talking with TAC, who is our developer who works on Tini USB and also has been helping maintain our SWD DAP, our like, you know, self-contained programming software for Cortex boards like Latin boards, program boards. And he basically said, look, you know, there's this multi-drop thing that the RP2040 is doing and it's like, I don't really have time to look at it and it's a little more complicated. But you know what we could do is we could use the USB host capability of the Tini C3.6 and mount this as a disk drive, like literally like put into bootloader mode and then access it like a disk drive and then like copy over USB to firmware. And then we actually got that working, which was pretty cool. So here I've got my Tini C3.6 and then I've got it connected to a USB socket over here. So you can see this USB socket and then I have a USB cable plugged into it. And the other side of the USB cable is over here. And all I have to do is connect the, all I have to do is I have, you know, just for now I've soldered the boot pin because you do need to use the boot pin to get into bootloader mode. I don't want people to have to press the buttons instead I just have a Pogo pin on the final tester wired to this GPIO. And then if we go to my computer, I have here this Tini C serial monitor and I'll just reload it. And you can see how the tester works. It uses the SD card on the Tini C. It loads a UF2 file from it called 4884test.uf2 and it literally copies it over the USB interface to the RP2040, thus loading the firmware. And in this case it does it about 100K per second which is like super nice and fast. And as a side benefit this also lets me really test well the bootloader and the USB interface. It's kind of like, I like it when tests test multiple things because it's like, I'm testing programming, I'm testing the flash, I'm testing the USB, I'm testing the bootloader. Like, ooh, so much testing going on. And then the test program itself was written in Pico SDK. I normally write them in Arduino but Arduino's available yet. When Arduino's available, I'll do that. So I thought I would show the feather test real fast. Looks like Xe-Mex doesn't let you make the text larger. When you're done with that. Yeah. Which is in question. Okay, well basically the test program, what it does, and maybe I'll publish this somewhere, is it actually has the NeoPixel swirl going until it receives a character on the UART, 0XAF. And if that character is received it breaks out of the NeoPixel swirly loop thing that's going on here which is just a NeoPixel color wheel. And then does some GPIO pin testing and what I do is like on the tester I have Pogo pins that connect different pads together and I verify that they're connected but nothing else is connected. Basically just making sure each pin is connected to the pad it's supposed to be. And if so it checks the ADC and makes sure the voltages are corrective well. And then on success it does a printf and I'm done. So what's interesting is that if I have this board and I connect the RX and TX pins and I send OXAF and you saw like it starts doing the test procedure. Like by after you have to trigger it with the special character and that tests the UART and then on the receiving the special character it stops the NeoPixel and then goes into test mode. So what's nice is that I can program this test and ship this test firmware and people who are getting it aren't gonna know it's actually test firmware. They're gonna be like, oh look the NeoPixel is rolling. Ooh, like this is the demo, it like runs. I know that the hardware is good. They don't need to know that that's also the test firmware so that's our little secret. But what's nice is that it means that I can program and test everything in about 0.64 seconds per board. So the entire test takes two seconds and that's like for me awesome. I love it when testing takes less than two seconds. Nobody shows how anything is tested, how any of the software is made, none of the things that you need to do testers make testers, we do that. And we've been asked in the past like to worried about your competitors being able to test things better, easier. And I say that is exactly why we release it because what if we all got better at this, what if we all shared stuff but someone has to go first. So that's the good stuff. So basically taking, getting rid of the raspberry, I love my favorite computers. But believe me, if you're making testers you want them to be a solid state as possible. You don't want Linux involved if you can avoid it. This is just a drop down and test in three seconds. It's gonna be awesome especially when we do the QT pies. I wanna keep those as cheap as possible. And test time is that counts towards how much it costs to manufacture something. It's how long it takes to test. If it takes three seconds, that's awesome, right? That it's a lot less expensive than two minutes. Okay, you wanna do some questions? Yeah, let's do some questions. All right. Yes, that's the RP2040 Feather from before. Yes. Next up, what's the diameter of the mounting holes I guess in the previous boards? I think they're 2.1 inch. Okay, so just in the look at the Blackmagic Probe SWD header, I guess they make smaller ones. I know that there's people who do like funky, cool, weird, like, ooh, it's tag connect. Okay, great. Classic 0.05 inch, two by five. Never go wrong with it. If you can fit it, I put it on there. Next up, does it aid for your testing program? The boards one by one? Yep. Yeah, we do. Every single board is hand tested. And how does the pick and place machine pick up one of those SWD connectors? Is there a tape over top like some of the USB and an insert? Looks like Jeff posted a photo, there's a little cap. And if you actually have a, if you have a board where we don't use the SWD for testing, you'll see the cap is still on. Let's just remove it. Okay. Okay, so. Longer question, don't you wanna leave this one for the end? No, no, this is just, let's get to it. Lemour talked about failure. It seems like it's a really important part of making, creating and Lemour makes it, making it seem really easy. It would be amazing to see her process for working through something that has had her stumped. Yeah, I don't, I don't usually do those kinds of streams. Mostly because they just take a little too long. They're actually really hard to do. So I usually try to get a bunch of things set up for Descaladiata. We do do blog posts about debugging process and at Scott's videos streams that are two hours long going to, you know, deep development and debugging at the same time. If you want to see like the development process where I make like mistakes, decisions and then I changed my mind, check out the Make It a Market series. We did a 10 video series where like I go through decisions, some of which are good, some of which maybe I made mistakes. Okay, any chance of a RP2040 with a BTLE ice grid system on board? You know, I think, you know, you could pair the RP2040 with an ESP32, you know, co-processor and us thinking about doing that for like my Metro design, but I just haven't had time to keep working on it. Yeah. Yeah, a couple of people honestly to test code, so if you want to post up there. Yeah, I'll publish, I don't, I'm not going to have time tonight, but I will publish, I will publish it somewhere. Okay. For sure. So one thing that I still have to work on, like, you know, I discovered this like today is my prototypes used a two megabyte flash and I was like, oh, you know, for the final version I want like four or eight megabytes. And then I realized that you can't just swap the QSPY flash turns out that each, I knew that each QSPY flash chip is different, but when you program in the Pico SDK, you do pick the, trying to remember the secondary boot stage, stage two file. And the stage two file is what defines, it's really annoying. Like QSPY flash is like almost completely identical from board to board, but like the status registering, the Q enable, the quad enable bit, it's like different and whatever, like you actually have to like do stuff to make it work. A little frustrating because it's like they're supposed to be, they're all pin compatible and they're kind of functionally compatible, but the details of them are not compatible. So turns out, you know, you have to set it into generic mode. And even in generic mode, the four megabyte didn't work. So bonus, the feather's gonna ship with eight megabytes instead of four, because I just don't feel like sitting down and figuring out why the four megabyte chip isn't working. I'll figure it out later. But for now, I think I'm just gonna bump the feather RP2040 up to eight megabytes because we stock the eight megabyte. I know it works. And you know, I got, you know, the test works under eight megabytes, but that was definitely, if you're working on your own hardware, watch out for like the Q-Spy flash chip. It isn't like drop anything you like. The default files do work with the GDQ25, which GD25Q32, which is a two megabyte giga device, flash is like one of my favorites. It's on a lot of boards. And so luckily, I think our default flash was also worked with the default files, but turns out that's not true. So the whole play I'll just save you like the two hours that I lost trying to figure out why the heck, you know, one board with two megabytes was working, but the board with four megabytes was not. All right, so that's, so expect the feathers, you know, unless something tragic happens, we'll at least be manufacturing them this week and they should be in the store soon. Especially if Scott can help me get the eight megabyte working in, circuit pipeline. Okay, and then last but not least, oh, we also got the, oh, can you go back to the overhead, please? We got the, it's easy to put these together. So this is Rev B, but I think this is the final rev. They need a nice silk screen, but these, these do have like, you know, the quote again, quote unquote standard, it's a bit see pin out, which means that the bottom has the SWD pins broken out like this. So nice little ground plan, not too bad, right? You know, good, nice, nice smooth continuous ground plan there as much as possible. And so yeah, this, this, this QD will also get wrapped up shortly, two buttons over here, got the little NeoPixel RP2040 in the middle, used by Flash over there, lots and lots of GPIO. And you know, if you want lipo, you can put a lipo backpack on top. And then the last thing I worked on is the Neo Trinkies. I don't know if people remember these from like a couple of months ago, but I finally just- Oh, well, we're here. Are we gonna keep the micro USB on the ITS-E, or are we gonna go to C? Yeah, you know, I want to keep it micro USB. And the reason being, all the ITS-Es are micro USB, and I kind of wanted to maintain consistency. I mean, I guess I could change it, but I sort of like, I kind of wanted to keep it the same. So like physically they're the same and use the same connector. The feathers I'm not as picky about because the feather, like so many people are making feathers that there's no guarantee of any component height or connector end. And people were starting to make USB-C, but for the ITS-E, I'm still kind of heading towards micro B land. I think I'm gonna stick with it. We'll see, you know, I can, I can always respin the board, right? That's something you have to keep in mind. So the Neo Trinkie, we want to make a very low cost circuit pad on board with four Neo Pixels and some capacitive touch pads, I was like inspired by like the FOMUs and the COMUs and the TOMUs and all those OMU boards where they're just like little USB keyboards with a little capacitive touch. So this is a split touch. You can kind of see here, there's a little line in the middle. So there's two pads left and right or top and bottom. And they're split, they're split in the middle. You can kind of see. So it'll be gold, not silver, but you can see how the middle is disconnected, but there's a little key chain slot. And then these are, standard thickness, so I just blobbed some solder on them to make them thick enough to work. And then one of these, yeah, so I got, this one has just, I think, green LEDs. One of these has a circuit Python demo on it, but I don't remember which, maybe neither, sorry. But you can kind of see the, maybe I broke them. I'm gonna ask. Anyways, so yeah, they got two capacitive touch pads. You can see the Neopixels, like now they're green, but they can of course be any color. Got a reset button, a little regulator, just enough capacitors and parts, like the minimal parts to get this SAMD 21 up and running. See if I can figure out, this is in bootloader mode. And use this one then. Oh, sorry, so this is the demo. This is one that can change the brightness. And then on this pad, when you touch this pad, it changes the color. So you can see a little color swirl and then this is like right. And then back down to not so bright. So this is a little color demo, just showing the two capacitive touch pads. So, you know, I just showed, you know, a bunch of RP2040 stuff. You know, why, why didn't I use the RP2040 for this board? Why stick with the SAMD 21? Well, I mean, when you want to talk about like minimal parts, not cost, right? Because the cost, you know, cost is, includes a lot of things, right? Manufacturability and, you know, access components. So, you know, the RP2040 does need a crystal and it needs external flash. It needs, you know, a lot of capacitors worse. What's nice about the SAMD 21 is, it's not as fast, but it has like flash built in. It can, has an internal resonator that can be phase up, looped off of the USB, so it can be more precise, process stuff for neopixels, which is all I really care about. It doesn't need like a pull up on the reset. There's a lot of parts that you can leave off on the SAMD 21. It's like a very like, happy with the least type of part, which I think is cool. And so, you know, there's place in my heart for both of them. And speaking, which I thought, you know, people will probably want to use the SAMD 21. They might be asking, well, there's so many different kinds. How, which one, how do I know what you want to pick and how do I find them on Digikey so I can include them in my design? So that sort of leads us into the great search. All right. Every single week, Digikey, I need a free, bring you to the great search. Lady, what is on the great search this week? Okay. So this week, I just showed off my Neo Trinky, it's a little SAMD 21 based board. I still really like the SAMD 21. It's a great chip. It's used, you know, in Arduino and circuit Python and Rust, Zephyr, free RTOS. I mean, like you can program the MicroPython, you can program this in like so many different languages, a really nice, solid chip. And it can run with really minimal hardware. I thought it would be cool to show all the different variants you can get on Digikey, how to select between them. I don't think we've done a microcontroller yet. So I thought it would be, this would be a good first microcontroller for the great search. So let's go over Digikey. And I'll do our search in. Okay. So, let's go around. Okay. So, the SAMD, so there's the SAMD 21, and then there's the SAMD 51. The SAMD 51 is the Cortex M4 family. It's kind of a different chip, although it shares a lot of the same peripherals. There's also the SAM C and SAM L series, which I won't be talking about. The SAM L is lower power, but it's very similar to the SAM D in many respects. The SAM C is interesting. I think the C stands for CAN bus. It's a five volt version of the SAM series, but it doesn't have USB. So just watch out for that. Like if people are like, oh my God, there's the SAM C 21. Why don't you use that? It's five volt compatible. Yeah, but it doesn't have USB. And one of the things I like about the SAM D 21 is it has USB. So let's look for at SAM D 21. So of course we sell tons of boards that have the SAM D 21. So you can get OLEDs, breakouts, whatever, evaluation boards. You'll get your feathers there. But we want the microcontrollers. Okay, so as usual, let's only look for active parts. And note that again, some of them don't have USB, maybe, I don't know. But the thing that you really want to look for, there's only three different things that you can pick. The size of the chip, the amount of RAM and the amount of flash. Now, like most microcontrollers, they're kind of tied together. The lowest pins usually also have the lowest flash and the lowest RAM. And usually you can't independently pick flash and RAM. Like the more flash, the more RAM. You can't have something with a ton of RAM and a little bit of flash. It doesn't work that way. They kind of go to lockstep. So it's just as long as you're aware of that. So you can pick the amount of RAM and program size, program flash you want from 32K up to 256K. And you're pretty much paying for those three things. You're gonna pay for more pins, you're gonna pay for more flash and you're gonna pay for more RAM. So a common thing is you design with the biggest version of the chip, figure out how much flash and memory your project took. And then when you go into production, you just round up to the closest one that will fit everything you need to do. But having it all be pin compatible makes it really easy to move between the families. And then of course, if your product or your project grows and increases, you can always start bumping it up. So that said, another thing to look at is the package and device. So in my case, I wanted the smallest, physically smallest one. So the WLCSP is gonna be literally the smallest one. This is tiny. It's only two by two millimeters. That said, it only has 32K of flash or 64K of flash, these two versions. And that's not enough for me. I actually want to be able to circuit Python on these. I need at least 32K of RAM and 256K of flash. But it's so interesting to see. Microchip, they kinda always have a teeny BGA version and then QFN and then go up to QFP. I kinda like QFN the most. Like if you had to pick one package, I like QFNs. Why do I like QFNs? Well, QFPs, they have more pins, but the pins get bent sometimes, especially during rework. I've always had the most yield success with QFNs. But talk to your manufacturer to decide what package you want. So of the QFNs, there is a bunch of different versions. There's the 21Gs, the 21Es, Js, 18s, and all those characters, you know, stand for different variables that you can adjust. So the E, when it says SAMD 21E, the E stands for 32 pin. And then the G stands for 48 pin and the J stands for 64 pin. So that's the biggest you can get at the SAMD 21s, the 64 pin chips. So, you know, physically the larger they get, usually the more pins you want. So you can pick the E subcode if you want just the smallest, physically smallest QFN. And then the digits afterwards, like the 17 and 18 and 15, those stand for, this is 15, the 15, those stand for how much flash and RAM. And like I said, the flash and RAM on these chips kind of go in lock steps. So let me close my, actually let me just open up the datasheet. In the datasheet, you will see a ordering information page and you'll wanna use this, this again tells you, pin count is this letter code and then flash memory is the number code. So in our case, we want 18, we want the most flash memory and we want E for the smallest size. So smallest, sorry, I'll scroll up, smallest, physically smallest, but memory wise, largest. And then there's Silicon revisions. You know, there's variants, but basically you're only gonna really see the A variant unless you, I guess they made them for some special customer or something. But in our case, the variants, we don't really see any other options. And then the dash MUT is for the package, again, QFN, QFP, or BGA. And then the package grading, how it's packaged and the temperature grade and then whether it comes in tape or in tray. So, but that doesn't matter for me as much. So again, I want to have the 256K of flash and I want the smallest number of IO. And that's what I got. So I've got a couple of different options. The MFT, you know, you're all SAMD 21E 18s, but they come in MFT, MU, MF, MUT and MF, right? So you're like, what's the difference? Go back to the datasheet and you can quickly tell, okay, they're all M, so that's QFN. And then the U or the F just stands for whether it's 40 to 85 or 40 to 125. Basically, is it industrial or is it commercial temperature level? You probably don't care, or at least I don't care, right? It doesn't matter to me whether I'm getting industrial commercial temp. So that's when I would basically put in, you know, okay, let's see the pricing at 1,000 pieces and I just picked the cheapest, right? Which in this case happens to be the commercial temperature range tray. This is the one I like, the MU series. But again, you can also get the MUT, which is like, you know, basically the same price. It's like six cents more. The nice thing about this version is it comes on tape and reel, which you might be beneficial for you depending on your manufacturing. Our machines take tray or tape very well. They don't like tube. So just, you know, if you go into manufacturing, just check with your manufacturer to see how they want the chips package. But that's it, the MU series. So this is the chip I like. I use this in the Gemma's, the Trinket's, Neo Trinky. This chip is awesome. It's really easy to use. It's easy to load it with our UF2 boot loader and there's boot loader protection. Once you put the boot loader protection, it's pretty much, you know, nearly indestructible. It's very rare that we've seen people frazzle their boot loader. It's nice and solid. And then you can load Arduino or circuit Python. And, you know, for the size and price, it's quite powerful. It's 48 megahertz, 256K of flash, 32K of RAM. So good enough to run some basic circuit Python code or quite a bit of Arduino or Rust or Zephyr code. So that's my great search at SanD21E18. All right, just to discover the idea, there was one question that was in the other chat. Was that the, can you show the Trinket? No, they joined late, they want to see it. Trinket. So it's got USB. It's got blob, solder to make it thicker because I got, you know, the non-thick PCBs. It's got the SanD21E18, a little power supply section here, capacitors. We have some of them. Love it in this store. It's either coming soon or it'll be for sale soon. Yeah, in the four NeoPixels. So, Neo Trinky. I don't know, he's a little catchy theme song. Yeah. Okay, that's all I've got. And that is Desk of Lady Eta for this week. We'll see everybody during week on our various shows and more. Thanks everybody.