#help-with-hw-design
1 messages ยท Page 43 of 1
It looks like Future has them for $4/per at 1+
I tried looking earlier and the website was down
Seems to be up, at least for me right now
Yep looks up now
Even at 10, you may end up finding that shipping costs of unifying your order outweigh per-part savings
@reef flicker what pressure range are you using?
and which form factor?
Perhaps something like this could work for you:
Metrodyne Microsystem Metrodyne Microsystem MPS-3117-006GC-A3 US$2.07 LCSC electronic components online Sensors Pressure Sensors leaded datasheet+inventory and pricing
$1.38 in 30+ quantities.
you'll need quite a bunch of analog circuitry for it as well though
LM324 for $0.08 should do it.
(per sensor)
check out the 3PEAK version, it is rail to rail:
3PEAK 3PEAK LM324A-TR US$0.14 LCSC electronic components online Amplifiers General Purpose Amplifiers leaded datasheet+inventory and pricing
hard to beat $0.08 for a 4ch rail-to-rail opamp.
those together, plus a few resistors and caps, would thus set you back around $15 for the 10 you need
something like this looks cool as well:
Goertek Goertek SPL06-007 US$1.09 LCSC electronic components online Sensors Pressure Sensors leaded datasheet+inventory and pricing
but your harmonica will store quite some spit. Not sure what you need to handle that ๐
I think that will be the hardest part of your project
making some kind of reasonable drain system, and a way to clean it.
GNDGNDGNDGNDGNDGND
Name the IC lol
I'll call him Manfred.
Good enough
SP3223?
close ๐
I had aspirations of a bunch of Arduino building modules
Then realizes how many people make them lol
hard finding a niche...
most obvious things are already crammed full of competition
This is what I most eagerly waited for
Going to build it up today and start writing the driver code for it
This one I hope to start selling mid/late January
what is it?
aha
capacitive touch sensor
I gotta finish up some USB code this week :/
Android Accessory Protocol
the STM32 USB stack code was also junk, we had to totally rewrite it.
No kidding?
yeah, was really junk
There are some people that swear by it
Iโm not familiar with it so I donโt have much of an opinion hehe
I need to get familiar with them though
perhaps it's just the way we're using it that uncovered some flaws
Makes sense
next up is firmware upgrades over USB
It probably works fine for most use cases, but fine doesnโt necessarily mean itโs great
we're using the device as an Android Open Accessory
What does that entail specifically?
it allows the device acting as a USB device, while at the same time charging a tablet connected over USB
and some other nifty things
plugging a fresh tablet into the device now, will make the tablet show a dialog "This device requires the application XXXX, please click here to download this application from Play Store"
makes the setup pretty neat.
the URL and text all passed by the firmware on the device itself
little preview of the thing in action
GUI is not finished yet, quite a bit of polish left
especially the voice detection popup should have all the commands listed
(besides the countdown spinner)
I might have to get more voice samples for the commands
I think it is a bit over trained on german accents.
I tried to get as many people as I could ๐
if you'd like to participate, I'd love it
Are you sharing this dataset?
nope
Too bad ๐
just using it for myself
additionally to this, I'm using all text-to-speech engine there are
and do various variations of every phrase. Length, pitch, intonation, background noises and so on
complete dataset for the model is pretty insanely large
I'm current doing some speech classification on Arduino but collecting the data is always a big hassle.
Yeah I'm doing those. There are also a number of free speech datasets.
yeah I know - but you usually want tons for specific words
a few thousand per word really
the data I collected through that website was pretty useful
you just get the most differing accents, microphones and whatever
The base dataset I'm using (speech commands from TF) was also collected that way.
yeah I know it
if you want really good data, you have to go Amazon MTurk.
and pay like $0.50 per participant
will still end up being $5k for 10k unique voices
and $0.50 might not be enough
would you sit for a few minutes, saying stupid words, for $0.50?
This is just for a hobby project ๐ I want my robot to wake up when I say a magic word.
I wouldn't ๐
what word do you want?
ok - so you only have a wakeup model then
no command model
"A magic word" is an interesting wakeword ๐
Yeah, just something that detects one word (perhaps two). The word is "joke", so I'd say something like "tell me a joke", and then the robot wakes up and tells a joke.
joke is a pretty decent wakeword
rather unique signature
just ensure you choose a wakeword that isn't ambiguous.
There is a reason she is called "Alexa"
Yet some people are named Alexa ๐
Alexandra
Alexander
quite different from Alexa, looking at the spectrogram
me saying "tell me a joke"
in Mel scale
So far I've been recording audio from the Arduino itself because I wanted to see how well that microphone worked. Was a good thing too because the sample code I was using (from the TensorFlow repo) was broken. Seems to work now.
if you only want it to listen to you, it is a ton easier
if you need a generic model, it is hard ๐
I wonder how many recordings of "Alexa" they have trained that model with...
true, but also: when you try to find a device/board that does what you need, quite frequently there isn't one, even though you would think it is an obvious need. Apparently "obvious" is in the eye of beholder
sure
for niche markets, that can always be the case
rarely is there a huge market gap left unfilled though
one that doesn't require significant investment, at least
and of course if there is a gap, you just fill it yourself....
yeah, if it is possible
for voice recognition, I always had trouble with these devices recognizing my Russian accent
29 years living in the US and I still have it
just makes you sound a bit scarier
Lol
Is it worth learning TinkerCAD if I'm too cheap to pay for the Fritzing download? Any major downsides?
@distant raven Was that for me? Not familiar with KiCAD but will check it out. Thx.
Itโs free and much better than tinker cad
slowly yet surely coming along with the C Driver code for my AT42QT1070 acorn
It doesnโt work yet, and I havenโt built up the board to test it on. Basically just writing as much as I can based on the data sheet
might be some smol help in my conversation with Tannewt in help with cp
oh it's C, this was about CP I2C
Yeah, I was looking at that
I am still a little confused but I've got the spirit
Well, Iโll be doing the CP soon
Which will be fun
Look at all the enum registers in the header
Lol
Ah it's been to long since I wrote any C++ it's greek to me now
It is burned into my brain for all eternity.
There was a video floating around of a 93 year old ballerina with Alzheimer's and they played some of Nutcracker and suddenly she was there for a bit.
When I'm senile, show me some C++
Ah I saw that
really inspiring stuff
I think it had to do with music being used though, so you might have to memorize an algorithm to Old McDonald or something
lol
hoping to have some pictures and whatnot to show for tomorrow for the capacitive touch board
I love those moments where you are trying to figure out the right way you want to ask a question about something and then end up answering your question before you even hit send.
Cool!
Hasl lead free and 3 coats of polyurethane!
anyone using powered vacuum pickup tool for populating PCBs?
if so, can you share which one?
aliexpress is full of cheap ones looking like slightly modified aquarium pump, and equally full of angry customer reviews saying that they work poorly...
I thought about using those little air pumps from Adafruit and rigging up a little system
@tough matrix
No idea if they would work well for it but ๐คทโโ๏ธ
Since they are for short burst use, it might be worth a shot
@tough matrix if you're trying to do anything at-scale or in-quantity, I think you will be quickly frustrated. If you checkout UnexpectedMaker's YouTube channel, you'll see him having endless issues with a less-expensive, lower-tier PnP, whereas his new Neodenium (sp?) is working great.
UnexpectedMaker also has a discord with other Makers you might want to pop in there as well.
Gotcha !
Which is why I suggested the little pumps adafruit sells as an option
Drive with a power transistor and a qtpy or itsybitsy
Yes, I've seen UM's videos. But I am nowhere near his level or scale - I need to occasionally populate and reflow 10 boards, not hundreds every week. So I am using regular tweezers for now, but grabbing large ICs such as TQFP48 with tweezers is quite inconvenient, so I am looking for better (but manual) options
i think the next step up from tweezers is ducktaping toothpicks to the end of chopsticks (wider grip than a lot of tweezers would allow)
you can't be serious ๐
Placing parts with chopsticks instead of tweezers sounds like a fun game, tho.
maybe stick a tiny dab of sugru on the end of the (toothpick, dental pick, jeweller's screwdriver, whatevs) to help grip the parts, but yeah. assuming good at chopsticks.
Is that like an EE version of beer-pong @elder peak?
I mean, soldering while drunk seems like overall a bad idea.
just gotta stick at the right point of the balmer curve
@elder peak It's worse if you are a mathematician: Don't drink and derive.
Sobriety is integral to proper math
So improper fractions are inebriated?
I feel like proper fractions are just being lowkey about it but that's just me
Chopstick technique can be very useful for probing across surface mount resistors, if you have good fine probes.
can anyone recommend a good PCB cleaning fluid for ultrasonic cleaners?
Yes let me find the one I use.
Elma Tec Clean A1
I personally need to figure out how to run minipots through the cleaner without removing their oil. Right now my process is just to solder the minipots on after everything else and manually clean after.
A1 works great and can be neutralized easily with vinegar.
So... the answer is still fluid?
got my $50 award from instructables - for runner up in one of their contests - and immediately spent it on buying a new "third hand" for soldering ๐
So, you almost got a big hand for your efforts but, instead, you got a little hand.
I still haven't found one I like, which means that the clear next step is becoming Doc Oc from Spiderman.
I got this one:
https://www.amazon.com/gp/product/B08BL2ZYSF
we will see how well it performs. For starters, I expect I will need to replace the plastic covering of the alligator clips with silicone tubing
I honestly hate all of those things.
I have two of these https://hakkousa.com/products/accessories/pcb-holders/omnivise-pcb-holder.html
And a bunch of masking tape.
I guess that's your only vice then.
No I have two ๐๐
dual-vise, dual-vise ... use them for holding your boooooards down
I have a stick vise, I saw it on a stream here
masking tape is quite useful, indeed
I have a Harbor Freight mini-vice thingie that's not circuit-board-specific and the 4-hands thing that Adafruit sells and mostly I just burn holes in my non-silicone ESD-safe mat on my soldering bench with no form of support, not even masking tape.
i made homemade playd'oh to hold some smaller parts still ... then left it out and it got disgusting
is this Edelweiss?
stuck in yer head now, innit
๐คฃ
sorry everyone, I know that some of you will now have one of these two songs playing in a loop. I can only apologize
Have-a Tequila, Have-a Tequila...
keep it PG, friends. ๐
what does it do?
the "antenna" is a bmp280 pressure sensor, you put a piece of silicone hose on it
and blow into it
how do you do the purple and white? that's so neat
I use pcbway's gerber viewer: https://www.pcbway.com/project/OnlineGerberViewer.html
it lets you set any colors
you can also download an svg
which is nice for docs
Oh and you just have to find a fab house that will do it?
many fab houses do purple
I meant purp and white at the same time, sorry, was not clear
I'm actually making the pcb white, because I want black silkscreen
Oh duh silkscreen
@misty escarp one color is the board, the other, silkscreen
What fab houses do you recommend for interesting colors? I've been using JLCPCB and they have red, white, black, blue, yellow, and green.
PCBway has purple alongside those, OSHPark does purple by default
PCBWay is my go to in general
We use their advanced service for PCBs that we assemble ourselves & front panels
Royal Circuits in the US will do any color you want, but they are pricey in low volumes.
I had several attempts at pcbway, but somehow it was always cheaper and easier to use one of the others, despite the fact that I won some free points in there at one time. They always added some extra payments when reviewing my designs.
I mostly use elecrow, jlcpcb and dirtypcbs โ the last one when I want to panelize something myself, because they don't charge you extra for more than one project
Oh, and oshpark if I want something small quick
OSHPark is great
I want to start making pretty boards with less right angles.
I just need more flexibility in volume.
(royal Circuits is one of the fabs OSHPark uses)
I only use Oshpark if I'm making 1 thing and want 3 boards in case I make mistakes
Gotta round all those corners!!!
So, it turns out that my workflow ends up being to draw it using Illustrator, then use svg2shenzhen in Inkscape, and then import it into KiCAD but I have been realizing that some nice PCB artwork really ties the design together.
Jlc has ridiculously low prices for 4 layer boards.
@elder peak that's similar to my process for front panels.
really responsive customer service in my experience too
Like, I have CS4 and I'm not giving Adobe more money but Inkscape just doesn't do it for me. ๐ฆ
I'm not a fan of inkscape either. I use Affinity Designer.
I've generally had really good interactions with PCBWay's engineers when doing my advanced PCB orders.
@elder peak I do the outline in Inkscape, and the routing in Fritzing
Oh that reminds me, I should kick some money over to KiCAD's org.
Fritzing is nice for artsy pcbs, because it supports SVG natively
and it's basically just drag-and-drop
I've had a hard time getting SVGs into eagle the "official" way, so I convert to DXF and import into a library, with mixed results
I might have to try out fritzing for my panels
Because the workflow from affinty to kicad is a nightmare
Anyone know where to get PCB design inspiration?
I'm thinking of maybe doing some kind of turning my company logo (a giant P ^ 2 with a resistor symbol) into a neon light display of sorts.
Since no one asked I'll share with yall while I research
^ That second link is very cool
You can look at some of the really cool panel designs that Eurorack folks do
Several companies (mine included) use PCBs for our panels.
@fast tundra I think OshPark is doing some of their own fab now
Oh yeah? Neat
I know they used Royal Circuits for a while but I believe they do own some of their own process now which is cool. They are experimenting with 6 layer too
If you want a free 6 layer board, Drew is filling up a panel currently
there seems to way to order transparent PCBs but you have to use lower temps to solder parts
transparent 6 layers
Lessons in not doing graphic design in EAGLE when you don't know what you're doing ๐ฆ
Eh, ship it
It's sooo close though
With black solder mask and gold plating? ๐ฅ
Oh I just thought of another way to do it, let me try that first before giving up
You should do after dark
It's not super smol but I might use Oshpark anyways, I think it's roughly at the point where it's the same cost to do it with them as JLC given the extra cost for ENIG-ROHS
I have some dork-friends I'm gonna gift it to I think, so 3 should be enough
I'm not sure how large this will be on the PCB, but you may well be in the range of CAD goggles (soda can and nuclear submarine look the same size on a screen)
In that it's easy to spend a lot of time tweaking something so it looks perfect in zoomed-in CAD view, only to be within the fabrication tolerances anyway
It's 2ish in in diameter
The circle is
Now the fun begins: importing this in 2 separate scripts and lying them on top of each other so they align and are centered on the origin
one for tPlace and one for tStop
but good call on the CAD goggles, I've been there before without realizing it
also I have to say I love seeing all the different username colors that show up in this channel. It's a rainbow here!
traded one problem for another: now have the not smooth spline but that's OK, it's much better
Have you ever written code and the realized someone made a library for the same chip youโre using. But their approach was way different
Found an Arduino library for the AT42QT1070 capacitive touch chip but they chose not to enumerate all the addresses for the device
A lot of people won't write code that sets registers on the MCU directly, so, yes. ;)
I was talking about the sensor registers
I used the MPR121 library as a reference
Though, Iโm thinking about refactoring to make it even easier to use
If the code was written by a hardware engineer it might not be the best code. ๐
I resemble that remark lol
Iโm not terribly good at coding
But I love hardware ๐
I work with a lot of code written by academics and researchers and it's often not pretty haha.
I do alot of LED designs and tend to jump to Aluminum core PCB but with the price of 4 layer boards coming down, I wonder if multiple layers of copper planes could do the same thing.
Has anyone seen a straight forward simulator to tell the thermal resistance of Al vs multilayer?
Or pre-made models for Solidworks. Those were bear last time I looked at it.
I think that's outside most of our areas of expertise, @pearl tapir ๐ฌ
Copper is certainly a better conductor, but the layers are probably thinner than the aluminum layers would be.
So, this blog post and the TI article linked was fascinating: http://www.brysonics.com/pcb-thermal-resistance-some-unexpected-results/
@fast tundra Of course. I typically use .062"/1.6mm for Al core and 2oz copper for FR4 PCBs when I have low power LEDs. 2oz copper doesn't conduct very much heat when the LEDs are running at, say, 1-2W.
For a UVC disinfector, the LEDs have to be quite cool or they will only last 5000hr. To disinfect an N95 mask at 60 seconds per cycle, 5000hr is fine.
To build an air purifier which needs to run 24/7 then 5000hr just won't cut it.
To me, I'd think that the extra copper from the two extra planes isn't going to add that much mass. But I am heavily speculating
@elder peak Agreed. It requires thermal vias under the LED that are filled with solder through all 4 layers. It still probably won't work based on my experience, but the only stupid questions is the one you don't ask.
I guess the advantage is that you have a fairly continuous power and ground plane on the outside, tho.
I think the "thermal vias under the LED filled with solder or otherwise" kinda implies that it's still easier to just clamp a good copper heat-sink on the back.
Buy a big piece of aluminum, expose a big copper fill on the back of your PCB, and mount the PCB on the aluminum with a piece of thermal adhesive between the exposed fill and the aluminum.
I literally just did this with three SSRs ๐
As @fast tundra and I frantically type almost the same thing in a race.
I guess that was able to rectify the situation?
loool
but seriously getting an aluminum heatsink is super cheap
https://www.oshcut.com/ can make you custom ones
@fast tundra To some extent, when you do that, you are paying for a PCB and aluminum. An aluminum core pcb might end up cheaper.
In designing some 250W high bay lights, I used Al core circuit boards, graphite filled thermal interface material, a custom extruded heatsink and it still needed a fan.
Custom extruded Al is actually fairly inexpensive if you are willing to buy 2000 pounds.
as the old adage goes, scale changes everything.
at medium scale I'd look for COTS heatsinks
I don't know yet what the scale of a COVID disinfector is. Most of what is seen on the market just doesn't work. To get certifiable viral reduction test results means a lot of very expensive LEDs.
Hospitals are claiming they don't have any money, so it is a hard sell.
I know someone who works in a hospital and works with budgets. This is just one hospital in one part of the US, but they are in trouble, so I believe it
If you're looking to produce a device like that for sale, I'd recommend hiring an industrial designer.
Why?
Look at the lower left bottom corner of this web site. That is my design being produced by a local sheet metal manufacturer. It works an I have results from a test lab. I'm just thinking through how to reduce costs.
http://www.metcam.com/
Metcam is a fabricator of precision sheet metal components and assemblies for OEMs and specializes in product finishing, assembly & integration, and logistics and inventory.
@pearl tapir ExpureLED thing?
It seems you've got this all figured out. I'm not sure how much help we can be!
@distant raven Yep
@fast tundra Just looking for what I haven't thought about. A lot of smart people here that have thoughts on cost reduction.
Oops. I'm in the process of uploading the gerbers and making sure I've got enough parts for my next three designs and I discovered that the pull-up resistor and protection diode weren't connected to Vcc and there were two vias that weren't necessary and were breaking up the power plane
Oh wow
Pulled the trigger on two new artsy boards!
Nice!
gotta figure out framing, need to fill out these walls they are too bare
Hello everyone, sorry for the noob question but what are some free pcb design software you recommend for learning with small projects? Thanks
Sweet, thanks I'll check those out ๐
Others may have other suggestions ๐
Can anyone confirm if the GD25Q16 sold in the Adafruit store (https://www.adafruit.com/product/4763) is the 150mil or 208mil variant? Looks like 150mil
I can't positively confirm that, but the "T" on the package code indicates 150mil, so that's what the product photos are showing, at least. If you wanted to be paranoid you could grab the QT Py board files and check the footprint.
E1 here
Excellent. Thank you all for the help!
It soldered up so nicely
woot
Iโve been working endlessly all day attempting to get the driver code to work
I ended doing an afterdark order from OSHPark btw. Different, smaller designt though
and I ran out of solder... time to order another spool
Oh so you are the YouTube chan author?
"YouTube chan" doesn't exactly have a nice ring to it
@fervent lance I am the one behind Riskable 3D printing, yes
@ivory jasper looks great!
Now can you type on that as it is shown in the photograph? how?
Better to call me, "Riskey-chan!" in some kawaii voice ๐
@fervent lance By pressing on keys
I just noticed my 1kg spool of lead free is looking slim, I gotta order some more too.
It'll work just like my Riskeypad numpad: https://gfycat.com/welcomeindolentcanary
hey @fast tundra any chance you can help me with something C Related?
well, arduino I guess
I posted about it in the arduino help
also I love the keyboard ๐
I want one!
all these keyboard projects have me wanting to make my own
All wiyht! Rho sritched mg kegtops awound?
it hurts us
@distant raven sure!
Wow I wouldn't buy solder from amazon
even brand name?
Nope, their "equivalent merchandise" policy can still bite you. That, and I loathe their corporate practices and refuse to do business with them.
yeah they're capitalist monsters
capitalism is all fun and games until someone becomes a corporatist and then it's all over
those pesky cronies
Digikey, on the other hand, will happily put on a pretty substantial markup for the convenience and brand-name. It can be worth at least checking the likes of findchips to see if there's another authorized reseller that's not as expensive.
e.g. https://www.digikey.com/en/products/detail/kester-solder/24-6337-0018/365505 vs https://www.masterelectronics.com/kester/2463370018-11629114.html?utm_term=24-6337-0018&ref=findchips&utm_campaign=findchips&utm_medium=inventoryrefferal&utm_source=findchips&utm_content=feed (though who knows about shipping)
McMaster-Carr has a hair trigger order button and (in my recollection) no way of cancelling an order. Quick as heck shipping though
I wonder if there are enough businesses with a limited number of approved vendors that this works?
Master is technically not an official distributor
ah gotcha
It seems to be listed on Kester's distribution page: https://www.kester.com/contact/distribution
(They may or may not be an official distributor of every product they sell, I can't speak to that, but at least for this specific product it appears legitimate)
I have found myself mostly shopping Digikey, though i've bought from Arrow before
when picking a 32.768 kHz crystal, should I be looking in Crystals or Oscillators in Digikey?
some JST 2 pin battery connectors that go out of stock frequently..
If you want a crystal specifically, you want the crystal. Oscillators generally are digital logic-level-output devices.
ok thought so, been having a hard time finding one in the package I want but I guess I can live with radial can
What parameters are you looking for beyond frequency? That's one of the more common frequencies, so there should be a good variety of parts available
Solderability with only an iron
You can solder the smt ones with an iron..
I tried to find one with an HC-XX/X footprint but couldn't somehow
I think 4-SMD is an okay size
algo como esto? https://www.digikey.com/en/products/detail/txc-corporation/9HT7-32-768KDZY-T/5822139
Order today, ships today. 9HT7-32.768KDZY-T โ 32.768kHz ยฑ20ppm Crystal 7pF 65 kOhms 4-SMD, Flat Leads from TXC CORPORATION. Pricing and Availability on millions of electronic components from Digi-Key Electronics.
though Adafruit uses some 2 pin ones for their cortex m4 that I want to try out
anything "wrong" with a radial can? I already have one, I'm just redoing this board again before ordering
Thought the Oscillator/Crystal question sounded familiar. Turns out it was last weeks "The Great Search" topic!
https://www.youtube.com/watch?v=AhRORYtAqsc****
The Great Search is a new series of videos focused on searching the Digi-Key catalog to narrow down to the parts you want to buy. This week:
If you have a microcontroller or audio chip, you're probably going to need a crystal to help it keep time. In this Great Search, we'll look at Crystals, Oscillators and even touch on Resonators. Which one ...
I just remembered that!
Glad it was one of those "I KNOW I saw that" that actually happened, and not "I'm pretty sure I didn't make it up in my mind"
hehe
mandela effect for electronics?
I just saw that the feather s2 has a little extending tab for the usb-c port to rest on. So neat
Just, some design concerns take longer to crystalize than others
Redoing my calcs too, to be safe: In this, CL is the Load Capacitance given in the datasheet, right?
And is it usually reasonable to say C1 = C2?
Yes and yes
I use a super tiny abracon one for Sol
๐ฎ
Though this isn't the formula I generally use since Cstray is generally not defined as such
I've used 18-22pF depending on the load capacitance
Well, that's one way to clean your clock
๐
Order today, ships today. ABS06-32.768KHZ-1-T โ 32.768kHz ยฑ10ppm Crystal 12.5pF 90 kOhms 2-SMD, No Lead from Abracon LLC. Pricing and Availability on millions of electronic components from Digi-Key Electronics.
Yeah, I was weirded out because a lot of the smaller crystals have a very low Cl and then the Atmel datasheets are like "don't use a capacitor that small"
From what I understand it's better to go a little higher than a little lower.
I use 22pF load caps with that crystal I just linked.
does my calc seem off? Should I pick a different cstray value?
the 48MHz crystal I use on the MSP432 feather wants like 28-32pF caps
this is un poco frustrating
In short: Your thing will probably work, especially if you're not trying to do super crucial timekeeping with that 32.768KHz crystal (e.g. in an RTC).
Sometimes they'll have suggested maximum ESRs or other parameters for the crystal in the clocked-device's datasheet
I feel like I should keep some chicken bones in my geekroom specifically so that I can print out any schematic with a clock circuit on it and do a ritual on it.
Isn't that the correct scientific method, @elder peak ?
I mean, quartz crystals are part of many shamanic traditions.
This is a pretty good app note on the subject, written specific to STM32 devices, but with principles that can be generally applied: https://www.st.com/resource/en/application_note/cd00221665-oscillator-design-guide-for-stm8afals-stm32-mcus-and-mpus-stmicroelectronics.pdf
thanks @heavy jasper I'll take a look
TI also has a 32kHz crystal selection guide that is nice
I should have my friend I made the PCB tarot card for do a reading for selecting these
Bonus points for chicken-bone powered, versus lemon/potato clock
They're a vegetarian ๐ฆ
not for taste
Back when Damien George was designing the original PyBoard, I backed the kickstarter and decided to design my own slightly more custom board for a robot I was working on. I started off wanting to use a smaller crystal than the one he was using, and went with one that looked like it should work. It didn't. The board still worked (using the internal oscillator) but the USB didn't work. I switched to the caps and crystal he was using for my next rev, and everything worked perfectly. Both the PyBoard and my board were using the same STM32F405 microcontroller. Crystals and caps are black magic, especially for the Cortex M4.
@pure fossil STM32F405 runs Dr. C.H. Ting's eForth unmodified.
https://github.com/wa1tnr/eforth-stm32f4x-a
http://forth.org/OffeteStore/OffeteStore.html
http://forth.org/OffeteStore/2165_stm32eForth720.zip
I have pretty much zero interest in running forth, so that's okay
;)
Back in the day my Apple 2 ran forth - I looked at it, didn't like it, and wrote 6502 assembler instead
It takes a special kind of brain to really appreciate RPN - although I like it from a conceptual perspective, and I've written compilers and parsers that essentially implement it (stack based computation), but I don't really like using it myself
I just dusted off and started my old HP48 and then I remembered it uses RPN ๐
Anybody else a bit uneasy about digikey's tendency to overpackage?
Just received an order, and I needed screws. They decided to individually package screws
;) I hear you.
@reef flicker it's usually because they have a lot of individual screws left in packages from previous orders so they keep them aside for suckers like us hehe
I quite often get my order of 100 capacitors in two packages, one with 49 and the other with 51
Pretty much, I use Digikey's overpacking as my orginizational system.
I use Sparkfun packing for my organizational system. ๐
just fixed the one defective board out of batch of 56 sent to me by fab house. Now I have 56 working boards.
I ordered and paid for 55, they sent one extra, and it turned that out of 56, one was defective - a simple solder bridge .
Well, that's something to do when your board and looking for something to play with
out of boardom
I sometimes think of getting the parts binders that Adafruit sells for SMT stuff but, eh, digikey bags work well enough
I got one, it's a lot smaller than I was expecting
You can export libraries and sometimes that has the info you need
Thx
Are you looking at an adafruit board?
In the Schematic or the brd
brd
Is it just a button or a file then a button
Sorry, just saying "Export this" was not very specific
make sure to save it as somefilename.txt or it will not open properly
It looks like it's a tsv so you could possibly save it as a tsv and then open it in excel/google sheets by converting it to the native format
Ok thx
having trouble opening in google sheets properly, not a tsv it seems
anyone knows what software people use for actually writing the datasheets of various chips? Please don't tell me they use Word...
A bunch of different things. Some companies do use word.
I wonder if there is a LaTeX template for such
It wouldn't be too hard to craft one.
sure, I am just lazy ๐
or failing that, a publicly available Word datasheet
that one can use as template
We used to do something similar, but our workflow was ... odd. We'd store the data in XML, run it through XSLT to format it as fop, then use fop to generate the PDF.
Worse, the XSLT transforms were huge, so they were generated by yet another XSLT transform from XML descriptions of what the final pages should look like.
I am not eager to go back to xml editors - last time I did that was 20 years ago
xslt to run xslt ๐
makes complete sense ๐
and in the darkness bind them
thanks for sharing yesterday @distant raven , congratulations! at least somebody had a good 2020...
HAPPY NEW YEAR ALL!
I think the nws surely made the year seem so much worse. And it was bad in many perspectives. But there was also a lot of good that happened this year, and not just for me.
I'm sure that is true. Not all the news was bad, saw a headline today "Nobody was killed by Murder Hornets in 2020 (as far as we know)" ๐
looks like I am not the only one interested in LaTeX template for datasheets: https://tex.stackexchange.com/questions/615/where-can-i-find-a-datasheet-document-class-or-a-tutorial-on-how-to-create-one/570808#570808
Good find!
I have totally seen Word metainformation on datasheet PDFs
I tend to eye that as an admission that the company settles for worst-of-breed tools
I'm not a terribly big fan of LaTeX
but i know i should probably get used to it lol
I have absolutely zero, zip, null, zilch, nada, rien, void experience writing up datasheets, but I've read some, and the ones I've seen seem like they could have quite cleanly been written in MarkDown (with Extras, for tables, footnotes, etc.). With some LaTex thrown in for special uses (formulae, maybe charts? I don't know LaTeX's capabilities or limitations), it should also be well suited for version control systems.
@supple pollen - that's because Notepad.exe doesn't put out any metadata for you to see! ๐
notepad might actually be better than word in many regards
I use Markdown for a lot of internal documents. Makes it easy to track changes in git, too.
I'm relearning markdown and whatnot
A long time ago, I'd written a utility that converted text with asterisks for bold, etc. into HTML. Markdown just seems like someone had a similar idea and fleshed it out considerably
yeah, there's a lot you can do
there's a whole website dedicated to talking about using markdown and I've glanced at about.... 1/4 of it
I do like the auto heading numbering, it works like in troff but even simpler
This one @distant raven ? https://daringfireball.net/projects/markdown/
that's probably it
It's incredibly well thought out and detailed, but some of the omissions, or "you can use plain HTML instead" seem odd, or "we don't want people using it, so we don't want to do it".
There's also CommonMark, which is an attempt at standardizing markdown conversion and rendering
https://commonmark.org/
I just know I was trying to reformat from github md file format into rst and it was a little bit of a headache at first
I've never dug into the difference between MD and rST. They both seemed pretty close to one-another
the differences are ever so slight
but enough to mess with things
like using # isn't permitted in rst for bold section names
Sounds part for the [technical] course
must use ------------------------------ under the word
Well, there's also asciidoc which is more oriented towards technical documentation
Which MD supports for H1 and H2
headline
======```
H1 header
------------
H2 Header
============
that works too
I dono, I've been using wikis as my primary notetaking and orginizational method for over a decade now so nothing about Markdown feels actually that weird.
But I prefer the pound (#) since the - and = can be interpreted as <hr> if your spacing is wrong
yeah
Not to mention, counting # helps know which level you're at with minimal effort, and follows naturally through <h6>
my rst looks like this
so my preview can look like this
I'm sure the bolding isn't necessary
I feel like it is, so I can spot a section at a glance. But that's my personal (and correct) opinion!
hehe
Refreshing a package tracking page causes the package to come sooner, right? asking for a friend
I wants my pcbs
Maybe just a bit slower than when you're tracking your Adafruit shipment! ๐
hehe, I want the bluetooth adapter and hdmi to DP cable I ordered from amazon 3 days ago that should have been here yesterday but hasn't moved one bit
I've decided to look on the bright side: Long shipping times means any covid in the package has a chance to die
assuming no additional handling in the time
It's coming wrapped in plastic and I always wipe down the outside. IDK if it does anything but it makes me feel better
Probably should put it in the microwave for 30-60 seconds, make sure any trace of living biological material is fried.
Hmmmm, should i also order a replacement microwave?
Probably safest. Order it alexa enabled, and make sure it's smaller than the current one, so you can nuke it before you use it. Safety first!
it's microwaves all the way down
The most complicated doc-writing format I worked with was DocBook which we used for GNOME documentation (with all the XSLT stylesheets needed to convert it to HTML/PDF). Compared to that, LaTeX is piece of cake. And I really love LaTeX.
I worked with a DocBook publishing pipeline and, yeah, it was painful.
I totally get what they were trying to do with DocBook, which probably made it worse.
ahah so DocBook I profitably overlooked. Been around for quite some time.
I honestly love RST
RST + Sphinx has been the Python community's secret weapon for a while
I accidentally let my soldering iron tip oxidize really bad, any way to save it?
I know scraping with abrasive materials is bad, but I tried to clean it using flux, rubbing on the sponge, nothing
Abrasive materials can be OK if you're gentle, then re-tin the tip. But most likely, you're going to need a new iron tip unfortunately.
there's a special tipcleaner flux that gets used for that, usually needs 300-350ยฐ. and a brass (not steel) "sponge", since brass will be softer than the tips.
The old soldering iron tips were pretty much a billet so you would just file the crud off. ;)
New ones have some kind of voodoo plating going on or something and I think most are hollow so filing exposes the air core to them. Maybe.
Guess I'll try scraping, nothing to lose anyways since at this state it's impossible to use
If you have some fine sandpaper or a knife sharpening stone, that'd also be a good controlled way to remove a bit at a time
Pretty excited, I ordered my PCB prototypes of my ESP32-s2
๐
And now begins the constant browser refresh cycle until they're at your door!
Haha email refresh while they fab ๐
Currently on the bank account refresh waiting on stimulus to hit
Mostly so I can pay bills ๐
I need to "surface mount" a Pi Zero to a larger board connected with the GPIO pins. I don't have clearance to add headers. Any suggestions or prior experience with this? ๐
ive never seen that done with a pcb not designed for surface mounting
I got the shipment notice for my most recent board set too.
Got my PCBs today! Of course it's the day after I ruin the soldering tip i need ๐
Well, I don't have any soldering tips for you there.
๐
To solder my packages, I'm faced with two options; using solder paste or pre-tinning the pads with solder
Both I would shoot with a hot air gun, and I have some flux if it makes sense when using the second option
Which route should I take?
what components are you soldering?
Tiny pressure sensors LPS33HW
The pads are underneath the actual sensor, no part of it is exposed
Last time I used solder paste but it wasn't an easy thing by any means
I had best success with such parts using solder paste, stencil, and hot plate. Using heat gun with the parts that have pads underneath is challenging but possible.
So solder paste over tinning the pads?
for me yes
but I've seen other people here give exactly opposite advice ๐
Guess i can try both and see
If I were to go pre-tin route, should I put flux between the package and the pads? I'm afraid it'll just stay there forever and possibly cause issues
Any advantage to pre-tinning with paste rather than solder?
Let's try it
Now you tin find out for yourself.
Curious for i2c device developers, if a device has multiple i2c addresses, should you just define the first listed address or all of them?
I was just going to do the first but I'm not great at anticipating how many of these a person would use in their setup
This was much harder than anticipated ๐
1 down, 9 to go
Turns out not enough solder is a thing, and it tends to behave kind of like no solder at all ๐
@distant raven The usual pattern I've seen is have the driver default to the first address but have a constructor option to let the library user pass in an alternate address to use.
Thatโs sort of what Iโm doing. I enumerated the other addresses
I figured it was probably better that way
The only lib I've really dived into does it the way ed mentions, are you then validating passed addresses against the enumerated addresses?
If you want to be absolutely positively explicit about it, you can make a typedef enum of the available addresses e.g. ADDR_LO_LO, ADDR_LO_HI, etc in the public section of the class, and then make the argument of the constructor that enum type.
Where then if you need to pass numbers at some point you can do static_cast<I2CAddress>(0x40)
but normally you'd do I2CDeviceDriver::I2CAddress::ADDR_LO_HI
(except with matched casing and all that)
Thatโs one way
anyone know of a feather s2 footprint? I'm making one but hoping to skip it if possible
alternatively does anyone know of a feather s2 brd file or dimensions?
Are you talking about the mini module @misty escarp
which board are you talking about?
there is this one from unexpected maker: https://www.adafruit.com/product/4769
and then just yesterday Adafruit have announced that they are making their own version of esp32-s2
in any case, all feather boards share the same dimension and footprint
Yeah
Ah shoot didn't realize there were two. The unexpected maker one. I've got it solved tho it seems Thanks everyone
I erroneously thought the spec allowed different pin spacings
Oh gotcha
Messing around with my poor hot hair skills appears to have stripped off some of the solder mask on top of some routes
Is this any concern?
Probably fine, but you can buy soldermask repair liquid.
@reef flicker "hot hair skills" ๐
This is slightly depressing, multimeter shows all nets shorted, even the ones the sensors don't touch
Hoping real hard it's not the board
Well, I don't think it'll work in the QT-Py form-factor. ๐
Hehe probably not
BLE+WiFi LOL
Unless it only has 1 antenna pin then probably could use a unified antenna
Today's design exercise: Can you make a plausable ATTiny84 in the Qt-Py form factor?
The answer may surprise you
It would be wickedly easier with the QFN part and 4 layers, LOL.
Yeah hehe
The thing that amuses me is that, other than the serial pins being connected to three places (ISP port, I2C, and SPI) all of the pins are able to be assigned sensibly with 2 GPIO pins left over (for a crystal oscillator that clearly won't fit)
Bottom?
Physically, yeah, but it would need 4 layers.
I feel like this would be fun to incorporate into a little breakout board
Order today, ships today. VEML6040A3OG โ Color Sensor 16 b Low Power Consumption 4-TFLGA from Vishay Semiconductor Opto Division. Pricing and Availability on millions of electronic components from Digi-Key Electronics.
Most forths can do SPI at best; few can do i2c because they never get past the initial development stage. ;)
USART nearly always the first peripheral to be utilized.
Thatโs a bummer, i2c seems pretty popular these days
This would be fun, but pricy to make a breakout for.
https://www.digikey.com/product-detail/en/MAX30102EFD+/MAX30102EFD+-ND/6166869
Order today, ships today. MAX30102EFD+ โ Oximeter/Heart Rate Sensor IยฒC Output from Maxim Integrated. Pricing and Availability on millions of electronic components from Digi-Key Electronics.
Well the thing is, programming an i2c driver is a step above - very different than the others.
I did one .. fifteen years ago, for 8051 .. that only served as master-transmitter role (running on an MCU) for talking to PCF8574 port expander. ;)
Can't find the code (lost it) and can't replicate it (too difficult). ;)
Was originally a port of a program written by the mfgr as an App Note, in the C language.
I imagine itโs pretty hard
indeed, a color sensor is nice to have - but is it better than https://www.adafruit.com/product/1334
I love when the reflow hits just right ๐ https://twitter.com/theavalkyrie/status/1345206590555750402?s=19
Did I make a big mistake with all of these decoupling capacitors?
Using the arduino analogRead to read the GND pin, it's not always at 0, it raises "randomly"
Reading the 3.3v IN of my board, I get this
To me, this looks like it's charging/discharging the caps very rapidly, but I don't know any better
Actually, creating an intentional short gives me identical waves, so I'm assuming there's a short somewhere on the board, how do I find where it is?
measure between grd and the 3.3v rail, how does it look?
Same pic as above
0v? or just the scale is off?
Nope, it's really at 0 with short bursts of slightly higher
This is using the analog read on the 3.3v in pin
on a 3.3v reference?
Disconnecting GND leads to clean 3.3v on the graph, disconnecting 3.3v in leads to clean 0
what does your multimeter say?
Multimeter measures 0.04mv and resistance of about 3ohm
ok unplug it that's a short
Alright
Good
How do I pinpoint where the short is? I'm not sure if multimeters would flag just a standard cap as a short
Just testing with another working board on a ceramic cap, it reads the same 0.04mv
sounds like the 3,3v rail is shorted to ground, it could be any component or trace on a custom PCB
the 100nF caps near the supply pins is a good idea
Even on each sensors?
it's a good idea to put small bypass caps near each IC or sensors Vcc input
because Vcc is shorted, could be anything
Right, but I mean on a different board where I know it's not shorted
you mean a different working board reads 0 volts?
Hold on I might be wrong
I don't have a continuity test mode on this multimeter, so I've been using DCV and reading anything near 0 as continuous when the board isn't powered, is that correct?
Maybe I should be using resistance measurements
DCV is volt mode so yes you'll read 0 if unpowered
if you do not power the board, set resistance to a low range, what does it read from 3,3v bus to ground?
0.3ohm
So definitely shorted then
The working board reads very high values, but they fluctuate
ok, time to start hunting for the short ๐
How do I do that?
i'd start visually, with a magnifying glass, look for anyplace Vcc is touching ground on the copper
I have 10 resistors mounted, all at 10kohm, for some reason halfway through the board they start reading as 0
there's a clue
I'm assuming the short is somewhere under my sensors, the problem is that I can't see underneath and it's very (very) difficult to solder / desolder them
So knowing which one is the problem is pretty crucial ๐
trace the trace between the 10Ks that read right and those that don't
solder blobs, wire in wrong place...
If one pullup reads as 0, shouldn't they all read as 0?
They're all between data line and the 3.3 rail
you don't want your sensor pins tied high, no
For reference, R9 is the first to read as 0ohms, anything above on the board reads as 0 as well
So all of these read as shorted, but not the ones below
Is it in theory a separate I2C bus per sensor?
It's a separate I2C bus per 2 sensors, since they have 2 possible addresses
So 5 buses for 10 sensors
If so it seems somewhat unlikely that a solder short on a single sensor would cause a mass short of nets to VCC. And even more so for that to be split where half of a bus is shorted and the other isn't.
It does however look like you have a bunch of vias located very close together underneath that TCA9548; have you checked your solder quality in that area to see if any solder wicked underneath the part while you were soldering and bridged vias?
I was pretty careful with that IC, but it's possible some got in there. I could try removing it to check
Depending on the size of the via you might be able to get a bright light and look from the bottom side of the board and see if you see any solder wicked into the vias.
If you don't see it, it doesn't necessarily mean there isn't bridging, but if you do see solder there that's generally not a great sign. Separately, just a close visual inspection of all the tracks on the bottom layer to see if there's some evidence of shorting.
I don't see it for this IC, but I did find 1 other via that does appear to have a tiny ball of solder in there, it's nowhere near other vias though
But even if those vias are shorted, could that really explain the GND VCC nets being short too? Since those vias are only data lines
Oh I missed that GND and VCC are shorted also
If you have a hard-shorted supply, definitely solve that first.
And that unfortunately is much more likely to be a badly soldered sensor
Sensors could def cause both shorts between VCC and GND, as well as the data lines
What's weird to me is that they all start being shorted starting from 1 pullup, any ideas?
Could be random, but could also point towards something
but 1 board (at least) is fully functional?
Nope! ๐
ohhhh
The other board I was referring to is a different design, I was just using it as reference to properly use the multimeter
Oh then all number of things could be wrong. Can you post your raw design files somewhere?
I want to run an independent drc
I'm off, good luck and Happy New Year!
Ideally schematic, brd, and the final gerbers you sent to manufacturing.
You should also try probing those resistances on a bare pcb
Mind if I send it over private message? Would rather keep those private ๐
To make sure theyโre what you expect and there isnโt a plane short.
Sure, whatever works.
So I got 3 boards of this, on the other boards none of the nets are shorted
Sent them ๐
I should've just went with different sensors, after all the trouble they've caused me
"on the other boards none of the nets are shorted" are those soldered boards, or bare boards?
Bare boards
This is the first time I try to assemble these boards, received them yesterday
Unfortunately, quite a few places by the sensors that things could be shorted; they're cramped and you did a bunch of layout within the footprint of the package so there's more room for easy bridging.
One thing I did notice: you didn't seem to put any orientation marks on the board; do you know that you soldered all of the sensors in the correct orientation and not 180 degrees off?
Despite being 10k resistors, for some reason R12 and R13 read as 5k ohms
Do you measure 0 ohms between SD4 and SC4?
That explains R12/R13 reading as 5K then, since they're currently wired in parallel
(so you're really just measuring 10K || 10K in each of your measurement cases)
Ah OK
What would you suggest I do then? There's no real way to find which sensor is shorted right?
Depending on your tolerance for PCB rework vs. soldering rework
So I just... remove them all, put them on 1 by 1 and test shorts after each one?
As in cutting the physical trace?
And then I dab solder on the cut traces to form them back?
Yeah it's not the easiest thing to do given there's also the plane on L1, but possible
SInce it's just one long string out, you could lessen the number of cuts by first cutting in the middle of the length-wise track, and probing each side
and binary-searching your way down
Oh yeah, didn't think about the plane
I'm assuming you don't have a thermal camera; that's the other way to do it
(let it short with power applied and see what gets hot)
I don't ๐ฆ
Assuming it doesn't get hot enough to tell by touch
Just so I know what I'm looking for, when the GND & VCC nets will not be shorted, what kind of resistance should my meter read?
Given that the caps will be in place
Caps on a plane look like an initially small resistance that then increases upwards
My guess is anything >1K ohms and you'd probably call it "open"
Ok, well I guess I have my work cut out for me then ๐
Yeah short-hunting is not my favorite activity
How could I avoid this kind of situation in the future?
I'm thinking I could have some kind of bridge that I link gradually, kind of like two empty pads for each sensors, that way I can "cut" it for testing
Or even put diodes to isolate the source?
There are some things you can tweak on the design for manufacture side of layout to make shorts less likely, or put in 0-ohm resistors in series with the supplies to each chip that you solder in after verifying that the chip isn't shorted, but really the best answer here is "have a professional assembly house assemble your board."
Stencils and controlled reflow ovens, ala what stargirl has, are a decent intermediate ground too.
But really nothing beats a robot, followed by X-ray and/or flying probe testing.
I could have access to a reflow oven, but I wanted to give it a shot without given current world's situation
Alright welp, thanks for all the help! I'm off to bed too
Guess we'll see tomorrow... Just hoping I didn't screw up all the sensors at this point
good luck. You might have a better time just pulling off sensors one by one, inspecting them, then soldering them to a new PCB and then probing the resistances after each one
Should've done that from the start, but was too lazy to pull out the multimeter. Learnt my lesson
@reef flicker did you try the alcohol pour technique?
just dab things with isoprop, the component that makes it evaporate the quickest, is the warmest
and likely the culprit.
I don't have a heatcamera, I just do that - and it's remarkably accurate
just set your bench supply to a low current limit, so you don't fry anything
That's a good idea
I don't have a bench power supply, but I can try just hooking it to the arduino and see if it's enough?
A low resistance tracer would help, but most of us don't have access to such and would have to build our own
I'm shaking ๐ got the first sensor to read
started populating some PCBs I just got for a new project and realized that I forgot one trace - an LED is unconnected...
Feel really stupid
@tough matrix Iโm not certain if that color sensor I shared is better than what Adafruit sells
if I had a penny for every time I forgot something on a PCB
my life is a bodge wire ๐
I just plan to throw away the first PCB, so then I'm not mad if I forgot something. Kinda a constant improvement system as I always find something I could have done better.
sounds like my term paper rewrites in high school, copied by hand.
It was literally impossible for me to not change a thing, but just render what was already written, in a presentation quality suitable for turning it in to the instructor.
No phone, no lights, no motorcar .. no word processor .. not a single luxury.
when ordering PCB assembly, what is the reasonable expectation of bad boards?
We did an order of 55 boards, and 4 of them have some problems, ranging from (seemingly) damaged NeoPixel to a short somewhere, causing the MCU to overheat within seconds.
It is not a particularly complicated board - a 48TQFP MCU, an LGA14 sensor, a bunch of passives, couple of SOIC8 motor drivers, a buck converter. I'd expect that modern fab houses should do it almost perfectly - I'd be OK with one bad board, but 4 out of 55 seems too many to me. Am I right?
a 7% failure rate? yeah that sounds kind of wild
where did you get them from @tough matrix ?
im not much of a professional in pcb design but that does sound quite poor
elecrow
really?
yes
elecrow is known for its pretty high quality...
i mean i tend to use pcbway but ive heard a lot of great things about them
i mean, maybe something is wrong with my design, but the remaining boards turned out ok
yeah i mean my best bet would be to contact them in that case
cause 4 malfunctions out of 55 is definitely grounds for discontent
i will. We ordered more than necessary, so we are ok with fulfilling orders, but it does make me uneasy
yeah no i def think they'll understand
ok, will try reaching out to them
there should be no shorts, they do a flying probe test
the only way you can have faults like that is if they are kind of on the edge, and get broken in transport or during the assembly
in theory, at least
If the board layout isn't designed for manufacturability, yield can suffer. I'm not saying this is the case with your board, but some component layouts/spacing and trace/pad geometry can be problematic.
sure, but they test the boards and mark the faulty ones
Given that one board had a clearly visible solder bridge, i guess they didnt do the proper inspection...
But of course, i will also take another look at my footprints etc
is there a black cross made with a black pen on the said board?
No
There may be other files you can send that would be helpful in inspection (e.g. a netlist). They may or may not do flying probe of assembled boards (it can become a very time-intensive process) but Iโd expect a bridge on exposed leads to get caught by AOI or 5dx. So itโs probably worth starting from the more basic โwhat outgoing quality control is done for boards ordered via online quote?โ
what is 5dx?
Automated X-Ray inspection
oh. Yes, I'd expect AOI and x-ray are standard.
Anyway, looks like people agree that 4 out of 55 is unacceptably high failure rate
Of course, it could also be that the problem is in the components - if the MCU is defective, with the best inspection you will get problematic boards...
Where was the bridge?
that's quite an apparent one
Hope you didnโt fry the mcu
This board I managed to save
Whoever assembled that really wasn't on the ball that day
as long as it didn't short a 400V input to an ADC input or something, I guess you're ok with just removing that bridge ๐
so far, my MCUs do not use 400v ... but that is an interesting idea to consider for my next design ๐
450V is useful for dekatron-based processors.
Hello all. I'm looking for resources to learn how to design a PCB. Think absolute newbie here. Suggestions?
A friend of mine pinged me asking, and I figured you folks were my best option to help out.
sparkfun had a nice series of tutorials. It mostly focuses on Eagle, but also explains general ideas such as layers, board structure, etc.
https://learn.sparkfun.com/tutorials/pcb-basics
https://learn.sparkfun.com/tutorials/using-eagle-schematic
https://learn.sparkfun.com/tutorials/using-eagle-board-layout
and follow-ups on more advanced topics.
Some technical details may be outdated (Eagle has changed somewhat since 2013), but much of it is still very useful.
Would you suggest trying to use the free Eagle version, or go with KiCad?
I am using education version of Eagle and haven't really used other software, so can't compare. Overall, I'd be happy with eagle (if you can get free education version) - the only thing which annoys me is that the component libraries are a mess; too many competing versions from different 3rd party providers...
If not the hobby version is good for starting out
the limitations could be good for stretching those mental muscles
I ended up cherrypicking and handcrafting a bunch of stuff into my own component library, which dodges library issues for a lot of my frequently used ones
Noted.
The request for learning resources stands for anyone else who has suggestions too. Thanks!
Kicad would probably be my recommendation, there are a variety of tutorials now available (e.g. https://contextualelectronics.com/courses/getting-to-blinky-5-0/)
EAGLE has forums/guides for most of the questions I've had about it if your friend ends up using that program
And there's a longer ramp
Also true.
The real answer: All CAD tools are bad in some way
I'm only recommending EAGLE because it's what I learned on, I hear great things about KiCAD
and changing between them is not as hard as learning the actual concepts of design
e.g. it turns out EAGLE is not meant to be used for graphic design, as I have learned to my chagrin
I started on EAGLE using the Adafruit and Sparkfun libraries for components
It seems like Kicad is now better at keeping a wider library up to date in a public-facing form.
Indeed, that was my understanding.
Unless your friend really likes library curation as a thing, the Kicad libraries are probably the best feature.
It lowers the barrier to entry to just start placing components on a board and going for it.
And also has more longer-term likelihood of multiplatform support (if that's important).
The caveat I'll give is that if your friend really really wants to do EE/ME codesign (boards and enclosures) and they're running windows, then the wider autodesk suite is probably the lowest cost/barrier to entry version of that workflow.
does KiCAD have good 3d modeling capabilities?
I think it does support 3d models of components and step export.
But EAGLE+fusion360 I think has a slightly more native co-working support
Yeah
You can even do most of what you do in Eagle in Fusion360
As far as I am aware, the Eagle product roadmap eventually ends with full integration into fusion360
I use the embedded version of eagle within fusion 360, since my 3d model heavily relies on the PCB
As it stands now, Eagle is tied to fusion360 license
It's pretty much fully integrated as far as I can tell, maybe there are some deeper features missing but I haven't had many problems
It's pretty useful to update the PCB and it waterfalls all the way down to updating your enclosure, although it's not as good as it sounds
I hate trying to get all the components together though
The 3D packages that is
Like I havenโt imported a model for usb yet ๐
I'm trying to talk him into joining because you folks are a wealth of knowledge and would be invaluable on his quest to learn. We'll see if he takes me up on it. ๐
Always here to help ๐
You know, at some point I'd had the impression that, because EAGLE was a fancy commercial package thingie and KiCAD was open source software that I didn't pay for, that somehow EAGLE would have a fastidiously awesome component library.
Honestly, I generally assume the opposite. The fancy commercial packages will happily upsell you on the subscription-based licensing for access to a well-curated component library, however.
I don't have enough time on the KiCAD libraries to know for sure how good they are, but they definitely seem abundant.
I'm more interested in frictionless ease of use than the component library (I can always make my own components). A big appeal of open source is that if I want to change something, I can, and submit my changes for consideration to incorporate in the mainline version.
Yeah, I was honestly working on the assumption that in terms of day-to-day frictionless use, these tools are basically all the same (they all have quirks, but can be managed).
And I think "can always make my own components" is something that seems (and generally is) pretty easy once you've already gotten started and have a sense of what you're doing.
But can be a ton of yak-shaving for someone who has a basic project in mind they want to start out with.
Well, the only time the KiCAD library has let me down was the APA102's had the wrong footprint, which I committed a patch for upstream.
That's sort of my point: they all have quirks, but the open source ones let me iron out the quirks so they work the way I want them to.
I do wish there were a redhat equivalent for KiCAD. Good paid support available for folks who can afford it, with general improvements made being fed back to the community.
CERN helping out helped a ton (e.g. features like length matching and differential pair tuning that didn't exist before) but it seems like they kind of petered out a bit.
That was good to see
But as it stands there's kind of a dangerous middle ground where small companies doing advanced boards can't really easily use KiCAD in its current implementation (especially on the layout side) but also can't really afford the countless engineer-hours of time it would take to build up the tool to the state where they could use it solely. (As much as I'd love to see a FOSS 2.5d or 3d field solver/transmission line extractor for doing SI work).
vs. just chucking a few K$ a year to the likes of Altium, or even ~10s of K$ to the likes of Ansys/Cadence for the super high-end simulation tools.
But these are very much commercial-focused arguments. I completely agree with you that for most folks (and especially those with more time than money) the FOSS solution is going to be great for them to make whatever small tweaks/usability improvements they like.
I love the open source world quite a lot but at the same time, I know too many burned out open source project maintainers to think that the model is anything close to a working one.
I'm kinda passionate about KiCAD (and Blender and FreeCAD) regardless because it's a tool people ought to have access to in this modern world.
What is SI work?
Signal Integrity
Ah, that makes sense.
In this case, doing extractions of the S parameters of the transmission lines (traces/vias/etc), and then combining them with transmitter/receiver models to make sure that signals are properly transmitted and received with low/no bit error rate.
Or in a more simple application, doing analysis of the propagation delay for different layers of PCB stackups (and the propagation delay of vias), so that 'length-matching' can become truly time-based-matching so that signals arrive within the right number of picoseconds of each other.
Wow, I haven't worried about picoseconds since the 1980s!
Does this count as getting a pico into your projects, electronic_harry?
Sadly, no comment. But these are well-known industry techniques.
Found myself a weekend project: https://www.digikey.com/en/product-highlight/t/texas-instruments/ads7138-q1-automotive-8-channel-adc
Texas Instruments' ADS7138-Q1 is a automotive 8-channel, multiplexed 12-bit ADC with an I2C compatible serial interface.
are you making a breakout board for it?
Yeah
Iโd laugh if this is the NPI for the week though ๐
Itโs the coolest thing on the new products page on Digi-Key
๐
PCB arrival day!
so PCBs are listening?
Okay, @distant raven is officially right about KSR223GLFG buttons.
They are my favorite
It's really super-nerdy but I probably need to write up a blog article about the buttons I like to add to projects.
I guess you could say it's a hot-button issue for me
Same, good buttons are a sacred topic
Some people just are like โthrow a button on itโ but Iโve got some issues with textures and tactile responses that borders along my hypersensitivity
Yeah, I kinda optimize for the "Will I be able to navigate this device by touch in total darkness" case.
I'm the same way about car dashboards, in that must have as much buttons as possible for me to be comfortable using it. I need the tactile feedback you don't get from a touchscreen
I hate a buncha things about Steve Jobs but I do feel like his obsessiveness for how a device feels was important.
sociopathic clocks are right twice a day
@elder peak you're holding it wrong
prototype delivery day!
finally have a batch of close-to-series (without branding) prototypes ready
All QC'ed serialized, readlocked firmware and ready to go!
No idea.
Free energy collectors?
They are dental motor systems
Combines with an android app on a 7โ tablet
(Connected over USB-C)
unlike the dentist, everybody loves the dental engineer ๐
thats awesome. are those custom enclosures or just off the shelf ones with a custom backplate to fit the connectors?
Itโs kind of a customized standard case yes
Custom plates, custom coating
Aluminium, with hexavalent chromium treatment
Series units will have powder coating and silkscreen additionally.
@cerulean crest Interesting load cell article made from standard resistors: https://hackaday.com/2019/01/27/quartet-of-smd-resistors-used-to-sense-z-axis-height/
Neat
I am wondering how repeatable it is. I have a customer that could use a system to measure the quantity of hand sanitizer remaining in a container and phone home when it is almost empty.
The article does state "did some repeatability testing with this sensor and it turned out to be surprisingly consistent."
Consistency is in the eye of the beholder.
One test unit that is consistent is not the same thing as 100 all giving a similar answer. I don't need huge accuracy but I don't intend to build a production device and hand calibrate each of them.
anything can be a load cell if you have an amplifier sensitive enough
what should I choose for C9? I was thinking either 0.1uF or 1uF but I'm not seeing any suggestions in the esp doc
maybe 1uF might be better since it's the "smoothing" capacitor
I just ordered 5000x 0402 0ohm resistors
nice
it was like... $4
should last me through the end of the year
I'd use the same as C8 just keep the number of parts in the BOM at a minimum.
C7 is 1ยตF so either value would do that
0ฮฉ? What tolerance? I saw some listed as plus or minus 50mฮฉ, but I'm guessing that the negative values don't actually occur?
Not sure on tolerance