|
Bad Munki posted:
Over-overhaul: Rotate the north-south cap 90 degrees, and move it on the north side of the existing caps. It will straddle the trace coming off of the other osc pin. Rotate the crystal 90 degrees and move it close to the caps. Rotate the (power?) header and scoot it up some. Move the resistor from the east side of the board to the west and flip it 180. To get the ground trace out, run it across pin PC0 (which is unused and can be connected to gnd no problem) so you don't have a ground trace These changes should be able to shave off a little bit of the board while shortening your gnd and osc traces. EDIT: Is that VCC running around the edge of the board? Your etched board lacks informative silkscreen 
Delta-Wye fucked around with this message at 17:44 on Oct 4, 2013 |
#
?
Oct 4, 2013 17:41
|
|
I suppose not. That position is just a holdover from the original layout, completely missed that. Before, most of the pins were coming from the near side of the atmega, but when I picked different pins to make the routing easier, I totally spaced the end result in that regard. The only possible problem is that vertical 4-pin header, I'll have to look at the rest of the setup to see if it's going to cause an issue (the wires for that header are fairly stiff, thick, and non-negotiable in their placement/orientation.) Thanks for pointing that out! 
|
|
| # ? Jun 4, 2024 01:30 |
|
|
Thanks a bunch for the help. I'm retooling it now, we'll see what I come up with. It bears mentioning that while the thickness of the board is limited, the width and height are tied directly to the keypad it's connecting to, so I can fully use the available 2x2.5 inches.Delta-Wye posted:EDIT: Is that VCC running around the edge of the board? Your etched board lacks informative silkscreen 
|
|
#
?
Oct 4, 2013 17:44
|
|
|
Okay, how do we feel about this: I added or added variations of or attempted to add the following: peepsalot posted:Is there a reason not to move that header at the bottom to the other end of the board? every single trace from that goes to the opposite side of that long IC. insta posted:Rotate the 4-pin header to the right 90 degrees, and move it closer to the ATMega. Delta-Wye posted:Over-overhaul: And sorry if the curvy traces bug the hell out of anyone. It'd be easier to do it with straight traces but I find it amusing. 
Bad Munki fucked around with this message at 18:49 on Oct 4, 2013 |
|
#
?
Oct 4, 2013 18:47
|
|
|
Now you just need a ground fill. Of course, that would effectively delete your smiley-face, spiral, and fetus (or whatever that is supposed to be). Edit: Also is there a reason you have a 9 pin header on it when only 7 of them are connected to anything? armorer fucked around with this message at 19:05 on Oct 4, 2013 |
|
#
?
Oct 4, 2013 18:59
|
|
|
Slanderer posted:These two presentations I read yesterday (by the same guy) explicitly make the opposite claim---different cap values don't help. I don't have the name of the book I read regarding this since it was on the work network but I'll try to remember to look it up on monday. It seems to me that the data presented doesn't thoroughly debunk multiple capacitors, and the summary of the first presentation even says that more capacitors are better and the value doesn't matter much. I do admit that I probably wouldn't actually place three nF range values of capacitor near a circuit, but in general placing multiple values of capacitor is sensible to cover different frequency ranges. This is basically stating the obvious, you put 100s or 1000s of µF at the power supply to provide bulk low frequency power, then put a few µ around to cover medium frequencies and finally 100n or smaller at each IC to cover the "spikes" i.e. the highest frequencies. On a board without a ground plane it might make more of a difference I think. A specific power supply filter I have worked on originally used 100nF X7R ceramic + 220µF tantalum for bypassing, adding a 2.2nF C0G capacitor reduced high frequency spiking a fair bit. I don't think I'll see much improvement from adding a 10nF in there, but certainly there was a benefit in adding different values, largely because they have significantly different resonant frequencies and so cover different frequency ranges. This board doesn't have a ground plane or a very optimal layout, we might find in the future that changing the board will remove the need for that particular cap.
|
|
#
?
Oct 4, 2013 19:10
|
|
|
armorer posted:Now you just need a ground fill. Of course, that would effectively delete your smiley-face, spiral, and fetus (or whatever that is supposed to be).  quote:Edit: Also is there a reason you have a 9 pin header on it when only 7 of them are connected to anything? Not really, except that that's what the keypad I'm connecting has as well. And it's not like those two end pins on the keypad are connected to ground or anything, they really don't go anywhere. I'm a sucker for things lining up nicely, though, so I included the two extra pins.
|
|
#
?
Oct 4, 2013 19:22
|
|
|
longview posted:This is basically stating the obvious, you put 100s or 1000s of µF at the power supply to provide bulk low frequency power, then put a few µ around to cover medium frequencies and finally 100n or smaller at each IC to cover the "spikes" i.e. the highest frequencies. On a board without a ground plane it might make more of a difference I think. No, it's the sort of thing that *seems* obvious, but lacks the technical proof that demonstrates that it's actually true. Having multiple values as filters makes sense with two assumptions: the resistance from the capacitor to the power pin is non-neglible, and the inductance IS neglible. So, RC filters with different time constants. However, if you actually look at the DC resistance from capacitor to pin it's way too small for this to be the case. So, we gotta ditch examining these things in the steady state condition and look at the impedance under transient conditions. One of his points is that inductance needs to be accounted for. The capacitance doesn't really change the transient response---there just needs to be enough stored charge to counteract a negative spike, or it has to be able to absorb the charge of a positive spike. His graphs this point are pretty clear, and were surprising even to me. The takeaway there is that the "standard" analysis doesn't properly account for the actual board impedance under transient conditions. quote:A specific power supply filter I have worked on originally used 100nF X7R ceramic + 220µF tantalum for bypassing, adding a 2.2nF C0G capacitor reduced high frequency spiking a fair bit. I don't think I'll see much improvement from adding a 10nF in there, but certainly there was a benefit in adding different values, largely because they have significantly different resonant frequencies and so cover different frequency ranges. This board doesn't have a ground plane or a very optimal layout, we might find in the future that changing the board will remove the need for that particular cap. Ok, to be fair, his analysis really depends on the existence of the ground plane, so I'll admit that without it using multiple capacitors might make a difference. However, you also need to consider that you aren't just stacking multiple capacitor values here, you are also mixing different capacitor types, both of which have different curves of capacitance vs. voltage, which complicates poo poo more. Slanderer fucked around with this message at 19:40 on Oct 4, 2013 |
|
#
?
Oct 4, 2013 19:36
|
|
|
Bad Munki posted:I'm looking at videos/guides on eagle and trying to make a ground plane but it's just not working. When I draw my big polygon, it doesn't look like in the videos/guides, it just makes a dashed outline and then renaming it to GND doesn't do anything.
|
|
#
?
Oct 4, 2013 19:38
|
|
|
FSMC posted:After you do that you need to run ratsnest, it will then show the ground plane how you would expect. You need to do it every time you open the project. Ah ha! Cheers.  That's with 12mil clearance on everything. e: fixed thermals, probably not even really visible at this resolution. One more question: all those unused pins, should I connect those to the ground plane, or is it better to just let them hang? Bad Munki fucked around with this message at 19:54 on Oct 4, 2013 |
|
#
?
Oct 4, 2013 19:46
|
|
|
I'm glad there doesn't seem to be an industry spec about decoupling caps that my education completely glossed over  If you guys agree on a good "all around use" case, I'd be interested in an idiot's guide to caps so I don't have to think about it when designing new pcbs. Right now I pretty much put them at random, a little bigger near the power sources, something small near every IC. Bad Munki posted:Okay, how do we feel about this: It looks like you might have clearance issues with some of the components. Especially with that header at the top, I might space out the cap a little more. edit: You should draw dicks in the empty space. Lots of dicks. edit2: unused inputs in most ICs get tied to low or high. Floating inputs are ESD sensitive and that is a Bad Thing. Most modern microcontrollers have internal pullups, though, it's probably okay there. ante fucked around with this message at 19:59 on Oct 4, 2013 |
|
#
?
Oct 4, 2013 19:56
|
|
|
Naw, the caps are just wee little ceramic disc caps, they take up about 30% of the actual silkscreen footprint.
|
|
#
?
Oct 4, 2013 19:59
|
|
|
Slanderer posted:The takeaway there is that the "standard" analysis doesn't properly account for the actual board impedance under transient conditions. It seems like your source is based on newer research than mine (about 10 years old), I have always made the assumption that a 1µ, a 100n and a 1n capacitor would all be different dielectrics since otherwise there wouldn't be much point I suspect we're also arguing about slightly different things, I'm more concerned with preventing external noise from digital circuitry from coupling into my opamp than the opamp causing a power supply dip. My source did show some measurements for these things, including the capacitor spacing from the pins and he concluded that it wasn't worth ripping up your design to fit them super close, even though it made a slight difference. longview fucked around with this message at 20:04 on Oct 4, 2013 |
|
#
?
Oct 4, 2013 20:02
|
|
|
ante posted:If you guys agree on a good "all around use" case, I'd be interested in an idiot's guide to caps so I don't have to think about it when designing new pcbs. Right now I pretty much put them at random, a little bigger near the power sources, something small near every IC. Yeah I have no idea what I started! It has been informative though, in so far as I will now use a ground plane and put something small at every VCC pin rather than just one. It seemed like nobody disagreed with that yet. I already put something largish by the power connection and something smallish by the IC.
|
|
#
?
Oct 4, 2013 20:03
|
|
|
This is good to know because I'm awful at using decoupling caps. Like I literally don't use them. But I will start now 
|
|
#
?
Oct 4, 2013 20:54
|
|
|
I really liked the curvy traces  You should fill up the extra space with LEDs or something dickbutt. Is the board being sent out for fabbing? I presumed with your 1 layer design you were etching it, but if you were going to get it manufactured you can do some cool things with empty space like that. If you're willing to spend some time, you can do at least a 3-tone image by importing a bmp onto the right layer. armorer posted:Yeah I have no idea what I started! Delta-Wye fucked around with this message at 21:54 on Oct 4, 2013 |
|
#
?
Oct 4, 2013 21:52
|
|
|
Delta-Wye posted:I really liked the curvy traces
|
|
#
?
Oct 4, 2013 22:48
|
|
|
longview posted:I don't have the name of the book I read regarding this since it was on the work network but I'll try to remember to look it up on monday. It seems to me that the data presented doesn't thoroughly debunk multiple capacitors, and the summary of the first presentation even says that more capacitors are better and the value doesn't matter much. I do admit that I probably wouldn't actually place three nF range values of capacitor near a circuit, but in general placing multiple values of capacitor is sensible to cover different frequency ranges. This is basically stating the obvious, you put 100s or 1000s of µF at the power supply to provide bulk low frequency power, then put a few µ around to cover medium frequencies and finally 100n or smaller at each IC to cover the "spikes" i.e. the highest frequencies. On a board without a ground plane it might make more of a difference I think. I only skimmed slanderer's first link but I've never understood it that multiple values were particularly important. What is important, and the first link didn't seem to address, are small package sizes for reduced L. The reason to have small values is because they come in smaller packages and smaller packages have reduced L and therefore better performance at high frequency. On the subject of power plane capacitance I spent several hours this week characterizing exactly that on a board designed to measure very low C (pico farads) but also very high current (50A). The plane capacitance of the large 50A traces was hurting us. It gets into the nF very easily.
|
|
#
?
Oct 5, 2013 01:54
|
|
|
longview posted:It seems like your source is based on newer research than mine (about 10 years old), I have always made the assumption that a 1µ, a 100n and a 1n capacitor would all be different dielectrics since otherwise there wouldn't be much point Most capacitors that are worth using are of the X5R/X7R type, and are primarily a barium titanate dielectric. They might have different internal construction, but the dielectrics are usually very similar. C0Gs are different, but those are only found in smaller values. asdf32 posted:I only skimmed slanderer's first link but I've never understood it that multiple values were particularly important. What is important, and the first link didn't seem to address, are small package sizes for reduced L. The reason to have small values is because they come in smaller packages and smaller packages have reduced L and therefore better performance at high frequency. Reduced inductance is a big deal, and it's not just dependent on the package size. The internal construction of the caps can play a large role too. Some of the big manufacturers (like Kemet and AVX) have tools to let you see the frequency response of their various caps, under different temperature/load conditions). There's a reason why an ATC 0.1 uF cap is good to 20 GHz and a standard Murata 0.1 uF isn't, despite both being 0402 X7R parts. Small values can be extremely useful when you have a limited board size, or use ferrites to isolate noise. Not all boards or layouts have significant interplane capacitance that can be relied upon. It can make a difference if you stick them on the backside of a board or directly adjacent to the pins.
|
|
#
?
Oct 5, 2013 12:27
|
|
|
longview posted:On a board without a ground plane it might make more of a difference I think. A while back when I read about a million datasheets to try and learn the basics of EMC I remember most of them saying that the very first step is to have a large, low impedance ground system (i.e. at least one dedicated PCB layer as a ground plane), because without that everything else like decoupling capacitors won't be as effective or more likely simply won't do anything at all. I think one of the easiest to understand resources on the basics of EMC I found was this website by Henry Ott, who is a apparently a leading expert on the subject. It looks kind of dated, but what do you guys think?
|
|
#
?
Oct 5, 2013 17:36
|
|
|
I'm a complete electronics novice, so if this is a dumb question excuse my ignorance. I'm wanting to make a few badges similar to this: http://www.thingiverse.com/thing:148244 The primary difference though is that I'd like to incorporate some sort of switch into the LED so it can be turned off easily. I've been looking for a simple how to tutorial for doing such a thing and they all pretty much incorporate soldering. Is there a way to do this without soldering? Any recommendations for parts?
|
|
#
?
Oct 6, 2013 21:42
|
|
|
If you're mechanically-inclined, you might be able to make some kind of system to push the LED in and out of contact with the battery to turn it on and off, but no, there aren't any better ways of adding a separate switch than soldering one in place. It's the best way of bonding metal to metal short of actually welding. Soldering really isn't too hard to pick up, though soldering wires to a coin-cell battery can be tricky. A $15 radio shack starter kit would probably do fine -- I wouldn't recommend one if you were wanting to get into electronics as a hobby, but a soldering iron is a good thing to add to your toolbox if you don't already have one.
|
|
#
?
Oct 6, 2013 21:50
|
|
|
sky shark posted:I'm a complete electronics novice, so if this is a dumb question excuse my ignorance. Yeah I wouldn't be afraid to solder here, you're soldering big things with big leads and big contacts so it's not like you're going to be doing anything super high-precision under a microscope or something like that. It's really pretty easy to pick up and will leave you with the satisfaction of learning a new skill  It's also the best way to go about what you're trying to do, so yeah.
|
|
#
?
Oct 6, 2013 21:57
|
|
|
Ok fair enough, thanks for the responses. The other option may be just to put a plastic tab under one lead and let them pull the tab when they want to activate it.
|
|
#
?
Oct 6, 2013 22:17
|
|
|
I'm building a capacitive sensing based MIDI keyboard out of nickels and pennies -- I call it the Nickelphone. I used an ATmega644 because it has tons of pins while still being DIP -- got it from Atmel as a free sample. Work in progress video here if anyone cares -- http://discspace.org/the-nickelphone-work-in-progress/ https://www.youtube.com/watch?v=kWq96TSM7XY
|
|
#
?
Oct 10, 2013 05:11
|
|
|
NKArRHH34m5ErF5VkgLA CnvOojjo0N7Vg8FilNi1 KMHJNyzZlHBwQAgJOOwy I6LaDdvmepHFNAqrqmXj ONP5jWoAL3DvkHWA2DEr u817CUzlirfkzThEvtx5 9kc4WysRYjpSxxkst4sc ottLcVgXQdS7YgBsEKQA AxsmmgwnEQRxXViYmAJ9 QpeRup7DQD4JkmN3TMRs Plasmafountain fucked around with this message at 21:13 on Feb 28, 2023 |
|
#
?
Oct 11, 2013 18:48
|
|
|
Can you provide us with a simple sketch of how you have the pot wired up? Edit: If the input is not wired to anything, its value will "float" and you'll get random readings like that. That's why I want to see how the pot is wired up - if it is wired incorrectly, that would explain the behavior you see. Edit2: Also you should divide by 1023.0, the range is 0-1023. armorer fucked around with this message at 18:58 on Oct 11, 2013 |
|
#
?
Oct 11, 2013 18:55
|
|
|
Zero Gravitas posted:Need your help, goons. This isn't a software thing at all, it's hardware. It's a simple 3-pin potentiometer, right? Left lead to +5v, right lead to GND, center lead to the analog input. The potentiometer functions as a voltage divider, and you're probing the center point between the "two" resistors it forms as you turn the dial. If the numbers still float in that case, then maybe you have a broken board, yes, but I bet it works fine in that case. e: and to expand on the above, if you need to convert the value from one range to another, try this int Res0 = 0; float Var0 = 0; Res0 = analogRead(0); Var0 = map(Res0, 0, 1023, 0, 100); //where 100 is the maximum value you want as an output Serial.println(Var0); also don't capitalize your variables 
Sagebrush fucked around with this message at 19:05 on Oct 11, 2013 |
|
#
?
Oct 11, 2013 19:01
|
|
|
3sd68jvItZz7fjJLh9fg 3ZKfpNTbEm9RFycVF80Z cgnVvgDMemVftsuxpswQ Do0F4hI6EfcRu4kEgVWQ 1QLV1KxpVkshCINoGail zOu66xwvtrAj418vOCMg 2SneaJcEe9tbTfprI9Iu Ym4eQ6asnkG3KyfiFfC1 QjWGUS7L2lJaujxdBeOM MJju2mcoyjHCwLCmLaHL Plasmafountain fucked around with this message at 21:13 on Feb 28, 2023 |
|
#
?
Oct 11, 2013 19:19
|
|
|
Check the potentiometer. Get your multimeter and set it to an appropriate range (probably the 2k or 20k range if it's not auto-ranging) and probe between the center pin and either one of the side pins. Turn the dial slowly and make sure the resistance goes smoothly from 0Ω to 10kΩ or whatever the maximum value is. It's a linear pot, right? Not an audio pot? That wouldn't cause the values to change randomly but it would give you an unexpectedly bizarre output.
|
|
#
?
Oct 11, 2013 19:26
|
|
|
Sorry to ask a really stupid question, but you're sure it's wired up to A0 right, and not (digital) 0? Also double check that the breadboard connections are firm. Sagebrush's advice is good in general - when diagnosing misbehaving electronics it is good to test each component in isolation until you find the problem.
|
|
#
?
Oct 11, 2013 19:40
|
|
|
vOp1xLOD3rjKvSiu6iNk gTsh3JfPlCZtyrD16yC8 CcgKjNIfyKW5LvDmF1AG R2y4KU8XvzNCictoiBoe C8LBvpX5hpq1RRMSDdlE guMA6e5w9QB1F5R9REW2 PkyoDrlQXbxjzGAkQp7c sqm1F866SyfYp2ZRX5fY p01US0BGKWQy5irnclEj LE6ZXb8QNFLvsUUePPFy Plasmafountain fucked around with this message at 21:14 on Feb 28, 2023 |
|
#
?
Oct 11, 2013 20:02
|
|
|
Zero Gravitas posted:Turns out the connecting wires I got from my supplier actually have a clear plastic coating over the apparently metal pins that (obv) insulate the pins from the board. What a dumb thing to do. So now I'm sitting here with a penknife scraping this loving coating off both ends of 75 connecting wires. Holy gently caress. Cheap wires from China or something? Please mention the supplier so I can 
|
|
#
?
Oct 11, 2013 20:16
|
|
|
iIbjqnUCzKb0JxiQtg94 FQIZYIpwwJ37Vl1REx9R UwvtwIf3VawIBxIhzxBu EzOisEgRfcyd3SEltNZZ CY0koMe920qHgRjWDD1W qpdsmRrxOensXqm76DPT FthN8764sFCSF8gpJcZq SVxqoVYRPrEByBQoDmQV PwbrX2Ufiq7unGbCvtPA 5Vdxdel1LU9MLBW87Vb2 Plasmafountain fucked around with this message at 21:14 on Feb 28, 2023 |
|
#
?
Oct 11, 2013 20:41
|
|
|
It's not enameled magnet wire, is it? That stuff behaves as you describe but it's supposed to be that way. If you mean they're the little patch cables with solid pins on each end, and the supplier put plastic insulation over each pin, wow. Just wow.
|
|
#
?
Oct 11, 2013 21:53
|
|
|
IKfsjt8Wdsjlh3EbyIst iDKThF3eV7SGgHa9ptyN m0TJ73zdsmDGxs4oXlog 6UAB2BbM0UtElVplEKBV I1IvSkm23YrKqyYuQi2n lIGlPhvk5nti6IqSxsfJ 008jgLtSHHq3O8bj4Gb1 2KsBnZRDZMD6BAUrHSUF 4JS7HfbE98fXZg07IKCv 75Ta4VgOUivqU5quJq8j Plasmafountain fucked around with this message at 21:14 on Feb 28, 2023 |
|
#
?
Oct 11, 2013 21:59
|
|
|
Zero Gravitas posted:These things. So now I need to try and find a nail file or sandpaper or something because gently caress scraping the plastic off each and every one. Are those from that store?
|
|
#
?
Oct 11, 2013 22:07
|
|
|
Zero Gravitas posted:These things. So now I need to try and find a nail file or sandpaper or something because gently caress scraping the plastic off each and every one.
|
|
#
?
Oct 11, 2013 22:49
|
|
|
Zero Gravitas posted:Just why? Why would anyone do this other than to be a complete dick? I keep thinking "forums poster Zero Gravitas must be an idiot, because who in the right mind would coat jumper wire ends with a non-conductive material?" but human stupidity/lazy Chinese manufacturing truly knows no bound.
|
|
#
?
Oct 11, 2013 22:59
|
|
|
You could try melting it off I guess, but I bought basically the same wires with no coating on a while back. Just find another supplier since there's no way it's worth your while to do rework on breadboard wires.
|
|
#
?
Oct 12, 2013 09:03
|
|
|
|
| # ? Jun 4, 2024 01:30 |
|
|
TbdcMqsn0YkLpa85HDlI S4u3uq0rifhQRFYzzlYX 7bS2fZO5t1zoPNS7Ufzr S4A2RzdZnQleKByyI7d4 voh39whyJvV8EZLloWkM a9JBrqWNydiRMpuvn2sW 48W062EkODhLC1dNVSYQ 3sp5k3idsZu6nfLZYmyJ WAVCkVxNRIVPZq4R6y87 6oSn129Ipq9dYx6rmLMy Plasmafountain fucked around with this message at 21:14 on Feb 28, 2023 |
|
#
?
Oct 12, 2013 09:59
|
|