>>>"There’s however one limitation – my vias are soldered through-hole wires." Why is this a limitation?
Anytime I lifted a trace or needed to repair a via, I just coiled up a 24-26 Awg single strand wire and stuffed it in the hole. This usually left enough room for the pin and also wicked the solder through.
My biggest issue with DIY boards is drilling the holes and in particular getting the bit in the center of the pad. This is such a pain, I now just use pre-drilled protoboards and hookup wire on one-off or prototypes.
Yeah, getting good quality circuit boards is so cheap and fast now that it's just not worth doing DIY ones in my opinion. Protoboard is good for one offs, but beyond that, having four layers so massively simplifies routing for anything moderately complicated, and you get proper solder mask, proper hole/via plating, proper coating (HASL, ENIG etc.)...
Not that a lot of that can't be done DIY, but to do really good quality boards at home you start to spend a lot more time refining manufacturing processes than actually doing electronics! That's fine if that's your jam, it's a valid hobby, but I'm much more interested in making more interesting projects!
Unless you want RF materials, or something other than G10/FR4. The cheap prices of aggregating everyone’s boards comes with the lowest common denominator of material.
Even the 4 layer board processes are not set up for good power integrity. They have a center core for layers 2 and 3, with 1 and 4 as foils. It ought to be two cores with prepreg between, giving the low impedance for internal power planes.
I remember a W1GHZ project from the 00s or maybe even the 90s where he eventually got hairpin filters for low microwave frequencies to work using cheap PCBs.
Generic fiberglass is NOT consistent across multiple mfgrs enough that you can spec it for mass production from any mfgr in any country at any time, you need the very low variation specified "RF PCB material" to mass produce precision untuned hairpin microwave filters.
HOWEVER, the cool part of the project was generic fiberglass is consistent enough across one mfgr over a short period of time, that you can characterize product X shipped from PCB house Y and then you can very consistently get boards at the proper resonant frequencies.
The meme was not overly successful in the amateur radio microwave community, sadly, but the project did work perfectly well at the time.
Isola 370 has good consistency, but I’d still only do resonant structures on Rogers laminates. The RO4000 laminates are low cost and work great, but you are paying full panel prices when ordering even the smallest PCB. Even if it were made in China, I’d never trust them due to counterfeit materials.
Ordering from china is inexpensive but it takes time.
I can send a pdf to a local printing house in the evening, get the negatives around the noon while the PCB is drilled. Then etch and mask it in about 2 hours and solder all the vias in next 3 or 4. (I wasn't able to get reliable plating.) Overall one day to have a prototype ready.
Where are you getting this kind of turnaround? Is this in the US? I purchase a couple dozen protoboards a year and am constantly looking for US board houses, but they all have lot charges of at least $500 and take a couple days just to get a quote back.
I use OSHPark for prototyping. I don't know where they send the boards, but they hit a reasonable compromise between cheap & fast: $5/sq inch for 3 boards. The quoting is immediate on their website. I usually have boards back in a week after ordering from OSH Park.
For production quantities, I use JLCPcb (China) with Airborne for shipping (so it doesn't take a month to get here). Their quality is very good. Pay the extra $20 and get a stencil to make paste application easy. It's worth it. I think my last order of 50 board roughly 2.5x3.5" with a metal stencil came to about $70 total with an additional $30 for 3-5 day shipping via Airborne.
In Poland I can get a board or 3 in a week but it will be expensive (about $150-$200 for a small one), normal, cheap delivery time is about 3 weeks. China is even cheaper but takes 4-8 weeks.
That's why I made a lot of boards myself as described in my comment - I get film exposed at professional house that does it for offset printing, and this is pretty fast and cheap (about $3-$4). I pick it up myself. The rest is done with photosensitive 3M dry film and Chinese wet soldermask. Drilling on DIY CNC.
If you pay for fast delivery, a Chinese board can be in your hands within a week of ordering. Sure, it's not instant, but you can do something else in the mean time.
What I usually do in the mean time is.... revise the board.
Sometimes I get 4 revisions in the pipeline where 3 versions I haven't even received yet are already obsolete and go right from the mailbox to the trash.
It's a limitation because this only works for the back and the front layer. You can not reliably connect to layers in between like this and you can not do 'blind' vias (vias between two or more layers not being back or front).
Back when I made my own boards, I used a modified pad shape (okay, I wrote my own layout program) with a tiny little hole in the middle, and after developing the board, the hole provided just enough of a dimple to guide the drill.
I made my last homemade board, some kind of electronic music circuit, in 1993. Today, there are so many interesting chips that are sold as breakout boards by Adafruit, Sparkfun, and the like, that it's easy to prototype by just stringing wires between boards. For the next level -- has to be nice but not commercial and not on my own nickel -- ExpressPCB is my friend.
WRT drilling problems the industry standard solution is make a little hole in the PCB layout so there's a ring of copper and use the stiffest sharpest carbide bits you can get but realize you can't hand old carbide the slightest side force will crack the bit, need to milling machine it.
WRT vias there is a circular problem going on with thermal vias. If you're building a shipping product with qty 10K+ and running into tight financial and thermal constraints the only way out sometimes is a field of vias perfectly flat with the silicon soldered to it and the vias dump the heat, neither air or PCB laminate are good enough thermal conductors.
So if you're building at home and want to pretend to be extremely constrained then you can manually perform the mass production steps necessary to use vias as heatsinks.
However the more rational less circular way to solve the heat problem for onesie-twosie construction projects at home is to think outside the box and bit and throw money at the problem. You can't afford a 50 cent heatsink on some consumer products but I certainly can afford fifty cents in some one-off thing I'm bodging together on my workbench at home. Or say you need squeaky clean power for a low level analog project, well, instead of crazy PCB tricks to dump all the heat in one 7805, dump some of the heat in a switcher running at 9 or so volts and have the linear reg turn noisy 8 V into clean 5V. The "CMRR" equivalent datasheet value for a boring old linear from the 70s is like 80 to 100 dB from memory so it'll be clean enough for analog work. If you try adding a $2.50 switcher to the BoM at work on a qty 100K project "to get out of needing thermal vias" your boss will have a quarter million dollar heart attack, but at home on a qty 1 project that $2.50 is like a rounding error and I just don't care. Or bodge in a beast of a heatsink if you need silence. Or bodge in a fan if you can afford the acoustic and possible electrical noise.
Now if you're trying to "Work at Home" and do a prototype, just while at home, then you have to work as if you're at work.
There are some nifty microwave ICs that as usual turn 99% of their input power into heat and they might "Need" to push all their heat thru a little grounded field of vias. Even that can be bodged using a milling machine and lathe if you make a roundish hole and plug it with a solid slug of copper. Remember labor cost at home for hobby projects is $0/hr, so something that would add ten minutes manual labor to a million product production run is perfectly acceptable at home but unthinkable at work if you can just use vias.
One thing to watch out for: you're correct on jellybean linear regulators having a PSRR of ~80dB, it is frequency dependent. It is usually good at low frequencies (couple kHz) bu usually falls off as you approach switching frequencies of modern DC-DC solutions (a couple hundred kHz to couple MHz). It is not uncommon to see PSRR of ~20-30dB at 1MHz (e.g. datasheet of L7805 doesn't even show PSRR beyond 100kHz, LM317 is around 20dB at 1MHz).
There are linear regulators that are better in this regard though (LT3042 is my favourite for high PSRR).
>> My biggest issue with DIY boards is drilling the holes and in particular getting the bit in the center of the pad.
Get a small drilling stand that can let you mount the PCB in a vise.
Now, I would still use pre-drilled prototyping board. Drilling all those holes is just such a hassle. I want to focus on designing circuits and boards, not drilling them.
I drill rarely, when the component must be mounted directly to the prototyping board but does not conform to standard 2.54mm pitch or will not fit the premade holes.
> My biggest issue with DIY boards is drilling the holes and in particular getting the bit in the center of the pad. This is such a pain, I now just use pre-drilled protoboards and hookup wire on one-off or prototypes.
I haven't done it with PCB, but for wood when I need a precise location and have no drill press, I use a nail to make an indentation and use my smallest drill bit to drill it through, then I can use progressively larger bits, as they will tend to self-center in the hole.
> My biggest issue with DIY boards is drilling the holes and in particular getting the bit in the center of the pad.
Trick 1: If you align all your holes on a 2.54mm pitch grid (which will be easy enough if you're doing a through-hole design) you can tape or clamp some spare protoboard on top before drilling. That will then act as a jig, getting all your holes nicely lined up with one another.
Trick 2: Drill the holes before you mask the board. This is particularly easy if you're drawing out the board by hand, with etch-resist pen.
Along the same lines, Favorit makes a press (along with tools / dies and rivets) that takes a lot of the guesswork out of through hole plating. I’ve been a long-time user of the press after getting frustrated with the “solder a paper clip” method of through hole plating, and its greatly increased the yield of my CNC PCB projects.
If you're wanting to do platted through holes idk why you don't just use the same elecroless plating chemical that board shops use. You could call them and ask for samples and would get enough for all your hobby projects.
They don't, at least not conductive ones. Volterra drills through holes but they aren't conductive. They recommend using rivets or less preferred is filling them with conductive ink and baking, but that's a huge pain and I imagine most people just use a piece of wire instead or try to avoid vias altogether. Othermill works the same way.
Usually you want to design your board around this limitation, which is why I generally don't recommend them for people looking to prototype PCBs with the intent of having them professionally manufactured later. IMO their niche isn't really as a replacement for a manufacturing service, more as a quick turnaround tool that lets you make one-off boards as needed.
Pretty much this. If you can solder though-hole resistors on one side of the board, and you can solder though-hole resistors on the other side of the board, you can line the holes up and solder a trimmed-off resistor lead to both sides. Sure it takes time, but so does soldering through-hole components in the first place.
I don't think this pays off for most people. Most circuits can be prototyped without printing boards. I use a combination of generic function boards, adapters for SMD components, DIP versions of components, etc. to be able to easily put together a working prototype on 100mil grid.
Worst case, I use dead bug technique for a non standard component I don't have adapter for.
For large projects I might be splitting the problem into multiple separate functions and have a prototype of each function separately, later to integrate all to a single board.
For example, I am currently working on a controller that will control a heater, pump and couple of valves based on input from a bunch of very sensitive temperature sensor. I have two separate boards one for temperature measurement and one for 1kW of AC through bunch of relays, triacs, SCRs and so on. Each relay, triac and SCR is mounted on its own generic small boards (something like this: https://pl.mouser.com/ProductDetail/Chip-Quik/PA0185?qs=gjT6...), soldered in turn onto larger generic board (for example https://pl.mouser.com/ProductDetail/BusBoard-Prototype-Syste...).
All these are hooked up to a Nucleo board where I develop my software.
Once I have working prototype it makes no sense to spend time trying to make a half-assed board. I would very much prefer to design much better looking, much tighter tolerances board based on tolerances accepted by the PCB manufacturer and wait for it for couple of days.
If I really needed to work on the circuit in the meantime I still have the prototype. While waiting for a nice shiny board I can still working on software, find bugs, etc.
For anyone unfamiliar with the term, dead bug construction is where you put the integrated circuits upside down (like a dead bug) and solder wires and other components directly to their leads. Here are some photos and discussion:
Most those photos are other techniques. What I call dead bug is specifically to solder chip to a prototyping board by inverting the chip and soldering wires directly to the chip pads.
> As bottom line – I am not going to start make all my PCBs at home. I don’t think it is worth it in 95 % of the cases. I tried this experiment for fun and for educational purposes.
Unless you value your time at < $5/hour, it's always more money efficient to just pay for a PCB for a prototype rather then getting into dead bug soldering. There is no such thing as a "quick" dead bug project, and the time taken is always more than the time to lay out a pcb plus its cost.
You probably misunderstood. Dead bug does not mean creating entire circuit out of PCB. Dead bug is a single chip you put on its back (hence, dead) so that you can get EASY access to its pad. This is only very small number of cases (I did TWO this year), for everything else I have selection of adapters and development tools to make my life easy.
It doesn't cost a lot more of effort to solder a single SMT chip this way. I already solder everything else to the board anyway so cutting couple pieces of wire to tack them one by one does cost maybe couple additional minutes for a small size chip, and then I can go to working on the rest of the prototype.
> When I soldered a pin header to the hole and rip it [off], I could nicely see the cylinder filled with solder covered by the copper from the hole. Again – just like you do it on a pro board.
This made me chuckle. Tearing out the vias is definitely something that happens, but never something I've meant to do...
I etched a few boards while in college, including one while I was in Australia preparing for a solar car race. Eventually the solar car team found a local PCB manufacturer excited to sponsor us, though. It was pretty great getting 1-2 day turn around on boards as a college student. I tried to keep everything standard 2 layer to make it easy to manufacture.
The sponsor kept saying, "Come on, give us a challenge. We specialize in heavy copper!" So... I designed a 40-pole 4-turn brushless motor stator meant to do 7KW+, with each phase consisting of 3 parallel traces to minimize eddy current losses. 32 layers of 8oz copper, 90%+ copper fill. We also pushed it to the physical limits of chemical etching in terms of trace spacing.
They tried their darnedest to manufacture it. They had to program their mills to peck drill for the first time. The plate was so heavy that it damaged several conveyor belts on the line. They managed to get me a 16 layer prototype, and we actually drove the car for a couple hours on it (but with an appropriately reduced torque limit due to the increased resistance). They couldn't get all the kinks out of the full stackup before our race deadline, but they eventually produced a 6oz 28 layer stackup that was pretty much perfect. I still have it as a momento. I think it weighs maybe 15lb.
36 comments
[ 2.8 ms ] story [ 77.8 ms ] threadAnytime I lifted a trace or needed to repair a via, I just coiled up a 24-26 Awg single strand wire and stuffed it in the hole. This usually left enough room for the pin and also wicked the solder through.
My biggest issue with DIY boards is drilling the holes and in particular getting the bit in the center of the pad. This is such a pain, I now just use pre-drilled protoboards and hookup wire on one-off or prototypes.
Not that a lot of that can't be done DIY, but to do really good quality boards at home you start to spend a lot more time refining manufacturing processes than actually doing electronics! That's fine if that's your jam, it's a valid hobby, but I'm much more interested in making more interesting projects!
Even the 4 layer board processes are not set up for good power integrity. They have a center core for layers 2 and 3, with 1 and 4 as foils. It ought to be two cores with prepreg between, giving the low impedance for internal power planes.
Generic fiberglass is NOT consistent across multiple mfgrs enough that you can spec it for mass production from any mfgr in any country at any time, you need the very low variation specified "RF PCB material" to mass produce precision untuned hairpin microwave filters.
HOWEVER, the cool part of the project was generic fiberglass is consistent enough across one mfgr over a short period of time, that you can characterize product X shipped from PCB house Y and then you can very consistently get boards at the proper resonant frequencies.
The meme was not overly successful in the amateur radio microwave community, sadly, but the project did work perfectly well at the time.
For production quantities, I use JLCPcb (China) with Airborne for shipping (so it doesn't take a month to get here). Their quality is very good. Pay the extra $20 and get a stencil to make paste application easy. It's worth it. I think my last order of 50 board roughly 2.5x3.5" with a metal stencil came to about $70 total with an additional $30 for 3-5 day shipping via Airborne.
That's why I made a lot of boards myself as described in my comment - I get film exposed at professional house that does it for offset printing, and this is pretty fast and cheap (about $3-$4). I pick it up myself. The rest is done with photosensitive 3M dry film and Chinese wet soldermask. Drilling on DIY CNC.
Sometimes I get 4 revisions in the pipeline where 3 versions I haven't even received yet are already obsolete and go right from the mailbox to the trash.
I made my last homemade board, some kind of electronic music circuit, in 1993. Today, there are so many interesting chips that are sold as breakout boards by Adafruit, Sparkfun, and the like, that it's easy to prototype by just stringing wires between boards. For the next level -- has to be nice but not commercial and not on my own nickel -- ExpressPCB is my friend.
You you can also make a simple probe for autoleveling and mill the pcb too, but imo it takes too much time. Etching has way better quality too.
WRT vias there is a circular problem going on with thermal vias. If you're building a shipping product with qty 10K+ and running into tight financial and thermal constraints the only way out sometimes is a field of vias perfectly flat with the silicon soldered to it and the vias dump the heat, neither air or PCB laminate are good enough thermal conductors.
So if you're building at home and want to pretend to be extremely constrained then you can manually perform the mass production steps necessary to use vias as heatsinks.
However the more rational less circular way to solve the heat problem for onesie-twosie construction projects at home is to think outside the box and bit and throw money at the problem. You can't afford a 50 cent heatsink on some consumer products but I certainly can afford fifty cents in some one-off thing I'm bodging together on my workbench at home. Or say you need squeaky clean power for a low level analog project, well, instead of crazy PCB tricks to dump all the heat in one 7805, dump some of the heat in a switcher running at 9 or so volts and have the linear reg turn noisy 8 V into clean 5V. The "CMRR" equivalent datasheet value for a boring old linear from the 70s is like 80 to 100 dB from memory so it'll be clean enough for analog work. If you try adding a $2.50 switcher to the BoM at work on a qty 100K project "to get out of needing thermal vias" your boss will have a quarter million dollar heart attack, but at home on a qty 1 project that $2.50 is like a rounding error and I just don't care. Or bodge in a beast of a heatsink if you need silence. Or bodge in a fan if you can afford the acoustic and possible electrical noise.
Now if you're trying to "Work at Home" and do a prototype, just while at home, then you have to work as if you're at work.
There are some nifty microwave ICs that as usual turn 99% of their input power into heat and they might "Need" to push all their heat thru a little grounded field of vias. Even that can be bodged using a milling machine and lathe if you make a roundish hole and plug it with a solid slug of copper. Remember labor cost at home for hobby projects is $0/hr, so something that would add ten minutes manual labor to a million product production run is perfectly acceptable at home but unthinkable at work if you can just use vias.
There are linear regulators that are better in this regard though (LT3042 is my favourite for high PSRR).
Get a small drilling stand that can let you mount the PCB in a vise.
Now, I would still use pre-drilled prototyping board. Drilling all those holes is just such a hassle. I want to focus on designing circuits and boards, not drilling them.
I drill rarely, when the component must be mounted directly to the prototyping board but does not conform to standard 2.54mm pitch or will not fit the premade holes.
I haven't done it with PCB, but for wood when I need a precise location and have no drill press, I use a nail to make an indentation and use my smallest drill bit to drill it through, then I can use progressively larger bits, as they will tend to self-center in the hole.
Trick 1: If you align all your holes on a 2.54mm pitch grid (which will be easy enough if you're doing a through-hole design) you can tape or clamp some spare protoboard on top before drilling. That will then act as a jig, getting all your holes nicely lined up with one another.
Trick 2: Drill the holes before you mask the board. This is particularly easy if you're drawing out the board by hand, with etch-resist pen.
https://www.vpcinc.com/Category/Favorit-Through-Hole-Rivets-...
If you're wanting to do platted through holes idk why you don't just use the same elecroless plating chemical that board shops use. You could call them and ask for samples and would get enough for all your hobby projects.
Usually you want to design your board around this limitation, which is why I generally don't recommend them for people looking to prototype PCBs with the intent of having them professionally manufactured later. IMO their niche isn't really as a replacement for a manufacturing service, more as a quick turnaround tool that lets you make one-off boards as needed.
Pretty much this. If you can solder though-hole resistors on one side of the board, and you can solder though-hole resistors on the other side of the board, you can line the holes up and solder a trimmed-off resistor lead to both sides. Sure it takes time, but so does soldering through-hole components in the first place.
Worst case, I use dead bug technique for a non standard component I don't have adapter for.
For large projects I might be splitting the problem into multiple separate functions and have a prototype of each function separately, later to integrate all to a single board.
For example, I am currently working on a controller that will control a heater, pump and couple of valves based on input from a bunch of very sensitive temperature sensor. I have two separate boards one for temperature measurement and one for 1kW of AC through bunch of relays, triacs, SCRs and so on. Each relay, triac and SCR is mounted on its own generic small boards (something like this: https://pl.mouser.com/ProductDetail/Chip-Quik/PA0185?qs=gjT6...), soldered in turn onto larger generic board (for example https://pl.mouser.com/ProductDetail/BusBoard-Prototype-Syste...).
All these are hooked up to a Nucleo board where I develop my software.
Once I have working prototype it makes no sense to spend time trying to make a half-assed board. I would very much prefer to design much better looking, much tighter tolerances board based on tolerances accepted by the PCB manufacturer and wait for it for couple of days.
If I really needed to work on the circuit in the meantime I still have the prototype. While waiting for a nice shiny board I can still working on software, find bugs, etc.
For anyone unfamiliar with the term, dead bug construction is where you put the integrated circuits upside down (like a dead bug) and solder wires and other components directly to their leads. Here are some photos and discussion:
https://www.google.com/search?q=dead+bug+construction
http://dangerousprototypes.com/blog/2012/11/15/fine-pitch-bg...
I don't put photos of my prototypes on the Internet, but this is how it, more or less, looks like (only I use board with holes):
http://dangerousprototypes.com/blog/2012/10/24/who-says-bga-...
It doesn't cost a lot more of effort to solder a single SMT chip this way. I already solder everything else to the board anyway so cutting couple pieces of wire to tack them one by one does cost maybe couple additional minutes for a small size chip, and then I can go to working on the rest of the prototype.
This made me chuckle. Tearing out the vias is definitely something that happens, but never something I've meant to do...
The sponsor kept saying, "Come on, give us a challenge. We specialize in heavy copper!" So... I designed a 40-pole 4-turn brushless motor stator meant to do 7KW+, with each phase consisting of 3 parallel traces to minimize eddy current losses. 32 layers of 8oz copper, 90%+ copper fill. We also pushed it to the physical limits of chemical etching in terms of trace spacing.
They tried their darnedest to manufacture it. They had to program their mills to peck drill for the first time. The plate was so heavy that it damaged several conveyor belts on the line. They managed to get me a 16 layer prototype, and we actually drove the car for a couple hours on it (but with an appropriately reduced torque limit due to the increased resistance). They couldn't get all the kinks out of the full stackup before our race deadline, but they eventually produced a 6oz 28 layer stackup that was pretty much perfect. I still have it as a momento. I think it weighs maybe 15lb.