/diy/ - Do It Yourself

>>327028

free quality CAD/CAM software at 4pcb.com put that home hobby to sleep for good for me.

>>327031
OP here
Yeah I use 4cb/advanced circuits a lot for one-offs.
BTW everybody, if you are student you can get a 2layer board up to 60 square inches for $33 (comes to about $50 shipped). That is a real good deal.

In retrospect between this and all the batch services out there there's no reason to piss away time etching your own at home, unless you like the challenge

Another one I use for 4layer protos is oshpark (former laen's PCB order)

6mil trace/space, 13mil minimum drill. I believe the rate as $10/sq in which is a good price if you have smaller-ish boards. I have done BGA designs and apart from the crappy silkscreen the board is USA good quality

>>327028
>anybody making some PCBs right now?
I'm halfway through CNCing 120 of these bad boys.

I print on single sided boards to save time and money. Also I make the traces as absolutely wide as possible because my machine can't cut thin lines like pro equipment.

>>327042
kinda hard to see with the toolpath, what is it?

>>327042
>120

but why? that's exactly the sort of thign that scales so well at places like 4pcb.com. if you consider tooling and wear adn time, precision, features (screens, resists, masks) it's cheaper too.

>>327034
>In retrospect between this and all the batch services out there there's no reason to piss away time etching your own at home, unless you like the challenge

I could not agree more. I've done a couple boards at home before and wouldn't go back to the time, stink, mess, and space required. The cost of getting a board shop to do it for you is money well spent in my humble opinion.

I've got a 24 layer class 3 board I'm working on for Aerospace, with over 12,000 components. Crazy shit

>>327085
You should post a section of the board if you can.
I'd be interested to know major components and stackup if possible, if not I understand

I'd post of a pic of the last board I did but the product is still in development

>>327085
ITAR prevents me form releasing any information other than general descriptions. Major components are FPGAs, buffers, SDRAM parts, and 3 switching power circuits. The discrete parts bulk up the component count like crazy.

>>327093
I'm an altera guy. I will post some pics of my little projects as I can

Pic related; Cyclone 3 on 4 layers with 2 banks of DDR

>>327097

Making a few boards atm. DACs, RS-485, BeagleBone. Easy stuff, I try to keep it 8/8 and two layer. Using KiCAD which I honestly think is pretty efficient despite some quirks.

>>327099
WOW purple solder mask, man if it wasn't so damn expensive....I'm tired of green all the time

PCB thread! I'm getting a chubby.

But seriously I have questions. To what lengths do go to design for and test for electromagnetic compatibility regulations?

And how do you assemble and solder SMT components for one-off boards/prototypes? BGA packages I'm particularly curious about. One of these days I might want to use a high density bga device if I come up with the right project.

>>327109
Metcal temp controlled soldering station, microscope, etc. for soldering.

For BGA parts, a ceramic heating element (see Ebay) can be used to warm up the board to reflow temp. Then if you lined up the part using solder paste or flux REALLY well, if you hit it with a heat gun (or hair dryer) it will flow itself down to the pads. Worked well for our senior design project

>>327109

EMC compliance -- basically design for minimal EM emissions, stuff like proper power filtering, signal termation, good grounds.

Check out "High speed digital design: A handbook of black magic" it's a book from 1993 but available if you know where to look. Though it mostly covers old-style card cage DIP boards, there is a lot of theory and rules of thumb that can be applied towards modern designs as well

I have soldered 324-bga with just a kitchen oven and SMD station, but I would HIGHLY getting a proper reflow oven. Seriously, after I got a real one, I would never look back. It's the only way to do bigger boards because you need even and precise heating, something you won't get with localized heating

(continued)

I use a Aoyue 968 combo soldering iron/rework station. I've used it to assemble probably over 30 boards and reflow several hundred over a 3 year time span. The iron cable is finally starting to get flaky, but the whole thing cost $160 and if you're doing this regularly I would spring for a separate Hakko iron and secondhand Metcal rework station.

>>327109
There's evidence here and there that you can solder BGAs with the hotplate method. Personally I'd go with a converted toaster oven. I dunno if I'd try anything smaller than 1mm or 0.8mm pitch though.

>>327118
>>327117
>>327115
>>327111
Awesome, thanks for the replies.
If the time ever comes for me to get into SMT I don't mind spending the money for good tools (and books).

I've already spent lots, but for a hobby this fun, it's worth it.

going through my bookmarks and cleaning out
anyone doing anything >1mhz on a pcb should read these

http://www.analog.com/static/imported-files/tutorials/MT-101.pdf

http://www.vagrearg.org/?p=decoupling

http://www.analog.com/library/analogDialogue/archives/39-09/layout.pdf

http://www.rickmiller.com/si.htm

http://www.analog.com/library/analogDialogue/archives/35-05/latchup/

I started a project earlier this year for an objective-two insipired portable amplifier and found that the voltage converter (3.7v to +/- 7v) will need a reflow oven and a 4-layer PCB. I've already got a case and batteries for it, but I need to finish up the schematic. It's such a pain making schematics using the sized parts you're planning on using for the project later even if you don't know what sized ones you're going to need. Someday I'll pick it back up, but until then I'll probably just make some arduino shields for practice with Eagle.

Thanks for the random tips OP

>>327887
pick another boost converter? there are tons out there

Anyone got any good links on designing car audio stuff?

>>327034
>>327072

I dont know if i agree with you.

I get a double sided A4 sheet for like 6usd, rough guess but you would get HCl2 and H2O2 for the same price. The chemicals would last for, i dont know, 10-20 A4 sheets, so the price is pretty negligible.
You can make your own A4 etched board for 7,2USD, that is pretty ridiculous.

But i have to agree with you on it being a great effort to produce a board at home. I myself find that really discouraging too. But i dont have tons of money to waste.

However its faster, you can be done in 10-20minutes.

On the other hand, the quality and durability of the pcb will be not even close to a manufactured one.

Note: This is my opinion only.

>>328039
>However its faster, you can be done in 10-20minutes.

This is the exact reason why I occasionally make my own boards. It's still a hassle and the results are shitty, so I tend to regret it afterwards
Ok, it takes much longer than 10-20 minutes for me, but still.

>>328039
Same here. I usually start doing a toner transfer for double sided PCB after dinner and usually populating the SMT board around midnight. The fastest turn around I could have is 2-3 days turn around from a board house.

As for durability, as long as I don't try to go below 7 mil tracks and do a lot of rework, what I have is fine. Stuff I built last for years. I have played with 64 pins 0.5mm pitch TQFP and even leadless MLF packages.

If I have to do a lot of bus routing for >100 pins 0.5mm device etc, a proper board from a board house is the way to go. Solder mask/through hole vias are very nice to have for soldering and help to stick thin tracks to the PCB.

>>328039
youre reasonable reply is reasonable.

it's all tradeoffs inn'it. it is also simply fun to make your own, ... until it isnt, and thats where i got to. more power to ya for doign your own.

i've also used the PCB system for surface etching metals.

what do you use for resist? i had good luck with the GC aerosol photoresists, but i dislike having all the chemistry about.

>>327887
O2 with a boost converter? I imagine that takes some careful filtering, if indeed it's not too much noise to deal with.

>>328050
Leadless MLF = QFN?
Havent tried those, but i have some components waiting for trying out.
It is easy with solder paste, i dont have any so the only way i could think is pre tinning the pads, and pressing the device down while reheating the pads and the side of the leads. Chances are high it wont work but meh.

I only order pcb when my design is flawless and was already tested on a homemade pcb. I usually wait and gather couple of designs so they can be merged together, that takes about half a year or so.

>>328053
I use toner transfer.
I print on a magazine paper or similar, non porous paper, so the toner is likely to stick to the copper rather than sucked in by the paper.

>>328053
I use HCl2 and H2O2 mix to etch it, it is certainly not healthy but with caution and a little breeze you can end up with healthy lungs and not vomiting blood after you are done.

I tried photoresist but failed, i was impatient at the time so did not gave it a second try.

I forgot to mention the drilling in the previous post.
For me it is the most risky part since a completely fine board can be ruined with a silly move. I tend to rush it and not position the eye perfectly under the drill and tear the copper off.

I have made a little drill press just for the job, not pretty but does a great job. I might make a pic of it some time later.

Has anyone used the Seeedstudio fusion pcb service? I've used it once and they quality was good. All holes plated through, silk screen on top and solder mask on both layers. It's the only time I've had PCBs made so I do not know if this is better or worse than other places. It cost 13 dollars for 10 boards 50x50 mm shipped to the EU.

>>328117
I have once, it came out pretty well. I think this was an 8/8 board with the exception of the MEMS stuff, and I assembled a bunch with no problems.

Generally I like to support domestic manufacturers but this was a student project where cheapness was a priority.

>>328109
I tried photo-resist with transparency, the result is slightly better than my toner transfer, but not enough to justify the extra trouble trying to cut presentized PCB in the bathroom under red light.

Back in school, the tech told me to making a film and helped me make PCB with proper equipment: vacuum plate to hold down the film, echant spray system etc. That actually works pretty well and I can see going down may be down to 5 mil track might be possible. I can see that more consistent.

http://en.wikipedia.org/wiki/Quad-flat_no-leads_package
MLF micro-leadframe Amkor Technology
QFN quad flat no-lead package

Yeah, like I said, I usually plate my copper with hot solder, so doing small leadless parts can be done with a soldering iron (need to extend the pads in the footprint) or hot air gun.

As for drilling, using mostly SMT means there are a lot less holes to drill -just the dozen or so of vias. I use large 50 mil square via pads as my drill bit is 0.032" One of these days, I might get around to getting a drill press.

I have used ferric chloride and ammonia persulphate and prefer the latter as it doesn't form a hard layer on the copper. I used a heated bath and cover the board in about 1-2mm deep in echant. Too much bubbles might rub off toners in some of the fine traces.

meta comment: this thread is filled with win; examplar 4chan thread. even the grousing and h8 was great.

oh and of course, fuck you all very much. and dicks, always dicks.

4chan, ftw!

>>328209
i spoze you might as well have page two.
(if you are really old you might remember that this ad actually ran in ELECTRONIC DESIGN or EE TIMES or something, in April, 1972?. I'm still waiting for my samples, fuckwads.)

>>328124

that may be the most beautiful homemade PCB I've ever seen.

>>328209

so beautiful... bwahahaha

i'd seen that one before but with a more careful look, it's full of gold

>>328215
You made me look at it again, and... oh god.

>>328214
Thank you for the compliment. I think I could still do better, though.

>>327028
man just switched to mixing my own etchant , and developer, so much cheaper and way more effective when dosed out properly, ammonium persulfate ftw

>>328124
hey what do you use for laminate and silkscreening, shit looks boss, and im looking to step my game up to that level

also general question, da fuck is the difference between say this fine homemade pcb and some multilayered factory made one?

>>327028
I have never designed a PCB before but I can make schematics in Eagle.

I need to make a somewhat simple (I think) pcb for about 6 through-hole components like dip-24s and dip-8s and some resistors and capacitors. Do you have any suggestions on where I should start?

>>328373

There's more to PCB cost than just the dollar amount you pay. For anything even remotely complex where I am taking lots of chances and going into unknown territory, how would you like to spend days debugging and tracking down what turned out to be 1 shorted trace on your homemade pcb? Or an intermittent connection, or separating vias from annular rings, but only after several thermal cycles?

It's a tradeoff you have to decide for yourself. In my case the money I spend on having my pcbs commercially made more than offsets the time and frustration/lost productivity.

When the boards are electrically tested via flying probe, it lets me focus on my actual design and not random physical gremlins.

If you're doing simple boards and your time is worth little, then by all means. But as you improve and become pro you can't afford to piss away valuable time anymore.

YMMV

>>328377
Typically in eagle you decide your physical package when you place down the part in the schematic. You can change it later "technology tool" but better to get it right the first time.

Once you have all your components laid down, do an ERC and check to make sure nothing is blatantly wrong. You may get pithy stupid errors, that's fine, I have shipping boards with 4000 DRC violations because i can't be bothered to spend 9000 hours tweaking the settings to make them go away

ANyway, once you File>Switch to Board, you'll get a bunch of parts crammed into the corner of thscreen. Now move them around the board so they make sense. Ergo, parts grouped according to function.

From there, post what you have and I'll get you started with routing it

>>328373
The person used seed studio which subcontract to Chinese PCB contractor, so not like it was done in the bathroom like some of us.

Someone have experience working with microstrip and special dielectric PCBs? i want to start making LNA in the band of 2.5GHz but i dont know how good you can tune a microstrip with conventional methods like photoresist, transfer and non professional cnc (like the lpkf with a set of rf bits), any suggestions?

>>328631

Paul Wade W1GHZ has had good results with regular FR4 PCB material. If you are designing the circuits yourself, you need to width of the microstrip to match the dielectric constant of the board material. The good news is that gain is relaitvely easy to get at 2.4GHz. Check out Mini-Circuits ERA and GALI series of MIMICs. You don't need Duroid at 2.4GHz.

>>328631

No idea about fancy dielectrics, but ordinary FR4 (and microstrips made on it) works reasonably well at 2.5GHz.

>>328534
im a poor college student so my time aint worth crap, but i've actually got some pretty sweet smd boards made

>>328566
oh well that takes away from the effect i gues :/

>>328631
http://www.sigcon.com/Pubs/news/4_5.htm

This article shows you what the effects of dielectric losses at high frequency does to the dielectric constant. Too much math to be quotes here.

You'll need to define the term "working well" by how much signal losses can you tolerate etc.

https://encrypted.google.com/url?sa=t&rct=j&q=dielectric%20losses%20fr4&source=web&c
d=1&ved=0CDAQFjAA&url=http%3A%2F%2Fgrouper.ieee.org%2Fgroups%2F802%2F3%2F10G_study%2Fpublic%
2Fjan00%2Fsayre_1_0100.pdf&ei=Vc2eUODEEMS_yQG1rYDABA&usg=AFQjCNFOSNB8C0XH27RKoECPEmtwZyD_cQ&
amp;cad=rja
>Limits of FR-4 in High-Speed Designs - IEEE.org

http://www.avagotech.co.jp/docs/AV02-0725EN
>PCB Layout Guidelines for Designing with
Avago SFP+ Transceivers

>>328648
>>328650
and how hard is to tune the input/output matching networks?, i am using an open stub at the ends but i dont know how well you can tune the strip by cutting little by little, searching in the web i still cant find an example of how to tune the strip, find a pdf about one guy doing a mixer and tunning by cutting with a dremel an a cutter bit.
About microcircuits and mmics, where i live unless you buy in digikey and import the circuits, you cant get those, and importing things here is too expensive so im using RF transistors and stubs basically

>>328687
Some people have used UV Curable PCB ink from ebay and some had used stain glass ink and bake their PCB.

Go google for "homemade solder mask"

>>328691
Never done it with stripline elements but at 2.5 GHz you're going to need to model the package/board parasitics. I don't suppose you have any test equipment.

>>328689
thank you, im ok in theory of design and, i think ,the deal is to get it real, in the course of Microwave amps we learn to design in base of the books of Pozar microwave engineering and Gonzales microwave amps, the old fashion, the thing with RF design is to make it work, those tricks in tuning and stabilizing irl only people with experience doing it know how, and here in the lab we have the equipment but almost nobody really make circuits they stay in paper, you know what i mean

>>328694

328689 here.

The fact that 802.11N dual band router exists tell you that someone managed to do 5GHz on FR4 on real life product that passes FCC too etc.

We went though playing with 3.125GHz on GETEK for 30cm of signals stripline and through 2 connectors and 2 vias.

Lot of debugging/modeling, expensive equipment - two 10GHz scope you name it we have it. 2-3 spins and 6+ months later all that we accidentally found the problem in the vendor's FPGA substrate (i.e. that green piece of PCB that their chip is bonded on) because of a assembly mistake. After some major hammering their corporate level, we get their substrate layout and confirmed what we knew.

Our simulation results agrees with reality and theory works. So been there done that. Unfortunately most of what I learn in routing 3GHz signals in a multilayer digital boards doesn't help you.

>>328691

You can certainly tune by old fashioned cut and try. Make the stub too long, measure the return loss AND gain. Use a caliper to record the length of the stub. Use a razor blade and slice off 10 mils or so and do it again. Eventually, you will be able to draw the curve you are after. If you need to add copper, use a cyanoacrylate glue like Super Glue to bond the foil to the board.

Another technique is called snowflaking. Cut a small square of copper and use a wooden stick (toothpick?) to slide it up and down the trace. If you have a network analyser you can look at the return loss easily while you do this. You might find places along the length of the trace that need more copper. Super glue a piece there. And do it again.

This is far easier in a modeling tool than it is on a physical board. If you can find it, Ansoft Designer SV (Student Version) is good. Older but still useful is HP AppCAD.

Bear in mind that there is probably a trade-off between the match and the gain.If you are stuck with discrete transistors you probably want to maximise gain, not the impedance match. Also, you might get better overall results if you do the match at the device and leave the microstrip at 50 ohms. Hard to say without building or modeling it.

Sounds like a great learning experience!

>>328691

I haven't done anything particularly impressive (small, relatively simple designs, not attempting to be state of the art), but dremeling has worked just fine. Depending on your requirements, you might be able to avoid microstrips and manage with lumped coils and capacitors.

>>328703
nice experience sharing, its what i have heard of guys who have done microwave design,
system design- ok
calculus-ok
stabilization-ok
optional Low noise coupling- ok
the simulation- ok
components- ready
and fabrication- omg wtf my design of a LNA at 2GHz its an oscillator at 4GHz irl
the primary approach is like that, a lot of time with the network analizer and guru-matching the circuit, i guess is the reason why nobodi here uses the equipment

>>328711
>Bear in mind that there is probably a trade-off between the match and the gain.If you are stuck with discrete transistors you probably want to maximise gain, not the impedance match. Also, you might get better overall results if you do the match at the device and leave the microstrip at 50 ohms. Hard to say without building or modeling it.
thank you, yes the cutt/add technique, nice to hear that im trying with a set of bits in a dremel but the edges end all burred.
first i want to make an simply max gain amp work and next, step in to a low noise, the first one (theoretically) you just need to tune the matching networks of microstrips to get the max of gain at the input of the transistor and tune the end to an 50 ohm network.
in the LNA approach you just need to tone the input network to an a precisely value (give to you by the manufacturer) to get the minimum noise figure at the first biport.

The thing is to get experience tuning this little lines the most precisely

>>328712
Almos all the coils and inductors stop working at about 2GHz because his resonant frequency, you have an inductor that works like an capacitor and a capacitor that works like an inductor and a potato that works like an antenna

things goes crazy

>>328729

There are 0402 size and smaller components which have their self resonant frequencies at 5GHz and higher. Certainly they aren't ideal at 2.5GHz, but you can live with that (depending on your application, of course).

>>328743
0402 mm they look small enough to a 2-3GHz wavelength, i am thinking seriously to buy a pack of digikey of microwave components and inductor/capacitor book type catalog to try the coupling and characterize the components (i am little short in cash right now), do you have a recommendation to an 0402 components to try to work at this frequencies?

>>327042
what program is that from? And what are the orange dotted lines?

The blue looks like the trace the tool follows

>>327034
4Pcb and Advanced circuits are one in the same right?

I've used them a lot at work and they seem pretty good, i guess for home usage i might go with their 133$ for 5 boards (2 layer) or 330$ for 5 boards (4 layer), but that's still kind of expensive for low run stuff.

i think joining a makerspace or something that has a CNC circuit bord plotter and then just a via punch kit would work better for that sort of stuff

>>327085
>class 3 board
What do the diffrent classes usually entail?

>>328749

I've used mainly Murata's LQG15HN series coils. Unfortunately I don't have info about the used capacitors on this machine.

Where did you all learn PCB design and layout? I find that colleges don't teach this sort of stuff much, if at all, in majors like EE/CE/etc.

>>328750
If your boards are small, Laen's PCB service is great.

>>328763
1. copying evaluation boards
2. reading app notes
3. looking at lots of boards and projects over the years

>>328749

ATC is the gold standard for microwave capacitors. At 2.4GHz you can make inductors with thin microstrip traces. Look at the W1GHZ transverters here http://www.w1ghz.org/MBT/multiband.htm If you want discrete parts, Murata make some good inductors but I've never used them as high as 2.4GHz. Read the spec sheets carefully - you may need to tune out the parasitics. Speaking of parasitics, the leads of through-hole parts are significant at these frequencies. Use SMD whenever possible. Personal experience with 1206 parts is OK at 2.4GHz. 0805 is better and in most cases perfectly fine.

you can also make test traces on a board which you measure to see what the dielectric constant/etc are and then adjust the traces as needed

>>328729

>thank you, yes the cutt/add technique, nice to hear that im trying with a set of bits in a dremel but the edges end all burred.

Cut the trace then heat the cut end with a soldering iron to remove it easily.

>>328692
hey thx ill have to check that out
>>328763
to be honest with you i started hand drawing pcb's and digitizing them and just kept going from their, i use kicad and reccommend it its FOSS and is a great alternative to eagle and has a pretty decent number of tutuorials on the interwebs

>>328763

I learnt [1] RF layout from my amateur radio mentor who is a retired RF engineer (30+ years). It is VERY different to digital layout. And microwaves (above 900MHz) is very different to HF.

[1] Learnt isn't exactly the right word. Some days it's more like slaughtering a goat and placing its entrails on the circuit at moonrise...

>>328754

class 3 designation is for very, very important systems. Namely, military and life support /medical equipment that must not fail.

They're clasiffied in terms of allowable downtime

>>328763

i taught myself. 5 years ago I made my first pcb, now I'm routing 533mhz ddr2 with fpgas

study every single you can get your hands on

read appnotes from mfg

lurk irc channels

contrib to open sores projects and pick brains

>>328903
Err right, i got that, but i mean what is done diffrently on the PCB manufacturing/design side to achieve that?

>>328919

i'm not a class 3 design engineer, but from what i gather:
1. prime-spec components all around, this means name brands and overspeccing where it makes sense. like using higher voltage rated caps, using tants/polymer caps instead of aluminum electrolytics, low tolerance resistors

2. very thorough testing procedures -- example would be extreme conditions, stress tests

3. basically every "best practice" ever taken to the extreme. many silicon parts come in an alternate automotive/military spec version with more generous temperature ranges and high MTBF.

>>328923
>like using higher voltage rated caps

wrong- you can always read about why if you're actually interested

>>328930
i have no idea what to google for because i am a flaming noob

please enlighten

>>328908

>open sores projects
good one

>>327899
Please respond

I was thinking of using a class A like an Aleph, but I think that could replace the heater...

bumb

>>329048
IMHO class A in a car is kinda pointless.

here have a pic
my attempt at lead-free bga

Can you recommend ANY software that will Autoroute simple single sided PCB's using jumpers?

Most PCB cad stuff is all double side with layers and vias errywhere.

The last software that had a "jumper" profile was OrCad/Cadence from over 10 years ago. Why no modern CAD that can auto route/place dat shit?

I like to etch my own boards at home.

>>331722

99% of autorouters are shit and don't come close to careful placement and routing by someone with a modicum of experience.

there's no magic solution here. just learn to route your boards more creatively, or break down and pay a few bucks to put your pcb on a batch panel.

>>331722
eagle works, set the autorouter to route on one side only.

>>332438
Yes, but it doesn't do the automatic jumper wires on the other side. At least last time I tried it? Farnell do the free version of Eagle. I am using Designspark from RS at the moment.

One way around is to place tracks with "0 ohm resistors" on the board, but then this cocks up the back-annotation to the source schematics.

I dunno, the Cadence/OrCad thing worked quite well, but I fucking hate the old clunky DOS based interface to it. I'm just surprised that none of the lower budget or student electronic cad packages do single side with jumpers. I mean surely everyone should learn how to make a PCB the old fashioned way?

>>332625

tho i haven't actually used it (i mean, i did to see what it did, but manually route anyway) the package 4pcb.com uses ("PCB Artist") is some high-end package locked to 4pcb.com.

their trick is, you can't get your CAM files (gerber, etc) until after you've uploaded and paid for a board. not a bad deal overall.

A small segment of my latest project

>>332647

fuk. propagation time compensation in copper? shits fast man! video?

>>332653
Just matching the trace lengths of each part, so that the timing matches up. 25MHz SPI bus to DACs, connected to a Spartan 3E. I just wanted the board to not be the problem if the timing doesn't work the first time. No video: nothing to show really, juts a board so far

>>332657

wow, cool pcb! no, 'vid' meant is it a video board. 25MHz doesn't seem *that* picky, but erm, yes, when overkill is easy overkill pays!! why add problems or do shit twice!! /respect

>>332661

I'm flattered thanks :)

>>328751
Program is CamBam. Blue lines represent where the tool is cutting through the copper or drilling so it goes at a slower speed. Orange line is it moving from one place to another as fast as possible.