Which level of assembly for newbie and prototype

I don’t know how best to formulate my question. But I’m motivated to have some experienced advice.

This is my fist PCB design and KiCAD design. Time to send for manufacturing. The design is a 12cm x 12 cm PCB. The design is about 700 pads, 140 devices around a TAS3251 TI audio chip (integrated amplifier and DSP). The design is largely inspired from the TI EVM. But I’m still not fully confident in my work. I need 2 boards. And as it is a hobby, I don’t want to waste too much money… I have an hot gun rework station.

There are promotions on Seedstudio for 20 types of componants (would represent 60% of my SMDs). PCBWay does not seems to charge a lot for assembly.

I hesitate between different options:

  • order the PCBs and stencil, source components by digikey or mouser, assemble myself using stencil. If it works, then next boards I will take full assembly,
  • order 5 PCB with assembly of all small/cheap SMDs and finalyse the expensive components on my own, so that I don’t loose too much if the design is not working,
  • order maxed assembled… and cross fingers,
  • other…

What seosonned professionals would advice ?

Best regards,
JMF

How confident of you of the design. Is it something you’ve had checked out by others on appropriate forums for advice on the schematic design and pcb layout in general?

EEVblog might be a good place to have the overall design reviewed if you are comfortable sharing. The biggest problem with just joining a forum like that is having a feel for who knows what they are talking about. :wink:

1 Like

What are you talking about? This is the internet, therefore I’m the world’s leading expert in all things. That includes things I have no experience in.
:wink:

2 Likes

How many and what parts would need adding? If you can get all your jellybeans from the list at JLCPCB, I would do this as it’s basically free.

However, only do this if all the parts to be added are easily soldered with an iron. Anything needing hot air, you will be better off just stencilling and hand placing the whole board.

1 Like

Well, OF COURSE I am right!! That - and my humility - are my best qualities!

Dale

1 Like

I don’t have an answer for your question but I can point to some factors to consider.

There are several significant unknown factors in your situation, and a few intangibles to consider as well. Are you, personally, motivated to acquire skills at soldering and board-level assembly? Do you have the time and treasure reasonably necessary to undertake that learning? Do you possess the temperament, patience, fine muscle control, visual acuity and spatial comprehension necessary for becoming moderately effective at soldering and assembly? (That laundry list of pre-requisites sounds intimidating but those abilities are fairly common.)

I generally estimate that it takes one build - often cobbled together on a breadboard - just to see if a design idea is possible, and a second one - usually the first PCB layout - to see if it might meet specs and requirements. One, maybe two, design iterations will probably be needed after the first board. And maybe a third to make it manufacturable. That’s the point where you start thinking about building a couple dozen or more at a time. Unfortunately, quick-turn board fabricators have minimum orders of 5 or 10 boards, so you get a lot of shims for propping up the legs of wobbly tables, etc.

I do not have personal, hands-on, experience with third-party assembly houses. A few incarnations ago I worked for a company that DID use contract manufacturers and outside assemblers, though I was never directly involved with that process. 15 - 20 years ago the contract manufacturer never got orders for less than about a dozen copies, and occasionally runs as large as 100 pieces were done by the in-house department that handled repairs, modifications and re-work. I know that times have changed and those numbers may now be meaningless.

Currently - at my job - I turn out commercial-quality assemblies, similar to what you describe, in batches of 10 - 25 units at a time and 4 or 5 batches a year. My equipment includes:

  • A pair of Hakko FX-888 and a Hakko FX-951 solder stations.

  • A drawer full of solder & assembly tools: a couple dozen soldering iron tips (many sizes and shapes), screwdrivers, pliers, tweezers (normal and reverse-action), push-and-poke tools similar to dental tools, solder suckers, draftsman’s knives, Panavise knock-offs, etc.

  • Lighted desktop magnifier (Mag-Lamp XL UN-1030). You don’t have to spend a fortune, but don’t bother with the US$20 variety. Some of the moderately priced units perform much better than others.

  • Binocular microscope, 10X magnification, about US$200 (new from AmScope). A little pricey for hobby applications, but definitely useful!

  • Stencils and solder paste (EP256 from CML Supply) is the only way to go if you do more than 3 - 4 boards of a given type. It’ll start to dry out after a year or so - resurrect it with a squirt of ROSIN BASED liquid flux.

  • An electric skillet from a thrift shop is my “tabletop reflow oven”. Get one with a glass top cover if you can. Took several experimental runs to determine a good process. The kits for converting a tabletop toaster oven to a reflow oven are probably worth their cost but I haven’t made that investment yet.

  • My “stencil frames” are pieces of scrap HDPE sheet (very common where I work) with pieces of old PC boards screwed into place as registration guides. The stencil hinge is a piece of duct tape. An inexpensive polyimide stencil has a lifetime of at least several hundred boards. “OSH Stencils” sends a solder-paste squeegee with every order, but an old credit card or ID card should work.

Our boards have a few thru-hole parts (mostly switches and connectors) but mainly SMT devices. Passives are in 0805 and 1206 packages. (20 years ago I could routinely handle 0603 but the superannuated, tri-focaled eyeballs need a special calibration and favorable astrological alignment to do so today.) I stick to SOIC and SOT-23 variants for semiconductor packages if possible, though I can be successful down to 0.65mm (0.025") pitch.

For the most part, electronic components are CHEAP in the grand scheme of things today. I won’t tell you what I paid for a uA709 opamp back in 1968, but it amounted to a couple hours’ worth of take-home pay from my after-school job. Today, for that same dollar amount (ignoring the general trend of rising prices), I can buy a whole tube of opamps that beat the pants off the uA709. I almost never buy fewer than 10 of anything from distributors like Mouser or Digikey - and often pad an order with extra stuff to get the break on shipping or processing charges. For home and hobby projects it doesn’t bother me (nor does it bother my credit card statement) to have a few dozen parts left over after the project.

These comments may give you an idea of what to expect if you do NOT use an assembly house. I hope somebody can tell you about the option of using an assembly house.

Dale

3 Likes

Thanks for all your feedbacks.
About the confidence, I had a check by a person on DIY Audio, knowleageable of that technology. He raised many layout and grounding improvements, some improvements, but no major mistake. No idea how deep he went in the control. But devil is in the details. fast to have a wrong footprint or other type of error. I don’t have prior experience. So let’s say 70-80% confident.

I would be happy to not reopen the pandora box with too many new people with different advices in the loop. My risks…

halachal comment made me dig more precisely on figures:
Total of real components to solder: 116
0805 and 1206: 76 = 67%
Through Holes: 28
Remaining: 10 (not so much) : 14-TSSOP, HSSOP(56), TO-252, SOT-223, Big SMD selfs, Coil LPS5030, some CP_Elec (maybe cheap to implement). The LM5010 14-TSSOP with the thermal pad down may not be the easiest to solder. The HSSOP(56) would be better with solder paste but should be OK with the soldering iron.

Major part of the cost is in the Though holes and remaining components.

Looks worth checking I think the assembly prices at JLCPCB

JMF

… I will say this with love. if… i could tell myself what to look out for… i … would ‘plot’ in 1:1 print out and triple check. youd be surprised how much you miss; simply by having it in your hands! if you look at your traces and … do not like it… your not done. id be more than happy to look things over, ill sign a NDA if need be. as for assembly… have all your boards out. choose one value of p/n. populate all boards with that value. then… move on. in other words… do all 10K resistors first… then 4.7k then 2.2k … the idea is to eliminate the part numbers… not the board. see??

@dchisholm : I know that you are right. I have access to the Hobbyist grad stuff… Good experience with Through Holes, Trained for SMD on SMD kits (could solder 0603, also 0402 with some sweat), no binocular or high quality magnifying glasses, limited sizes to 0805 with hand soldering pads (out of bigger chips), and would have benefited from a less ambitious first SMD project. So full assembly is achievable, but with a risk (as always in life). However thanks for the sound warning.

@SU_PYROW: Triple checking never hurts. Think twice and cut once (know it, but too often I remember it after cutting too short). My issue is that I don’t really know yet where to look. I believe that KiCAD workflow and DRC checks is a huge improvement compared to my initial attempt 20 years ago. This shall help to eliminate tons of errors.

I will be happy to share the full design (without NDA :-)) to anybody interested in having a look at it. It is a stripped down version of http://www.ti.com/tool/TAS3251EVM, simplifing the digital part to restrict this to interfacing to an external MCU for configuration and I2S input. With audiograde components.

If interested, just PM.

Best regards,

JMF

No one will complain if you just add a link to the project, or even attach it here (you can drag a file from your file manager to this forum when you write a message).

I had in mind that the PW way was a reasonable intermediate between intellectual property paranoïa and granting access to all source files without any copyright.

Do you think that there are inconveniences proceeding this way?

I would be fully OK with saying that it is Open Source Hardware (not working yet :slight_smile:
JMF

I am not assembling myself PCBs since 25 years or more, but it happens from time to time that I have to spend few days changing many times few 0603s.
I would not see a special problem with assembling your PCB.
I don’t have any special equipment - only soldering iron with sharp tip and old 110W trafo-solering iron.
I am using also lamp with increasing glass (but 10 years back I would not need it).

One notice about 14 TSSOP with thermal pad. I have never done that, but I have heared that amateurs to solder that pad like to made a big (few mm) plated hole under IC and solder it from backside.

In practice copyright laws say that you own all legal rights (supposing you have created it and own the copyright) and don’t give any rights to others unless otherwise stated, so just sharing it without a licence is an implicit permission to view and use it for personal/private purposes only. If you care more about licencing, you can for example choose a restrictive CC licence (https://creativecommons.org/licenses/by-nc-nd/4.0/) and change it later to be more permissive, or choose a more generous licence now. Saying it’s OSHW doesn’t mean anything in itself, it’s not a licence.

So short update:

  1. I think that I will go the JLCPCB way. Assembly is really cheap, the majority of my components are basic 0805 components on top side, there is no issue to add remaining components afterward. So I’m looking at their equivalents. I will check unsure equivalences on EEVblog,

  2. Let’s make it simple: if willing to help do a crosscheck (or simply interested in the project), here it is: https://www.dropbox.com/s/vyw69wo5cwg8qoo/TAS3251JMF_ExecLM5010v2.zip?dl=0

  3. I would be really intersted in a “checklist” about what to triple check before launching manufacturing (a sort of workflow). If you have a link or detail it or can detail it here, that would be really helpful.

Best regards,

JM

Mike has is own SMD Pick and place machine, but for small production runs he still prefers to do the pick & place by hand.

In this excellent 16 minute video he gives an overview of how he does it.
From simple tips such as using double sided tape and hand vacuum picups to a short overview of the ovens he uses for soldering.

https://www.youtube.com/watch?v=pdGSFc7VjBE

The video is worth watching for anyone interested in home assembly of PCB’s. If you have more time to burn then check out his other video’s with quite interesting teardowns or sometimes a bit of reverse engineering or other electronics related stuff.

For 2 boards, where most components are plain 0805 or 1206 SMDs, I would go for manual assembly without stencil. Take good solder flux paste, soldering iron and even cheap hotair and you can have it assembled in 2-3 hrs. Good assembly drawing showing component values help a lot in the process.
I can easily do 0603, 0402 requires some more patience (especially those wirewound inductors, hate to solder them!). But ordinary 0805, 1206? Piece of cake :slight_smile:

Watched the video: very informative !

Thanks

By the way, below some checklists I found on the net:

  • MAKE SURE THE BOARD LAYOUT MATCHES THE SCHEMATIC!

  • Breadboard the circuit exactly as the schematic is drawn.

  • If changes are needed, make the change on the schematic.

  • Make the board layout from the exact finished schematic.

  • Don’t be in such a hurry that you are sloppy.

  • That circuit board can be exactly right, or exactly wrong, the mfg. will send you what you send them.

  • Don’t pay for useless boards made for half thought out plans

  • Zoom way out, then select all to make sure there aren’t any forgotten or stray objects outside of your board outline.

  • Don’t choose so small of a font size it is unreadable.

  • Print essential notes or instructions on the board. (Will I know how to use it by just looking at it a year from now?)

  • Treat it as though you are stranger going to have to troubleshoot it. Label voltages at strategic locations.

  • Mark calibration points and their settings.

  • Don’t make the board so small you have no room for labels, test points, etc.

  • Label e-b-c, d-s-g etc….component outlines and descriptors on both sides of board.

  • Put all part outlines on both sides.

  • Don’t print on top of any solder connection. Solder is contaminated by ink.

  • Label pin description on top of and following the trace near the component.

  • Use larger pads if wires are to be directly soldered to board. Possibly both a solder pad and a header connector.

  • More than one wire soldered to a connection? Make another pad for it to connect to.

  • Make generous sized traces, especially those with higher current flow.

  • Include a board title describing what it is. Among many others a year from now it may not be so obvious.

  • Add a logo for a little polish and professional look to your personal boards.

Schematic check

erc

look for cross-sheet references that aren't connected

add 2 fudicials, sparkfun lib-FIDUCIAL-1X2

4 mounting holes

add ground test points with holes for probe gnd

check notes.txt file for todos

create readme.txt with board name, zip content desc and fudicial placement, example below

check that all operational output lines have pull up/down during reset

make sure I2C SDA/SCL have pull-up resistors and solderable test points

board check

put a silk square down for sharpie id marking

add dimensions to tDocu

load dru file, this sets spacing for pours, etc

ratsnest

run zoom-unrouted

drc

add company name to tNames

add board name and version to tNames

run normalize-text:

    set to 40 & 8 mils

    smashes all components

place names with top, bottom off, tcream on

place logo with import-bitmap, or use lib part logo-oshw

check all the version numbers, board, schematic, readme and zip filename (to be created)

board submission

print layout

check connector placement

create readme.txt with fudicial placement

cam 4 layer

    gerb274x-4layer.cam

        make sure layer 51 tDocu is in plc, side silkscreen CMP

        make sure layer 200 bmp  is in plc, side silkscreen CMP

    excellon.cam

gerbv *.crc *.plc *.stc *.cmp *.ly2 *.l15 *.sol *.sts *.pls

run Centroid_ScreamingCircuits_smd.ulp

update version and rev in readme.txt, add version/rev to sheet 1 of sch

zip gerber.zip *.cmp *.crc *.crs *.drd *.dri *.gpi *.l15 *.ly2 *.plc *.pls *.sol *.stc *.sts readme.txt *centroid.csv *.dim

zip -sf gerber.zip

    verify that sch and brd are NOT included in gerber zip file

mv gerber.zip to board-v0.0-gerber.zip

print sch and brd as pdf

zip board-v0.0-release.zip board-v0.0-gerber.zip *.sch *.brd *.pdf *.pdf *.csv

    mv 2 zip files to ../release dir

freedfm

board submission itead

download cam file

cam iteadstudio*.cam

gerbv   *.GTO *.GTS *.GTL *.GBL *.GBS *GBO

zip a.zip *.GTL *.GBL *.GTS *.GBS *.GTO *.GBO *.GBP *.GML *.GTP *.do *.TXT *.dri *.gpi readme.txt

mv a.zip to boardname-v?.?.zip



rm -i *.GTL *.GBL *.GTS *.GBS *.GTO *.GBO *.GBP *.GML *.GTP *.do *.TXT *.dri *.gpi

pre-assembly

send connector drawing showing pin 1 indicator

verify IC pin1 indicators

I don’t know about you, but I may not remember how to use it by the time first run of the boards arrive from the fab house… :wink:

1 Like

There seems to be an accelerating trend among board fabricators to “fix errors” by the board designer, without informing the designer. From scanning fabricators’ web pages I have identified two particular concerns:

  • Solder mask clearance is set to a uniform, fabricator-chosen, value across the whole board. If I specify a clearance value that is smaller than what the vendor routinely supplies, I want to know that so I can adjust parameters in KiCAD. If I specify a value larger than the fabricator’s default value, it is likely an engineered value (e.g., for RF characteristics, or related to safety requirements, or a place where I intend to apply adhesive or paint, etc) and I don’t want it changed!

  • Silkscreen - both text and graphics - is removed from areas where there is no soldermask. The intent is commendable: you DO NOT want silkscreen printed over places that will receive solder. Silkscreen is often the last item to receive attention in a board design. The KiCAD DRC has not checked for silkscreen violations (it’s coming in V6!) so it is easy to overlook, say, a component designator that overlaps an SMT pad. But chopping out bits and pieces of silkscreen that lop over into the silkscreen clearance leaves a ratty-looking board. It is ELECTRICALLY correct, but gives a bad impression to anybody who happens to view it. In the worst case, the chopped-up silkscreen is illegible, leading to assembly errors, troubleshooting difficulty, etc.

If a vendor is going to deviate in any way from the Gerber files I sent, i expect to at least get a notice to the effect, "We changed the following things, for the reasons shown . . . "

We often repeat this admonition on this Forum: The GERBER FILES are the final authority. The 3-D modeling is irrelevant. The KiCAD display on your computer screen doesn’t matter. There is no appeal if KiCAD, through some program error, creates a Gerber file different from what you intended. The board will be fabricated according to the Gerber files.

Fabricators have specs for minimum character size and minimum stroke width, though they are sometimes difficult to find. Many fabricators will attempt to print characters that violate these specs, even though the result may be illegible blobs of silkscreen ink.

So you think a BC560 is an acceptable substitute for a 2N3906? Check the pinouts! The TO-92 package may be the worst offender for lacking a universally recognized pin assignment, but the SOT-23 isn’t a shining example of standardization.

Ummmm . . . . . OK. This primarily affects troubleshooting techs, hence is more important for prototypes than production. Until recently, silkscreen on both sides was costly to both price and delivery schedule.

Just because your vendor doesn’t charge more for feature sizes down to, say, 6 mils (0.15mm), doesn’t mean it’s a good idea to do your whole board with 6 mil traces and 6 mil spacing. Don’t force him to work at the very edge of his process capability.

KiCAD includes footprints for the KiCAD logo, Lead-free logo, static-sensitive logo, RF emitter logo, hazardous voltage logo, and others. Look in the “Footprints” library under “Symbol.pretty”; and in the symbol library under “Graphic.lib”.

I wish KiCAD had an easy way to create tables of text information! I’d use 'em for “General Notes”, “Revision History”, “Alternative Part Values” (e.g., some versions get a 110V transformer and 2 amp fuse; others get a 220V transformer and 1 amp fuse), etc.

“Comments”, “Design Notes”, “Theory of Operation” are additional examples where text fields would be useful.

A year from now? Some of us need remedial training after a lunch break. Several incarnations ago I was doing software maintenance for some custom production test stands. On one occasion, my supervisor, a manufacturing engineer, some co-workers and I were studying a software module and wondering, “Why did some idiot code this module so elaborately? It only checks and displays a few straightforward parameters, that aren’t even used anywhere else in the test process.”. Eventually, we turned to the “Change Log” to see if it contained any clues (though Change Log entries were usually quite brief and cryptic - “What was changed”, not “Why was it changed.”). The last entry was something like “Major rewrite.”, and guess whose initials were attached to that entry? And I didn’t even remember working on that module!

Dale

I like to put both the schematic and pcb revision numbers on the board. (Often my schematic and pcb revision numbers drift a bit from each other.) I have to remember to manually update these. I kinda wish it was more automated, but apparently not enough to put the effort into submitting a feature request…

1 Like