Thanks Bob et al,

This is about what I expected, but I had to ask.  I wonder how long it'll take for that several thousand bucks for a pick-n-place machine to become a couple hundred?  That would be the final hurdle for the tiny electronics business.

Anyway, I've had my say and we can let this die.  Thanks for the responses!

Bob

GFS GPSDO list:
groups.yahoo.com/neo/groups/GFS-GPSDOs/info

On Thu, 6/23/16, Bob Camp kb8tq@n1k.org wrote:

Subject: Re: [time-nuts] OT stuffing boards: was GPS interface/prototyping board
To: "Bob Stewart" bob@evoria.net, "Discussion of precise time and frequency measurement" time-nuts@febo.com
Date: Thursday, June 23, 2016, 5:27 PM

Hi

Around here, assuming:

1. You supply all the parts on full reels with
   leaders
2. There is no hand assembly
   work
3. You already have framed stencils
   that are the correct size for their gear

You have multiple proper solder and placement fiducials  on
both sides
5) The boards are designed to
mount on their gear
6) Your parts and design
rules fit their gear and rules.
7) No
electrical test, visual inspect only.
8)
Best effort only, If the part does not solder etc, you
replace it on your time.

You can get various places to look at a batch
for $500 to $1000. If your stencils !=
their
stencils figure $100 to $200 each.

If you want to ship things a ways, you can save
a bit of money. Shipping plus packing
always seems to be a bit expensive.

By far the best approach is to
get all of their rules before you start a board layout.
Then
do it in whatever arrays / panel size
they are set up for and all the other little details.

This all starts to make a lot
more sense to the local outfits when you are talking a few
hundred boards.
Even more so if it is a few
hundred boards a month, every month for a few years.

====

Some math:

120
parts on 10 boards is 1200 parts. A good machine will do
that in < 6 minutes. Setting up the machine,
loading and unloading the machine, pulling
boards on and off the machine, programming the whole thing,

validating everything ….. that’s an
afternoon’s worth of work (maybe more) and maybe an hour
of down time
on the machine.

====

Of course for a few thousand dollars you can
buy your own pick and place machine ….

Bob

bob@evoria.net
wrote:

related question before this topic dies, if you don't
mind.  What about the other side of building: stuffing the
boards.  My GPSDOs have about 120 parts per board, plus
some custom work on the SMA connectors.  Is there a service
out there that will populate boards with SMT components for
small orders at a reasonable price?  Small is 10 boards.

groups.yahoo.com/neo/groups/GFS-GPSDOs/info

wrote:

Re: [time-nuts] RS232 / GPS interface/prototyping board

frequency measurement" time-nuts@febo.com,
"Nick Sayer" nsayer@kfu.com

easier

service,  I have a pile of

from them which are far greater

than anything I could

to have.

past.

that,  I did hand

single sided boards from

to completed product in one

using a Dremel drill

hand soldering

2016 00:01, "Nick

time-nuts" time-nuts@febo.com

Hi

You can indeed get a pick and place for under a thousand dollars. I wold not use one of them, but they do exist. It all depends on how much of an “advantage” you want over a hand place approach. A half way decent screen printer will run $500. Some sort of reflow setup will be a couple hundred. You can go cheap on the printer and get it down to $100 or so. A rebuilt toaster oven will run $20 or less. It all is a matter of how much hassle / how tight pitch you want to deal with.

Bob

This is a very important topic for anyone who wants to build state of
the art electronic today.  You can't continue to live in the 1970s
using DIP parts with 0.1 inch leads.  So how to make small batches of
custom designs.

The pick and place machine could be very inexpensive if you are
willing to let it run very slow using only one or two really of parts
at a time and work on small boards.  The RapRap type 3d printers
don't cost much to build.  A pick and place is not much more than a 3d
printer with a different nozzle.    You can find people doing this on
other email lists that deal with robots

For most projects these SBCs (arduino, Pi 3, BBB,...) allow you to
build almost anything without need of a custom PCB.

On Thu, Jun 23, 2016 at 3:38 PM, Bob Stewart bob@evoria.net wrote:

--

Chris Albertson
Redondo Beach, California

I'll second this, and suggest you consider:

1. Pick and place machines use a lot of floor space (even for the
   "small" ones are more than 1/2 a bench.)
2. Even the best ones require pretty continuous tuning. If you aren't
   using them continuously each new run is a new and different
   experience.  Often unpleasant for the first few scrapped boards.
3. You can only place a limited list of parts for a run.  If you have
   one more part than the machine will accomodate, its a second (or
   third, or fourth pass.)
4. They are all high maintenance in addition to requiring tuning. A lot
   of the maintenance is based on calendar, not operation time.  Even
   and idle machine requires time if you actually want to use it
   eventually.
5. Most are closed software loops. You work around their poor (or un)
   documented formats and bugs.
6. There are really cheap small batch assembly houses coming online
   that will do under 10 units. See Macrofab, PC:NG, Small Batch
   Assembly are fairly quick turns.

If all you are doing is protos, hand placement, mylar solder stencils
(see Oshstencils and others) and a hacked toaster oven are a good
solution. The $500 Chinese reflow ovens seem to require more (re)work
that a $50 toaster oven.  If you use stencils to place the solder, part
placement is as fast (or faster) than through hole parts. I have to use
a microscope.  I'm shaky enough that  may need to built some Waldoes
soon.  ;-)

I just did six moderately complex boards (no fine pitch parts) and that
was 2-3 too many for me.

Solder stencils make all the difference.

Oz, in DFW

On 6/23/2016 6:21 PM, Bob Camp wrote:

--
mailto:oz@ozindfw.net
Oz
POB 93167
Southlake, TX 76092 (Near DFW Airport)

We use
"Advanced Assembly" in Colorado for prototype assembly.
http://aa-pcbassembly.com/

For just one or two boards, it is faster to hand solder the parts, as long
as no BGA's or similar.

If complex soldering like BGAs, or more than three boards, we use a proto
assembly house.

--- Graham

==

On Thu, Jun 23, 2016 at 7:58 PM, Chris Albertson albertson.chris@gmail.com
wrote:

Hi

The gotcha with “really slow” is that once you print the solder paste on the board, it has a very
limited “open air” life. If you don’t get the board done fairly quickly, your soldering quality can
suffer quite a bit.

Bob

On 6/23/2016 9:36 PM, Bob Camp wrote:

For most of the paste formulations I've had no trouble with several
hours of working time. So you need to get at it, but really don't end up
hurting yourself, but can't leave it overnight. .

--
mailto:oz@ozindfw.net
Oz
POB 93167
Southlake, TX 76092 (Near DFW Airport)

On Thu, 23 Jun 2016 20:16:34 -0500
Oz-in-DFW lists@ozindfw.net wrote:

The trick is to use semi-manual pick&place machines for low volumes.
Ie machines that you do not program, but guide by hand. This allows
faster and more accurate placing of components than would be possible
with a pure manual process, but does not have any of the complexity
of the fully automated solutions. The components do not need to be
100% exactly centered, as the surface tension of the molten solder will
pull the parts into place (which is also the reason why the copper inside
the solder resist mask should be symmetric).

These machines are still all pretty expensive (IMHO, the cheapest
start from around 2kusd IIRC), but with the continuous growth of the hobbyist
market, and that market becomming more and more professional/proficient,
the production volumes of these  machines will for sure rise and thus become
cheaper. I am pretty sure that we will see hobbyist marketed pick&place systems
build upon open source based control systems in the next couple of years.
There are already a couple of DIY systems out there, that look quite good.
e.g http://vpapanik.blogspot.de/2012/11/low-budget-manual-pick-place.html
http://www.briandorey.com/post/Diy-Manual-Pick-and-Place-Machine-part-1

Oh, yes! Please, do not try syringe dispensers! These fail more often than
they work. Also pay the additional couple of bucks to get a steel stencil
instead of a kapton one. Especially if you make more than one or two boards
or those with fine pitch.

		Attila Kinali

--
It is upon moral qualities that a society is ultimately founded. All
the prosperity and technological sophistication in the world is of no
use without that foundation.
-- Miss Matheson, The Diamond Age, Neil Stephenson

I do not see why a small pick and place assist system could not be built on a 3-d printer.
Don

Felix qui potuit rerum cognoscere causas.
Lucky is he who has been able to understand the causes of things.
Virgil

"Noli sinere nothos te opprimere"

Dr. Don Latham, AJ7LL
Six Mile Systems LLC, 17850 Six Mile Road
Huson, MT, 59846
mailing address:  POBox 404
Frenchtown MT 59834-0404

VOX 406-626-4304
CEL 406-241-5093
Skype: buffler2
www.lightningforensics.com http://www.lightningforensics.com/
www.sixmilesystems.com http://www.sixmilesystems.com/

On 6/24/2016 9:32 AM, Attila Kinali wrote:

Laser cut Kapton are fine for a few boards - up to 5 or so.  You start
seeing the effects of use after six or eight and steel is clearly a
value. . I use them for most protos.

If I'm doing anything really fine that would drive me to steel on
resolution alone, I pay someone else to do it.

Oshstencils are cheap, and they will do 4 mill Stainless Steel for 2X
Kapton cost.  Deal if I'm doing more than 5 boards.

--
mailto:oz@ozindfw.net
Oz
POB 93167
Southlake, TX 76092 (Near DFW Airport)

Hi

If you are doing a board with a few hundred parts on it (as earlier mentioned). And have designed
a “3D printer” pick and place that runs one reel at a time. And are running the printer slow to keep
everything from going all over the place. And have a manual reel advance (no feeder) ….

I can easily see you getting into a “put it away for today” situation. You can beat a simple system like that
(speed wise) with a good manual vacuum pickup and some modest pre-organization of parts.

Bob

Don Latham said
to Discussion
I do not see why a small pick and place assist system could not be built on
a 3-d printer.

Lots of problems to be solved...

How do you take loose parts or cut tape or tape reels and get the right
part out, and into the chuck, oriented in the right direction?
How many different kinds of parts, sizes, shapes, pin counts, IC
footprints, can you handle at once?
How do you know it is the correct part?
How do you know where the "+" end, or "pin 1" is?
How do you know that there actually is a part in the chuck?
How do you know the part in the chuck is oriented the way you expected it?
How do you know where the footprint on the circuit board is located? (to a
few thousandths.)
How do you know the part left the chuck and ended up where you intended it
to be?

Commercial systems have answers to all of these questions. In many cases
involving intelligent vision systems.

Once they are all answered, then you have a PCB unique set-up / programming
effort to instruct the placement machine all about today's specific board
and parts list to be assembled.

I can usually build one or two boards by hand about as fast as an engineer
can do the placement machine specific PCB programming and verification.

--- Graham

==

On Fri, Jun 24, 2016 at 11:52 AM, Don Latham djl@montana.com wrote:

Hi

A “low cost” auto pick and place does have some things in common with a 3D printer:

Drive is by stepper motors
Micro stepping stepper motor drivers are used
Belt drive is pretty common
Tube support and circular ball bearings are generally used
You have X, Y and Z drive
Very custom software to drive them that may or may not be open source
A setup and calibration process that may not be a lot of fun
A requirement to do some trial runs before you do anything important

There are generally a few things different between my 3D printer and a simple pick and place:

The drive is on both sides of the X / Y region
The X/Y is much larger than the Z
There is a Z rotation (4th axis) on the pick and place.
My 3D printer has one head, two heads is about as low as they go on pick and place
The parts feeders are a big part of the pick and place machine and there are a lot of them (even on a simple one).
The base plate is a lot heavier duty (more rigid)
Vacuum to do the parts pickup
Pressure to blow off a stuck nozzle
Some means to detect that a part has been picked up
Interchangeable pick heads (the head size needs to match the part size to some degree)
A “small” (“desktop”) pick and place is a meter square. Most 3D printers are much smaller.

— As you move up, the pick and place might have —

A camera to check parts on the placement head to properly align them (“up camera”)
A camera to check the board and align to it (“down camera”)
Four or more placement heads.
Enough variety in feeder widths to match the size of the reels you have :)

No matter what you do for a pick and place, you will need:

Some way to mount the board on the machine
Mounting holes (or whatever) on the board to match up with the machine
A design allowance that keeps parts out of the area contacted by the mounting setup
Fiducial marks to line things up that are of a type the machine can work with

There are lots of details….The 3D printer takes cad files and generates motion information.
The pick and place take the output of a PCB layout program and generates motion information.
It goes on and on ...

Bob

On Fri, Jun 24, 2016 at 3:32 PM, Attila Kinali attila@kinali.ch wrote:

For what it's worth, I have a manual / guided pick and place machine
available. UK pickup only unless you're really keen ! Off-list please if
anyone's interested.

-adrian

On Fri, Jun 24, 2016 at 9:52 AM, Don Latham djl@montana.com wrote:

You are not the first to think of this.  But the problem is

1. moving the reels of parts,
2. programming the machine.  Translating the PCB design files into g-code.
3. Building a vision system.  With very small parts you really can't
   run open loop

--

Chris Albertson
Redondo Beach, California

On Fri, 24 Jun 2016 13:59:58 -0500
"Graham / KE9H" ke9h.graham@gmail.com wrote:

Most of these problems are easy:

You don't. No loose parts with any kind of pick&place machine.
As for cut tape, these can be taped on an empty reel to make
them compatible. Everything has to be in a tray, reel or similar.

Orientation is defined by the reel/tray the parts come in.
This is also documented in the datasheet, usually.

As many as there is space around the machine :-)

You put it manually in the right feeder and double check that it
fits the programming.

This comes with the orientation of the part in the reel/tray.

Your trays are guaranteed to be non-empty by manually loading them.

The manufacturer guarantees that the reels/trays are loaded correctly.

This is provided by the pick&place file. Usually its 3-5 digits after the
decimal point, when using mm. But as I wrote before, you don't have to
place part hyper exact. Being within 0.1-0.3 of the pitch of the part
is usually enough. Surface tension does the rest.

You dont :-)

The way how this is checked is either a pre-solder and/or post-solder visual
inspection. This is either done manualy or using a camera system where
computer compares the PCB to the picture of a known-good PCB.
As this is ment for a small volume and hobbyist system, doing the visual
inspection manualy is good enough and more than fast enough.

		Attila Kinali

--
Malek's Law:
Any simple idea will be worded in the most complicated way.

The ideal hobby use pick and place machine would be very different
from a commercial machine.  Lets say I want one board made.  What I
want to minimize is my time.  With a conventional machine by FAR most
of my time is spent setting the machine up.  In fact setup is so slow
that for smaller PCBs I could do it with tweezers in a fifth of the
time needed to set up the machine.

So a hobby machine must be designed such that you could get it going
in nearly zero time.  In the ideal case you drop the parts all mixed
up, (but right side up) in a small tray.  They are mixed and in random
orientation.  then you give the machine your PCB design file (not a
special pick and place file) and then a vision system IDs the parts.
Today vision is dirt cheap.

But the 3D printer needs one more degree of freedom.  It must be able
to rotate the part (or the PCB) as it is unlikely the part on the tape
or tray only needs translation to the PCB, likely ration is required
in almost all cases.

I think a hobby machine would only be successful if it could reduce
the setup time to nearly zero and for that it would need a really good
vision system that could hunt down randomly placed parts.  It would
have to work pretty much like you or I would do the job manually.  But
we have software like openCV and good "board cams" with M7
interchangeable lenses for $35.  A vision system actually saves a ton
of money because the machine need not be so precise as vision closes a
feedback loop.

Also how many hobbyists are going to have reels of parts?  I might buy
some parts by the dozen but most no more than about 4 or 6 at a time.
I don't want a large machine.  It should have a working surface, a
white melamine table about 12 inches square and I place the PCB to be
stuffed and all the parts on the same foot square table at any random
location then press the "go" button.  The camera scans the table.
This kind of machine would be horrible for production work but a one
foot cube machine that required zero setup is what most of us want.

Going a little farther.  I'd like this SAME machine to actually make
the PCB too.  A 3D printer could route the copper and drill holes and
print the solder resist plastic too.

On Fri, Jun 24, 2016 at 8:56 PM, Attila Kinali attila@kinali.ch wrote:

--

Chris Albertson
Redondo Beach, California

Many parts can't be recognised visually. Capacitors are the obvious example.

On Sat, Jun 25, 2016 at 6:11 AM, Chris Albertson albertson.chris@gmail.com
wrote:

Hi

If the part is oriented top-bottom and you have it on the board both
top-bottom and left-right .. you need to be able to re-orient it after
you have picked the part.

If it is a diode or semiconductor , you have four obvious orientations for the part. Doing
four reels for every diode or semi just isn’t the way it is done. You
buy one reel and let the machine deal with it.

If you add in things that are oriented at a 30, 45, or 60 degree angle, you
simply can not have a reel for all the orientations.

The Z axis needs to rotate.

If it is a moving table design (as many if not most are) that really does
not count. You can’t get the head there.

Bob

And try to tell transistor A from transistor B from diode C when they are all upside down.

A moving head design can be made to pick up parts off of reels on all four sides. But it takes more table space. Which is money.

As someone else said, you need Z rotation, which isn't as easy as it sounds when using pneumatics to pick up the part.

Moving table design sounds like a recipe for shaking the parts off the solder pads. I've not had good luck with solder paste staying tacky for long. In which case you are dispensing paste then sticking part. Not a deal breaker but slow.

Hi

Even on a “real design” desktop pick and place, vibration is an issue for parts moving around. The same
is true of parts “squirming” on a pick head as it does it’s thing. Part’s aren’t as flat as you might think they
are and you can only get just so much vacuum.

Lots of grubby design details.

====

One thing that has not been covered so far is (at least to me) the best way to do manual pick and place.
A lot of people try to do it with tweezers. I find that to be a massive pain. A vacuum pickup with a foot
switch goes much faster for me. It also is a much better way to keep your hands out of the solder paste.
You can get some nice long pickups. A setup is not very expensive, even with an illuminated  magnifier to help see
what you are doing.

As an added bonus, some systems will let you reverse the pump and use it to dispense solder paste. The
net result is that you have an effective way to both put down solder and place the parts. I prefer stencils for the
solder side of things, but that is very much a mater of debate.

Bob

Hi

Might be a little more fun if it  showed the proper way to use
the 150B (or any similar device).

Bob

There is a "hobbY" pick and placer called LitePlacer that meets a lot of
your criteria.

http://www.liteplacer.com

It works on cut strips, not reals, it will handle random placed parts on
the table top, but so far I don't think it automatically figures out
what they are. But you can put a bunch of one part in one area, another
bunch of parts in a different area etc. But for most parts you just
leave them in the cut strips taped to the table.

The cameras read the holes in the tape and use that to figure out where
the parts are. The first one it grabs from a tape it takes it over to an
upwards facing camera to make sure it gets rotation correct and knows
exactly where the pins are.

It even has the white melamine table.

I haven't bought one yet, but I'm strongly considering it.

John S.

On 6/24/2016 10:11 PM, Chris Albertson wrote:

excellent vid, Bryan!!!!!

On 2016-06-25 03:48, Bryan _ wrote:

--
Dr. Don Latham
PO Box 404, Frenchtown, MT, 59834
VOX: 406-626-4304

On Sat, 25 Jun 2016 02:48:04 -0700
Bryan _ bpl521@outlook.com wrote:

Some additional info:

The Q of a crystal decreases with frequency. For a given crystal slap
and production process, Q*f is almost constant, with a peak around 5 to 10MHz.

The highest Q I remember seeing were BVA's that reached 2e6 to 3e6 @ 5MHz

		Attila Kinali

--
Malek's Law:
Any simple idea will be worded in the most complicated way.

Hi

Every paper I have ever read on the intrinsic Q of quartz makes the claim that Q * F
is a constant ( Q goes up as frequency goes down).  Unless blank diameter gets in the way,
this has been true for any crystals I have ever used. Q does change as overtone changes,
but that is not related to the Q of the material. A given blank design may (or may not) be
limited by the Q of the quartz at any specific frequency. That is a function of a lot of things.
The material’s properties set a maximum Q you can achieve no matter how good your
blank design is and how big the blank. Done properly, the best 5 MHz resonator you can
do will have 2X the Q of the best 10 MHz resonator.

Bob

Hi Bob,

Can you tell those of us with those cool Saunders 150B crystal impedance meters how to do it right?

/tvb

Hi Attila,

Don't forget about University of Western Australia's multi-decade gift to the world -- whispering gallery Cryogenic Sapphire Oscillator (CSO) -- which have Q near 1e9.

http://arxiv.org/pdf/1504.02711.pdf

http://www.nict.go.jp/publication/shuppan/kihou-journal/journal-vol57no3_4/journal-vol57no3-4_0204.pdf

http://www.armms.org/media/uploads/ARMMS_Oxborrow.pdf

/tvb