8-head Pick and Place Machine - JSM

[awjennin]

1/21/2023
Ive convinced myself that I want to build a pick and place machine, and potentially transition into a small batch manufacturing business as a side hustle. In any case, i'm a few months in to this project which has taken a fair amount of effort. Therefore I wanted to start documenting my work in the event that it offers help to others!
I've made too much progress to document in a single post, but I will start adding notes, design files, pictures, etc. as i go. Lets start with requirements and what ive used 

Requirements
Capable of populating a 16"x22" PWB panel
Facilitate enough feeders to build a modern PWB side in one pass (no reel changes)
  58 lanes of 8mm tape
  16 lanes of 12/16mm tape
  2 lanes of 32mm tape
  2 lanes of 44mm tape
  1 lane of 72mm tape
  10% room for expansion
Optimize Cost and Quality, sacrifice schedule - achieved via extensive use of ebay and scrap yards
Camera vision
  Placement/Top
  Component/Bottom
Automatic calibration (via feedback sensors and routines)
Modular - will assemble in the basement, Built for upgrades / future disassembly and relocation
Open Source SW
8-pick/place heads

HW Components
------------------
Frame
  Lower Frame - UniStrut cut/welded into left/right/middle bolt-together sections
  Gantry - Heavy Extruded 6061T6 Aluminum, 80mmx120mmx83", discovered at metal scrap yard years prior
  PlastiChain 4210-series flexible wire-routing chain
  Loads of surplus milspec STP/shielded twisted pair; 18awg + single braided shield
------
X-Axis
  Two Bosch/Rexroth 65-series bearing blocks (roughly 48mmx108mmx30mm) on a 82" Bosch linear guide rail (roughly 20mmx25mm)
  One NSK 20973 ballscrew, 20mm pitch 84" and 71.75" stroke
  One IMS/Schneider MDM-series Mdrive45+microstepping stepper motor + integrated controller
  One US-Digital H6S-2000 (2000 CPR/cycles per revolution - 8000 quadratures per rev, SE-output)  optical encoder on right side
  Two Turck BI5-S18-AN6X inductive sensors (one for max + and one for min - movement on right and left sides, respectively)
------
Y-Axis
  Two THK KR-46 (20mm pitch, 480mm stroke, 740mm outer rail length, 840mm overall length, precision grade 1440mm/s)
  Two IMS/Schneider MDM-series Mdrive23+microstepping stepper motor + integrated controller (I modified, adding a not populated programming connector) 
  One US-Digital H6S-2000 (2000 CPR/cycles per revolution - 8000 quadratures per rev, SE-output)  optical encoder on right side, modified to directly mate to stepper motor
  Three Turck BI5-S18-AN6X inductive sensors (one for max + movement on left side, one for max + and one for min - movement on right side, helps square the gantry)
------
Head - 8-needle MyData Hydra head from a dismantled MyData Pick and Place machine. This includes:
  MyData Logic CCA
  Z-axis Maxon DC-Brushed motor
  A-axis Maxon DC-Brushed motor
  8 PnP needles with solenoids
  Vacuum and position sensors on each needle
  Hengstler RI36 optical encoders (i believe 900 counts / 3600 quadratures per rev) for both Z-Axis and A-Axis
  (i added) One Turck BI3-Q06-AP6X2 inductive sensor to sense when the needle actuator was in a "safe" position
  (i added) Two GeckoDrive G320X DC Motor Drivers (accepts step/dir inputs and optical encoder inputs)
  (i replaced the ZiLog microcontroller on the Logic CCA with a) PWB i cut; has an Atmel ATmega32A + burned Arduino bootloader
------
Feeders
  Qty 23 (Siemens / SiPlace / Schultz) 00131098-02 3-lane x 8mm tape, 2mm/4mm feed capable
  Three Siemens "Trolleys" - disassembled for the Feeder mounting plates and the Lemo-connector interface boxes
  Three home-grown PWBs to replace Siemens control with Arduino
  Three Arduino Nano Every (ATmega4809 processors)
  Gore USB bulk wire
------
Control Cabinet
  One IBM/Lenovo M710s (i7-6700 + 16G DDR4 + Two Dell 2xUSB3.0 PCIe cards)
  One PV PVS-USOPTL4 Isolated RS422/485 USB Adapter
  One DynoMotion KFLOP
  One DynoMotion Konnect
  One home-grown PWB to interface KFLOP with system IOs
  One IMS ISP300-7 125Vin 68 VDC Switching UnRegulated power supply (for the Mdrive34 on X-Axis)
  One MeanWell SP-200-48 48VDC Switching Regulated power supply (for the Mdrive23's on Y-Axis, amd GeckoDrives for Z-Axis and A-Axis)
  One MeanWell SP-200-24 24VDC Switching Regulated power supply (turned up to 28VDC, for the feeder banks which want 30V)
  One MeanWell MDR-100-24 24VDC Switching Regulated power supply (for inductive sensors and MyData Hydra solenoids)
  One MeanWell MDR-06-05 5VDC Switching Regulated power supply (for DynoMotion and MyData Hydra digital systems and optical encoders)
  Three Amphenol CPC Series 1 circular plastic connectors (23-shell size, 37-contacts)
  Three Neutrik XLR connectors for 30VDC to each of the three feeder banks
------
Camera + Lighting
  One Banner LEDWRV62X62M white ring light (top)
  One PointGrey/Flir Flea2 USB3 camera (top)
  One MegaPixel 4-12mm 1:1.4 Aspherical cctv lense (top)
  One Banner LED 80X80mm red ring light (bottom)
  One iDS UI149xLE-C USB2.0 camera (bottom)
  One (unknown) cctv lense (bottom)
------
Softwares
  OpenPNP (https://openpnp.org/)
  Dynomotion KMotion (https://www.dynomotion.com/)
  NotePad++ (https://notepad-plus-plus.org/)
  IMS SEM SPI (programming Mdrive motors + MD-CC300-000 USB to SPI cable) (https://imshome.com/downloads/discontinued_products.html)
  Arduino IDE (https://www.arduino.cc/)
------
References
  https://github.com/openpnp/openpnp/wiki/SchultzFeeder-and-SlotSchultzFeeder
  https://groups.google.com/g/openpnp/c/zPgehj7d0r0
  https://github.com/bilsef/SchultzController
  https://youtu.be/_snAjJtib24
------

(thats enough for tonight - i will start adding design notes pics etc over the next days)

[bidrohini]

Is your pick and place machine fully functional now?

[awjennin]

The current item i've been working to integrate is the feeders. I find it difficult to describe much of the process as it has occurred over many weeks. The road has been long and there have been many sideroads, backtracks, and other excursions! I am close to done, but still need to repair the solenoids and reattach the side and tape guide.

I selected the Siemens 0014109x series feeders as they appeared to be plentiful (on ebay), mechanically solid, and included the solenoids/control systems. This reduced the design complexity on my part, as all I needed to do was provide power and a data interface... or so I thought. While I was at work, I looked at some circuit card assemblies (CCA's) and created a notional requirement (seen in first post) on how many lanes of each tape-width was needed. That many feeders required a bit of a financial commitment, so i decided to break the order into tranches; a preliminary order of half the feeders - and a final purchase if this project actually comes to life. Modern CCA designs use lots of 0402 parts with 0201's and X2SON's becoming more common, so I decided to focus on purchasing 8mm feeders first.

As I investigated the feeders, i realized that it would require some effort to adapt them to the machine. The mounting system had to be robust and repeatable; the electrical interface needed to be flexible. Without getting into obnoxious detail on how I rationalized the decision, i decided it was cheaper to buy the trolleys for the feeder mounting plates, and the Lemo-connectors / communication box (Siemens 00316265).
I committed to ordering qty 24 of the Siemens 3x8mm feeders (siemens/schultz/siplace, 00141098-02) along with three "trolleys" (00316265-04 N1-M5-0165). These items came together on three pallets's and were in very bad condition. All items were physically damaged from a long life and abuse, but also filled with a fine dust from either the semi-truck or sitting outside..


To relieve the space constraints of all this nonsense sitting in my garage, I completely disassembled the trolleys. I created piles of screws, steel, stainless, mounting plates, communication boxes, and etc., then took a lot of the total materials to the metal recycling yard, and kept the rest.

Chapter 1, Feeders Part 1
I spent weeks taking every bit of the feeders apart. I cleaned and degreased, repaired and sorted. This was an exhausting step as each feeder is an extremely complex system. I did this whole process in an assembly-line styled process... but i estimate that it cost roughly 4 hours per feeder. I'm close to completion but hit a snag with the shutter solenoids. From what I can tell, the solenoids have been thermally stressed which resulted in very minor deflection of the plastic bobbin.. long story short my plan is to cut a square broach to press through each one to trim ~ 0.005" of the sidewalls to reduce drag. I just need to get the surface grinder working first... A story for another day.







(note to self - learn how to appropriately attach inline images)

[awjennin]

Chapter 2, Communication Boxes
The communication boxes are nice; tey use Lemo connectors which mate to the dangling cable assembly on the feeders, they are made from heavy stainless, and came with right-angle mounting brackets from the trolley. Internally, they are made of a power supply+Processor card, and a connector breakout card. I removed the Power-supply+Processor card and threw it on a hot plate - removing all the components and connectors for my use on other projects. I was going to leverage the work of Bill Ruckman who's created adapter cards which use an Arduino to control the feeders. However i had file issues and couldn't review/verify the 96-pin backplane version of his card, so i decided to do my own schematic/layout heavily referencing his work. The PWB is a 2-layer design with no special routing requirements (en lieu of the low freq's). I came up with dimensions which would fit within the stainless chassis, and used 0204 MELF resistors as I had just scavenged many from the power-supply+Processor card. I also reused the BC856 BJT from the same card. Therefore my digikey order only consisted of two additional resistor values, and the BC846 and 2n7002 transistors. I used Sunstone to procure the PWBs. I used ebay for the Arduino Every's.
[ Specified attachment is not available ]
[ Specified attachment is not available ]
[ Specified attachment is not available ]
[ Specified attachment is not available ]
The 96pin_Interface.zip file attached has my Eagle-based design files, BOM, and PDF's (layout/assembly drawings and schematics)

I got the cards in and started building them up. The process went well enough, and i've now switched to integration mode. There's a lot of work remaining, but Bills effort has given me a huge jump start.  see:  https://github.com/bilsef/SchultzController

My PWB uses a slightly different mapping, so i had to make the following changes:
feeder.h -   Commented out/replaced const uint8_t TXPortPin[20] array in BOARD96PIN with the following:
                  const uint8_t TXportPin[20] = { 13, 2, 3, 4, 5, 6, 7, 8, 9, 12, 11, 10, 21, 20, 19, 18, 17, 16, 15, 14 };
config.h -   Commented out #define BOARD4PIN, uncommented #define BOARD96PIN
                  Updated #define LANES_PER_PORT 3   (was previously set to 2)

Now i'm breaking into te next level of integration which includes getting OpenPNP configured to see the feeders, and perform "feeds". There is a comprehensive google group conversation on this topic. While the effort is primarilly focused on the 2x8 feeders, i'm hoping some of these folks can continue their efforts on the 3x8's which use slightly different command structure. I am very comfortable in HW land, but very weak in the SW domain.
reference: https://groups.google.com/g/openpnp/c/hBWwKGdBlgE

Kind regards and more to come later!

[awjennin]

I am restarting work on the PnP head this weekend, however I want to share some pics of the structure first. I have a little CNC Mill i built many years ago, and it has worked wonderfully to cut parts for all my other various projects!
Below, you will see the views of the gantry system. If there is interest, I can share my models, however i dont imagine there is a great desire as its very specific to the HW ive wrangled up across ebay (as listed in my first post)
Here are the models:




And a few pictures of the actual HW:

[awjennin]

[awjennin]

And here are some additional pictures of the mechanical assembly of the Hydra head. The include some pictures of:
- The front of the unit
- The back of the manifold including purge solenoid and individual needle-solenoids
- The removal of the optical sensors to determine whether needles are up or down
- Addition of a Turck inductive sensor for initial calibration and also to ensure the head is up before X-Y machine moves
- The Z-axis motor and encoder
- The back of the head where you can see the A-axis motor and encoder

[Neepa]

Very impressive undertaking.

Never would have imagined that there is such a relatively big second hand market for PNP machine components. Quite reasonable prices as well for what you're getting.

The Feeders are the units responsible for the mechanical tape advance so the suction pippetes can take the components off of it right? If so can you briefly explain how it does so?


Now I'm also questioning why the sole bigish open source/diy/kit PNP machine (LumenPNP) had so much trouble designing their own if they could have simply bought some off ebay to base their own dumbed down design on? They went through a lot of iterations to get somewhere.


In the end your PNP will use a portal gantry with a moving head to place the components. With the feeders themself mounted to the gantry too? I hope you've thought about how much inertia the gantry will have in the end zipping over the board. While the weight may dampen any vibrations coming from the motors and spindles the whole frame flexing and trembling in place from the rapid acceleration and decelartion could negatively impact your positioning or throw off your optical part orientation correction.

That is jsut my initial caution without having seen the whole machine frame myself. So pictures when?

[boB]

Such awesome work you are doing here !

WoW !

[H.O]

As one who has also been down the path of the DIY pick and place machine I applaud your efforts this far. I happen to have a couple of Maxon DC-motors with those black anodized heatsinks and the exact same Hengstler encoder shown here (bought used for the encoders) so I can't help thinking they once originated from one of those heads.

Me, I bought I bunch of FUJI dual lane electric feeders, couldn't find any information on how to talk to them so designed a new board and wrote firmware to run them talking to OpenPNP. All this, >5 years ago and not much else has happened since then. So, keep it up!

[jmelson]

I notice you talk about 72 mm tape!  The head pictures seem to have the nozzles insanely close together.  How can the nozzles turn large parts without them hitting their corners?  My machine has 3 nozzles separated by ~ 60 mm so the parts have no chance of hitting each other.
Also, I don't see how those tiny nozzles can safely hold large parts without slippage.
My Quad QSA30A had changable nozzles and can swap them from a rack (automatic nozzle changer).  This seems like the best way to do a wide range of component sizes.  Maybe I am missing something, as I have never seen a Mydata machine.
My QSA30A machine has 3 separate stepper motors (5 phase) for rotation of each nozzle, and separate Z motors (brushless servo) each.  This is so that the vision system can gauge size, rotation and centering of all parts simultaneously while moving from feeders to board position (flying vision).  This all works amazingly well! 
Jon

[jmelson]

As one who has also been down the path of the DIY pick and place machine I applaud your efforts this far. I happen to have a couple of Maxon DC-
Me, I bought I bunch of FUJI dual lane electric feeders, couldn't find any information on how to talk to them so designed a new board and wrote firmware to run them talking to OpenPNP. All this, >5 years ago and not much else has happened since then. So, keep it up!

Quad electronic feeders just take DC power from a PC board in the machine (+5 and +12V) and when the nozzle comes down to pick the part, it breaks a light beam, when the nozzle lifts, the feeder advances by a programmable amount (4, 8 12mm etc).  The downside is the Quad feeders are fairly wide, wider than most other mechanical feeders.  But, they are fairly plentiful in the US at reasonable prices.  A couple years ago I could get them for $70 each, lately they seem to be going for $100 each.  Be sure to get the 12 V feeders, the 5V ones are supposed to be very weak.
Jon

[awjennin]

Good evening all, and THANKS for the comments/thoughts on this little endevor    As I go up against these struggles, i become keenly aware that had i known the sheer amount of effort required ahead of time I would have likely never started this nonsense! haha!!

Neepa,  That's exactly right. The feeders accept the component tape and has a sprocket in front that engages with the tape and advances it via a stepper motor and ratcheting cam (for precise advancing). As they do so, a shutter actuates where the tape foil is being removed - this shutter is closed during movement helping restrain components from 'popping' out of the tape. This is via an solenoid.  In the background, there is a gear assembly that grabs the discarded foil and packs it into a little trashcan on back. The gear motors are enabled when a little little tension arm indicates that the foil is loose via photobeam. Please see this vid, the gentleman does a very nice job explaining:
I don't have any experience outside of the Siemens feeders i selected - however i grew a big appreciation for the level of precision that was evident in their design. Tolerances were tight, gear meshing was precise, component advance is exact, foil tension is optically monitored, and the chassis is very rigid. From what i've learned rebuilding mine, and what i observed looking at the LumenPNP that you mentioned, it does seem like they would need to address a lot of mechanical shortcomings with good SW and good visual control loop from camera on PnP head.
Regarding your question about the setup, I will have 4 "quads" or banks of feeders; Front Left (feeders), Rear Left (feeders), Rear Right (feeders), Front Right (trays). The PWB will exist in the middle of the quads - which allowed my design to have many feeders but also allow common parts to all have a similar distance to the PWB. I've attached two pictures that might help illustrate.
Your comments are appreciated; what i've noticed so far is that high speed movements on the X-axis result in oscillation of the long ball-screw. They can get really bad when i make it run fast, so i've limited the velocity in my SW which runs on the Dynomotion KFLOP machine controller. Additionally, to minimize abrupt movements, i make adjustments in the acceleration and jerk values too. I also spent some time looking at the control loop (step response, within Dynomotion) to adjust and minimize step-error. The frame is 3 sections of welded Unistrut with gussets/triangles to help stabilize, but i expect i will need to add some more stability as the project matures


H.O, I would like to see a pic of your machine! i generally don't publicly document my work - but appreciate viewing the designs of others to see how they approached problems. Therefore I figured it would be great to share some of my projects as its helpful to others. Hopefully I can assist others as they review my success and many failures.
I am currently struggling BIG TIME trying to control the A-axis and Z-axis motors. There is excessive dithering when operational, as well as loads of heat developed in the motors... I need to educate myself a little more on those because i'm doing something wrong. How did you integrate your Maxon's? which motor drivers did you use?
Nice work on the Fuji's too! The Siemens have a socketed Atmel chip as their main processor - your comment made me consider that it may be more valuable to just come up with new SW for those chips than continue to go down this road of reverse engineering.... i will have to think on that!!

jmelson, My tentative plan was that during 72mm picks, i would only use needles 1 & 8 (and not use any of 2-7). 1&8 are roughly 72mm apart. That being said, anything heavier than a SMD connector would likely need a larger needle anyways. I've left some room for that and potentially a solder-extruder on the right side of the "carriage assembly" for future dev.  The needles i have also require a rubber tip / nozzle like yours - i admit i need to investigate the different sizes/capabilities.  unfortunately, i fear it will be a long time before i get to that point. This whole project feels quite overwhelming from time to time. I did look in to the Quad feeders your spoke of - Those seem great. Width/density aside, the photo-break concept is great - that requires so much less integration. Generally i appreciate a simple solution - complicated is too easy until its not! (did that make sense?) I'm worried that these siemens feeders are overly complicated...

[H.O]

H.O, I would like to see a pic of your machine!

I don't mean to polute your thread with my photos but since you ask I'll attach an overview. This is where it all kind of stalled... OpenPNP talking to a a Smoothieboard. X/Y uses ballscrews and DC-servo motors driven by Granite Devices VSD-A servo drives. Z-axis is beltdriven by a stepper and the two nozzles are hollow shaft NEMA8 steppers. You can see the new control PCB inside the FUJI feeder, PCB on the head interfacing Z-axis reference switches, vacuum sensors, constant current LED driver for camera light and likely something I've forgot.

Regarding the Maxon motor perhaps you misunderstood. I don't use them on the machine, I bought them second hand mainly for the encoders fitted to them (but you won't find DC-motors much nicer than Maxon). Seeing your myData head and the fact that it says myData Sweden AB (and I'm in Sweden) I recon my motors originally came from one of those heads.

As for your problem I MIGHT have an idea... I see you're using Gecodrive G320X drives, these specify a minimum motor inductance of 1mH. Depending on the exact winding configuration of the motor it's possible that it has a very (VERY!) low inductance causing large ripple currents which will lead to motor heating and can cause issues with the control loop (drive current limiting). Searching for that specific Maxon P/N doesn't bring up much today so it's either an old P/N, a custom one or one for a specific combination of motor, encoder etc. If you have local Maxon representative, give them a call, perhaps they can look up the details for you.

Maxons own drives generally use a 50kHz PWM frequency while the G320X uses 20kHz. You might want to try adding an external inductor in series with the armature to see if that improves things.

[awjennin]

Good evening H.O
Fantastic machine you've assembled! I particularly like the dimensions/proportions of the machine - appears to be structurally sound and well designed! Additionally, having the feeders arranged like that make it appear very easy to load and manage. Excellent work! I would encourage you to restart your work on it - much too nice to quit on.

Your comment on the inductance struck me; I had noticed that 1mH inductance called out too. As I reviewed your VSD-A drives, they had a similar spec: 1mH min, 100mH max. I am not strong in the power-electronics - however your description regarding the PWM frequency requiring a certain inductance to manage the current transients is starting to make sense the more i think about it. Additionally, as you mentioned, those Maxon motors are a tiny inductance. By pulling specs of a similar dimensioned modern Maxon motors, i see values in the 0.05mH ballpark! I appreciate your idea, and will continue to investigate over the next few days!

As I goofed around the other day, i wasn't playing close attention to how long i was testing the motor. In any case it got hot (thermal cam said 250F ) and i thought i saw a ribbon of smoke come from it. Dismayed, I pulled the motor apart thinking i had nuked it... It visually appears to be in good shape, however it smells of burned windings. Upon reassembly I decided to attach it to a DC bench supply. The motor has continuity (which manifests as a voltage drop/current draw) regardless of where I manually rotate the armature. The commutator has 4 contacts, and the brushes and contacts are very clean. When energized, the motor does not want to spin (even as i put a light oil into the bearings). I would gradually increase the current limit of my bench top supply until the motor is dissipating about 2Watts of power - at which point it springs into motion. Once spinning it will continue to run, but i'm suspecting theres a bad winding or similar. I am unsure if this is new damage from myself, or what ive been fighting all along! The former is easier to believe, however the latter fits the failure mode which i was experiencing initially. That is, I have to use a very high current setting on the G320 servo drive before the motor would rotate... but then the motor would also get hot. Here's a few pictures to go along with my rambling story. Thank you for your comments and perspective! i'm going to hunt down a new motor, some power inductors in the single-digit milliHenry range, and report back. Kind regards!


EDIT: found the Maxon datasheet - the -980 is ~0.5mH   Not as bad as what I referenced above, but still 50% under the min spec. Will continue to peruse
https://www.maxongroup.com/medias/sys_master/root/8807091437598/13-365-EN.pdf

[H.O]

Ah, I thought we were talking about the larger, heatsinked motor, my bad.
The 980-winding is designed for 12V. What voltage are you feeding the G320X's with? Higher voltage will cause even larger ripple current since there's more "oomph" pushing current "into" the inductance.

I'm not sure about your poor little motor. I have a couple of 980's and tested one of them. Winding the voltage up from 0V it starts spinning by itself at about 750mV. At 2V it's pulling 55mA (with the encoder on the back attached). Without the encoder dragging it down it starts at 250mV and at 2V it pulls 40mA. My LCR meter measures the inductance as 398uH@1kHz and 142uH@10kHz.

The G320X has some DIP switches for the current gain but the manual is very vague on what they actually do. I suspect that the G320X might be not sutiable for that small motor even WITH external inductor.

As I said, I have a couple of 980's (less encoder) that I don't see me using. I'd be happy to send you one for the price of shipping if you like.

Thank you for the kind words on the machine.
It's not how I would design it today, should I redo it but it was designed around extrusions and linear rails that I had available.
It's quite solid but the gantry is a bit on the heavy side. I have 10mm pitch screws, would like 20mm pitch. I would like to finish it but I obviously don't have the stamina. We'll see...

[jmelson]

Good evening all, and THANKS for the comments/thoughts on this little endevor    As I go up against these struggles, i become keenly aware that had i known the sheer amount of effort required ahead of time I would have likely never started this nonsense! haha!!

  Yes, that's why I went with a working production machine.  In THEORY, it was all supposed to work.

jmelson, My tentative plan was that during 72mm picks, i would only use needles 1 & 8 (and not use any of 2-7). 1&8 are roughly 72mm apart. That being said, anything heavier than a SMD connector would likely need a larger needle anyways. I've left some room for that and potentially a solder-extruder on the right side of the "carriage assembly" for future dev.  The needles i have also require a rubber tip / nozzle like yours - i admit i need to investigate the different sizes/capabilities.  unfortunately, i fear it will be a long time before i get to that point. This whole project feels quite overwhelming from time to time. I did look in to the Quad feeders your spoke of - Those seem great. Width/density aside, the photo-break concept is great - that requires so much less integration. Generally i appreciate a simple solution - complicated is too easy until its not! (did that make sense?) I'm worried that these siemens feeders are overly complicated...

I do up to 20mm TQFP FPGA chips.  The nozzle I use for that is about 5mm wide with a rubber seal, kind of like an O-ring with a squared-off cross section.  I did some boards with larger capacitors in the 1210 size but fairly thick, they were going on the board rotated.  I had to switch to a larger nozzle because they were slipping on the nozzle during the vision measurement and rotation for placement.    So, I try to always specify the largest nozzle that doen't risk the part pulling up into the hole in the nozzle.  Other than the one large chip nozzle with the rubber seal, all my nozzles are steel and most are the nozzles that came with the machine.  They are quite durable.
My old Philips/Yamaha machine used the rod that lowered the nozzle to poke a lever on the feeder to advance the tape after the rod lifted back up.  The issue with that one was the springs that advanced the tape were fairly weak, and didn't provide enough pull to peel the cover tape reliably.  That was a CONSTANT issue with that machine. Those Yamaha feeders were about as simple as you could possibly make them.  Later ones used an air cylinder on all sizes of feeder, and then the springs could be made stronger.  Now that I have a machine with flying vision, I can really appreciate what vision measurement of each part can do.  I really would not want to be without it, as the pickup location of many parts as they sit in the component tape is not tightly controlled.
Jon

[jmelson]

As I goofed around the other day, i wasn't playing close attention to how long i was testing the motor. In any case it got hot (thermal cam said 250F ) and i thought i saw a ribbon of smoke come from it. Dismayed, I pulled the motor apart thinking i had nuked it... It visually appears to be in good shape, however it smells of burned windings. Upon reassembly I decided to attach it to a DC bench supply. The motor has continuity (which manifests as a voltage drop/current draw) regardless of where I manually rotate the armature. The commutator has 4 contacts, and the brushes and contacts are very clean. When energized, the motor does not want to spin (even as i put a light oil into the bearings). I would gradually increase the current limit of my bench top supply until the motor is dissipating about 2Watts of power - at which point it springs into motion. Once spinning it will continue to run, but i'm suspecting theres a bad winding or similar. I am unsure if this is new damage from myself, or what ive been fighting all along! The former is easier to believe, however the latter fits the failure mode which i was experiencing initially. That is, I have to use a very high current setting on the G320 servo drive before the motor would rotate... but then the motor would also get hot. Here's a few pictures to go along with my rambling story.

The G320 is a synchronous antiphase drive, which has the problem of sending 20 KHz square waves of the DC supply voltage into the motor at idle.  Thus the motor sees a triangle wave current of up to several amps even whn doing nothing.  You may need to put an inductor in series with the motor to limit that idle current.    If you have a scope, connect the motor input to the scope and turn it by hand.  If you see a steady voltage with two clear notches per revolution that is a sign of a shorted winding.  Otherwise, apply a very low voltage with a power supply and turn the motor by hand.  Apply enough current so you feel torque from the motor.  If the torque jumps up and down twice/rev, same as above.
These motors can take a fair amount of abuse.  So, maybe the motor was bad from the beginning.
Jon