Electronics > Manufacturing & Assembly

Tronstol E1 experience

(1/17) > >>

Sort of surprised nobody else has talked about this yet. This was purchased for prototype and small (like a dozen or two boards) builds.  I use local contract assembly places for bigger builds.  I've already made some obvious mistakes (like sending a nozzle across the floor because I put it in the auto-changer holder wrong) and probably will make more as I figure out this machine, but hopefully some of my experience will help others.  Or others out there that have one of these - my understanding their aren't many of us yet - can offer assistance.

I've recently uncrated it and started to set it up. It arrived a day after I got access to new office space so things will change in the background too. There was also a 3 alarm fire that closed the building for a while (nobody hurt, my things were OK).

The unit was purchased from Neoden USA (New Jersey). Tronstrol and Neoden are now two different companies but there's some things that seem to share commonality between them. Neoden USA sells both. They offer support and factoring in shipping/tariffs made it kind of a no-brainer to buy from them if you're in the US. The unit is in the wooden crate, the big box was the stand, and the other box was a FP2636 stencil solder printer (frameless).

The E1 was securely bolted to the shipping crate, the tool kit includes a socket wrench and other tools needed to get it off the crate base and mounted to the stand.  It is a 2 person job to move/lift the unit.  Getting it set up mostly consists of removing packing materials and tie downs used for shipment.  The only assembly needed is to bolt it to the stand (which is optional) and screw in the status indicator light and plug it in.  Took about 2 hours.

It has a touch screen and a keyboard & mouse can be connected, but IMHO much easier to VNC in to it, as well as then you can move files directly to it (it's a RPi running Linux) instead of the USB stick.  It only has WiFi available.

The manual that came with it was out of date (OTOH they should get points for including a printed manual) but Neoden USA was able to supply me with the latest manual.  The machine comes with a test board but no noob-level directions on the steps needed to actually get parts on to it.  There's example files on the machine that are for other machines which is confusing.  In the end the way to get it working was to start with a blank file, import the pick and place .CSV (which you might want to trim/make a new version with just one component for initial tests), and then go through each step and get all of the settings right.  This is where the support from Neoden USA was great, as the manual isn't quite complete to explain what exactly the settings are doing and/or how to figure out what they should be.  A screen share resolved a lot of things.  Probably took about 8 hours from first power up to getting parts to place.

Here's the machine in action:https://youtu.be/AkTbvYJ7PnM

Placement seems to be off a bit for different areas.

Placing one part at a time, the UI shows where it thinks it wants to place it (red box) and you can see where the part ends up is off. I've adjust some of the speeds, delays and accuracy settings but doesn't seem to change what's happening.

There's some more things it can do for calibration and that's what I'll work on next and report back when there's progress.  After that there's things that are (so far) unclear about the "materials" and "footprints" libraries and how to incorporate those in to workflow where one bank of feeders/parts is going to be permanent in the machine.

Update in case someone else is hitting these things. It has V 1.01 software on it (April 2022) and I never trust version one of anything.  So far, with two exceptions, the bugs have been tolerable and shouldn't be show stoppers.

The minor bugs:

* CN40 nozzle tolerance needed to be increased. Or maybe the size is wrong?  Leads to head errors if not fixed.
* A head error can lock up the software
* Selecting a head pickup test with a nozzle that doesn't map to the selected footprint causes nothing to happen vs. a warning message.
* Bug(s) in swapping columns after .CSV pick and place file import. Possible weird "they don't stay swapped next time the board file is opened" bug. Suspect if I give the columns the exact names the software wants, and in the exact order, this probably goes away. OTOH since there's the ability to assign/swap columns, make it work so I don't have to muck with the .CSV?
* Yellow status on light doesn't work. Just green or red.
The "that's odd/hope they can improve this" :

* Forward compensation. It turned out to be the cause of parts ending up skewed (when placed) as a results of the pickup position drifting (see example).  (Though this does not solve the parts being shifted when placed, see below.)  Set to 0.5 now for the parts I was working with, otherwise the tape slowly drifts and the pick up goes badly.  What concerns me is this seems to be a global setting for all parts using a specific footprint, and I'm not convinced this is ideal, i.e. maybe some 0603 need more forward compensation but maybe not all?

Semi-related is when changing a reel it's easy to disturb adjacent components, meaning rechecking the alignments.

Seems to me a pin feed system, to be reliable and efficient for use, needs to auto align on the sprocket holes before every pickup.
Big problems:

* Have not figured out the offset between where the machine thinks its placing the part and where it ends up.  I suspect when reflowed the parts will pull back correctly, but the machine should be more accurate than what I see. See the example of an 0603 not where it thinks it is.  There was a suggestion to adjust the place height but have not been able to try that yet.
* Removing the feeder assembly (for example to put in a new reel, etc) when the system is powered up bricks the software when you unplug the peeler controller cable.  If you power up the machine without the peeler(s) plugged in the software will run again, and let you know it doesn't see the unplugged side.  But powering up with everything plugged in just causes the controller software to hang more or less when the app opens.  Log file shows something being retried but no info as to what it means.  This problem is now in Tronstol's hands to resolve.  It also raises questions about how to put more parts on the machine during a run.

... Interesting.

As a non "Chinese PnP" user my input may not be valid but, what should be happening with picking a part such as your 0603 is, you define a pick offset for that feeder after loading based on where the pick position seems to be (center of pocket not device in pocket). Subsequent picks will adjust the offset based on the vision results to keep the pick consistent and correct for feeder inaccuracy. A global package based offset is not to correct an inaccurate pick position, its purpose would normally be to pick a package off-center because of its symmetry e.g a DPAK.

Correctly defining the height of a component has a significant effect on accuracy, it changes the perceived dimensions of the device to camera and also means the device might meet the board sooner (or later) than the defined height might suggest. Many machines will still be moving in XY while lowering the nozzle, unless its placement parameters define otherwise. I have noticed on the examples I have seen of chinese PnP that how they handle component & place height can be somewhat odd & long winded.


First, clean the position feedback ruler with alcohol and microfiber cloth as it gets dirty easily and it causes drift or failed initialization problems. To clean it, turn off the machine, move the head to back manually, spray the cloth with alcohol and clean the ruler in one gentle wipe from front to back. Do this regularly, if you're in a dusty environment or whenever you get a precision problems.

Second, do the calibration in settings:
- leave mark points with each head and needle and center them. Double sided transparent adhesive tape, painted with whiteboard marker will give you best results. Use CN040 nozzle to get as small circle as possible.
- set the nozzle exchanger centers
- place TQFP100 with supplied demo project and change the DCAM rotation angle accordingly, so the TQFP100 will be in line with pads. If you reuse the same TQFP100, always clean it as any dust affects the recognition and dust sticks to double-sided tape easily. Be sure that pins are not bent upwards as they will make the component look bigger. Recognition frame should be right on pin edges, there shouldn't be any gap.

Now you're ready to place the demo project. Fill in the 0201(preferably) or 0402 feeders and go to EDIT/material. Select the 0201 feeder, that you want to place and follow these steps to set the pickup and drag location. Order is VERY important:
- pull the tape so the pocket to be picked would be as far as possible from the edge but it doesn't expose the next component.
- click needle and set the center of the tape hole. Watch the tape to see, if it will move and click feed. Click feed again just to be sure dragging is repeatable. Needle doesn't need to be in center anymore, it just needs to be at repeatable position of the hole. These steps are very important as they will allow the hole and needle to self adjust.
Now, when feeding is repeatable, switch to pick and set the center of the pocket.
Now you can switch back to needle, click feed, switch back to pick and check if next pick is still in the center of the pocket.
This procedure by selecting needle+feed first and pick second is very reliable as it self adjusts the tape. Otherwise you will always have some offset as needle diameter is smaller than the hole and it needs some travel, before needle starts to pull the tape.
- place 0201 components in circle and TQFP100 and check positions. If placement is ok, you know that machine is well calibrated and precise.

Create a new project and import CSV. Be sure to be precise with angle, fiducial and first component settings. Set the pick height to just touch the component(use picking test function) and set the pick vacuum delay to 0.1s. Set the place height to i.e. 0.5 mm and place vacuum delay to 0.1s. Set all the speeds to 10, so the machine will move slowly and you will be able to see, what it's doing. Place the project and check the results. If they are not ok, use a slow motion camera on your phone and record the nozzle during pick up and placement. Check if nozzle goes down to far, spring shouldn't squeeze by more than 0.5 mm. If heights are ok, you haven't setup the project properly, because sample project was placed correctly.
Whenever you have problems, set all the speeds to minimun and use a slow motion camera, it will help you a lot. In most of the cases it's a user problem, machine is very precise, it can place 0201 easily.
Check Tronstol's youtube videos as they are very useful for programming and setup:

Hope this helps.


--- Quote from: SMTech on March 02, 2023, 08:43:25 am ---
Correctly defining the height of a component has a significant effect on accuracy, it changes the perceived dimensions of the device to camera and also means the device might meet the board sooner (or later) than the defined height might suggest. Many machines will still be moving in XY while lowering the nozzle, unless its placement parameters define otherwise. I have noticed on the examples I have seen of chinese PnP that how they handle component & place height can be somewhat odd & long winded.

--- End quote ---

I've now played with heights but that doesn't seem to be affecting the placement position on the board. I've also added delay but one of my next tests will be to slow everything down with really large values and see what, if anything changes.


[0] Message Index

[#] Next page

There was an error while thanking
Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod