Mdc 7722FV alignment problems

[Backerly]

I have recently purchased a 7722FV from manncorp. The idea was to get a machine to do 0402 and BGA work accurately. I currently own an entry level machine from APS (CS40) that although has worked well for me, it is stepper based and doesn't have the precision for this sort of work. I chose the 7722FV over an upgrade to the APS LE40.

But my first impressions with the 7722FV have been somewhat mixed.

I have been having a lot of problems getting it working and after two solid and frustrating weeks with it i still have not succeeded in loading a compete board. The main problems I am now having now are with the machines accuracy.

Getting it set up went pretty smoothly. But out of the box it wouldn't load anything. The manual was very hard to follow and there wasn't any comissioning procedure or anything.

The list of issues I found trying to load a simple board were as follows:
1) machine kept coming up with errors trying to place 0805 and 1206 parts. Rejection was more than 50%. It would not place 0603 resistors at all. 100% rejection.
2) The "F" command in teach mode always showed the head at different random heights. Often well out of the red outline box and would then abort. I then discovered the top of the spindle was loose. After tightening the nuts the results were more consistent. But now it rejected everything!!!
3) changing the spindle height constant from 807 to 870 now allowed me to reliably place parts using LV=1. But the machine still seems to randomly loose this setting.
4) 0402 feeders would not advance. And other feeders would not advance reliably. I found the nylon heads on the advance solenoids were screwed right down. Careful readjustment and tightening of the locking nuts now allows all feeders to advance reliably.
5) placing 0402 parts and even some 0603 parts would be well off the pads. After several emails and a phone call to manncorp I had an alignment procedure for the spindle to camera offset. This improved the issue but some 0402 parts were still missing pads or being placed up to half a component width off.
6) dropping the nozzle on carbon graph paper at various rotations showed significant nozzle offsets at rotations other than 0 degrees. Manncorp helped me with an alignment procedure for the placement rotations. This fixes the problem for most jelly bean parts but placement still seems to be a bit variable, especially as I move along the x direction. And I am using panel fiducials.
7) QFP and BGA packages are still placed with significant and unusable offsets. 0.5mm BGAs and fine pitch QFPs using LV=26 or LV=27 are placed up to a whole ball or pin off.
I notice that without learned fiducials if I place a metric rule in place of the PCB and align the origin to be 0mm, when I get to 300mm the difference between the rule and the navigation window in the x-direction is more than 1mm. If I do the same for the Y direction the error is not as great but is offset the other way.
9) as has already been mentioned by others SOT23 and SOT323 packages with LV=1 that are not rejected are always rotated and offset. I have had only random success with LV=26 and 27. And the manual is pretty useless in this area. All complex parts that do manage to get placed need manual tweaking. Kind of pointless machine pacing them in the first place!
10) I regularly get run time errors that crash the software while setting up parts with no explanation and no chance to save work.

What I am basically wanting to know is if these sorts of commissioning and operation problems with the 7722FV are normal or is it possible that a significant step has been missed in the manufacturing process. Did anyone else get a commissioning or alignment instruction with their machine?

I had some issues with the APS machine initially, but nowhere near as much as this. Please, any comments from others experienced with machine would be appreciated.

[adafruit]

hi there, sounds like you're having set up problems that require a bit of support. this isnt really a support forum for mdc machines...you may want to contact manncorp since they're supposed to help you get the machine running. we have some tutorials here
http://www.ladyada.net/wiki/mdcpickandplace

[Backerly]

Thanks but I was just outlining the issues I have had as background.

I just wanted to know from others that also had this machine (some of whom I thought had posted elsewhere in this forum) if these kind of issues were normal.

[ktownsend]

It was a pretty painful process for me at first getting the machine setup, and I had a lot of problems with SOT23 as well, but I did eventually get it placing 0603 parts pretty reliably as well as QFN/QFP. I do still sometimes have some small placement issues with QFP, though, maybe being offset 0.2mm or so, but it's rare that it's so far off it won't move into place during reflow.

I haven't had any stability issues with recent releases of the SW myself. Not that it never crashes, but it's been fairly good once I found a working setup and figure out some of the many quirks in the system. I didn't have any sort of training on it, though, other than the blog here and from Akiba over at www.freaklabs.org who also has an MDC machine.

If you bought the machine from MDC, though, they should really be providing some support here since you're paying a hefty premium buying through them for the support and repair guarantees.

The machine isn't perfect, but it should be able to reliably place at least 0603 (it does for me and for Limor at Adafruit anyway), though I haven't tried 0402 much myself. It's a pretty steep learning curve, though, and maybe there are some config and alignment issues with your machine.

Kevin

[vincentp]

@backerly : just in case this thread needs to be revived, I've fully described the calibration process in the upcoming 7722 handbook I've written, packing together the former MDC manual and my personal experience, things I've been taught etc.

we're now 7 months after your purchase, how things are going ?

In my case, I noticed that :
- my magnification of the side vision was incorrect
- vertical axis of the crosshair was not aligned in flying vision
- flying vision and bottom vision general offsets were not calibrated

From a technical & mechanical point of view, all those are "normal". I see shipping as the biggest problem as you'll alway have things moving a bit whatever you do, so the final calibration must be done once the machine is installed. It's the case for any CNC. However, what's lacking is the proper explanation of the calibration process. It's not difficult at all, it just has to be described (that's what I've just finished).

The rotation on some parts (with flying vision) is "normal" too and inherent to the vision method. There are additional global angle offset split by part type (TR, SOP/CHP, and chi), and also global placement offset that compensate the lack of concentricity between the nozzle shaft and the nozzle itself (because it's removable, concentricity cannot be perfect). Once you have all that calibrated, 90% of your parts will not require tweaking with part specific offsets. For special parts (big parts) checked with flying vision, the camera distortion will create some lack of accuracy to estimate the depth of the part, so the placement offset will be wrong, but generally that's the only thing you need to tweak. This is detailed in the manual

the manual went for proof reading by a native speaker. It should be online in beta version in about a week if my friend has the time to read the... 170 pages

[Backerly]

Vincent,
Yes, thanks! I would reallly like to take a look at your new manual!! (The one that comes with the machine as you know lacks a lot of details in many areas. and there is practically nothing on calibration.)

I am still having lots of problems with this machine. Mainly to do with accuracy and consistency. Even after a number of 3-4 hour remote maintenance sessions with Manncorp.

I did notice the crosshairs on the top vision camera were rotated and have had a go at trying to realign it.

Biggest problem still seems to be with the bottom vision camera. I have to run everything using LV=226/227 (with the controller board level adjusted to flat out) with thresholds as low as 0.3 to get the thing to reliably recognise pins. I have tried refocusing it, but it still doesnt seem to show the pin edges as sharply as some pictures I have seen on here. It keeps coming up with error codes that dont exist in the manual so I have no idea what to "tweak" to rectify the problem. It seems to take hours to get a part to recognise reliably if at all!

I really look forward to having a go at the thing again with the aid of your new manual. At the moment I am still resorting to using an APS Novastar CS40 which is not capable of the accuracy the 7722FV promised (but to date still hasnt quite delivered...)

[vincentp]

you're welcome !

your camera focus should have been working as calibrated in the factory. In your hardware documents (blueprints, if you have them) you should have a sort of check list with what you ordered and the constants that were set for the machine.

My top camera has a minor minor rotation offset too. This isn't affecting anything. Part centering should occur with a target / marking in the 1/3 central area of the picture to avoid distortion (it's a VERY GOOD quality wide angle optic but it's still a fish eye). I have a PCB which is missing the centering crosshair for my TQFP, I can't use the extension of the crosshair to match the pin on top/bottom/left/right, it's not accurate. The proper method is to go in a corner (in teaching mode), press 1, opposite corner, press 2, then C (center). Same applies to tray programming with large pockets. For your top camera, what is important is the calibration of the magnification constant so that your Fids are recognized properly and not ignored for size reasons. Also, what is capital is the nozzle to top cam offset. This is the FIRST thing to diagnose / adjust. In the upcoming manual (sorry to keep you guys waiting, I'd like to shoot up as many typos / grammar mistakes first).

Some constants in K10 are controlling the focus, basically I believe it's the height of the Z shaft to bring the nozzle in focus. Mine is set to 360, that's for the param referred as scrub in the quick settings, there's also another name focus and I have 80. Part focusing is then using the part thickness to elevate that ref of the thickness so that the bottom of the part is in focus.
How did you refocus ? My picture aren't so "sharp" but my nozzle tip seems in focus (I'll check again)

Checking that the focus is good is easy. Even without a part, go into the nozzle exchanger menu then edit mode, select the posi2 and teaching mode with F8. That will make sure you have both nozzle to bottom camera aligned and in focus. Choose a large nozzle (#3 or 4) and check that the edges are not blurred by offsetting a bit the nozzle centering from the camera (to move the crosshair and target circle away).

lighting : I admit I didn't check those settings. Keep in mind that absolute lighting or grey levels aren't required for bottom vision. The system makes an average based on min and max shades found in the picture. If you light up too much, you'll wash out the picture.

Attached my pot settings. VR1 is the brighter lighting (22x). VR2 is the lower lighting (12x). VR4 is the medium lighting (2x). MDC indicated that the difference between the medium and the extremes is about 20-30%. Ideally, you use a "standard" part to define the medium, then adjust the lower and the higher ones. The pic (sorry about the quality, maglite in one hand to light up under the smoked plexiglass, phone in the other hand) reflects this (VR1 is on the right) :

20120920_133041.jpg (110.8 KiB) Viewed 2586 times

VR1 set to 3/4, VR2 almost at the min, VR4 in the middle. Those settings work.

here's a test with a "medium-rare" part (That's part #19 in my part datalist below). The part has finish problems on the leads (bumpy). Too much lighting was causing reflection problem with the center pad, so I used 126 to test the vision. Still getting some reflections, so I increase the level to filter everything but whitish shades. Level isn't an absolute level if you keep it under 50. Level is supposed to be a modifier for the calculated (dynamic) threshold calculated on the picture. In this case I increased it by 50%. I could have used regular vision 26 and an even higher Level like 2.0 but I preferred to reduce the lighting.
As you see on the picture, it's getting the lead edges perfect.

checking-SSOP8.png (229.26 KiB) Viewed 2586 times

my parts definition

parts.png (27.22 KiB) Viewed 2586 times

I generally don't check with 22x, it's not needed, I don't have parts with so much problematic leads contrast. The QFN16 is checked also like that because the contrast is poor and the pad finish is oxidized (I have those parts for quite a while).
The LGA however is perfect with regular lighting. While I spent a good hour on the first 2 SSOP (I didn't understand how Level worked at that time). The LGA took 10 minutes (plus I was in cheat mode with the adafruit tutorial for that very package). The TQFP (#15) worked like first place or almost.

when you say flat out, are you talking about the adjustable pot. ? VR1 should set the brighter lighting. Checking the exact meaning of flat out, it's "full throttle", right ? then with the level at 0.3, you basically accept most of the darker shades as white (pic is already white washed and you accept the dark pixels => you pic is almost white). That could be your problem.

ideally post a screen cap of a vision test of a SSOP or SOIC like above. Also post a pic of your potentiometers setup so that we can compare !

[bmw330i]

ERror 301

Error301.jpg (145.23 KiB) Viewed 2529 times

Hello,

Here's my vision problem. It's a QFN24 (ITG3200). The vision tests all give a 301 error. I've tried Vincents settings for QFN32 and QFN16 without success.

Any ideas?

[vincentp]

I asked a few days ago to MDC the complete list of vision errors (as this one indeed isn't in the manual) so that I can include it in the handbook. I believe in your case it's a "no part found" or "pad error" error, as from what I see you illumination is WAY too high. If your dimensions are incorrect (see below), it could be also the problem.

did you check your potentiometer settings to compare with mines ? that's not absolute as the relationship between the light amount and the potentiometers position probably varies a bit from a machine to another but the ratio between the 3 pots (VR1-2-4) will give you an indication.

I don't know if it's calibrated in factory with a light meter. Still, the background plate is almost pure white so I'm sure it appears in the background. Check pot levels, and eventually use this part as a reference to setup your regular lighting (VR4) which should be in middle position.

when you said you tried my settings, did you put the correct measurements of *your part and just used my settings for vision type & level ?
which vision did you use ? 126, 26 or 226 ?
whatever, it's way to bright IMO. If you compare with the SSOP screen cap below, you barely see the nozzle plate in the back, and that's vision 126 (darker). My QFN32 (brand new parts, not oxidation) works fine with vision 26 (standard lighting) and not even a size filter (Pt, W) as there's no asymmetric dot or pad to perturb the vision. Even in vision 26, my nozzle plate remains quite dark.

For many parts, the dynamic calculated threshold should be ok as it is (hence Level = 1.0) IF (and only if) the stock lighting (Lv=26) is correct (and of course if the part isn't complicated). Your ITG looks very standard and its package should be dark enough to provide a good contrast, plus QFNs usually have a very flat finish on their pads. I would say start by a lighting calibration.
Review VR4 wiper position. Pickup your part and bring it permanently to bottom cam (use the LOW CAM button now it's working in the most recent 2.1.23) and test it there with Shift+P to bring it to focus (make sure your part is defined with Lv=26). Adjust VR4 accordingly. Don't forget to take a picture of your current pots settings as a "backup".

[Backerly]

First, here are my pot settings.....

(This is the first time will have tried adding an attachment so hopefully this will work...)

7722FV pot settings

IMG_0840a.jpg (566.92 KiB) Viewed 2501 times

[vincentp]

Compare with mines : they seem to be really high. I won't be surprised that your part vision image is washed out (white). Post a vision test picture (screen cap) with regular lighting Lv=26 so that I can see for myself.

Now, it really depends as V1,2,4 are set properly in terms of ratio. VR1 to the max, VR4 in the middle and VR2 in the middle of those 2.
BUT
if they are all too high, vision is incorrect.

[Backerly]

Vincent, here are a few shots from the machine:

First using the LV=26 setting:

LV26

msop8_lv26.png (43.91 KiB) Viewed 2499 times

Then using LV126 (just before it reports "error 64" of which there is no explanation in the manual):

LV26

msop8_lv26.png (43.91 KiB) Viewed 2499 times

And LV=226:

LV226

msop8_lv226.png (164.97 KiB) Viewed 2499 times

As you can see, for a smiliar part its nowhere near as sharp as your image - at any light level!. And the crosshairs never seem to find the pin edge. The part often just rocks back and forth.

[Backerly]

sorry it seems to hase stuffed up the attachments and posted LV26 twice .. which mucks up the association of the comments and i cant fix it.....

But what I still dont get is that apart from being blurry the image I have appears to be much smaller than yours. What am I doing thats different?

[vincentp]

armchair PnP operator here. I'm only going to suggest hypothesis as I don't know for sure what's happening.

- lighting : looks totally correct. dimmer lighting leads to improper centering => normal I'd say (vision can't see). My machine is from 2010, I just don't want people to think my pot settings are "the ones". It depends on the led driver, current, led kind etc. For now, the balance of your settings looks correct (VR2 low < VR4 medium < high VR1)

- the part appears illuminated correctly. Your part has the same issue as mine, ie the finish is bumpy / grainy so it's hard to get the edges BUT the contrast with the package is correct. Any rough thresholding should be able to find edges.

- BUT that darn camera looks FAR from the part. Indeed. So far that the nozzle plate and the part are both in focus. Wondering if it's calibrated at all.

Something to test for you: test your chip like you did (vision Lv=26). Let it fail. Click in the live video window, on the edge of the chip. This will reset the X axis measurement. Go to the other edge (don't click) and look at the values displayed in the live window. You should read a distance in mm. Check that it matches the size of the chip. If not, the magnification scale of the camera is wrong.

Still, it's too far IMO. You need to have it closer, in focus and with the right pixel to mm scale.
Go in the nozzle changer menu, and teach mode for posi2.
I have this (centering is perfectible BUT with only 2 decimal digits, I don't think I can do better).

bottom cam centering & scale

bottom-cam-centering.png (219.93 KiB) Viewed 2488 times

That's nozzle #3, it's 2mm OD. I checked the scale by clicking on the left side of the nozzle OD then move the cursor to the other side/edge. I read 1.97mm something, which means my scale is correct. Here I can't use the nozzle base which is 5mm OD because it's not in focus, it's slightly behind hence scale will be wrong (if I measure it I read 4.5mm-ish which is correct as it's smaller because seen further).

I would contact MDC to ask. BMW got his machine from MDC directly, you got yours from manncorp (?), I don't know who is doing the bottom cam calibration. What is your camera "focus" (K10 #1 and #2) ?
I believe the camera focal point is fixed (I couldn't see any control wire, just the power supply and the video signal) so I believe the "focus" (or the scrub) is just the distance / stroke at which you need to present the nozzle tip to be in focus and big enough. From there, the scale (pixel / mm) is adjusted.
let compare our system constant values. I'll ask MDC about this.

[ADD-ON]
I tested and indeed K10 #2 is the Z height that bring the nozzle tip in focus with the camera. Mine is set to 360, I would probably need to suck a piece of paper with a super sharp pattern to adjust it but the current value is fairly well adjusted. When you go in teaching mode for Posi2 (bottom cam location), you'll notice the stroke depth is set to a certain value, that's your "focus" setting. You can play with it using the U / Shift+U key (change the increments with keypad / and *) to visually adjust the focus of your nozzle tip. I used nozzle #2 and #3. It should allow you to get closer and be in focus.
Finally, use the technique above (nozzle tip OD) to diagnose then adjust the magnification / pixel to mm scale (K7 #1)

thank you very much for having so many problems: this is going in the manual. The management.

[bmw330i]

Hi Vincent,

What I meant by your settings was proper X, Y from the itg3200 manual and your FQN32 and then FQN16 values.

I have not tried messing with the POTS since I would like to first try other options that do not require modifying the machine.

The room was pretty dark. Maybe with more light in the room the setting for 126 may work.