My machine was manufactured about three years ago I think and didn't include needle valves on the cylinders. The cylinders would create enough of a shock wave to pop nearby 0402 parts out of the tapes.
I found these excellent flow control valves, SMC part number SMC AS1201F-M5-04A (with 5mm threads) that have really tamed down the cylinder actions. I haven't had a chance to check if they will entirely solve the 0402 problem but will get to that soon, and hopefully add a reply to this post. The machine makes a much calmer sound on feed advance and it's a comfort to know that I can apply the required amount of power and no more. The valves also have a really clever locking mechanism where the top pops down to lock the control so vibration can't change the setting.
First off all it has needle valves inside machine i think. Second parts can’t pop out if you use your parameter settings well. The open tape parameter need to get adjusted to keep the part pocket closed during pushing! Check the winder calibration values if adjusted according to your mechanics. Everything you find inside the manual. And finally reduce push feeder air pressure it will work from 1,5bar. Maybe you check the chapter about push feeder in our manual to compare if your top springs well adjusted.
The problem is on 0402 parts. The push feeder can only push 4mm so the tape has to open two parts I believe. Is this not the case? Is it possible to get tape to just reveal one part. Does the 0402 feeder setting trigger a second tape takeup (but not push feed on second part)?
I found that 0402 caps were pretty easy to keep in the tape as they are thicker, but resistors ended up bouncing out from shock wave from nearby feeds. Again this is the second part - after two have been fed and then tape removed.
There is no option to trigger the winder on second part. You need to adjust the top spring pressure and I think it will work more reliable. It could be possible to decrease the winder speed too.
The needle valves on the cylinders worked perfectly and I can slow things down to wherever I want them. 0402 problem is totally cured. 0402s are still slower than 0603 because we do optical recognition on all of them. I haven't tried turning off optical recognition now that problems are solved but my guess is that any time saved while placing would be spent later on fixing unplaced parts, cleaning nozzles, reworking boards etc. But it's always worth an experiment.
I tried an experiment in two head placement with an 0603 only board and it kind of ended up that way. By the time I got done fixing errors, it was easier to just use one head. This obviously might not be the case if I had longer runs and got all bugs worked out.