PushPullFeeder alpha release

[Toby Dickenson]

Hi all,

I continue to use a PushPullFeeder in production based on Mark's
design on my lumenpnp machine, and have been gradually improving it
over the last year. This alpha release is a significant improvement
since I last mentioned it here. It is smoother, parts are less likely
to jump, and the mechanism is less likely to jam and break. It has
much improved printability on modern 3d printers. I dont think it is
faster than the previous release, but it is a significant quality of
life improvement.

This is an alpha release that I am testing in production over the next
few weeks:

https://github.com/tobydickenson/PushPullFeeder/tree/test

[Satya Gupta]

Sorry if i am asking this question in the wrong post. 

But why not add a simple spring for pull action? Use header only for pushing not pulling. 

[Toby Dickenson]

On Fri, 6 Mar 2026 at 03:41, Satya Gupta <3.s...@gmail.com> wrote:
>
> Sorry if i am asking this question in the wrong post.

This is a good place to ask.

> But why not add a simple spring for pull action? Use header only for pushing not pulling.

I guess the PushPullFeeder would be better named a PullPushFeeder. The
first action is the Pull, so a spring return action would leave the
feeder in the wrong idle position.

But maybe you can spot a mechanical solution that I have overlooked.
Feel free to share a sketch of any ideas and I might just test it.

Toby

[Satya Gupta]

True that, this is first time i am using this. Did a test print to understand how it works and yes indeed it is a PULL and PUSH action.

The CAD is in OPEN SCAD which is not my cup of tea, i will try to modify or add some automation in it. 

[andrew...@gmail.com]

Toby, What are the cut strip feeders in your videos on the machine (F23 to F31)

[Toby Dickenson]

Hi Andrew,

A common problem with strip feeders is that you have to peel back the film to expose multiple parts, but then you have the risk of them jumping out of the pockets. BlindsFeeder was one solution but that did not really work for us. This design actively clamps the tape edges to prevent vibration and is quite effective. We made lots of boards using just those feeders before adopting PushPull, and still use them for short ends of tapes.

We had always intended to publish the design, but the menlu feeder published their design first with a very similar clamping principle and better DFM (on a modern 3d printer) so I  recommend looking at that.

Our old design has some advantages over the menlu feeders; higher density, and less wasted Z space, so maybe I will do a best-of-both remix at some point.

--
You received this message because you are subscribed to the Google Groups "OpenPnP" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openpnp+u...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/openpnp/4eeb1cfc-8e99-476c-a1a5-a64473b76774n%40googlegroups.com.

[andrew...@gmail.com]

Thanks - i'll try some menlu ones for my short strips.

I am just printing some of your push/pull design to replace a couple of problematic original Mark design ones

Marks original has been fantastic for many tapes I have tried, but a couple of tapes I have just not getting great results. 

[Toby Dickenson]

Thanks,

I would be grateful for any feedback.

If you have some of Marks original design and like his machine mounting scheme then I believe my current version of the model still supports it - just disabled by default.

To view this discussion visit https://groups.google.com/d/msgid/openpnp/a6db0a75-0646-4c1c-b636-7c72a766fca9n%40googlegroups.com.

[andrew...@gmail.com]

Still have to print a few more to use them in anger - but playing with them so far

Things I like.

1, I love the spiral groove to catch the printing seams.  My original prints needed so much reworking from seam pimples.

2, The friction springs are much better.  Original ones I needed to heat and bend them to get the friction just right.

3, Spring under the paper tape is great.

Problems I have

Sometimes the lever gets into a funk where the spike/peg on the front won't engage the tape holes on thicker paper tape.

I am going to try drill a hole and put a brass peg in there like the original dining philosopher ones.

Unless of course that is a problem you have encountered and you can just tell me I am printing something wrong.

For there lever I am using no-name PETg in a Bambu P1S.  0.12mm layer height , Inner and Outer wall set to 0.3mm as recommended.

[andrew...@gmail.com]

I increased the "Thorn Length" to 1.1 and "Pointedness" to 0.72 and the pointier peg is working much better

[andrew...@gmail.com]

"Thorn Sideways Tension" had to be set to -0.4 to move the peg sideways enough to have it always drop into the hole on the thicker paper tapes I have.

[Toby Dickenson]

One of the changes in this release is the "bumper cam": this lump.

The idea behind this feature is that it is a 3d cam. When the pusher dog moves backward, it takes the blue path over the top of the cam which quickly withdraws the thorn from the tape hole to avoid having it drag along the top surface of the tape. When it pushes forward it takes the green path to the right of the cam which lowers the thorn into the hole before pushing forward. 

That means the pusher dog (and its thorn) are reciprocating slightly in X and Z direction, while making their large movement in Y direction.

If you benefit from adjusting the thorn position, then maybe this bumper cam mechanism is not working quite right for you. But I cant imagine exactly what is going wrong. Another parameter that might be useful is "Dog Bumper Tension". The big hole for the lever axle is not quite perpendicular, to add a small force on the dog to return it from the green path onto the blue.

To view this discussion visit https://groups.google.com/d/msgid/openpnp/58cf6866-b1cd-43bb-8cbf-8d276025befen%40googlegroups.com.

[andrew...@gmail.com]

All the thin paper tapes and plastic tapes worked fine.  The dual-path-cam mechanism worked as expected.  Back to the left/up forward to the right/down.

However with some thicker paper tapes the thorn/peg was preventing the dog-spring from pushing to the right/down path along the cam.

I didn't realise the axel hole was not perpendicular.  That might be another parameter for me to play with later.

I have now printed 24 of them and am about to make a few hundred small boards - I already have some feedback - but will wait a week or two till I have finished getting a good feel for them.

Thanks again - keep up the good work :)

[andrew...@gmail.com]

More feedback.

16 out of the 25 feeders I printed worked like a dream.  No jumping parts like the older version.

Nine of the thicker paper tapes and some of the larger embossed plastic tapes did not work great and had problems pealing the cover.

I modified the ratchet/pawl to have slightly longer teeth (1.4mm depth) and this solved four of the nine.  The last five are still having problems pulling the cover tape.

Now that I am having a break in making boards I will see if I can work out what the issue is.

The combination of the two improvements in the last run (Duet > Smoothie and Your-Feeder > Old-Feeder) has made my openpnp machine really feel worthwile.

[Toby Dickenson]

Thank you for the feedback.

On Sun, 29 Mar 2026 at 04:49, andrew...@gmail.com
<andrew...@gmail.com> wrote:
>
> 16 out of the 25 feeders I printed worked like a dream.

We can fit 20 feeders on the machine, and this week have been testing
with half of them swapped to the new version.

> No jumping parts like the older version.

We had no jumping with mostly 0402 parts, even using Multiplier=2
(which means 4 parts get exposed for picking).

Pick position was super-predictable too, but fractionally different to
the previous version. Not a problem for 0603, but I had to adjust the
pick window position for 0402.

> Nine of the thicker paper tapes and some of the larger embossed plastic tapes did not work great and had problems pealing the cover.

Yes, we saw the same problem. Too much peeling force causes the
problem where tape gets advanced by film tension faster than with the
pusher. But the initial alpha release was set with the peeling force
too low.

Do you have a public git for your changes?

> I modified the ratchet/pawl to have slightly longer teeth (1.4mm depth) and this solved four of the nine.

I made a somewhat equivalent change of increasing the angle of the
toothed arms (parameter friction_tension) from 0.16 to 0.30.

https://github.com/tobydickenson/PushPullFeeder/commit/583eb6d0a76b44c07e3b9feec88959613e11e046

For wider feeders the toothed section does not run the full width of
the friction drum. The 12mm was significantly under-peeling so the
toothed section was made a little wider:

https://github.com/tobydickenson/PushPullFeeder/commit/5d30e0c2f34c301dc244cfe2f28f88125b7b2dc9

This test also found a paper tape thinner than I had previously
tested, so the compliant spring had to be set a little higher to grip
it.

Another tape was thicker, and catching on the reverse blocking spring
mounting feature. So the lower surface of that feature has been shaved
down a little.

https://github.com/tobydickenson/PushPullFeeder?tab=readme-ov-file#changes-since-the-first-alpha-release-of-this-branch

> The last five are still having problems pulling the cover tape.
>
> Now that I am having a break in making boards I will see if I can work out what the issue is.

Did you have any levers break? On the previous designs the level would
break if the feeder jammed. I guess they break with fatigue over time
too. This week we had one break for no obvious cause.

> The combination of the two improvements in the last run (Duet > Smoothie and Your-Feeder > Old-Feeder) has made my openpnp machine really feel worthwile.

Thats good to hear. We are seeing that several small changs all stack
up, and this week the machine was 20% faster than 6 months ago.

[andrew...@gmail.com]

Sorry  - I do not have a GIT repository I am just making a change - rendering the part - saving the STL as a file name with the change so I can keep track of them.  I am not printing a full plate at a time of a feeder but individual parts with slight changes at present.

My machine has space for about 30x 8mm feeders.  The outermost 4 on each side can only be accessed by one nozzle.  I have not yet tried any >8mm tapes.  The couple of parts I have that are >8mm are on cut strips.

Attached is a picture of the front of my machine.  The black sharpie dots on feeders levers is the ones that I have replaced/modified the tape-holders/ratchets to correct issues.  The ones with little blobs of bluetack are ones with problems still.  Today I am trying friction wheels with higher tension to see if I can get them more reliable.

I never had any lever-springs break on the older push-pull design - however I only placed a few 1000 parts max on any one reel as there where not in use for too long.

I did have levers break on dining-philosophy feeders which I used for much longer.

The small incremental improvements all add up.  I think my biggest quality of life improvement was going from smoothie (which had constant drift over time with my C-axis choice) versus the Duet which has both CAN for the tool head and does not have the faulty firmware causing drift.

However all the little things like not having miss picks from parts jumping gets the machine closer to perfect.

My last two things that bother me are "inconsistent Z home" and "ELP camera resetting auto exposure"

I am building a new hall effect Z end stop for the former.  The latter maybe scripting can help but maybe I just need to find an alternative to the ELP cameras.

[Artem Stanchak]

With such a feeder, the assembly probably takes a long time? For example, all my feeders are 0816 SMT Automatic. And the assembly speed is 1200CPH.

ср, 1 квіт. 2026 р. о 02:57 andrew...@gmail.com <andrew...@gmail.com> пише:

--

You received this message because you are subscribed to the Google Groups "OpenPnP" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openpnp+u...@googlegroups.com.

To view this discussion visit https://groups.google.com/d/msgid/openpnp/4de4a6d6-4422-4d5b-a21b-d2bd3bfd754bn%40googlegroups.com.

--

З повагою, Артем Станчак

менеджер "Bitwire", SMT Engineer

tel: +38(067)3117978

"Ми втілюємо Ваші ідеї у життя"

[Toby Dickenson]

It wont be winning any sprint races, but it is not the slowest feeder
either. We get 1100 cph on a typical job.

This is quite an old video now; I should do a new one:
https://www.youtube.com/watch?v=HmRUPP_7lOk

We tried one of the 0816 variants and found it impossible to tune the
film peeler tension. The PushPullFeeder never really gives us any
trouble.

Toby

> To view this discussion visit https://groups.google.com/d/msgid/openpnp/CAN9j%3D9W5a3iop_aT9SA5en_tADnY6VQ2ry3Ur9%2Bw2VzvfN%3DmsA%40mail.gmail.com.