Feedback rotary encoder with a separate wheel?

[T J]

Hi,

Regarding the DIY PnP machines with belts, like the Anthony Webb design, has anybody installed separate rotary encoders with small diameter wheels to X and Y axis? The wheel would travel to the true location, because it would be connected to the moving part after the belt. The feedback would go back to the motor controller. So backlash would be compensated. Such a rotary encoder would be easier to attach to the frame than the typical linear encoder. A suitable encoder component would be for example this one (2k quadrature pulses per rotation, 8192 individual PPR):

https://odriverobotics.com/shop/cui-amt-102

I would imagine that this type of feedback would greatly reduce the location accuracy problems?

[Michael Anton]

Usually the belt is attached to the moving part, so I'm not sure how this could work if the encode was also attached to the moving part.

[T J]

It would be normal feedback which would include the belt's accuracy problems. But in this case the requirement for the PID controller would be to also handle the non-linear backlash region without oscillation. And the extra encoder wheel needs to be tightly pressed against the aluminium extrusion so that it would not slip during the movement accereration. This would result a position offset error.

I wonder if somebody here has tried this with belt drive?

[Marek T.]

The "typical linear encode" is realy dificult to be installed?

It's the one of the most trivial operation that I've needed to do with my machine. If it's scaring...

W dniu niedziela, 8 marca 2020 00:50:45 UTC+1 użytkownik T J napisał:

(...) Such a rotary encoder would be easier to attach to the frame than the typical linear encoder. (...)

[Jim]

On Sunday, March 8, 2020 at 1:38:55 AM UTC+1, T J wrote:

It would be normal feedback which would include the belt's accuracy problems. But in this case the requirement for the PID controller would be to also handle the non-linear backlash region without oscillation. And the extra encoder wheel needs to be tightly pressed against the aluminium extrusion so that it would not slip during the movement accereration. This would result a position offset error.

I wonder if somebody here has tried this with belt drive?

Each rotating element generates errors.

An incremental linear encoder serves as a reference here, to be able to compensate e.g. backlash and linearity errors.

Installing a self-adhesive magnetic stripe with a small reader is not very complicated.
You only need a motor driver with the appropriate inputs.

[Marek T.]

Magnetic is relatively expensive, Chinese optical scales are cheap and works good enough.

But it's interesting what do you think about the feedback (encoder) where it connect to. In my case, my drivers for dcservos have the encoder inputs. But what in case of stepper motors? I've not seen the steppers' drivers with encoder inputs. I'm almost sure that Marlin 2.0 has the feature to encoders serve. But not the Smoothie. So what Marlin only? Or are there some stepper drivers with inputs? (I know about smart steppers but it's other story).

[Jim]

I'm almost sure that Marlin 2.0 has the feature to encoders serve.

Are sure you did not mix it up with handling for rotary push button encoders?

[jdlv]

I'd asked a quote for magnetic encoders, ~1500€ for 3!
Chinese scales are certainly less accurate but still accurate enough for
a pnp. And much cheaper!

Do you have link about the Marlin encoders inputs? Couldn't find it in
the official documentation :

https://marlinfw.org/docs/configuration/configuration.html#configuring-marlin

Joël

[Jarosław Karwik]

1500??!

What size do you need ? I paid around 200...250 Euro for the head and 50cm tape (5um i think). Let me know the size and accuracy  i can send you link

--
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 on the web visit https://groups.google.com/d/msgid/openpnp/09305305-44a6-b428-a4ba-164ee2d613c6%40gmail.com.

[Marek T.]

Hmmm, now I'm not sure.

My adventure with 2.0 is short, in last days I've compiled and changed 1.0 into 2.0 in my printer because needed firmware with added backlash compensation (works great btw) and that's all. I've seen there in the code a sections with encoders, and focused on motors problem assumed it's just axis encoder not the encoder for the control panel :-(.
I'll look into the code where it is to make sure what encoders it really is and write.

[Marek T.]

Quick search says I was wrong and official Marlin really does not have it. But I'll make sure later what I've seen yet.

But I've found this that may be interresting: https://blog.aus3d.com.au/magnetic-encoder-modules/
https://wiki.aus3d.com.au/Magnetic_Encoder#Using_with_Marlin

[Bill Ruckman]

Official Marlin does have it.  Search for  I2C_POSITION_ENCODERS in configuration_adv.h.

--
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 on the web visit https://groups.google.com/d/msgid/openpnp/21e72ae8-6b47-4007-af34-6f6dfb53324c%40googlegroups.com.

[T J]

There is also an encoder company called Posic. They have encoders and PCB based strips. Prices online.

https://www.posic.com/products/linear-encoders.html

[Marek T.]

Nice Swiss product, very small (miniaturized) and friendly to mount tlooking on space saving. But unfortunately also Swiss price - almost $200 together with scale, case, holder.

And 10m/s, 30m/s, crazy low resolutions - it's rather overkilling looking on this what we really need in PNP machine.

But dimensions are really priceless!

Yes Bill, now I'm at the PC and I see I was not wrong :-).

It's really added and it is just the service for these Aus3D forum encoders which I have linked above.

So it's worth to think about it and try on some day!

[jdlv]

The quote was for this:

https://www.lika.it/eng/products/linear-encoders/incremental-1/sme21#

Joël

[Jarosław Karwik]

Well, if you need 1um...

I bought 10um version

https://www.eldar.biz/enkoder-inkrementalny-liniowy-lika-smk-l-1-10-i-l2

Costs ~ 130 Euro netto with 8um tape - 50 Euro / m

>> email to ope...@googlegroups.com.

[ma...@makr.zone]

What are you guys picking and placing?

_Mark

[jdlv]

Quote was for 2µm but you are right 10µm is enough.

Joël

Le 08/03/2020 à 18:26, Jarosław Karwik a écrit :
> Well, if you need 1um...
>
> I bought 10um version
> https://www.eldar.biz/enkoder-inkrementalny-liniowy-lika-smk-l-1-10-i-l2
> Costs ~ 130 Euro netto with 8um tape - 50 Euro / m
>
>
>
>
>
> W dniu niedziela, 8 marca 2020 18:04:07 UTC+1 użytkownik jdlv napisał:
>>
>> The quote was for this:
>>
>> https://www.lika.it/eng/products/linear-encoders/incremental-1/sme21#
>>
>> Joël
>>
>> Le 08/03/2020 à 12:40, Jarosław Karwik a écrit :
>>> 1500??!
>>> What size do you need ? I paid around 200...250 Euro for the head and
>> 50cm
>>> tape (5um i think). Let me know the size and accuracy i can send you
>> link
>>>
>>> niedz., 8 mar 2020, 12:29 użytkownik jdlv <jdlv...@gmail.com

>> <javascript:>> napisał:

>>>> email to ope...@googlegroups.com <javascript:>.

[Jarosław Karwik]

I use these encoders for my industrial ink printers .

 I have also spare set to put into my Chinese pick & place - currently I do not go below 0603, but I do have some 0.4mm picth LGA / BGA

[Jarosław Karwik]

Help me find a machine for retorfit - then I finish and share my single axis RS485/CAN stackable controllers which support encoders directly :-)

[Marek T.]

I've mailed to CriS but he didn't answer, as far as I know him it means he doesn't have this machine already.

[Jim]

On Sunday, March 8, 2020 at 7:10:03 PM UTC+1, ma...@makr.zone wrote:

What are you guys picking and placing?

Hi Mark,

Currently i do not go below 0201 and 0.4mm pitch.

My lika encoders are 2µm (2MHz) and it makes really fun to make panels (i do panelization outside of openpnp) with only three fids from panel frame.

But i'm sure that for most cases 10µm resolution would more than enough.

Jim

[T J]

Hi Jim,

What kind of machine do you have?

[ma...@makr.zone]

Hi Jim

So you do place 0201 using OpenPNP?

Do you have a video/photo?

_Mark

[Jim]

@T J  

It's a simple scratch build OpenPnP machine. Belt driven, unfortunately.

@Mark

Yes. I've got to ask my boss for permission.

Jim

[Airhead Bit]

Actually the price isn't bad, I purchased 500mm of the 5um 'tape' (linear mag strip with stainless steel back) and have been testing is with some of the latest generation of encoders that produce the signals needed for TMC drivers. I can see a 400mm axes with encoder at around $40.00, full roll of tape would be less of course.

All the software is in Marlin, Aus3d started it then was told (by AMS) the EOL for the encoder which has since been reborn, AMS is all over the place with their product line. The current issue is getting the ABI signals to the drivers/controller, reliably, cheaply, seems everyone wants to use I2C and SPI, neither of which travel nicely and then there is the fact there is no displayable feedback, that requires the sensor sending ABI to the TMC drivers and a MCU reading the encoder and sending that data to the presentation device, in PnP that means a sub-controller USB channel to OpenPNP?

But it's spring and my Cellular CAT M1 project has priority, amazing technology in this generation of IOT Cellular, that and Verizon finally took their head out, they are selling SIM cards directly to the IOT market that work!