Shop OBEX P1 Docs P2 Docs Learn Events
Prop inside Servo (getting closer) — Parallax Forums

Prop inside Servo (getting closer)

RaymanRayman Posts: 13,892
edited 2019-03-01 16:52 in Propeller 1
I'm finally getting close to having a Prop powered hobby servo.
The control wire will now be a 1-wire communication bus with bidirectional coms...

I went down this path before, but found I didn't have the skill set or the tools to do it: https://forums.parallax.com/discussion/124175/serial-servo-making-a-servo-more-intelligent

But now, I've learned Eagle, so I can make circuit boards thin enough to fit and my PCB design skills have improved.
Also, I have a Shapeoko CNC router so can mill the boards myself.

So, I'm gearing up to take another stab at this.

Current plan is to also add in a 3-axis accelerometer and a WS2812B in there.
This is all too much to fit on one PCB of same size as the one in this HXT12K servo.
But, there's plenty of room under the board. So plan is for a stack of two boards.
Got the layout almost done.
Top board has QFN Prop, eeprom, 3.3 v regulator, crystal, accelerometer, and prop plug port.
Bottom board has the motor driver h-bridge, transistors and resistors and soon a big capacitor.

One new thing I'm trying out is using a milled out piece of acrylic to give a hole on the back of the servo for mechanical support, turning it into a "robot servo".
IMG_1701.JPG
3024 x 4032 - 2M
Top_2b.png
586 x 651 - 77K
Bottom_2b.png
516 x 650 - 50K
IMG_1702.JPG
3024 x 4032 - 2M

Comments

  • AwesomeCronkAwesomeCronk Posts: 1,055
    Wow! With this, you can have self-debugging servos, telling you if they have power, are stalled, or whatever else, all thanks to the WS2812B LED chip!
  • PropGuy2PropGuy2 Posts: 358
    Glad to hear that you are using Eagle PCB layout and have a ShapeOKO CNC milling machine. I am using Sprint PCB Layout 6.0 (from Germany). It works very well and is super simple to do design work with. I t is very much like PCBexpress in simplicity.
    I also have a ShapeOKO machine and have modified to to mill PCBs. Basically I spent a lot of time squaring it up in all axis and making sure it was absolute FLAT with a steel baseboard and T-slot rails on the baseboard ( no cheap warp prone MDF board for me!)
    I use FlatCAM to do the Gerber to GCODE translation and it works very well. For the actual SpapeOKO software I use G-Code Sender Ver1.09... Again very stable and predictable, with good results every time.
    Lastly, there are a few tricks of the trade (PCB layout rules, cutter tool selection, feed and speed) to make everything work - and I am willing to share if you are interested.
  • RaymanRayman Posts: 13,892
    PropGuy2: You're using Shapeoko to make your PCBs? I've seen you can do that. I'm just going to use it to cut out the PCBs from the outline.
    That way, I don't have to pay for panelization… Also, I can populate a larger board with several of these on it and then cut it out...
  • jmgjmg Posts: 15,148
    edited 2019-03-03 02:02
    Rayman wrote: »
    Top board has QFN Prop, eeprom, 3.3 v regulator, crystal, accelerometer, and prop plug port.
    I think the QFN package is tagged EOL - will that be an issue ?
    You can get oscillator modules, 5MHz ~ 6.5MHz that come as small as 2.0 x 1.6mm, or even (1.54mm x 0.84mm) ! if you can work BGA packages
    If you are just going board-board, 2mm or even 1mm pin sticks can take up less area.

    Addit: for a proven oscillator (used in FLiP), Digikey stock the SIT8918BEL13-33E-5.000000 2.5x2.0mm
    - see also https://forums.parallax.com/discussion/166754/flip-here/p4 for talk around the slew options, and radiation effects. the 'BEL' is the slowest edge choice.

    Smaller and well stocked looks to be DSC6001HI1A-005.0000T.
    I see that DSC60xx family supports dual-clock choice, so it could be possible to use a DSC6023 with two MHz choices ?

    These two MHz pairs seem popular/commonly stocked, - I wonder what uses that choice ?
    DSC6023HI2A-00ADT Microchip XO 1.71V-3.63V 6.1298MHz, 6.1679MHz ±25ppm -40°C ~ 85°C 1.3mA (Typ) (1.60mm x 1.20mm) $0.90640/1k
    DSC6021CI2A-00A3T Microchip XO 1.71V-3.63V 6.1298MHz, 6.1679MHz ±25ppm -40°C ~ 85°C 1.3mA (Typ) (3.20mm x 2.50mm) $0.69010/1k
  • RaymanRayman Posts: 13,892
    edited 2019-03-03 00:44
    regarding EOL, see your comments from 2015 here:
    https://forums.parallax.com/discussion/160864/end-of-life

    and somebody pointed to Parallax's statement on this on the QFN product page: 
    Long-term supply notification from company owner: The Propeller chip is a full-custom, Parallax design in entirety, unencumbered from any third-party intellectual property obligations. Parallax endeavors to supply Propellers as long as we have demand (i.e., foreseeable future, a minimum 20 to 25 years). As an example of our commitment, Parallax still manufactures the BASIC Stamp I and it was designed over 20 years ago. For these reasons, customers should have no concern about the long-term supply of Propeller chips. If you have any concern about long-term supply please feel free to contact us for personal communication.
  • RaymanRayman Posts: 13,892
    This crystal is a gorilla in the design. I'm kinda surprised the original PIC version uses it.
    I have a tiny crystal I could use instead.
    But, as of now, I have enough space...
  • jmgjmg Posts: 15,148
    Rayman wrote: »
    regarding EOL...
    IIRC it was a comment from Ken, mainly around high MOQ's needed - maybe they have since added more customers ?

    Ah, here it is ...
    https://forums.parallax.com/discussion/comment/1292522/#Comment_1292522
    " When these QFNs are gone, they're gone. I'd be happy to be in a shortage situation with QFNs because they move very slow. We will unfortunately EOL this part. The MOQ is too high for the inventory turns we have with them.
    If somebody wants to order 12-25K at a time we'll build them; otherwise we can't afford to stock these.
    QFNs are not preferred by contract manufacturers due to a lower yield during manufacturing.
    Ken Gracey"
  • Mark_TMark_T Posts: 1,981
    Rayman wrote: »
    This crystal is a gorilla in the design. I'm kinda surprised the original PIC version uses it.
    I have a tiny crystal I could use instead.
    But, as of now, I have enough space...

    Are all of the 0.1" through-hole pads actually needed? I reckon single 4-layer board would be enough,
    perhaps a tag-connect for programming?
  • RaymanRayman Posts: 13,892
    Re. EOL: Ken also says this later in same thread: 
    As for redesigning your module to switch to QFP, don't bother. We have plenty in stock for several years.

Ken Gracey

    Five years later, there's still 3,000 in stock...
  • RaymanRayman Posts: 13,892
    edited 2019-03-03 17:14
    Smaller through-hole pads probably is a good idea.
    I don't think 4-layers would help though.
    (ok, I just ran into a problem where 4-layers would help... Can't get 5V over to WS2812, have to try it at 3.3v...)
  • TubularTubular Posts: 4,622
    The WS2812B's work pretty well at 3v3. The blue and green are a little starved, which results in a pink color cast, but that can be compensated for in software

    The new WS2812C-2020's are designed to work in 3v3 systems.

  • RaymanRayman Posts: 13,892
    Using sigma-delta with 2 external caps and 2 external resistors to read the pot position.
    Not 100% sure this is good enough...
    If not, I'll have to drop in a digital ADC chip...
Sign In or Register to comment.